Current trends in positron tomography are toward instrumentation which will provide resolution finer than 7 mm FWHM and a sensitivity of 75,000 events per second, per transverse section, for 1 pCi per cm3 of activity in a 20 cm³ diameter phantom. Multiple stationary layers of tightly packed crystals with widths of 6 mm or less in circular arrays can provide adequate sampling for thorax or head PET devices. The major instrumentation problem is to achieve efficient optical coupling between crystals and photo detectors without a sacrifice of sensitivity. Sampling and coupling schemes, an analysis of the time-of-flight gains and aspects of data acquisition and display are presented.

A multipurpose nuclear imaging system is under development at Brookhaven National Laboratory, featuring two large field of view scintillation cameras on a rotating gantry. The system can image single photon as well as positron emitting nuclide distributions. The energy of all events in a wide range (approximately 50 KeV to 1 MeV) can be digitized for event by event (list mode) acquisition. This allows distributions of multiple isotopes to be imaged simultaneously, or optimization of sensitivity for single isotopes by weighting of off-peak events. In static mode the detectors can be mounted with seven pinhole collimators with perpendicular axes to permit dynamic analysis such as T1-201 uptake in the myocardium. An important aspect of the UNICON is the data processing system, including a PDP 11/34 with image display system, and a VAX 11/780 with an array processor for image processing.

A Monte Carlo simulation was developed to study scatter contributions from a 140 keV point source at various depths and for different energy windows in finite water phantoms. Photoelectric and Compton interactions were considered. Scatter fractions, energy spectra, and radial spread functions of three approximately patientsized phantoms (rectangular prism, elliptical cylinder, and a sphere) were examined as a function of point-source depth and detector energy-window width. For a 100% energy window, energy spectra are characterized by a high energy region, a backscatter peak region, and a low energy, multi-scatter region. Depth dependent spatial limitations to the radial spread functions occur with decreasing window width. Scatter fractions for the sphere are much smaller than those of the other two phantoms, but approach their values as the size of the energy window decreases.

A Monte Carlo program has been developed to simulate the response of a single photon emission computed tomography (SPECT) system. Phantoms are represented by any number of spheres and cylinders which are placed inside a single, larger cylinder. The phantom, source distribution, and system parameters are specified in an input file. The code permits data acquisition by a stationary or rotating gamma camera. The quality of the images is degraded by scatter and attenuation of radiation by tissue which, in turn, is dependent on object size, shape, composition, and density. The simulation enables control of components which govern the emission and transport of radiation through the source and attenuating medium. Thus, the simulation is useful for analyzing and designing algorithms to compensate for scatter and attenuation. The accuracy of the SPECT simulation is illustrated by measuring the scatter fraction, resolution, ARMS noise, and image contrast. These results are then compared with experimental results to validate the accuracy of the simulation. A method of compensating for the degrading effects of scatter and attenuation is examined.

In order to quantitatively study cellular and nuclear properties, we have assembled a low cost video/microscope densitometric system. The system consists of a conventional light microscope coupled to a low light level TV camera whose output is sequentially processed by two microcomputers, one for image acquisition and one for data analysis. The system allows us to a) acquire and store microscope images, b) make spatial measurements such as size and shape, and c) make grayness measurements needed to compute optical density and texture descriptors. Despite the inherent limitations of a video micrometry system (e.g. poor signal-to-noise ratio, automatic gain compensations, etc.) we have been able to collect reproducible data with satisfactory accuracies.

" Analytical Ion Microscopy " is a method of micro analysis proposed and developped by Castaing and Slodzian in 1960 (1, 2). With this method applied to the study of biological material it is possible to obtain images of distribution of a given stable or radioactive element in a tissue section (3). The spatial resolution is 1/2 μm and the minimum detectable concentration varies from 1 ppm to 1 ppb.

