This PDF file contains the front matter associated with SPIE Proceedings Volume 6913, including the Title Page, Copyright information, Table of Contents, In Memoriam Sam Dwyer, and the Conference Committee listing.

Imaging tumor angiogenesis in small animals is extremely challenging due to the size of the tumor vessels. Consequently, both dedicated small animal imaging systems and specialized intravascular contrast agents are required. The goal of this study was to investigate the use of a liposomal contrast agent for high-resolution micro-CT imaging of breast tumors in small animals. A liposomal blood pool agent encapsulating iodine with a concentration of 65.5 mg/ml was used with a Duke Center for In Vivo Microscopy (CIVM) prototype micro-computed tomography (micro-CT) system to image the R3230AC mammary carcinoma implanted in rats. The animals were injected with equivalent volume doses (0.02 ml/kg) of contrast agent. Micro-CT with the liposomal blood pool contrast agent ensured a signal difference between the blood and the muscle higher than 450 HU allowing the visualization of the tumors 3D vascular architecture in exquisite detail at 100-micron resolution. The micro-CT data correlated well with the histological examination of tumor tissue. We also studied the ability to detect vascular enhancement with limited angle based reconstruction, i.e. tomosynthesis. Tumor volumes and their regional vascular percentage were estimated. This imaging approach could be used to better understand tumor angiogenesis and be the basis for evaluating anti-angiogenic therapies.

A prototype physiologically gated micro-computed tomography (micro-CT) system based on a field emission micro-focus x-ray source has been developed for in vivo imaging of small animal models. The novel x-ray source can generate radiation with a programmable waveform that can be readily synchronized and gated with non-periodic physiological signals. The system performance is evaluated using phantoms and sacrificed and anesthetized mouse models. Prospective respiratory-gated CT images of anesthetized free-breathing mice are collected using this scanner at 100msec temporal resolution and 10 lp/mm of 10% system MTF.

During the last eight years our group has developed radial acquisitions with angular undersampling factors of several hundred that accelerate MRI in selected applications. As with all previous acceleration techniques, SNR typically falls as least as fast as the inverse square root of the undersampling factor. This limits the SNR available to support the small voxels that these methods can image over short time intervals in applications like time-resolved contrast-enhanced MR angiography (CE-MRA). Instead of processing each time interval independently, we have developed constrained reconstruction methods that exploit the significant correlation between temporal sampling points. A broad class of methods, termed HighlY Constrained Back PRojection (HYPR), generalizes this concept to other modalities and sampling dimensions.

Both mammography and ultrasound imaging have sub-optimal sensitivity for the detection of solid breast lesions. We investigate the feasibility of detecting occult solid breast lesions using ultrasound propagation velocity differences between malignant lesions and surrounding breast tissue in a phantom study using 3D ultrasound. In this technique, the breast is placed on a flat backplate and imaged using a 3D ultrasound system. The backplate serves as a reference plane in the resulting 3D image. The difference in ultrasound propagation velocity through solid lesions versus surrounding breast tissue produces a shift in the apparent elevation of the backplate in the resulting image. Visible backplate distortion may therefore indicate the presence of an occult solid breast lesion. We provide a mathematical model to predict backplate distortion as a function of lesion thickness, and validate this theory experimentally using multi-compartment agar phantoms with embedded lesions ranging from 2 mm to 16 mm in thickness. Two-dimensional backplate elevation maps were constructed, and visible backplate distortions were observed for lesions of thickness 4 mm and greater. Measured backplate distortions were in excellent agreement with our theoretical predictions for these lesions. This study suggests that ultrasound propagation velocity differences between solid lesions and surrounding tissue may be useful as an additional screening tool for detection and localization of otherwise occult solid breast lesions.

Preliminary results for a new CT scanning device with dose-reduction potential were presented at the SPIE Medical Imaging conference 2007. The new device acquires the Radon data after the X-ray beam is collimated through a special mask. This mask is combined with a new and efficient data collection geometry; thus the device has the potential of reducing the dose by a factor of two. In this work, we report the first complete proof of the idea using the same simplified mask of 197 detectors as last year, and a clinical C-arm with a flat panel detector to simulate the gantry. This addition enables the acquisition of two independent and complementary data sets for reconstruction. Moreover, this clinical set-up enables the acquisition of data for clinically relevant phantoms. Phantom data were acquired using both detector sets and were reconstructed with the robust algorithm OPED. The independent sinograms were matched to a single one, and from this a diagnostic image was reconstructed successfully. This image has improved resolution, as well as less noise and artifacts compared to each single independent reconstruction. The results obtained are highly promising, even though the current device acquires only 197 views. Dose comparisons can be carried out in the future with a more precise prototype, comparable to current clinical devices with respect to imaging performance.

The multi-prism lens (MPL) is a refractive x-ray lens consisting of two rows of prisms facing each other at an angle. Rays entering the lens at the periphery will encounter a larger number of prisms than will central ones, hence experiencing a greater refraction. The focusing effect of the MPL can be used to gather radiation from a large aperture onto a smaller detector, and accordingly to make better use of the available x-ray flux in medical x-ray imaging. Potential advantages of a better photon economy include shorter acquisition times, a reduced tube loading, or an improved resolution. Since the focusing effect is one-dimensional it matches the design of scanning systems. In this study we present the first images acquired with an MPL instead of the pre-breast slit collimator in a scanning mammography system. According to the measurements, the MPL is able to increase the flux 32% at equal resolution compared to the slit collimator, or to improve the resolution 2.4 mm^-1 at equal flux. If used with a custom-made absorption filter in a clinical set-up, the gain of flux of the MPL is expected to be at least 45%, and the corresponding improvement in resolution to be 3 mm^-1.

Neutron stimulated emission computed tomography (NSECT) is being proposed as a non-invasive technique to diagnose iron overload in humans. It uses inelastic scatter interactions between incident neutrons and iron nuclei to stimulate gamma-ray emission from iron. Tomographic detection of the emitted gamma-rays yields information about the concentration and spatial distribution of iron in the liver. Early proof-of-concept experiments have shown that NSECT has the potential to quantify clinical quantities of liver iron overload through single-position spectroscopy. However, a tomography application for patient diagnosis has never been tested. This work uses a Monte-Carlo simulation of a tomographic NSECT system to investigate the feasibility of imaging the spatial distribution of liver iron through tomography. A simulation of an NSECT system has been designed in GEANT4 and used to tomographically scan a simulated human liver phantom with high-concentration iron lesions. Images are reconstructed with the MLEM algorithm and analyzed for pixel values within iron regions to determine the statistical significance of detection. Analysis results indicate that a wet iron concentration of 3 mg/g can be detected in surrounding liver tissue with p-value ≤ 0.0001 for neutron exposure corresponding to a radiation dose of 0.72 mSv. The research performed here demonstrates that NSECT has the ability to image clinically relevant distributions of iron through tomographic scanning.

Most electronic portal imaging devices (EPIDs) developed so far use a Cu plate/phosphor screen to absorb x rays. The main problem with this approach is that the Cu plate/phosphor screen must be thin (~ 2 mm) in order to obtain a high spatial resolution, resulting in a low quantum efficiency (QE) for megavoltage (MV) × rays (typically 2-4%). In addition, the phosphor screen contains high atomic number (high-Z) materials, resulting in an over-response of the detector to low energy x rays in dosimetric verification. Our overall goal is to develop a new high QE MV x-ray detector made of a low-Z material for both geometric and dosimetric verification in radiotherapy. Our approach is based on radiation-induced light (Cherenkov radiation) in optical fibers to convert x-ray energy into light. With our approach, a thick (~ 10-30 cm) fiber-optic taper (FOT) consisting of a matrix of optical fibers aligned with the incident x rays is used to replace the thin Cu plate/phosphor screen to dramatically improve the QE. In this work, we demonstrated that the predominant light source in optical fibers under irradiation of a MV beam is indeed Cherenkov radiation, and thus validated the feasibility of using Cherenkov radiation as the primary light source in our proposed Cherenkov detector. A prototype Cherenkov detector array was also built and images were obtained.

Analytically predicting photon paths in real-high scattering-anisotropic tissues is extremely complex, due to the significant random scattering events that photons suffer as they traverse the tissue, especially at boundaries between areas with different optical properties. An statistically correct optical and anatomical model of photon trajectories inside laboratory animals will therefore improve considerably our understanding about how light diffuses within the animals, and therefore help us designing efficient experimental setups and reconstruction algorithms for fluorescence mediated tomography (FMT). Here, we present new simulations of photon propagation and fluorescence emission in anisotropic media using realistic models of laboratory animals and a Monte Carlo (MC) based approach. We compare the MC simulation results with an approximation of the solution of the diffusion equation using finite differences and discuss the different behaviour of the two methods.

The forward problem of optical bioluminescence and fluorescence tomography seeks to determine, for a given 3D source distribution, the photon density on the surface of an animal. Photon transport through tissues is commonly modeled by the diffusion equation. The challenge, then, is to accurately and efficiently solve the diffusion equation for a realistic animal geometry and heterogeneous tissue types. Fast analytical solvers are available that can be applied to arbitrary geometries but assume homogeneity of tissue optical properties and hence have limited accuracy. The finite element method (FEM) with volume tessellation allows reasonably accurate modeling of both animal geometry and tissue heterogeneity, but this approach is computationally intensive. The computational challenge is heightened when one is working with multispectral data to improve source localization and conditioning of the inverse problem. Here we present a fast forward model based on the Born approximation that falls in between these two approaches. Our model introduces tissue heterogeneity as perturbations in diffusion and absorption coefficients at rectangular grid points inside a mouse atlas. These reflect as a correction term added to the homogeneous forward model. We have tested our model by performing source localization studies first with a biolumnescence simulation setup and then with an experimental setup using a fluorescent source embedded in an inhomogeneous phantom that mimicks tissue optical properties.

In this paper, we describe a new 3D light propagation model aimed at understanding the effects of various physiological properties on subcutaneous vein imaging. In particular, we build upon the well known MCML (Monte Carlo Multi Layer) code and present a tissue model that improves upon the current state-of-the-art by: incorporating physiological variation, such as melanin concentration, fat content, and layer thickness; including veins of varying depth and diameter; using curved surfaces from real arm shapes; and modeling the vessel wall interface. We describe our model, present results from the Monte Carlo modeling, and compare these results with those obtained with other Monte Carlo methods.

In this study we apply a total variation (TV) minimization algorithm to image reconstruction in magnetic resonance imaging (MRI). This algorithm is particularly effective for underlying images that are approximately piecewise constant. While the underlying proton spin density in MRI can satisfy this condition under certain circumstances, it is often distorted by unavoidable physical factors that alter the phase of the complex image. In this work, we employ a known method of removing this slow phase variation resulting from magnetic field inhomogeneities to obtain a spin density distribution that is piecewise constant. After the phase removal, we apply the TV minimization algorithm to obtain images from 20% of the full MRI data set.

A novel method for highly-undersampled Magnetic Resonance Image (MRI) reconstruction is presented. One of the principal challenges faced in clinical MR imaging is the fundamental linear relation between net exam duration and admissible spatial resolution. Increased scan duration diminishes patient comfort while increasing the risk of susceptibility to motion artifact and limits the ability to depict many physiological events at high temporal rates. With the recent development of Compressive Sampling theory, several authors have successfully demonstrated that clinical MR images possessing a sparse representation in some transform domain can be accurately reconstructed even when sampled at rates well below the Nyquist limit by casting the recovery as a convex ℓ₁-minimization problem. While ℓ₁-based techniques offer a sizeable advantage over Nyquist-limited methods, they nonetheless require a modest degree of over-sampling above the true theoretical minimum sampling rate in order to guarantee the achievability of exact reconstruction. In this work, we present a reconstruction model based on homotopic approximation of the ℓ₀ quasi-norm and discuss the ability of this technique to reconstruct undersampled MR images at rates even lower than are achievable than with ℓ₁-minimization and arbitrarily close to the true minimum sampling rate. A semi-implicit numerical solver is presented for efficient numerical computation of the reconstruction process and several examples depicting the capability for accurate MRI reconstructions from highly-undersampled K-space data are presented.

Breathing motion causes severe artifacts in Magnetic Resonance Imaging (MRI), which are usually avoided using either breath-holding during the acquisition or gating or triggering techniques while the patient is breathing freely. In MRI with continuously moving table the patient is moved through the magnet during the data acquisition, thereby allowing for time efficient coverage of arbitrary large body regions. Due to the continuous table motion standard breathing motion compensation techniques are not applicable here. In this work a retrospective compensation technique for abdominal breathing motion in moving table MRI is proposed, which reconstructs artifact free images from data acquired during free breathing. For each slice position multiple motion consistent snapshots of arbitrary breathing states are acquired and subsequently combined consistently using a combination of rigid and deformable slice-to-volume registration. Typical breathing motion induced artifacts such as ghosting, blurring or misalignments of organs occurring if data from different breathing states is inconsistently combined are avoided. The proposed method significantly extends a previous approach which aimed only at the suppression of ghosting artifacts in MRI with continuously moving table.

Spin noise is inherent in magnetic resonance. It is caused by incomplete cancellation of spin moments when the external static magnetic field is absent or by their small but finite fluctuations when the magnetic field is applied. Spin noise is viewed as the variation of thermal equilibrium macroscopic magnetization (TEMM), and can be described statistically. For MRI, TEMM is shown to be characterized by a Binomial distribution and is well approximated by a Gaussian. Parameters of this Gaussian distribution are determined by the spin density and the ratio of population difference over the total population of spins in a unit volume of the sample. Statistics of spin noise not only confirm Bloch's prediction of spin noise in the absence of the external magnetic field, but also give a more accurate account of its behavior under various conditions. These statistics also provide a new insight into the limits of spatial resolution in magnetic resonance microscopy (MRM) and are consistent with Glover and Mansfield' corresponding conclusions based on their experiments.

Conventional active matrix flat-panel imagers (AMFPIs), employing amorphous silicon (a-Si:H) semiconductors, are based on a relatively simple pixel architecture, commonly taking the form of a single, thin-film transistor (TFT) coupled to a pixel storage capacitor. Although this semiconductor-architecture combination has led to the successful creation of x-ray imagers for many applications, a variety of significant performance limitations related to DQE, frame rate and charge trapping have also become apparent. While prospects for designing solutions to these restrictions based on a-Si:H TFTs are uncertain, progress in the development of high-quality polycrystalline silicon (poly-Si) TFTs is opening up new possibilities for large area x-ray imager design. Recently, initial prototype imagers have been developed using poly-Si TFTs in the form of 1-stage and 2-stage pixel amplifiers-ircuit architectures that can generally be referred to as active pixel sensors (APS). The insight gained from empirical evaluations of such prototypes, coupled with theoretical studies, can inspire increasingly sophisticated APS architectures that overcome the limitations, while preserving the advantages, of conventional AMFPIs. In this paper, cascaded systems analysis and circuit simulation are used to explore potential performance improvements enabled by APS architectures based on poly-Si TFTs. These studies suggest that it is possible to achieve significant improvements in DQE at low exposures or very small pixel sizes, higher maximum frame rates, and reduced charge trapping effects through implementation of such architectures.

We describe a new type of a sensitive semiconductor photodetector that could be used in medical imaging applications. The photodetector, based on the mechanism of discrete amplification, has performance parameters comparable to, and for some applications exceeding, those of the vacuum photomultiplier tubes. High amplification gain achieved at very low levels of excess noise is accompanied by the fast speed and high dynamic range of the photodetector. Comparison of the technology with classic arrays of Geiger-mode APD arrays is also performed.

The solid-state x-ray image intensifier (SSXII) is an EMCCD-based x-ray detector designed to satisfy an increasing need for high-resolution real-time images, while offering significant improvements over current flat panel detectors (FPDs) and x-ray image intensifiers (XIIs). FPDs are replacing XIIs because they reduce/eliminate veiling glare, pincushion or s-shaped distortions and are physically flat. However, FPDs suffer from excessive lag and ghosting and their performance has been disappointing for low-exposure-per-frame procedures due to excessive instrumentation-noise. XIIs and FPDs both have limited resolution capabilities of ~3 cycles/mm. To overcome these limitations a prototype SSXII module has been developed, consisting of a 1k x 1k, 8 μm pixel EMCCD with a fiber-optic input window, which views a 350 μm thick CsI(Tl) phosphor via a 4:1 magnifying fiber-optic-taper (FOT). Arrays of such modules will provide a larger field-of- view. Detector MTF, DQE, and instrumentation-noise equivalent exposure (INEE) were measured to evaluate the SSXIIs performance using a standard x-ray spectrum (IEC RQA5), allowing for comparison with current state-of-the-art detectors. The MTF was 0.20 at 3 cycles/mm, comparable to standard detectors, and better than 0.05 up to 7 cycles/mm, well beyond current capabilities. DQE curves indicate no degradation from high-angiographic to low-fluoroscopic exposures (< 2% deviation in overall DQE from 1.3 mR to 2.7 μR), demonstrating negligible instrumentation-noise, even with low input signal intensities. An INEE of < 0.2 μR was measured for the highest-resolution mode (32 μm effective pixel size). Comparison images between detector technologies qualitatively demonstrate these improved imaging capabilities provided by the SSXII.

Digital flat panel a-Si x-ray detectors can exhibit image lag of several percent. The image lag can limit the temporal resolution of the detector, and introduce artifacts into CT reconstructions. It is believed that the majority of image lag is due to defect states, or traps, in the a-Si layer. Software methods to characterize and correct for the image lag exist, but they may make assumptions such as the system behaves in a linear time-invariant manner. The proposed method of reducing lag is a hardware solution that makes few additional hardware changes. For pulsed irradiation, the proposed method inserts a new stage in between the readout of the detector and the data collection stages. During this stage the photodiode is operated in a forward bias mode, which fills the defect states with charge. Parameters of importance are current per diode and current duration, which were investigated under light illumination by the following design parameters: 1.) forward bias voltage across the photodiode and TFT switch, 2.) number of rows simultaneously forward biased, and 3.) duration of the forward bias current. From measurements, it appears that good design criteria for the particular imager used are 8 or fewer active rows, 2.9V (or greater) forward bias voltage, and a row frequency of 100 kHz or less. Overall, the forward bias method has been found to reduce first frame lag by as much as 95%. The panel was also tested under x-ray irradiation. Image lag improved (94% reduction), but the temporal response of the scintillator became evident in the turn-on step response.

An indirect flat-imager with programmable avalanche gain and field emitter array (FEA) readout is being investigated for low-dose x-ray imaging with high resolution. It is made by optically coupling a structured x-ray scintillator CsI (Tl) to an amorphous selenium (a-Se) avalanche photoconductor called HARP (high-gain avalanche rushing photoconductor). The charge image created by HARP is read out by electron beams generated by the FEA. The proposed detector is called SAPHIRE (Scintillator Avalanche Photoconductor with HIgh Resolution Emitter readout). The avalanche gain of HARP depends on both a-Se thickness and applied electric field E_Se. At E_Se of > 80 V/μm, the avalanche gain can enhance the signal at low dose (e.g. fluoroscopy) and make the detector x-ray quantum noise limited down to a single x-ray photon. At high exposure (e.g. radiography), the avalanche gain can be turned off by decreasing E_Se to < 70 V/μm. In this paper the imaging characteristics of the FEA readout method, including the spatial resolution and noise, were investigated experimentally using a prototype optical HARP-FEA image sensor. The potential x-ray imaging performance of SAPHIRE, especially the aspect of programmable gain to ensure wide dynamic range and x-ray quantum noise limited performance at the lowest exposure in fluoroscopy, was investigated.

The output response characteristics of an X-ray photon counting detector are measured experimentally and simulated using a Monte Carlo method in order to quantify the loss of statistical information due to pile-up. The analysis is applied to idealize counting detector models, but is adaptable to realistic event processing that is not amenable to analytic solution. In particular, the detective quantum efficiency (DQE) is calculated as a function of flux rate and shown to have an intermediate zero for the paralyzable case at the maximum periodic rate. The progressive degradation of the spectral response as a function of increasing flux rate is also modeled. Analogous metrics to DQE are defined in regards to the detector's ability to resolve atomic number and enhance image contrast based on atomic number differentiation. Analytic solutions are provided for the output and linearized response statistics and these interpolate well across the Monte Carlo and experimental results.

We developed an efficient, depth- and energy-dependent Monte Carlo model for columnar CsI detectors. The optical photon, electron/positron Monte Carlo package MANTIS developed by our group, was used to generate optical photon response and collection efficiency as a function of the x-ray/electron interaction depth for a realistic scintillator geometry. The detector geometry we used for the simulations was reported in the past and is based on a 500 μm thick columnar CsI scintilator. The resulting depth-dependent optical photon responses were fit to a parametrized Gaussian mixture model. The model parameters were the depth-dependent radial shift of the response peak, the depth dependent widths of the Gaussians, and the depth-dependent magnitude of the Gaussians in the mixture. The depth-dependent optical spread has a maximum spatial shift of 53 μm. The optical collection efficiency at the photo-diode layer followed a power law varying from 90% for interactions at the scintillator exit surface to 20% for interactions at the detector entrance. The responses were consequently incorporated into penMesh, a PENELOPE based Monte Carlo x-ray, electron/positron transport simulation package for generating clinically realistic images of triangular mesh phantoms. The resulting detector responses from this empirical model were compared against the full x-ray/electron/optical photon simulation using the package MANTIS, showing good agreement. The simulation speed, using the optical transport model in penMesh, increases by two orders of magnitude compared to MANTIS.

A hybrid pixel detector based on the concept of simultaneous charge integration and photon counting will be presented. The second generation of a counting and integrating X-ray prototype CMOS chip (CIX) has been operated with different direct converting sensor materials (CdZnTe and CdTe) bump bonded to its 8x8 pixel matrix. Photon counting devices give excellent results for low to medium X-ray fluxes but saturate at high rates while charge integration allows the detection of very high fluxes but is limited at low rates by the finite signal to noise ratio. The combination of both signal processing concepts therefore extends the resolvable dynamic range of the X-ray detector. In addition, for a large region of the dynamic range, where counter and integrator operate simultaneously, the mean energy of the detected X-ray spectrum can be calculated. This spectral information can be used to enhance the contrast of the X-ray image. The advantages of the counting and integrating signal processing concept and the performance of the imaging system will be reviewed. The properties of the system with respect to dynamic range and sensor response will be discussed and examples of imaging with additional spectral information will be presented.

Despite its obvious advantages, well known CsI:Tl scintillator has two characteristic properties that undermine its use in clinical and high speed imaging: the presence of an afterglow component in its scintillation decay, and a hysteresis effect that causes drift in the scintillation yield after exposure to high radiation doses. We have previously reported that the addition of a second dopant, Sm²⁺, to the CsI:Tl crystals, significantly suppresses both afterglow and hysteresis. Here we report on the fabrication and characterization of the Sm co-doped CsI:Tl microcolumnar films to examine if these properties are preserved in films as well. Our preliminary data suggests that the Sm co-doped CsI:Tl films significantly improve temporal response relative to their CsI:Tl counterpart, and that the newly developed films demonstrate excellent spatial resolution. Various aspects of these effects and their consequences for imaging performance are discussed in this paper.

Sophisticated radiotherapy techniques lead to more conformal dose distributions but increase treatment complexity. Image guidance allows for varying degrees of accuracy in patient set-up. However, the consequences of inaccurate set-up and/or patient motion during treatment become more serious when treatment doses are increased and treatment margins are decreased. Thus, the need to know if the dose has been delivered as planned has driven the development of plastic scintillation detector systems for accurate measurements in real time with high spatial resolution. We have developed a clinical prototype comprising 29 scintillating fiber detectors 1 mm in diameter and 2 mm in length. The detectors are coupled to clear optical fibers that collect the scintillation photons and transport them to a CCD for detection. Open field profiles and depth-dose profiles in water-equivalent phantoms were compared to ionization chamber measurements in water. The maximum relative in-field difference was 1.6%. With a standard deviation for in-field measurements smaller than 1%, this prototype array was found to be accurate, precise and practical. Monte Carlo simulations were also used to evaluate the response of the scintillation detector to proton beams and to optimize the light collection efficiency. The Monte Carlo code Geant4 was used to simulate dose deposition, the production of scintillation photons and the propagation of those photons inside the scintillation detector. Further development of the system will allow thousands of measurement points distributed in a three-dimensional volume per single irradiation, therefore producing a rapid evaluation of complex dose distributions emanating from these new complex treatment modalities.

Permanent breast seed implant (PBSI) brachytherapy technique was recently introduced as an alternative to high dose rate (HDR) brachytherapy and involves the permanent implantation of radioactive ¹⁰³Palladium seeds into the surgical cavity of the breast for cancer treatment. To enable accurate seed implantation, this research introduces a gamma camera based on a hybrid amorphous selenium detector and CMOS readout pixel architecture for real-time imaging of ¹⁰³Palladium seeds during the PBSI procedure. A prototype chip was designed and fabricated in 0.18-μm n-well CMOS process. We present the experimental results obtained from this integrated photon counting readout pixel.