First performance tests of an ultra-fast laser scanning microscope are reported. The objective lens has been perfected to reach the performance level specified in the design with respect to the spread function, freedom from spherical aberration and coma. The attained dimensional tolerances of the polygon spinner are given. The test results on operational pyramidal error suggest that the critical 4 arc sec limit is reached at 36,000 rpm. Excellent linearity and uniformity for the slow slide translation is attained with the PMI linear force motor. Over-all system performance is demonstrated with test scans of a bar pattern and biologic imagery.

The photon energy spectrum has played a relatively minor role in classical radiographic imaging. Radiologists have used different regions of the spectrum for different studies. For example; lower energies are used for mammography to emphasize calcifications, and higher energies are used for chest studies to minimize rib shadows. These effects, however, are relatively subtle compared to those of selective energy imaging to be described in this paper. We will describe the application of energy selectivity to computerized tomography and projection radiography. In the case of computerized tomography the technique eliminates the non-linear spectral-shift artifact, and provides the average atomic number and density of each pixel. In the case of projection radiography, any intervening material can be removed to facilitate visualization of some desired structure. In addition, as with CT, an unknown lesion can be identified based on its average atomic number.

Several aspects of iodine sensitivity of digital fluorographic systems are discussed as applied to digital subtraction angiography. Expected iodine signal magnitudes are reviewed and shown to be in the range of 1-10 mg/cm². An experimental example of spurious contrast in iodine sensitivity measurements is2 provided. Imaging of a 1 mm wide 0.25 mg/cm simulated iodinated vessel is demonstrated. Such iodine amounts are thought to be near the limit of the sensitivity of digital fluorographic systems. The concept of matched filtering is proposed for generating the equivalent of temporal subtraction images but using low intensity continuous x-ray exposures. The first experimental matched filter canine image is presented using a total exposure for the procedure about one-sixth of that used for pulsed exposure images of comparable quality.

The assessment of stenotic arteries is an important parameter for the diagnosis of the coronary heart disease and for the indication of a bypass operation. A system has been developed to digitize and process frames from 35 mm- cinecardiographic films on the computer. Vessels filled with contrast medium are recognized automatically by their typical density profiles. The contour finding takes into account the special edge form of the projected boli. When the bolus contours are determined the degree of diameter narrowing can easily be calculated. A further step to evaluate an obstruction is to integrate over the vessel width and to re-transform the mainly exponential imaging process in order to reconstruct the original cross-sectional area of the bolus or its diameters, respectively. In this way even with only one projection a semi-quantitative prediction of an excentricity becomes possible. The correlation of computer measures with known model dimensions and with clinical findings is discussed briefly.

A scanning beam system with feedback equalization is described. This offers the reduction in scattered radiation with a pair of moving slits and the regional manipulation of incident x-ray exposure to compress large variations in trans-mitted film exposure to fit within the useful exposure range of radiographic film. When applied to chest radiography, this allows maximum soft tissue contrast in the retrodiaphragmatic, retro-cardiac, mediastinal and lung field regions. The scan time to expose a 14" x 17" chest film is 5 seconds with either a one-dimensional slit scan or a two-dimensional raster scan. By manipulation of the feedback scheme, considerable variation in low frequency contrast can be obtained.

A new radiographic system has been developed for use in mapping three-dimensional epicardial motion. The new radiograph consists of an array of detectors arranged such that they monitor 1 cm strip of scattering photons along a contour across the cardiac surface. Synthesis of the detector data permits immediate display of the beating cardiac contour on a CRT screen for observation and analysis. The entire radiographic system is under microprocessor control. Location of the cardiac surface and initial cardiac contour display is possible within ten seconds of a scan's initiation. The system has been carefully validated on mechanical models. Studies with anesthetized canines demonstrate the new radiograph in a physiological environment. Eventual clinical use is expected to provide the physician with accurate data on ventricular function at patient costs and radiation dose rates lower than those of other available techniques.