This work aims at defining an information-theoretic quality assessment technique for cardiovascular X-ray images, using a full-reference scheme (relying on averaging a sequence to obtain a noiseless reference). With the growth of advanced signal processing in medical imaging, such an approach will enable objective comparisons of the quality of processed images. A concept for describing the quality of an image is to express it in terms of its information capacity. Shannon has derived this capacity for noisy channel coding. However, for X-ray images, the noise is signal-dependent and non-additive, so that Shannon's theorem is not directly applicable. To overcome this complication, we exploit the fact that any invertible mapping on a signal does not change its information content. We show that it is possible to transform the images in such a way that the Shannon theorem can be applied. A general method for calculating such a transformation is used, given a known relation between signal mean and noise standard deviation. After making the noise signal-independent, it is possible to assess the information content of an image and to calculate an overall quality metric (e.g. information capacity) which includes the effects of sharpness, contrast and noise. We have applied this method on phantom images under different acquisition conditions and computed the information capacity for those images. We aim to show that the results of this assessment are consistent with variations in noise, contrast and sharpness, introduced by system settings and image processing.

The objective performance evaluation metrics, termed Generalized Modulation Transfer Function (GMTF), Generalized Noise Power Spectrum (GNPS), Generalized Noise Equivalent Quanta (GNEQ), and Generalized Detective Quantum Efficiency (GDQE), have been developed to assess total imaging-system performance by including the effects of geometric unsharpness due to the finite size of the focal spot and scattered radiation in addition to the detector properties. These metrics were used to evaluate the performance of the HSMAF, a custom-built, high-resolution, real-time-acquisition detector with 35-μm pixels, in simulated neurovascular angiographic conditions using a uniform head-equivalent phantom. The HSMAF consists of a 300-μm-thick CsI(Tl) scintillator coupled to a 4 cm diameter, variable-gain, Gen2 light image intensifier with dual-stage microchannel plate, followed by direct fiber-optic coupling to a 30-fps CCD camera, and is capable of both fluoroscopy and angiography. Effects of focal-spot size, geometric magnification, irradiation field-of-view, and air-gap between the phantom and the detector were evaluated. The resulting plots of GMTF and GDQE showed that geometric blurring is the more dominant image degradation factor at high spatial frequencies, whereas scatter dominates at low spatial frequencies. For the standard image-geometry and scatter conditions used here, the HSMAF maintains substantial system imaging capabilities (GDQE>5%) at frequencies above 4 cycles/mm where conventional detectors cannot operate. The loss in image SNR due to scatter or focal-spot unsharpness could be compensated by increasing the exposure by a factor of 2 to 3. This generalized evaluation method may be used to more realistically evaluate and compare total system performance leading to improved system designs.

This paper presents an information-entropy based metric for combined evaluation of resolution and noise properties of radiological images. The metric is expressed by the amount of transmitted information (TI). It is a measure of how much information that one image contains about an object or an input. Merits of the proposed method are its simplicity of computation and the experimented setup. A computer-simulated step wedge was used for simulation study on the relationship of TI and the degree of blur as well as the noise. Three acrylic step wedges were also manufactured and used as test sample objects for experiments. Two imaging plates for computed radiography were employed as information detectors to record X-ray intensities. We investigated the effects of noise and resolution degradation on the amount of TI by varying exposure levels. Simulation and experimental results show that the TI value varies when the noise level or the degree of blur is changed. To validate the reasoning and usefulness of the proposed metric, we also calculated and compared the modulation transfer functions and noise power spectra for the employed imaging plates. Results show that the TI has close correlation with both image noise and image blurring, and it may offer the potential to become a simple and generally applicable measure for quality evaluation of medical images.

A perfusion phantom with unique features and a wide variety of applications in MR and other imaging modalities is presented. The phantom is especially suited for tissue perfusion simulation with diffusible and non-diffusible MR tracers. A network of micro-channels in the scale of actual capillaries replicates the blood flow in tissues. Using microfabrication techniques, networks with any desired pattern can be generated. Since the geometry of networks is known, flow rate, delay, dispersion and other fluid parameters can be exactly calculated using finite elements numerical methods. These calculated results can be used to investigate the accuracy of experimental measurements and the precision of mathematical models.

For optimisation in diagnostic medical imaging it is important to consider the relation between diagnostic image quality and patient dose. In the past, schematic representations of the human body were commonly used for dosimetric simulations together with Monte Carlo codes. During the last two decades, voxel models were introduced as an improvement to these body models. Studies performed by various research groups have shown that the more realistic organ topology of voxel models constructed from medical image data of real persons has an impact on calculated doses for external as well as internal exposures. As a consequence of these findings, the ICRP decided to use voxel models for the forthcoming update of organ dose conversion coefficients. These voxel models should be representative of an average population, i.e. they should resemble the ICRP reference anatomical data with respect to their external dimensions and their organ masses. To meet the ICRP requirements, our group at the Helmholtz Zentrum München (formerly known as GSF-National Research Center for Environment and Health) constructed voxel models of a male and female adult, based on the voxel models of two individuals whose body height and weight resembled those of the male and female ICRP reference adult. The organ masses of both models were adjusted to the ICRP reference anatomical data, without spoiling their realistic anatomy. The paper describes the method used for this process and the resulting voxel models.

Dual-energy contrast-enhanced breast tomosynthesis has been proposed as a technique to improve the detection of early-stage cancer in young, high-risk women. This study focused on optimizing this technique using computer simulations. The computer simulation used analytical calculations to optimize the signal difference to noise ratio (SdNR) of resulting images from such a technique at constant dose. The optimization included the optimal radiographic technique, optimal distribution of dose between the two single-energy projection images, and the optimal weighting factor for the dual energy subtraction. Importantly, the SdNR included both anatomical and quantum noise sources, as dual energy imaging reduces anatomical noise at the expense of increases in quantum noise. Assuming a tungsten anode, the maximum SdNR at constant dose was achieved for a high energy beam at 49 kVp with 92.5 μm copper filtration and a low energy beam at 49 kVp with 95 μm tin filtration. These analytical calculations were followed by Monte Carlo simulations that included the effects of scattered radiation and detector properties. Finally, the feasibility of this technique was tested in a small animal imaging experiment using a novel iodinated liposomal contrast agent. The results illustrated the utility of dual energy imaging and determined the optimal acquisition parameters for this technique. This work was supported in part by grants from the Komen Foundation (PDF55806), the Cancer Research and Prevention Foundation, and the NIH (NCI R21 CA124584-01). CIVM is a NCRR/NCI National Resource under P41-05959/U24-CA092656.

Dual-Energy Contrast Enhanced Digital Breast Tomosynthesis (DE CEDBT) is a promising technique for breast cancer detection, which combines the strengths of functional and 3D imaging. In the present study, we first focused on the optimization of the acquisition parameters for the low and high-energy projections, which leads to a trade-off between image quality in the recombined slices and the Average Glandular Dose (AGD) delivered to the patient. Optimized parameters were found and experimentally validated on phantom images. Then, we addressed the problem of iodine quantification in the recombined slices. In DE CEDBT, iodine quantification is limited by the z-resolution, due to the restricted angle acquisition inherent to tomosynthesis. We evaluated the lesion thickness above which determination of iodine volumetric concentration is possible. For lesions below this thickness, estimation of iodine concentration is possible if a priori information or a model on the shape of the lesion is available. Iodine quantification for lesions located near the breast boundary is also challenging, due to scatter border effects and variation of the breast thickness in this region. A scatter correction algorithm based on a deconvolution scheme and a thickness compensation algorithm were applied on the low and high-energy projections. Corrected images showed a more accurate quantification of iodine.

Dual-energy subtraction imaging (DES) is a method to improve the detectability of contrast agents over a lumpy background. Two images, acquired at x-ray energies above and below an absorption edge of the agent material, are logarithmically subtracted, resulting in suppression of the signal from the tissue background and a relative enhancement of the signal from the agent. Although promising, DES is still not widely used in clinical practice. One reason may be the need for two distinctly separated x-ray spectra that are still close to the absorption edge, realized through dual exposures which may introduce motion unsharpness. In this study, electronic spectrum-splitting with a silicon-strip detector is theoretically and experimentally investigated for a mammography model with iodinated contrast agent. Comparisons are made to absorption imaging and a near-ideal detector using a signal-to-noise ratio that includes both statistical and structural noise. Similar to previous studies, heavy absorption filtration was needed to narrow the spectra at the expense of a large reduction in x-ray flux. Therefore, potential improvements using a chromatic multi-prism x-ray lens (MPL) for filtering were evaluated theoretically. The MPL offers a narrow tunable spectrum, and we show that the image quality can be improved compared to conventional filtering methods.

The relationship between theoretical descriptions of imaging performance (Fourier-based cascaded systems analysis) and the performance of real human observers was investigated for various detection and discrimination tasks. Dual-energy (DE) imaging provided a useful basis for investigating this relationship, because it presents a host of acquisition and processing parameters that can significantly affect signal and noise transfer characteristics and, correspondingly, human observer performance. The detectability index was computed theoretically using: 1) cascaded systems analysis of the modulation transfer function (MTF), and noise-power spectrum (NPS) for DE imaging; 2) a Fourier description of imaging task; and 3.) integration of MTF, NPS, and task function according to various observer models, including Fisher-Hotelling and non-prewhitening with and without an eye filter and internal noise. Three idealized tasks were considered: sphere detection, shape discrimination (sphere vs. disk), and texture discrimination (uniform vs. textured disk). Using images of phantoms acquired on a prototype DE imaging system, human observer performance was assessed in multiple-alternative forced choice (MAFC) tests, giving an estimate of area under the ROC curve (A_Ζ). The degree to which the theoretical detectability index correlated with human observer performance was investigated, and results agreed well over a broad range of imaging conditions, depending on the choice of observer model. Results demonstrated that optimal DE image acquisition and decomposition parameters depend significantly on the imaging task. These studies provide important initial validation that the detectability index derived theoretically by Fourier-based cascaded systems analysis correlates well with actual human observer performance and represents a meaningful metric for system optimization.

Dual-energy (DE) X-ray computed tomography (CT) has been proposed as an useful tool in various applications. One promising application is DECT with low radiation doses used for attenuation correction in positron emission tomography (PET). In low-dose DECT, conventional methods for sinogram decomposition have been based on logarithmic transformations and ignored noise properties, leading to very noisy component sinogram estimates. In this paper, we propose two novel sinogram restoration methods that are statistically motivated; penalized weighted least square (PWLS) and penalized likelihood (PL), producing less noisy component sinogram estimates for low-dose DECT than the conventional approaches. The restored component sinograms can improve attenuation correction, thus allowing better image quality in PET. Experiments with a digital phantom indicate that the proposed methods produce less noisy sinograms, reconstructed images, and attenuation correction factors (ACF) than the conventional one, showing promise for CT-based attenuation correction in emission tomography.

Fast kV-switching is a dual energy acquisition technique in CT in which alternating views correspond to the low and high tube voltages. Its high temporal resolution and its suitability to a variety of source trajectories make it an attractive option for dual energy data acquisition. Its disadvantages include a one view mis-registration between the data for high and low voltages, the potential for poor spectrum separation because the fast kV-switching waveform may be more like a sine wave than the desired square wave, and the higher noise in the low voltage data because of the technical difficulty of swinging the tube current to counter the loss of x-ray production efficiency and loss of penetration at lower tube voltages. These issues are investigated with a recently developed pre-reconstruction decomposition method by the authors. Results include that symmetric view matching eliminates streaks from the view mis-registration, a sinusoidal waveform swinging between 80 and 135 kV gives sufficient spectrum separation, and that contrast-to-noise for the simulated imaging task maximizes at monochromatic energy of 75 keV.

The goal of this work is to create a detailed three-dimensional (3D) digital breast phantom based on empirical data and to incorporate it into the four-dimensional (4D) NCAT phantom, a computerized model of the human anatomy widely used in imaging research. Twenty sets of high-resolution breast CT data were used to create anatomically diverse models. The datasets were segmented using techniques developed in our laboratory and the breast structures will be defined using a combination of non-uniform rational b-splines (NURBS) and subdivision surfaces (SD). Imaging data from various modalities (x-ray and nuclear medicine) were simulated to demonstrate the utility of the new breast phantoms. As a proof of concept, a simple compression technique was used to deform the breast models while maintaining a constant volume to simulate modalities (mammography and tomosynthesis) that involve compression. Initial studies using one CT dataset indicate that the simulated breast phantom is capable of providing a realistic and flexible representation of breast tissue and can be used with different acquisition methods to test varying imaging parameters such as dose, resolution, and patient motion. The final model will have a more accurate depiction of the internal breast structures and will be scaleable in terms of size and density. Also, more realistic finite-element techniques will be used to simulate compression. With the ability to simulate realistic, predictive patient imaging data, we believe the phantom will provide a vital tool to investigate current and emerging breast imaging methods and techniques.

Dedicated x-ray computed tomography (CT) of the breast using a cone-beam flat-panel detector system is a modality under investigation by a number of research teams. As previously reported, we have fabricated a prototype, bench-top flat-panel CT breast imaging (CTBI) system and developed computer simulation software to model such a system. We are developing a methodology to use high resolution, low noise CT reconstructions of fresh mastectomy specimens for generating an ensemble of 3D digital breast phantoms that realistically model 3D compressed and uncompressed breast anatomy. These breast models can be used to simulate realistic projection data for both breast tomosynthesis (BT) and CT systems thereby providing a powerful evaluation and optimization mechanism.