A new method to obtain simultaneously two or three radiographs with a wide dynamic range was studied. This is to divide the transmitted X-ray energy spectra through a human body into lower and higher parts than K absorption edge by a metal foil (Pb, Ta, Gd) and give radiographs using two or three pairs of an one-side coated film and an intensifying screen. The backward film has the informations filtered by the metal foil. The forward film before the metal foil, if the film density is same, relatively contains the informations of lower parts of the transmitted X-ray spectra through a human body. Secondly, a metal foil can make shadows of thin parts and thick parts of a human body displace on high region of film, respectively and separatedly. These radiographs of thin parts were useful to be observed superposing two films with a wide dynamic range. As to thick parts it was useful to view two films hanging side by side. This technique was appreciated to be applied to the organs such as extremities, knee and elbow, head and neck, lung and etc.

The final step of the medical imaging process is image display to a human for interpretation. In many cases the detectability of the visual signals is limited by statistical fluctuations. Under these circumstances it is possible, in principle, to calculate the best possible signal detection performance. We find that under some conditions humans can closely approach this limiting performance.

An experiment that tested our computer sys-tem for tumor detection in chest radiographs is described. The dat for this experiment consisted of forty-three chest films, twenty-four of which are normal and nineteen are abnormal. The abnormal films contain images of twenty-two verified lung tumors, We describe the performance of this system on receiver operating character-istics obtained by varying a set of design para-meters, and compare this performance to the individual and average performances of a group of eight radiologists examining the same set of films. On one of our choices of design parameters, our computer system missed 14% of the lung tumors, and produced an average of three false positives per film. The average performance among the radiologists was a miss rate of 30% with an average of 0.2 false positives per film.

We have noticed that observer-display conditions are not usually controlled when conducting Contrast-Detail-Dose (CDD) experiments. Here we report our preliminary investigation into the visual mechanisms involved in the interpretation of CDD experi-ments and the importance of some variations in the image-observer-display conditions.

Digital radiography (DR) utilizes subtraction techniques to provide improved low-contrast detectability. A recently reported dual-potential imaging technique has demonstrated that it is possible to achieve selective tissue and bone visualization in resultant images. This paper describes some results of an ongoing project in dual-potential imaging. Three different types of digital radiographical equipment are used in this project: (1) a fan beam system, (2) an area beam system, and (3) conventional x-ray films plus an image processor. Step wedges and a Lucite® phantom with cavities for contrasting material were constructed for the experiment. Images of this phantom and of a lung-chest phantom using these three types of equipment are presented and compared.

Serial periapical dental radiographs are produced using a constant exposure geometry made possible by a simple occlusal template. Selected radiograph pairs are placed in register with the aid of a micromanipulator and a closed-circuit video system. The same video system, interfaced with an A to D converter, is used to store the respective radiographs as 6 bit images in a frame buffer. A computer is then used to subtract the two images. Gamma correction is accomplished with special software to assure that any non-linear effects attributable to contrast differences between radiographs are minimized. Blurring of the difference image, in an area known to be unchanged, provides a means for thresholding out differences attributable to registration artifacts. Preliminary data on both patients and radiographic phantoms demonstrate that diagnostic accuracy from subtracted images is markedly improved when compared to that obtainable from the undigitized control radiographs displayed conventionally.

During the first decade of CT scanner development, continuous progress has been made in scanning speed and resolution. This trend is expected to continue in the next decade as new source and detector technologies are introduced. CT has the theoretical potential to eventually rival fluoroscopy in both dynamic speed and resolution. This paper discusses the recent progress in development of the CVCT scanner, a millisecond electron-beam device, intended for high-speed multislice CT. Extensive evaluations using an electron-beam testbed device illustrate the feasibility of this approach.

The use of any ECG-gated CT heart scanning technique which uses a single scan must necessarily have gaps in the data. In third generation scanners these are view gaps, while in conventional fourth generation scanners they create incomplete views. It is possible to re-sort the raw data from a fourth generation scanner so that the views are defined in a third generation manner. This paper addresses two potential difficulties with this technique: (1) loss of resolution and (2) sensitivity to inequalities in detector gains. It is shown how the reconstruction algorithm can be modified to avoid these problems.

The magnitude and texture of statistical noise in images of positron emission computed tomography are evaluated analytically. A simple approximation method is presented for evaluating the variance and autocovariance function of noise for objects having given distributions of radionuclides and of attenuation coefficient. The comparison between the approximation method and the accurate estimation shows excellent agreements at various points in the images of uniform disc sources having constant attenuation. The dependence of noise amplification on convolution filters in image reconstruction and on the additional image processing (smoothing) is given. Considerable anisotropy of the autocovariance is observed near the periphery of the image even for a uniform disc source. This fact suggests the usefulness of spatially variant, anisotropic smoothing in some cases.

A new method for image reconstruction from smaller view angle projections less than 180 degrees, is reported. This method does not include any missing projection data estimating processes, which are often seriously jeopardized by the presence of a slight fluctuation component. In this algorithm, the projection function is expanded by using complete functions, which can compose an orthogonal set on the projection allowed data variable region. These orthogonal functions are analytically extended to the outside variable area, where the projection data are not available, and the whole object functions are obtained. Computer simulations were carried out to demonstrate the effectiveness of the method. Some practical algo-rithms for the real data calculations are also presented.

Limited-angle computed tomography was studied in a project to develop algorithms for a limited-angle scanner. Both the ART algorithm and an orthogonal function algorithm were investigated. The artifact produced by the different methods was very similar. A method is presented for producing model limited-angle artifact in phantom images. Filters are investigated for reducing the artifact, and it is shown that substantial improvement in subjective quality of images can be obtained. This method can be incorporated into limited-angle convolution back-projection algorithms.

In order to extract the information content of cardiac computed tomography scans we have devloped quantitative image processing methods for the delineation of myocardial boundaries. The computer generated images are displayed as iso-CT contour plots and CT number line section plots to allow for quantitative assessment of the rate of contrast medium perfusion and washout. From statistical criteria established in computer processed images, objective and reproducible sizing of myocardial infarctions can be made. Another application of the computer to cardiac CT scans enables myocardial wall motion measurements to be made as a function of position.

The problem of estimating an unknown density from its lower dimensional projections in a limited range of views can be put in the framework of "time and band limited functions." In some instances this leads to a highly favorable and exceptional situation: the singular value decomposition of the corresponding "Finite Radon Transform" can be accomplished and the degree of illconditioning fully analyzed. The reason for this accident is poorly understood; we give a number of examples which include the cases of X-ray as well as NMR tomography.

Image reconstruction algorithms implemented on existing CT scanners require the collection of line integrals that are evenly spaced over 360 degrees.1 In many practical situations requirements for high temporal resolution or the presence of an x-ray opaque structure prevent the measurement of all the line integrals. Attempts to use existing algorithms in this "limited data" situation result in images with severe streak artifacts.2 This paper formulates the limited data image reconstruction problem as an optimization problem. An estimate of the missing data is sought which is consistent with the measured data and any a priori knowledge about the object. An iterative procedure computes a set of error signals at each step and uses these errors to improve the missing data estimate. A variety of iterative algorithms can be derived using different methods of updating the estimate. These algorithms have been implemented on a commercial CT scanner. Examples of images with reduced streak artifact generated from limited data are presented.

We consider the application of various low-pass (noise reduction and/or smoothing) and high-pass (image restoration/enhancement) filters to reconstructed CT images. Limitations of some traditional smoothing filters are described and modified filters which overcome these disadvantages are presented. A second class of filter attempts to compensate for image degradation effects introduced by the system transfer function (influenced primarily by detector and source aperture sizes). We specifically investigate the Wiener filter as a convenient means of implementing the image restoration procedure.

The problem of image reconstruction using a minimal complete set of projection measurements in conjunction with a standard divergent reconstruction algorithm is reviewed. It is shown that by optimally weighting the projection measurements, unimpaired image quality can be obtained. Boundary conditions are developed which minimize boundary discontinuities and resulting streaks in the final image. A simple application of calculus of variations is used to find a set of weights which satisfy the boundary conditions and which minimize second order changes in the weights. Image quality comparable to reconstruction from a full 360° data set is obtained using these weights.