The use of mammography techniques for the screening and diagnosis of breast cancers has been limited by the overlapping of cancer symptoms with normal tissue structures. To overcome this problem, two methods have been developed and actively investigated recently: digital tomosynthesis mammography and cone beam breast CT. Comparison study with these three techniques will be helpful to understand their difference and further might be supervise the direction of breast imaging. This paper describes and discusses about a technique using a general-purpose PC cluster to develop a parallel computer simulation model to simulate mammograms and tomosynthesis imaging with cone beam CT images of a mastectomy breast specimen. The breast model used in simulating mammography and tomosynthesis was developed by re-scaling the CT numbers of cone beam CT images from 80kVp to 20 kev. The compression of breast was simulated by deformation of the breast model. Re-projection software with parallel computation was developed and used to compute projection images of this simulated compressed breast for a stationary detector and a linearly shifted x-ray source. The resulting images were then used to reconstruct tomosynthesis mammograms using shift-and-add algorithms. It was found that MCs in cone beam CT images were not visible in regular mammograms but faintly visible in tomosynthesis images. The scatter signal and noise property needs to be simulated and incorporated in the future.

Breast density is an independent factor of breast cancer risk. In mammograms breast density is quantitatively measured as percent density (PD), the percentage of dense (non-fatty) tissue. To date, clinical estimates of PD have varied significantly, in part due to the projective nature of mammography. Digital breast tomosynthesis (DBT) is a 3D imaging modality in which cross-sectional images are reconstructed from a small number of projections acquired at different x-ray tube angles. Preliminary studies suggest that DBT is superior to mammography in tissue visualization, since superimposed anatomical structures present in mammograms are filtered out. We hypothesize that DBT could also provide a more accurate breast density estimation. In this paper, we propose to estimate PD from reconstructed DBT images using a semi-automated thresholding technique. Preprocessing is performed to exclude the image background and the area of the pectoral muscle. Threshold values are selected manually from a small number of reconstructed slices; a combination of these thresholds is applied to each slice throughout the entire reconstructed DBT volume. The proposed method was validated using images of women with recently detected abnormalities or with biopsy-proven cancers; only contralateral breasts were analyzed. The Pearson correlation and kappa coefficients between the breast density estimates from DBT and the corresponding digital mammogram indicate moderate agreement between the two modalities, comparable with our previous results from 2D DBT projections. Percent density appears to be a robust measure for breast density assessment in both 2D and 3D x-ray breast imaging modalities using thresholding.

The purpose of this study is to evaluate the recently proposed variable dose (VD) acquisition scheme that has been hypothesized to overcome the limitations of microcalcification detection in breast tomosynthesis. In this acquisition methodology, approximately half of the total dose is used for one central projection. This central projection view is similar to a conventional mammogram and used to detect microcalcifications. The other half of the total dose is split among the rest of the projection views. These variable dose projection data are then reconstructed and the 3D slices are used for detection of masses. This novel acquisition methodology can potentially overcome the current limitations with microcalcification detection in breast tomosynthesis (BT) and may result in faster and more accurate detection of both microcalcifications and masses. Having access to both a conventional mammogram (i.e., the central projection) and tomosynthesis slices would also act as a bridge for radiologists who are used to viewing single projection images. In the current study, a comparison of microcalcification detection accuracy obtained using VD and conventional BT was conducted. A realistic computer simulation was used to model the realistic noise and blur encountered in BT systems. The simulation used a compressed breast phantom, modeled using CT images of compressed mastectomy specimens. Localization receiver operating characteristic (LROC) analysis was performed for detecting microcalcifications of size ranging from 147 microns to 178 microns. The results suggested higher microcalcification detection and localization accuracy using the VD technique. The complete study will also consist of evaluating detection of masses for the two strategies.

A stationary digital breast tomosynthesis (DBT) system using a carbon nanotube based multi-beam field emission x-ray (MBFEX) source has been designed. The purpose is to investigate the feasibility of reducing the total imaging time, simplifying the system design, and potentially improving the image quality comparing to the conventional DBT scanners. The MBFEX source consists of 25 individually programmable x-ray pixels which are evenly angular spaced covering a 48° field of view. The device acquires the projection images by electronically switching on and off the individual x-ray pixels without mechanical motion of either the x-ray source or the detector. The designs of the x-ray source and the imaging system are presented. Some preliminary results are discussed.

We studied the use of the mammography contrast detail phantom (CDMAM) with tomosynthesis to evaluate the performance of our system as well as to explore the application of CDMAM in 3D breast imaging. The system was Hologic's 1st generation tomosynthesis machine. CDMAM phantom plus PMMA slabs were imaged at 3 cm, 5 cm, 7 cm, and 9 cm PMMA-equivalent thickness with 11 projections per scan and the scan angle selected from 0, 15 and 28 degrees. CDMAM images were reconstructed using the back projection method, and were scored with the CDCOM automatic analysis program. The threshold thickness of each disk size was obtained with psychometric curve fitting. We first studied errors and variability associated with the results when different numbers of images were used in contrast detail analysis, then studied factors that affected CDMAM results in tomosynthesis, including the x-ray dose, the scan angle, the in-plane reconstruction pixel size, the slice-to-slice step size, the location of the CDMAM inside the PMMA slabs, and the scatter effect. This paper will present results of CDMAM performance of our tomosynthesis system, as well as their dependence on the various factors, and the comparison with 2D mammography. Additionally we will discuss the novel processing and analysis methods developed during this study, and make proposals to modify the CDMAM phantom and the CDCOM analysis program to optimize the method for 3D tomosynthesis.

European Guidelines for quality control in digital mammography specify minimum and achievable standards of image quality in terms of threshold contrast, based on readings of images of the CDMAM test object by human observers. However this is time-consuming and has large inter- and intra-observer error. To overcome these problems a software program (CDCOM) is available to automatically read CDMAM images. After some further analysis the automated measurements can be used to predict the threshold contrast for a typical observer. The results of threshold contrast determination by human observers at three different centres were compared against automated readings. These data provide a means of predicting average human performance using the automated reading software. The coefficient of variation in automatically determined threshold gold thickness was about 4% for detail sizes from 0.2 to 1.0mm when 8 images were analysed. The coefficient of variation was about 10% at a detail size of 0.1mm. Using larger numbers of images improved reproducibility for all detail sizes. A change in phantom design could greatly improve reproducibility for the smallest detail sizes. Greater consistency of phantom construction would also be desirable as one of the four phantoms tested was significantly different from the other three. Despite some limitations automated reading of CDMAM images can provide a reproducible means of assessing digital mammography systems against European Guidelines.

Preliminary evidence has suggested that computerized tomographic (CT) imaging of the breast using a cone-beam, flat-panel detector system dedicated solely to breast imaging has potential for improving detection and diagnosis of early-stage breast cancer. Hypothetically, a powerful mechanism for assisting in early stage breast cancer detection from annual screening breast CT studies would be to examine temporal changes in the breast from year-to-year. We hypothesize that 3D image registration could be used to automatically register breast CT volumes scanned at different times (e.g., yearly screening exams). This would allow radiologists to quickly visualize small changes in the breast that have developed during the period since the last screening CT scan, and use this information to improve the diagnostic accuracy of early-stage breast cancer detection. To test our hypothesis, fresh mastectomy specimens were imaged with a flat-panel CT system at different time points, after moving the specimen to emulate the re-positioning motion of the breast between yearly screening exams. Synthetic tumors were then digitally inserted into the second CT scan at a clinically realistic location (to emulate tumor growth from year-to-year). An affine and a spline-based 3D image registration algorithm was implemented and applied to the CT reconstructions of the specimens acquired at different times. Subtraction of registered image volumes was then performed to better analyze temporal change. Results from this study suggests that temporal change analysis in 3D breast CT can potentially be a powerful tool in improving the visualization of small lesion growth.

The feasibility of using the dual-resolution cone beam breast CT technique to obtain high-resolution images inside a selected volume-of-interest (VOI). The spatial resolution improvement, dose saving and scatter reduction with this technique are studied and demonstrated with simulations. With the dual-resolution cone beam CT technique, the breast is first scanned with a low resolution detector at a lower exposure level. A selected volume-of-interest (VOI) in the breast is then scanned with a small field, high-resolution detector at a higher exposure level. The two image sets are then combined together to reconstruct high-resolution 3-D images for the VOI. The spatial resolution that can be achieved was estimated by obtaining 3-D images of point objects and use them to compute the MTFs for evaluation as a function of the geometric magnification, detector blurring function and focal spot size. Monte Carlo simulation based on the Geant4 package was used to estimate the degree of dose saving and scatter reduction for a cylinder shaped breast phantom. The VOI images generated with the dual-resolution cone beam CT technique demonstrated the same visibility of micro-calcifications as those generated with the full-breast, high resolution image acquisition. It has been shown that the spatial resolution can be increased by factor of 1.2 with smaller focal spot size and larger magnification. With the exposure level outside the VOI reduced by a factor of 15, scatter components can be reduced by a factor of 5.5 or greater in and outside the VOI. Dose can be reduced by a factor of 5.5 inside the VOI and up to 20 outside the VOI. We have demonstrated that high spatial resolution inside the VOI may be achieved with a high-resolution detector (e.g. CCD/CsI), reduced focal spot size (0.3 mm), and optimized geometrical magnification (e.g.1.63). Exposure reduction outside the VOI has been shown to reduce the scatter components in the high-resolution projection image data for the VOI. This has also led to significantly lowered doses inside and outside the VOI.

Image guidance during cardiac interventional procedures (IP) using cardiac C-arm CT systems is desirable for many procedures. Applying the concept of retrospective electrocardiogram gating (ECG) to the acquisition of multiple, ECG-triggered rotational acquisitions using a C-arm system allows the 3D+t reconstruction of the heart. The process of retrospective gating is a crucial component of 3-D reconstruction. The gold-standard in gating is still ECG based. However, the ECG signal does not directly reflect the mechanical situation of the heart. Therefore an alternative gating method, based on the acquired projection data is required. Our goal is to provide an image-based gating (IBG) method without ECG such that already acquired projection data from a multi-sweep acquisition can still be used for reconstruction. We formulate the gating problem as a shortest-path optimization problem. All acquired projection images build a directed graph and the path costs are defined by projection image similarities that are based on image metrics to measure the heart phase similarity. The optimization is additionally regularized to prefer solutions where the path segment of consecutive selected projections acquired along a particular forward or backward C-arm sweep is short. This regularization depends on an estimated average heart rate that is also estimated using an image-based method. First promising results using in-vivo data are presented and compared to standard ECG gating. We conclude that the presented IBG method provides a reliable gating.

Since the advent of multi-slice CT, helical scan has played an increasingly important role in cardiac imaging. With the availability of diagnostic volumetric CT, step-and-shoot scan has been becoming popular recently. Step-and-shoot scan decouples patient table motion from heart beating, and thus the temporal window for data acquisition and image reconstruction can be optimized, resulting in significantly reduced radiation dose, improved tolerance to heart beat rate variation and inter-cycle cardiac motion inconsistency. Multi-sector data acquisition and image reconstruction have been utilized in helical cardiac imaging to improve temporal resolution, but suffers from the coupling of heart beating and patient table motion. Recognizing the clinical demands, the multi-sector data acquisition scheme for step-and-shoot scan is investigated in this paper. The most outstanding feature of the multi-sector data acquisition combined with the stepand- shoot scan is the decoupling of patient table proceeding from heart beating, which offers the opportunities of employing prospective ECG-gating to improve dose efficiency and fine adjusting cardiac imaging phase to suppress artifacts caused by inter-cycle cardiac motion inconsistency. The improvement in temporal resolution and the resultant suppression of motion artifacts are evaluated via motion phantoms driven by artificial ECG signals. Both theoretical analysis and experimental evaluation show promising results for multi-sector data acquisition scheme to be employed with the step-and-shoot scan. With the ever-increasing gantry rotation speed and detector longitudinal coverage in stateof- the-art VCT scanners, it is expected that the step-and-shoot scan with multi-sector data acquisition scheme would play an increasingly important role in cardiac imaging using diagnostic VCT scanners.

The Scanning-Beam Digital X-ray (SBDX) system performs rapid scanning of a narrow x-ray beam using an electronically scanned focal spot and inverse beam geometry. SBDX's ability to perform real-time multi-plane tomosynthesis with high dose efficiency is well-suited to interventional procedures such as left atrial ablation, where precise knowledge of catheter positioning is desired and imaging times are long. We describe and evaluate techniques for frame-by-frame 3D localization of multiple catheter electrodes from the stacks of tomosynthetic images generated by SBDX. The localization algorithms operate on gradient-filtered versions of the tomosynthetic planes. Small high contrast objects are identified by thresholding the stack of images and applying connected component analysis. The 3D coordinate of each object is the center-of-mass of each connected component. Simulated scans of phantoms containing 1-mm platinum spheres were used to evaluate localization performance with the SBDX prototype (5.5 × 5.5 cm detector, 3° tomographic angle) and a with new SBDX detector under design (10-cm wide × 6 cm, 6° × 3°). Z-coordinate error with the SBDX prototype was -0.6 +/- 0.7 mm (mean+/-standard deviation) with 28 cm acrylic, 24.3 kWp source operation, and 12-mm plane spacing. Localization improved to -0.3 +/- 0.3 mm using the wider SBDX detector and a 3-mm plane spacing. The effects of tomographic angle, plane-to-plane spacing, and object velocity are evaluated, and a simulation demonstrating ablation catheter localization within a real anatomic background is presented. Results indicate that SBDX is capable of precise real-time 3D tracking of high contrast objects.

Rational and Objective: In CT systems, blurring is the main limiting factor for imaging in-stent restenosis. The aim of this study is to systematically analyze the effect of blurring related biases on the quantitative assessment of in-stent restenosis and to evaluate potential correction methods. Methods: 3D analytical models of a blurred, stented vessel are presented to quantify blurring related artifacts in the stent diameter measurement. Two correction methods are presented for an improved stent diameter measurement. We also examine the suitability of deconvolution techniques for correcting blurring artifacts. Results: Blurring results in a shift of the maximum of the signal intensity towards the center position of the stent, resulting in an underestimation of the stent diameter. This shift can be expressed as a function of the stent radius and width of the point spread function. The correction for this phenomenon reduces the error with 75 percent. Deconvolution reduces the blurring artifacts but introduces a ringing artifact. Conclusion: The analytical vessel models are well suited to study the influence of various parameters on blurring-induced artifacts. The blurring-related underestimation of the stent diameter can significantly be reduced using the presented corrections. Care should be taken into choosing suitable deconvolution filters since they may introduce new artifacts.

One of the key measures of response to treatment for patients in multicenter clinical trials is the lung density measured in Hounsfield Units (HU) from Computer Tomography (CT) scans. The purpose of this work is to determine the dependence of CT attenuation values on scanner type by using in vivo measurements made from homogeneous anatomic areas. In vivo measurements were made in areas within the trachea, aorta, fat and muscle regions of CT scans obtained from subjects scanned as part of a multicenter treatment trial. Scans were selected so that exams from all four major manufacturers were included in the study. For each anatomic region of interest, the mean and standard deviation values were computed to investigate attenuation dependence on scanners. For example, trachea mean (standard deviation) measurements for exams from GE, Siemens, Philips and Toshiba scanners were -986 HU(±15), - 993 HU(±9), -988HU(±8), -1046(±10) respectively. Inter-scanner variability was observed for each scanner showing significant differences (all p-values <0.005). Previous work in examining attenuation dependence on scanners has been performed using anthropomorphic phantoms. The novelty of this work is the use of in vivo measurements from homogeneous regions in order to examine scanner effects on CT attenuation values. Our results show that CT attenuation values for the anatomic regions vary between scanners and hence, dependence of CT attenuation values on scanners is observed.

The purpose of this study is to evaluate the effect of reduced tube current, as a surrogate for radiation dose, on lung nodule detection in pediatric chest multi-detector CT (MDCT). Normal chest MDCT images of 13 patients aged 1 to 7 years old were used as templates for this study. The original tube currents were between 70 mA and 180 mA. Using proprietary noise addition software, noise was added to the images to create 13 cases at the lowest common mA (i.e. 70 mA), 13 cases at 35 mA (50% reduction), and 13 cases at 17.5 mA (75% reduction). Three copies of each case were made for a total of 117 series for simulated nodule insertion. A technique for three-dimensional simulation of small lung nodules was developed, validated through an observer study, and used to add nodules to the series. Care was taken to ensure that each of three lung zones (upper, middle, lower) contained 0 or 1 nodule. The series were randomized and the presence of a nodule in each lung zone was rated independently and blindly by three pediatric radiologists on a continuous scale between 0 (definitely absent) and 100 (definitely present). Receiver operating characteristic analysis of the data showed no general significant difference in diagnostic accuracy between the reduced mA values and 70 mA, suggesting a potential for dose reduction with preserved diagnostic quality. To our knowledge, this study is the first controlled, systematic, and task-specific assessment of the influence of dose reduction in pediatric chest CT.

Computed tomography (CT) has been well established as a diagnostic tool through hardware optimization and sophisticated data calibration. For screening purposes, the associated X-ray exposure risk must be minimized. An effective way to minimize the risk is to deliver fewer X-rays to the subject or lower the mAs parameter in data acquisition. This will increase the data noise. This work aims to study the noise property of the calibrated or preprocessed sinogram data in Radon space as the mAs level decreases. An anthropomorphic torso phantom was scanned repeatedly by a commercial CT imager at five different mAs levels from 100 down to 17 (the lowest value provided by the scanner). The preprocessed sinogram datasets were extracted from the CT scanner to a laboratory computer for noise analysis. The repeated measurements at each mAs level were used to test the normality of the repeatedly measured samples for each data channel using the Shapiro-Wilk statistical test merit. We further studied the probability distribution of the repeated measures. Most importantly, we validated a theoretical relationship between the sample mean and variance at each channel. It is our intention that the statistical test and particularly the relationship between the first and second statistical moments will improve low-dose CT image reconstruction for screening applications.

Metallic objects severely limit diagnostic CT imaging because of their high X-ray attenuation in the diagnostic energy range. In contrast, radiation therapy linear accelerators now offer CT imaging with X-ray energies in the megavolt range, where the attenuation coefficients of metals are significantly lower. We hypothesized that Mega electron-Voltage Cone-Beam CT (MVCT) implemented on a radiation therapy linear accelerator can detect and quantify small features in the vicinity of metallic implants with accuracy comparable to clinical Kilo electron-Voltage CT (KVCT) for imaging. Our test application was detection of osteolytic lesions formed near the metallic stem of a hip prosthesis, a condition of severe concern in hip replacement surgery. Both MVCT and KVCT were used to image a phantom containing simulated osteolytic bone lesions centered around a Chrome-Cobalt hip prosthesis stem with hemispherical lesions with sizes and densities ranging from 0.5 to 4 mm radius and 0 to 500 mg•cm^-3, respectively. Images for both modalities were visually graded to establish lower limits of lesion visibility as a function of their size. Lesion volumes and mean density were determined and compared to reference values. Volume determination errors were reduced from 34%, on KVCT, to 20% for all lesions on MVCT, and density determination errors were reduced from 71% on KVCT to 10% on MVCT. Localization and quantification of lesions was improved with MVCT imaging. MVCT offers a viable alternative to clinical CT in cases where accurate 3D imaging of small features near metallic hardware is critical. These results need to be extended to other metallic objects of different composition and geometry.

The recent multidetector-row CT (MDCT) systems yield high spatial resolution in all directions of volumetric images in clinical routine. The quantitative characterization of the performance of MDCT systems is important for comparing the effects of different scan and reconstruction parameters, for comparing between different CT systems, and for evaluating the accuracy of size and density measurements of fine details in MDCT images. This paper presents a method to characterize the performance of MDCT using a three-dimensional (3D) point spread function (PSF) obtained from MDCT images. With separability assumed, the 3D PSF is decomposed into 2D PSF in the scan plane and SSP on the z-axis. The 2D PSF and SSP are modeled on the basis of the symmetric Lèvy function that generalizes Gaussian and Lorentzian functions. The descriptions of 3D PSF are defined from the Fourier analysis of MDCT images of wire (diameter 0.049mm) and micro-disk (thickness 0.05mm) phantoms. Experimenting with a MDCT system, we demonstrate the method for 3D PSF measurement.

Tumor motion induced by patient breathing decreases the effectiveness of radiation treatment. Image guided radiation treatment (IGRT) is an advanced approach for cancer radiation treatment. The success of IGRT is largely dependent on the accurate localization of tumor in real-time. There are two major imaging approaches currently in use to localize a tumor: internal imaging and external imaging. Internal imaging determines the tumor locations by directly x-ray of the tumor area. It is accurate however radiation dose is a big concern. External imaging derives the internal tumor locations through an external mark on the patient surface. It is radiation dose free however the insufficient accuracy limits its wide application. Integrating the internal and external signals together is necessary for reliable radiation treatment and acceptable patient radiation exposure. Our work tries to identify the correlation patterns between internal/external signals and the influential factors so that the hybrid signal will give desire accuracy in dose delivery while limiting radiation exposure to the patients. Both theoretical simulation based on sinusoidal functions and statistical analysis on real patient data are performed. The sinusoidal simulation will identify the potential influence factors of different correlation conditions. The results have demonstrated the various correlation patterns with amplitude various, frequency changes (duration changes), phase shifts, and baseline drift. The results will aid the statistical analytical on real-patients to identify the dominant factors of the internal/external motion signals for a specific patients. The described work is very useful in advanced IGRT to update the internal/external correlation in real-time for better cancer patient care.

The statistical properties of medical images are central in characterizing the performance of imaging systems. The noise in cone-beam CT (CBCT) is often characterized using Fourier-based metrics, such as the 3D noise-power spectrum (NPS). Under a stationarity assumption, the NPS provides a complete representation of the covariance of the images, since the covariance matrix of the Fourier transform of the image is diagonal. In practice, such assumptions are obeyed to varying degrees. The objective of this work is to investigate the degree to which such assumptions apply in CBCT and to experimentally characterize the NPS and off-diagonal elements under a range of experimental conditions. A benchtop CBCT system was used to acquire 3D image reconstructions of various objects (air and a water cylinder) across a range of experimental conditions that could affect stationarity (bowtie filter and dose). We test the stationarity assumption under such varying experimental conditions using both spatial and frequency domain measures of stationarity. The results indicate that experimental conditions affect the degree of stationarity and that under some imaging conditions, local descriptions of the noise need to be developed to appropriately describe CBCT images. The off-diagonal elements of the DFT covariance matrix may not always be ignored.

In the framework of Spectral Computed Tomography (Spectral CT), scattered X-ray radiation is examined for its spectral composition and spatial distribution by means of Monte Carlo simulations. A reliable material (e.g. bone / contrast agent) separation and quantification requires a precise knowledge of the transmitted X-ray spectrum especially for low energy photons. Unfortunately, for lower energies the primary intensity is increasingly covered by scattered radiation. The detected scattered radiation can be classified into two main categories with respect to their scattering history. The first category contains purely Rayleigh or one-time Compton scattered photons which typically have small scattering angles and an energy spectrum similar to that of the transmitted primary radiation. The second category comprises multiple Compton scattered photons with a spectral composition which is typically softer than that of the transmitted primary photons. In regions of strong beam attenuation (i.e. in the X-ray shadow of a scanned object), the scattered radiation is mainly composed of multiple Compton scattered photons. As a consequence, the spectrally resolved scatter-to-primary ratios strongly increase at low energies. High-quality anti-scatter grids can be used to reduce especially the detection of multiple Compton-scattered photons. A quantitative evaluation of measured photon energies below a certain limit between 30 keV and 50 keV (depending on the phantom geometry and the applied anti-scatter grid) is challenging, since primary photons are superposed by a significantly higher amount of scattered photons.

Crucial to understanding the factors that govern imaging performance is a rigorous analysis of signal and noise transfer characteristics (e.g., MTF, NPS, and NEQ) applied to a task-based performance metric (e.g., detectability index). This paper advances a theoretical framework for calculation of the NPS, NEQ, and DQE of cone-beam CT (CBCT) and tomosynthesis based on cascaded systems analysis. The model considers the 2D projection NPS propagated through a series of reconstruction stages to yield the 3D NPS, revealing a continuum (from 2D projection radiography to limited-angle tomosynthesis and fully 3D CBCT) for which NEQ and detectability index may be investigated as a function of any system parameter. Factors considered in the cascade include: system geometry; angular extent of source-detector orbit; finite number of views; log-scaling; application of ramp, apodization, and interpolation filters; back-projection; and 3D noise aliasing - all of which have a direct impact on the 3D NEQ and DQE. Calculations of the 3D NPS were found to agree with experimental measurements across a broad range of imaging conditions. The model presents a theoretical framework that unifies 3D Fourier-based performance metrology in tomosynthesis and CBCT, providing a guide to optimization that rigorously considers the system configuration, reconstruction parameters, and imaging task.