Measurements of low contrast resolution are most important. Xyron material .( stylene grafted poly-phenyl ether made by Asahi Medical Co.) can have an arbitrary CT number near to CT number of water. Phantoms with xyron-bar-in-water were made. But these phantoms showed different CT number between several CT scanners. They had energy dependency of CT number and were not suitable to measure low con-trast resolution. Then phantoms with xyron-bar-in-xyron were made. These phantomsshowed better differences between CT scanners than the former phantoms. These phantoms are most suitable to measure performance of low contrast resolution. CT number of xyron material decreases as X-ray tube voltage decreases. But differences of xyron bar contrast do not change so much.

A new dual rectangular photomultiplier tube has been developed for the positron CT. It has two independent segments in a single 24mm square glass envelope. The coincidence time resolution with BG0 scintillators equal to 2.9ns in FWHM and 6.2ns in FWTM was obtained for 511 KeV positron annihilation ,γ-rays. The experimental results promise the possibility of a new PCT system design-ing of a high packing ratio and spatial resolution with good optical coupling to small rectangular BGO scintillators.

An abundant number of different CT scanner models has been developed in the past ten years, meeting increasing standards of performance. From the beginning they remained a comparatively expensive piece of equipment. This is due not only to their technical complexity but is also due to the difficulties involved in assessing "true" specifications (avoiding "overde-sign"). Our aim has been to provide, for Radiation Therapy Treatment Planning, a low cost CT scanner system featuring large freedom in patient positioning. We have taken advantage of the concurrent tremendously increased amount of knowledge and experience in the technical area of CT1 . By way of extensive computer simulations we gained confidence that an inexpensive C-arm simulator gantry and a simple one phase-two pulse generator in connection with a standard x-ray tube could be used, without sacrificing image quality. These components have been complemented by a commercial high precision shaft encoder, a simple and effective fan beam collimator, a high precision, high efficiency, luminescence crystal-silicon photodiode detector with 256 channels, low noise electronic preamplifier and sampling filter stages, a simplified data aquisition system furnished by Toshiba/ Analogic and an LSI 11/23 microcomputer plus data storage disk as well as various smaller interfaces linking the electrical components. The quality of CT scan pictures of phantoms,performed by the end of last year confirmed that this simple approach is working well. As a next step we intend to upgrade this system with an array processor in order to shorten recon-struction time to one minute per slice. We estimate that the system including this processor could be manufactured for a selling price of $210,000.

A 16-plane multiwire proportional chamber has been developed to accurately map intensity pro files of heavy ion beams at the Bevalac. The imaging capability of the system has been tested for reconstruction of 3-dimensional representation of a canine thorax region using heavy ion beams.

The speed of ultrasound in soft tissues is about 1 500 m s-1, and different soft tissues have similar characteristic impedances. Reflexion and scattering occur as ultrasound travels through tissues, and the received frequency may be shifted by the Doppler effect if the target is moving. The attenuation is about 1 dB cm^-1 MHz^-1. Ultrasound is generated and detected by piezoelectric transducers. Pulse-echo techniques, giving resolution cells of the order of a few wavelengths in size, are used for A-scan, B-scan and time-position (M-mode) displays. Real-time scanners are based on mechanical or electronic beam steering, and arrays are often used. Tissue-equivalent phantoms have been developed. Doppler techniques, continuous wave and pulsed, are used for the detection of, structure motion and for blood flow studies. Pure Doppler and duplex pulse-echo/pulsed Doppler systems are very useful clinically. Transmission techniques have limited applicability. Ultrasonic diagnostic methods appear to be safe.

Diffraction and refraction effects on pulsed ultrasound were studied using simple models such as finger cots filled with saline or alcohol. The scattered ultrasound was detected with a hydrophone and A/D converter. The resulting set of digitized signals were analyzed for diffraction and refraction terms. Simple methods for backward propagating the received pressure field are described and results presented. The efficacy of diffraction tomography methods for solving the problems associated with transmission ultrasonic tomography is discussed.