Currently all CT scanners collect the projection images sequentially, one at a time. The serial approach demands high x-ray power which in turn limits the scanning speed of the CT scanners. To overcome the limitations of the current CT scanners, the concept of stationary CT canners has been proposed to completely eliminate the need for gantry rotation. In such multi-pixel x-ray system, multiple x-ray sources and detectors are distributed around the scanning tunnel. Based on the multi-pixel x-ray system, we have recently demonstrated the feasibility of multiplexing radiography that enables simultaneous collection of multiple projection images through multiplexing. A drastic increase of the speed and reduction of the x-ray peak power can be potentially achieved without compromising the imaging quality. In this paper we demonstrated novel Hadamard multiplexing radiography based on Hadamard transform technique using a carbon nanotube based multi-pixel x-ray source. The combination of the multi-pixel x-ray and multiplexing technologies has the potential to lead to a new generation of stationary CT scanners that have drastically increased throughput at reduced cost.

Multiplexed x-ray imaging was recently proposed as a way to possibly reduce x-ray source power requirements while maintaining temporal resolution and imaging speed. Rather than measuring projections sequentially, with multiplexing multiple sources send photons to the same detector corresponding to different projections. Data for the multiple projections that are measured by the same detector are separated by energizing the sources with different temporal sequences. This concept could be used for radiography, tomosynthesis, or CT imaging. Multiplexed measurements are used very successfully in other modalities. For example, in magnetic resonance imaging (MRI) data from multiple voxels, perhaps even the entire object, can be measured through a single channel. In MRI, the simultaneous interrogation of multiple regions has no SNR penalty. It is important to examine the noise impact of multiplexing/demultiplexing in x-ray imaging. We examined the propagation of noise due to the quantum statistics of the measured x-rays. The analysis showed that for quantum limited acquisitions, the noise penalty of multiplexing can be severe. Indeed, if both the multiplexed and sequential methods are quantum limited, the simpler sequential method always outperforms multiplexing.

Image reconstruction of dynamically deforming objects from projections and known time-dependent motion field is of interest for x-ray computed tomography. Recently, three analytical exact methods have been developed based on DFBP or DBPF algorithms which compensates for time-dependent standard affine or relaxed affine transformation. In contrast, an empirical algorithm has been proposed by Schafer, et al., which merely "trace" the motion of each voxel during the backprojection process. The method is known to be an approximation; however, it has not been discussed how good or bad the level of approximation is. In this paper, we present that a slightly modified Schafer's method (FBPx) is exact if the motion of the object can be described by a class of time-dependent affine transformation - isotropic scaling (contraction and expansion), rotation, and translation. We show mathematically and experimentally that Schafer's method is a good approximation.

In dual source CT (DSCT) with two X-ray sources and two data measurement systems mounted on a CT gantry with a mechanical offset of 90 deg, cross scatter radiation, (essentially 90 deg Compton scatter) is added to the detector signals. In current DSCT scanners the cross scatter correction is model based: the idea is to describe the scattering surface in terms of its tangents. The positions of these tangent lines are used to characterize the shape of the scattering object. For future DSCT scanners with larger axial X-ray beams, the model based correction will not perfectly remove the scatter signal in certain clinical situations: for obese patients scatter artifacts in cardiac dual source scan modes might occur. These shortcomings can be circumvented by utilizing the non-diagnostic time windows in cardiac scan modes to detect cross scatter online. The X-ray generators of both systems have to be switched on and off alternating. If one X-ray source is switched off, cross scatter deposited in the respective other detector can be recorded and processed, to be used for efficient cross scatter correction. The procedure will be demonstrated for cardiac step&shoot as well as for spiral acquisitions. Full rotation reconstructions are less sensitive to cross scatter radiation; hence in non-cardiac case the model-based approach is sufficient. Based on measurements of physical and anthropomorphic phantoms we present image data for DSCT systems with various collimator openings demonstrating the efficacy of the proposed method. In addition, a thorough analysis of contrast-to-noise ratio (CNR) shows, that even for a X-ray beam corresponding to a 64x0.6 mm collimation, the maximum loss of CNR due to cross scatter is only about 7% in case of obese patients.

A common problem arising in medical imaging is the suppression of undesired image artifacts with the simultaneous preservation of salient clinical information. Often the proposed processing "cure" introduces its own artifacts in other parts of the image that confound reliable diagnosis. A canonical example is the suppression of artifacts from hyperdense objects, such as metal and calcium. In this paper we propose a new decomposition-based approach to the combined image formation and the suppression of localized image artifacts which is motivated by recent results on image inpainting. The approach, which we term Model-Based Algebraic Iteration (MBAI) processing, decomposes an image into a collection of homogeneous components, each of which can be reconstructed in the manner most appropriate to its underlying nature. Because each component is localized, the effects of processing on that component do not contaminate other areas of the image. Our specific motivation is the mitigation of artifacts in cardiac multi-detector computed tomography (MDCT) images. Recently, MDCT has offered the promise of a non-invasive alternative to invasive coronary angiography to evaluate coronary artery disease. An impediment preventing its utilization as a routine clinical replacement for angiography is the presence of image "blooming" artifacts due to the presence vascular calcium. We develop MBAI for the purpose of ameliorating artifacts in cardiac images and thus increase the applicability of MDCT for the evaluation of at-risk patient population. We demonstrate preliminary results in the reduction of the calcium blooming-effect in software simulation, phantom, ex-vivo, and in-vivo MDCT data.

In flat detector cone-beam computed tomography (CBCT), scattered radiation is a major source of image degradation, making accurate a posteriori scatter correction inevitable. A potential solution to this problem is provided by computerized scatter correction based on Monte-Carlo simulations. Using this technique, the detected distributions of X-ray scatter are estimated for various viewing directions using Monte-Carlo simulations of an intermediate reconstruction. However, as a major drawback, for standard CBCT geometries and with standard size flat detectors such as mounted on interventional C-arms, the scan field of view is too small to accommodate the human body without lateral truncations, and thus this technique cannot be readily applied. In this work, we present a novel method for constructing a model of the object in a laterally and possibly also axially extended field of view, which enables meaningful application of Monte-Carlo based scatter correction even in case of heavy truncations. Evaluation is based on simulations of a clinical CT data set of a human abdomen, which strongly exceeds the field of view of the simulated C-arm based CBCT imaging geometry. By using the proposed methodology, almost complete removal of scatter-caused inhomogeneities is demonstrated in reconstructed images.

X-ray in-line phase-contrast imaging is a technique that aims to reconstruct the projected absorption and refraction properties of an object. To achieve this, phase retrieval algorithms are employed. The statistical properties of the reconstructed images in phase-contrast imaging remain largely unexplored. In this work, the covariance structure of the absorption and refractive index images is derived analytically, to characterize the noise texture in quantitative in-line phase-contrast imaging. This information is utilized to investigate how object detectability is affected by specification of imaging geometry.

Metrics of system performance are used to assess the abilities and safety of x-ray imaging systems. The detective quantum efficiency (DQE) is used as a measure of "dose efficiency" but, when applied to fluoroscopic systems, requires a measurement of the temporal modulation transfer function (MTF) to account for the effects of system lag. It is shown that the temporal MTF is exposure-rate dependent, and hence must be measured under the specific exposure conditions of interest. We develop a small-signal approach to temporal MTF measurements using a semi-transparent moving slanted edge. Using an x-ray image intensifier-based bench-top system, we show that there is a 50% overstatement of the DQE when not properly accounting for lag. The small-signal approach is used to calculate a lag-free fluoroscopic DQE that agrees with a radiographic DQE measurement under the same exposure-rate conditions. It was found that the temporal MTF did not change within measured precision over normal fluoroscopic conditions, and the radiopaque falling-edge results were consistent with the small-signal temporal MTF. This approach could be implemented in a clinical setting with access to raw (linear or linearized) fluoroscopic image data and could be generalized for use on pulsed-exposure systems.

In this paper, we present shift-invariant filtered backprojection (FBP) cone-beam image reconstruction algorithms for a cone-beam CT system based on a clinical C-arm gantry. The source trajectory consists of two concentric arcs which is complete in the sense that the Tuy data sufficiency condition is satisfied. This scanning geometry is referred to here as a CC geometry (each arc is shaped like the letter "C"). The challenge for image reconstruction for the CC geometry is that the image volume is not well populated by the familiar doubly measured (DM) lines. Thus, the well-known DM-line based image reconstruction schemes are not appropriate for the CC geometry. Our starting point is a general reconstruction formula developed by Pack and Noo which is not dependent on the existence of DM-lines. For a specific scanning geometry, the filtering lines must be carefully selected to satisfy the Pack-Noo condition for mathematically exact reconstruction. The new points in this paper are summarized here. (1) A mathematically exact cone-beam reconstruction algorithm was formulated for the CC geometry by utilizing the Pack-Noo image reconstruction scheme. One drawback of the developed exact algorithm is that it does not solve the long-object problem. (2) We developed an approximate image reconstruction algorithm by deforming the filtering lines so that the long object problem is solved while the reconstruction accuracy is maintained. (3) In addition to numerical phantom experiments to validate the developed image reconstruction algorithms, we also validate our algorithms using physical phantom experiments on a clinical C-arm system.

In 3^rd generation CT systems projection data, generated by X-rays emitted from a single source and passing through the imaged object, are acquired by a single detector covering the entire field of view (FOV). Novel CT system architectures employing distributed sources [1,2] could extend the axial coverage, while removing cone-beam artifacts and improving spatial resolution and dose. The sources can be distributed in plane and/or in the longitudinal direction. We investigate statistical iterative reconstruction of multi-axial data, acquired with simulated CT systems with multiple sources distributed along the in-plane and longitudinal directions. The current study explores the feasibility of 3D iterative Full and Half Scan reconstruction methods for CT systems with two different architectures. In the first architecture the sources are distributed in the longitudinal direction, and in the second architecture the sources are distributed both longitudinally and trans-axially. We used Penalized Weighted Least Squares Transmission Reconstruction (PWLSTR) and incorporated a projector-backprojector model matching the simulated architectures. The proposed approaches minimize artifacts related to the proposed geometries. The reconstructed images show that the investigated architectures can achieve good image quality for very large coverage without severe cone-beam artifacts.

It is known that x-ray projections collected from a circular orbit of an x-ray source are insufficient for accurate reconstruction of a 3D object. For each local region of the object (except in the plane containing the source trajectory) there is a conical volume in the object's spatial frequency space that is unmeasured due to the circular geometry. The Feldkamp, Davis and Kress (FDK) algorithm based on filtered backprojection (FBP) involves a 3D backprojection step so that these unmeasured spatial frequencies are set to zero, resulting in cone beam artifacts for certain objects. We present a new type of cone beam CT reconstruction algorithm based on the Fourier rebinning (FORE) framework of Defrise et al. The cone beam x-ray projection data are rebinned into a set of in-plane sinograms using the FORE rebinning approximation, followed by 2D FBP to reconstruct each axial slice. The algorithm is able to extrapolate data into the missing region of the object's frequency space in a computationally efficient way, allowing for a reduction of cone beam artifacts for certain objects. Unlike FDK, the algorithm is exact for an impulse object located anywhere along the axis of rotation. Reconstruction errors are dependent on the radial distance, cone angle, and the second-derivative of the projection data in the longitudinal direction. Finally, an extension to the algorithm is presented that permits reconstruction in regions of the object that are not seen by the detector in every view.

Some of the recently developed image reconstruction algorithms for cone-beam computed tomography (CBCT) involve the computation of the finite Hilbert transform. We have previously studied noise property of the finite Hilbert transform and observed that it can be used for potentially improving the image noise property within a region of interest (ROI) in IGRT. Imaging radiation dose is one of the critical issues in IGRT, and in addition to existing dose-reduction schemes by use of ROI imaging, it is possible to achieve further patient dose reduction through modulating beam intensity so that a sub-ROI in the ROI be exposed by high flux of x-ray photons and the rest of the ROI be exposed by low flux of them. In this work, we investigate the technique for obtaining sub-ROI images, which is supposed to include the target under treatment, with high contrast-to-noise ratio (CNR) and the images within the rest of the ROI with low CNR. Numerical studies have been conducted as a preliminary in this work.

We present Computed Tomography (CT) acquisition and reconstruction schemes for low-dose neuro-angiography based on the method of HighlY constrained back PRojection (HYPR). Simulated and experimental low X-ray radiation dose scans were prepared using the techniques of interleaved view angle under-sampling and tube current reduction. Dynamic CT Angiograms (CTAs) were produced for both standard and low dose images sets. The spatial correlation coefficient, r, between the two reconstruction approaches was determined for each time frame and the SNR and CNR values in arterial ROIs were calculated. The undersampled HYPR reconstructions produced r values of > 0.95 at undersampling and dose reduction factors of 10 and SNR and CNR were more than doubled using HYPR techniques at a tube current of 25 mA. HYPR approaches to contrast enhanced neuro-imaging provide not only volumetric brain hemodynamics but also the ability to produce high quality maps of standard perfusion parameters. The synergy of volumetric hemodynamics and assessment of tissue function provides the medical imaging community with high quality diagnostic information at a fraction of the radiation dose in a single contrast-enhanced scan.

X-ray differential phase-contrast tomography (DPCT) is a method for reconstructing the spatial distribution of the X-ray refractive index within an object from knowledge of differential projection data. Assuming geometrical optics wave propagation, these data describe the angles by which the probing optical beams are deflected by the object due to refraction. Phase-sensitive X-ray imaging methods such as diffraction enhanced imaging can measure the required beam-deflection data, and are being actively developed for medical imaging applications. In this work, we investigate and demonstrate the applicability of algorithms recently developed for conventional tomography for obtaining region-of-interest images in DPCT from knowledge of truncated differential projection data. A preliminary numerical study is conducted to validate and demonstrate the proposed reconstruction algorithm.

Digital breast tomosynthesis (DBT) is a rapidly developing imaging modality that gives some tomographic information for breast cancer screening. The effectiveness of standard mammography can be limited by the presence of overlapping structures in the breast. A DBT scan, consisting of a limited number of views covering a limited arc projecting the breast onto a fixed flat-panel detector, involves only a small modification of digital mammography, yet DBT yields breast image slices with reduced interference from overlapping breast tissues. We have recently developed an iterative image reconstruction algorithm for DBT based on image total variation (TV) minimization that improves on EM in that the resulting images have fewer artifacts and there is no need for additional regularization. In this abstract, we present the total p-norm variation (TpV) image reconstruction algorithm. TpV has the advantages of our previous TV algorithm, while improving substantially on the efficiency. Results for the TpV on clinical data are shown and compared with EM.

Optimal noise control is critical for dose reduction in CT. In this work, we investigated the use of a locally-adaptive method for noise reduction in low-dose CT. This method is based upon bilateral filtering, which smoothes the projection data using a weighted average in a local neighborhood, where the weights are determined according to both the spatial proximity and intensity similarity between the center pixel and the neighboring pixels. This filtering is locally adaptive and can preserve important edge information in the sinogram, thus without significantly sacrificing the spatial resolution. It is closely related to anisotropic diffusion, but is significantly faster. More importantly, a CT noise model can be readily incorporated in the filtering and denoising process. We have evaluated the noise-resolution properties of the bilateral filtering in a phantom study and a preliminary patient study with contrast-enhanced abdominal CT exams. The results demonstrated that bilateral filtering can achieve a better noise-resolution tradeoff than a series of commercial reconstruction kernels. This improvement on noise-resolution properties can be used for improving the image quality in low-dose CT and can also be translated to substantial dose reduction.

In classical tomosynthesis, the x-ray source generally is moved along a curve segment, such as a circular trajectory, within a plane that is perpendicular to the detector plane. Studies suggest that when the angular coverage and number of projection views are limited, it can be difficult to reconstruct accurate images within planes perpendicular to the detector plane in classical tomosynthesis. In this work, we investigate imaging strategies in tomosynthesis using trajectories that are not confined within a plane perpendicular to the detector plane. We expect that such trajectories can increase data information and thus lead reconstructed images with improved quality. Numerical studies were conducted for evaluating the image-reconstruction quality in classical tomosynthesis and tomosynthesis with trajectories that are not confined within a plane perpendicular to the detector plane. The results of the studies indicated that, with the same number of views, (or equivalenntly, the same amount of image radiation), data acquired in tomosynthesis with the trajectories that are not confined within a plane perpendicluar to the detector plane generally contain more information than that acquired with classical tomosynthesis and can thus yield images with improved quality.

In this paper we propose a novel scatter correction methodology for X-ray based cone-beam CT that allows to combine the advantages of projection-based and volume-based correction approaches. The basic idea is to use a potentially non-optimal projection-based scatter correction method and to iteratively optimize its performance by repeatedly assessing remaining scatter-induced artifacts in intermediately reconstructed volumes. The novel approach exploits the fact that due to the flatness of the scatter-background, compensation itself is most easily performed in the projection-domain, while the scatter-induced artifacts can be better observed in the reconstructed volume. The presented method foresees to evaluate the scatter correction efficiency after each iteration by means of a quantitative measure characterizing the amount of residual cupping and to adjust the parameters of the projection-based scatter correction for the next iteration accordingly. The potential of this iterative scatter correction approach is demonstrated using voxelized Monte Carlo scatter simulations as ground truth. Using the proposed iterative scatter correction method, remarkable scatter correction performance was achieved both using simple parametric heuristic techniques as well as by optimizing previously published scatter estimation schemes. For the human head, scatter induced artifacts were reduced from initially 148 HU to less than 8.1 HU to 9.1 HU for different studied methods, corresponding to an artifact reduction exceeding 93%.

There exists a strong need to reconstruct computed tomographic (CT) images with practically useful quality from a small number of projections in image-guided radiation therapy: for lowering radiation dose delivered to the subject, for shortening the imaging time, and for reducing the imaging-configuration complexity. We have recently developed an iterative image reconstruction algorithm based on total-variation (TV) minimization from incomplete projection data in CT. In numerical studies with a variety of incomplete projection-data sets including truncated data, reduced scan range, and sparse sampling, the developed algorithm seems to yield reasonable reconstruction, as compared to some of the existing algorithms, such as algebraic reconstruction technique (ART) and expectation minimization (EM). The TV-based algorithm begins in general with a uniform image as an initial guess, and goes through iteration steps to minimize the image TV subject to satisfying the given incomplete projection data. In image-guided radiation therapy (IGRT), a patient usually undergoes CT scanning for treatment planning, which can provide the reference image for image guidance. Therefore, we propose a TV-based algorithm with a priori information in few-view CT for IGRT, in an attempt to further reduce the number of projections needed for image reconstruction from what the TV-based algorithm uses when no a priori information is included. In this work, we report the initial results of a preliminary numerical study that we have conducted to demonstrate this approach.

Recently, foundational mathematical theory, compressed sensing (CS), has been developed which enables accurate reconstruction from greatly undersampled frequency information (Candes et. al. and Donoho). Using numerical phantoms it has been demonstrated that CS reconstruction (e.g. minimizing the ℓ₁ norm of the discrete gradient of the image) offers promise for computed tomography. However, when using experimental CT projection data the undersampling factors enabled were smaller than in numerical simulations. An extension to CS has recently been proposed wherein a prior image is utilized as a constraint in the image reconstruction procedure (i.e. Prior Image Constrained Compressed Sensing - PICCS). Experimental results are demonstrated here from a clinical C-arm system, highlighting one application of PICCS in reducing radiation exposure during interventional procedures while preserving high image quality. In this study a range of view angles has been investigated from very limited angle aquisitions (e.g. tomosythesis) to undersampled CT acquisitions.

Truncation of CT projection data is always coupled with incomplete angular sampling and can lead to severe image artifacts in clinical CT. Extrapolation of projection data is needed to restore CT values inside and outside the scan field of view (SFOV). We present three types of extrapolation schemes. The first type (M1) is characterized by extrapolation of projection data using a virtual object of constant attenuation. For multi-slice helical CT this extrapolation scheme is applied in a row-wise manner. The second type (M2) utilizes consistency conditions of parallel projection data. The conservation of m^th order moments of non-truncated projections can be utilized for the extrapolation of truncated projection data by fitting extrapolation functions of variable length. The third method (M3) extrapolates truncated data by sinogram decomposition and completion. For each voxel in image space the corresponding trace in the 3D-sinogram is computed. The minimum signal within each trace is extrapolated to the extended sinogram parts, which represent the extended FOV. Based on the evaluation of both simulation data of an anthropomorphic thorax phantom and clinical data, we evaluate the three reconstruction techniques. Sinogram decomposition proofs to be better than the other techniques, but is computationally very demanding.

To improve the speed and quality of ordered-subsets expectation-maximization (OSEM) SPECT reconstruction, we have implemented a content-adaptive, singularity-based, mesh-domain, image model (CASMIM) with an accurate algorithm for estimation of the mesh-domain system matrix. A preliminary image, used to initialize CASMIM reconstruction, was obtained using pixel-domain OSEM. The mesh-domain representation of the image was produced by a 2D wavelet transform followed by Delaunay triangulation to obtain joint estimation of nodal locations and their activity values. A system matrix with attenuation compensation was investigated. Digital chest phantom SPECT was simulated and reconstructed. The quality of images reconstructed with OSEM-CASMIM is comparable to that from pixel-domain OSEM, but images are obtained five times faster by the CASMIM method.

Chord-based algorithms can eliminate cone-beam artifacts in images reconstructed from a clinical computed tomography (CT) scanner. The feasibility of using chord-based reconstruction algorithms was evaluated with three clinical CT projection data sets. The first projection data set was acquired using a clinical 64-channel CT scanner (Philips Brilliance 64) that consisted of an axial scan from a quality assurance phantom. Images were reconstructed using (1) a full-scan FDK algorithm, (2) a short-scan FDK algorithm, and (3) the chord-based backprojection filtration algorithm (BPF) using full-scan data. The BPF algorithm was capable of reproducing the morphology of the phantom quite well, but exhibited significantly less noise than the two FDK reconstructions as well as the reconstruction obtained from the clinical scanner. The second and third data sets were obtained from scans of a head phantom and a patient's thorax. For both of these data sets, the BPF reconstructions were comparable to the short-scan FDK reconstructions in terms of image quality, although sharper features were indistinct in the BPF reconstructions. This research demonstrates the feasibility of chord-based algorithms for reconstructing images from clinical CT projection data sets and provides a framework for implementing and testing algorithmic innovations.

Electronic noise compensation can be important for low-dose x-ray CT applications where severe photon starvation occurs. For clinical x-ray CT systems utilizing energy-integrating detectors, it has been shown that the detected x-ray intensity is compound Poisson distributed, instead of the Poisson distribution that is often studied in the literature. We model the electronic noise contaminated signal Z as the sum of a compound Poisson distributed random variable (r.v.) Y and a Gaussian distributed electronic noise N with known mean and variance. We formulate the iterative x-ray CT reconstruction problem with electronic noise compensation as a maximum-likelihood reconstruction problem. However the likelihood function of Z is not analytically trackable; instead of working with it directly, we formulate the problem in the expectation-maximization (EM) framework, and iteratively maximize the conditional expectation of the complete log-likelihood of Y. We further demonstrate that the conditional expectation of the surrogate function of the complete log-likelihood is a legitimate surrogate for the incomplete surrogate. Under certain linearity conditions on the surrogate function, a reconstruction algorithm with electronic noise compensation can be obtained by some modification of one previously derived without electronic noise considerations; the change incurred is simply replacing the unavailable, uncontaminated measurement Y by its conditional expectation E(Y|Z). The calculation of E(Y|Z) depends on the model of Y, N, and Z. We propose two methods for calculating this conditional expectation when Y follows a special compound Poisson distribution - the exponential dispersion model (ED). Our methods can be regarded as an extension of similar approaches using the Poisson model to the compound Poisson model.

The back-projection filtration (BPF)algorithm is capable of reconstructing ROI images from truncated data acquired with a wide class of general trajectories. However, it has been observed that, similar to other algorithms for convergent beam geometries, the BPF algorithm involves a spatially varying weighting factor in the backprojection step. This weighting factor can not only increase the computation load, but also amplify the noise in reconstructed images The weighting factor can be eliminated by appropriately rebinning the measured cone-beam data into fan-parallel-beam data. Such an appropriate data rebinning not only removes the weighting factor, but also retain other favorable properties of the BPF algorithm. In this work, we conduct a preliminary study of the rebinned BPF algorithm and its noise property. Specifically, we consider an application in which the detector and source can move in several directions for achieving ROI data acquisition. The combined motion of the detector and source generally forms a complex trajectory. We investigate in this work image reconstruction within an ROI from data acquired in this kind of applications.

Most recently, a novel data acquisition method has been proposed and experimentally implemented for x-ray differential phase contrast computed tomography (DPC-CT), in which a conventional x-ray tube and a Talbot-Lau type interferometer were utilized in data acquisition. The divergent nature of the data acquisition system requires a divergent-beam image reconstruction algorithm for DPC-CT. This paper focuses on addressing this image reconstruction issue. We developed a filtered backprojection algorithm to directly reconstruct the DPC-CT images from acquired projection data. The developed algorithm allows one to directly reconstruct the decrement of the real part of the refractive index from the measured data. In order to accurately reconstruct an image, the data need to be acquired over an angular range of at least 180° plus the fan-angle. Different from the parallel beam data acquisition and reconstruction methods, a 180° rotation angle for data acquisition system does not provide sufficient data for an accurate reconstruction of the entire field of view. Numerical simulations have been conducted to validate the image reconstruction algorithm.

In order to improve reconstructed image quality for cone-beam collimator SPECT, we have developed and implemented a fully 3D reconstruction, using an ordered subsets expectation maximization (OSEM) algorithm, along with a volumetric system model - cone-volume system model (CVSM), a modified attenuation compensation, and a 3D depth- and angle-dependent resolution and sensitivity correction. SPECT data were acquired in a 128×128 matrix, in 120 views with a single circular orbit. Two sets of numerical Defrise phantoms were used to simulate CBC SPECT scans, and low noise and scatter-free projection datasets were obtained using the SimSET Monte Carlo package. The reconstructed images, obtained using OSEM with a line-length system model (LLSM) and a 3D Gaussian post-filter, and OSEM with FVSM and a 3D Gaussian post-filter were quantitatively studied. Overall improvement in the image quality has been observed, including better transaxial resolution, higher contrast-to-noise ratio between hot and cold disks, and better accuracy and lower bias in OSEM-CVSM, compared with OSEM-LLSM.

In this paper we extend our previous evaluation of 2D tomographic image reconstruction using a content-adaptive mesh model (CAMM) for emission tomography (EM). First we present a model for the projection operator calculation using a ray-tracing algorithm based on mesh modeling for image representation. The proposed calculation allows us to incorporate a non-uniform attenuation and distance-dependent spatial resolution of the imaging system. Next the proposed approach was tested using realistic simulations with the NCAT phantom and SIMIND Monte Carlo simulation of a SPECT system in which the mesh model is calculated using images from pre reconstruction by filtered-back projection. Thereby this paper establishes a necessary step for the future development of 3D reconstruction images and spatio-temporal processing using deformable mesh modeling.

Single photon emission computerized tomographic (SPECT) images often suffer from low resolution and low count density. To improve spatial resolution of SPECT it is possible to use a pinhole collimator; however, this further reduces the system sensitivity. A potential solution to this problem is to use coded apertures, which offers increased sensitivity by using multiple pinholes, at the cost of increased image reconstruction time. A generic reconstruction algorithm has been developed which allows for arbitrary acquisition geometry via affine transforms (translation and rotation). The reconstruction process uses a (Siddon) ray projector, the expectation maximization (EM) algorithm and a 1 to n pinhole position matrix. Iteration times scale as a function of the number of pinholes in the collimator. Resolution recovery has also been incorporated into the reconstruction algorithm. The algorithm developed allows for the investigation of optimal imaging settings for small animal imaging. Simulated acquisitions of an ex-vivo rat heart with 1, 5 and 8 pinholes, over 360 degree acquisition, showing that multi-pinhole imaging can be successfully applied to rat cardiac imaging. Further refinement of the acquisition parameters, such as image overlap, collimator pinhole configuration and geometrical imaging configuration, will predict the theoretical settings for quantitative cardiac multi-pinhole SPECT imaging.

A method is proposed for a 3D reconstruction of coronary networks from rotational projections that departs from motion-compensated approaches. It deals with multiple views extracted from a time-stamped image sequence through ECG gating. This statistics-based vessel reconstruction method relies on a new imaging model by considering both the effect of background tissues and the image representation using spherically-symmetric basis functions, also called 'blobs'. These blobs have a closed analytical expression for the X-ray transform, which makes easier to compute a cone-beam projection than a voxel-based description. A Bayesian maximum a posteriori (MAP) estimation is used with a Poisson distributed projection data instead of the Gaussian approximation often used in tomography reconstruction. A heavy-tailed distribution is proposed as image prior to take into account the sparse nature of the object of interest. The optimization is performed by an expectation-maximization like (EM) block iterative algorithm which offers a fast convergence and a sound introduction of the non-negativity constraint for vessel attenuation coefficients. Simulations are performed using a model of coronary tree extracted from multidetector CT scanner and a performance study is conducted. They point out that, even with severe angular undersampling (6 projections over 110 degrees for instance) and without introducing a prior model of the object, significant results can be achieved.

Forward and Backward projections are two computational costly steps in tomography image reconstruction such as Positron Emission Tomography (PET). To speed-up reconstruction time, a hardware projection/backprojection pair has been built following algorithm architecture adequacy principles. Thanks to an original memory access strategy based on an 3D adaptive and predictive memory cache, the external memory wall has been overcome. Thus, for both projector architectures several units run efficiently. Each unit reaches a computational throughput close to 1 operation per cycle. In this paper, we present how from our hardware projection/backprojection pair, an analytic (3D-RP) and an iterative (3D-EM) reconstruction algorithms can be implemented on a System on Programmable Chip (SoPC). First, an hardware/software partitioning is done based on the different steps of each algorithm. Then the reconstruction system is composed of two hardware configurations of the programmable logic resources (FPGA). Each one corresponds mainly to the projection and backprojection step. Our projector/backprojector has been validated with a software 3D-RP and 3D-EM reconstruction on simulated PET-SORTEO data. A reconstruction time evaluation of these reconstruction systems are done based on the measured performances of our projectors IPs and the estimated performances of the additional simple hardware IPs. The expected reconstruction time is compared with the software tomography distribution STIR. A speed-up of 7 can be expected for the 3D-RP algorithm and a speed-up of 3.5 for the 3D-EM algorithm. For both algorithms, the architecture cycle efficiency expected is largely greater than the software implementation: 120 times for 3D-RP and 60 times for 3D-EM.

Tomography is an important technique for non-invasive imaging, with applications in medicine, materials research and industry. Tomographic reconstructions are typically gray-scale images, that can possibly contain a wide spectrum of grey levels. Segmentation of these grey level images is an important step to obtain quantitative information from tomographic datasets. Thresholding schemes are often used in practice, as they are easy to implement and use. However, if the tomogram exhibits variations in the intensity throughout the image, it is not possible to obtain an accurate segmentation using a single, global threshold. Instead, local thresholding schemes can be applied that use a varying threshold, depending on local characteristics of the tomogram. Selecting the best local thresholds is not a straightforward task, as local image features (such as the local histogram) often do not provide sufficient information for choosing a proper threshold. In this paper, we propose a new criterion for selecting local thresholds, based on the available projection data, from which the tomogram was initially computed. By reprojecting the segmented image, a comparison can be made with the measured projection data. This yields a quantitative measure of the quality of the segmentation. By minimizing the difference between the computed and measured projections, optimal local thresholds can be computed. Simulation experiments have been performed, comparing the result of our local thresholding approach with global thresholding. Our results demonstrate that the local thresholding approach yields segmentations that are significantly more accurate, in particular when the tomogram contains artifacts.

This work is a feasibility study of a phase-contrast cone-beam CT (CBCT) system for ROI (region of interest) reconstruction in breast imaging that incorporates the in-line holography technique into a cone-beam CT system. A conventional CBCT scan is done first to find any suspicious lesion, followed by a phase-contrast CBCT scan of the ROI for detailed characterization. The phase-contrast in-line holographic images are generated using Fresnel theory through computer simulation, and the projected phase maps, as line integrals of the phase coefficient of the scanned breast, are retrieved using phase-attenuation duality theory. In this way the object's phase coefficient, as the object function, can be reconstructed using these projected phase maps through FDK algorithm. The reconstruction error is calculated to evaluate the accuracy of this approach. The noise property of this approach is investigated as well by adding Poisson noise to the holographic images. The projected phase maps are retrieved and the object function is reconstructed in the presence of noise. The results show that the object's phase coefficient can be reconstructed with very small reconstruction error, and the noise level can be greatly reduced compared to the conventional CBCT system. In conclusion, the phase-contrast CBCT breast imaging approach is very promising to provide better image quality and to lower x-ray dose level for tumor characterization.

With the increasing reliance of doctors on imaging procedures, not only visualization needs to be optimized, but the reconstruction of the volumes from the scanner output is another bottleneck. Accelerating the computationally intensive reconstruction process improves the medical work flow, matches the reconstruction speed to the acquisition speed, and allows fast batch processing and interactive or near-interactive parameter tuning. Recently, much effort has been focused on using the computational power of graphics processing units (GPUs) for general purpose computations. This paper presents a GPU-accelerated implementation of single photon emission computed tomography (SPECT) reconstruction based on an ordered-subset expectation maximization algorithm. The algorithm uses models for the point-spread-function (PSF) to improve spatial resolution in the reconstruction images. Instead of computing the PSF directly, it is modeled as efficient blurring of slabs on the GPU in order to accelerate the process. The algorithm for the calculation of accumulated attenuation factors that allows correcting the generated volume according to the attenuation properties of the volume is optimized for processing on the GPU. Since these factors can be reused between different iterations, a cache is used that is adapted to different sizes of the video memory so that only those factors have to be recomputed that do not fit onto graphics memory. These improvements make the reconstruction of typical SPECT volume near interactive.

Tomographic acquisition uses projection angles evenly distributed around 2π. The Mojette transform and the discrete Finite Radon Transform (FRT) both use discrete geometry to overcome the ill-posedeness of the inverse Radon transform. This paper focuses on the transformation of acquired tomographic projections into suitable discrete projection forms. Discrete Mojette and FRT algorithms can then be used for image reconstruction. The impact of physical acquisition parameters (which produce uncertainties in the detected projection data) is also analysed to determine the possible useful interpolations according to the choice of angle acquisitions and the null space of the transform. The mean square error (MSE) reconstruction results obtained for data from analytical phantoms consistently shows the superiority of these discrete approaches when compared to the classical "continuous space" FBP reconstruction.

Helical cone-beam CT is widely used nowadays because of its rapid scan speed and efficient utilization of x-ray dose. HCT-FDK is an effective reconstruction algorithm on Helical CT. However, like other 3D reconstruction algorithms, HCT-FDK is time consuming because of its large amount of data processing including the convolution and 3D-3D back projection. Recently, GPU is widely used to parallel many reconstruction algorithms. The latest GPU has some nice features, such as large memory, lots of processors, fast 3D texture mapping, and flexible frame buffer object. All these features help reconstruction a lot. In this paper, we present a solution to this problem with GPU. First, we bring a lookup table into HCT-FDK. Then, both convolution and back projection are implemented on GPU. At last, the reconstruction result is directly smoothed and visualized by GPU. Experimental results are given to compare among CPU and two generations of GPU: Geforce 6800GT and Geforce 8800GTX. The comparison was applied both on simulation data and real data. We show that, GPU-accelerated HCT-FDK gets result with similar levels of noise and clarity but gains a speed increase of about 10-100 times faster than using CPU only. With its newer feature, Geforce 8800GTX can get a similar quality like Geforce 6800GT and about 20 times faster.

Modern CT systems have advanced at a dramatic rate. Algebraic iterative reconstruction techniques have shown promising and desirable image characteristics, but are seldom used due to their high computational cost for complete reconstruction of large volumetric datasets. In many cases, however, interest in high resolution reconstructions is restricted to smaller regions of interest within the complete volume. In this paper we present an implementation of a simple and practical method to produce iterative reconstructions of reduced-sized ROI from 3D helical tomographic data. We use the observation that the conventional filtered back-projection reconstruction is generally of high quality throughout the entire volume to predict the contributions to ROI-related projections arising from volumes outside the ROI. These predictions are then used to pre-correct the data to produce a tomographic inversion problem of substantially reduced size and memory demands. Our work expands on those of other researchers who have observed similar potential computational gains by exploiting FBP results. We demonstrate our approach using cardiac CT cone-beam imaging, illustrating our results with both ex vivo and in vivo multi-cycle EKG-gated examples.

Clinical demands of image-guided procedures present technical challenges in X-ray 1K×1K fluoroscopy and cone-beam CT on a mobile C-arm. Performance-per-watt and performance-per-dollar are other major considerations in a search for an optimal computational platform. Real-time constraints of processing high-resolution fluoroscopic images currently necessitate the use of highly specialized proprietary image processing hardware, which cannot be easily repurposed for acceleration of other computing tasks. In our previous studies, we were investigating heterogeneous computing architectures and suitable hardware/software components to assist in time-critical surgical applications. Through those studies, it has been shown that Graphics Processing Units (GPUs) can provide outstanding levels of computational power utilizing the Single Instruction Multiple Data (SIMD) programming model. In the present study, we expand our research in the domain of real-time processing and continue to explore the feasibility of GPU acceleration for both fluoroscopic and tomographic imaging. Current emphasis is being placed on applicability of NVIDIA's novel Tesla computing solutions and Compute Unified Device Architecture (CUDA). The results of this pilot project comprise the Cg/OpenGL and CUDA algorithm implementations, benchmark evaluations, and examples of processing image data acquired with use of anthropomorphic phantoms.

In today's tomographic imaging, there are more incomplete data systems, such as few-view system. The advantage of few-view tomography is less x-ray dose and reduced scanning time. In this work, we study the projection distribution in few-view fan-beam imaging. It is one of the fundamental problems in few-view imaging because of its severe lack of projection data. The aim is to reduce data redundancy and to improve the quality of reconstructed images by research on projection distribution schemes. The reconstruction algorithm for few-view imaging is based on algebraic reconstruction techniques (ART) and total variation (TV) constraint approached by E. Sidky et al in 2006. Study of few-view fan-beam projection distribution is performed mainly through comparison of several distribution types in projection space and reconstructed images. Results show that the distribution called short-scan type obtains the best image in five typical distributions.