Data acquisition and image reconstruction in acoustical imaging often requires high bit rates and complicated calculations in order to achieve good quality images in reasonable time. Therefore reducing the bit rates and simplifying the reconstruction algorithms are of great importance in this technology. It is possible to produce high-quality images by using only the phase information of the detected signal. The phase-only technique constitute a way of performing high-quality imaging at lower bit rates. This paper analyzes phase-only reconstruction, considers bit-rate reduction both in sampling and quantization, and develops the resolution enhancement algorithm associated with the phase-only technique.

Present medical ultrasound systems are based on energy detection methods and therefore only utilize echo intensity information. However, phase, as well as spectral information, is recorded by the transducer, which is a pressure sensitive device, but is not utilized in present display or measurement schemes. This additional information may be of diagnostic significance since the interaction between sound and tissue is exceedingly complex, since many types of tissue can be categorized in terms of their acoustical properties, and since changes in tissue acoustical properties can be correlated with specific pathological states. Thus, in principle, in vivo techniques could be devised which would extract and separate the medically significant features of the ultrasound interactions with tissue and would display ultrasonic tissue signatures appropriate for a differential diagnosis.

Piezoelectric transducers with an electrode pattern in the form of a Fresnel zone plate (FZP) on the air side and a full, planar electrode on the water side have been used in many acoustical imaging systems to focus the acoustic beam. The goal is to get an acoustic pressure pattern in the shape of an FZP. The acoustic pressure pattern, however, is never an exact replica of the FZP electrode pattern because of mechanical aberrations introduced by the piezo-electric plate. In a previous paper we derived the focal-plane acoustic field distribution of such a transducer. The distribution just in front of the transducer, however, was not given in that paper. In this paper we use a linear spatial piezo-electric model to construct a spatial transfer function for the piezoelectric plate. We show that elastic-wave generation and propagation inside the piezoelectric plate cause the acoustic field pattern just in front of the transducer to differ substantially from the electrode pattern. A plot of amplitude versus position on the piezoelectric plate shows that the sharp edges associated with the electrode pattern are severely rounded off by these effects and the amplitude is greatly reduced in the region of the outer zones. The aberrations also cause a vari-ation with the position of the phase of the out-put field.

The opto-acoustic transducer (OAT), an essential element in a real-time acoustic imaging system, is a sandwich-type structure which consists of a planar transparent indium-tinoxide electrode, a sputter-deposited CdS photoconductive layer, a piezoelectric layer, and another planar electrode. In order to make the OAT operate in real-time, the photoconductive layer must have rise and decay times of less than 3.3 psec with a suitable switching ratio of dark to illuminated resistivities. A fast-reacting photoconductive layer of CdS has been fabricated by sputtering with a gas of 99% Ar and 1% H₂S. Rise and decay times of about 3 psec have been measured. By controlling the temperature of the substrate holder, the dark and illuminated resistivities can LI made to have switching ratios of 10² Ω-cm to 10 Ω-cm. This is the correct range of values needed for real-time operation.

A comparative account is given of methods of nuclear magnetic resonance imaging for medical purposes. Following the classification scheme of Brunner and Ernst, the various methods are classified as sequential point, sequential line, sequential plane and simultaneous methods. For the purpose of imaging a single specific plane in the human body sequential plane methods are most efficient. The best examples of NMR whole body images at the present time have in fact been obtained by such methods with imaging times of a few minutes. For substantially faster NMR images the echo planar imaging method shows most promise and has already demonstrated real time images of moving objects. For the purpose of providing images of a series of planes in the object, three-dimensional simultaneous methods may be advantageous.

The requirements for spatial uniformity of the radio-frequency magnetic field used in three-dimensional MAR imaging are discussed and an improved winding distribution for a saddle-shaped single transmitter-receiver coil has been developed and tested by computer simulation of the rf mag-netic field pattern. The use of flat local or "surface" coils for NMR imaging is also proposed. A. method for correcting such images for the apparent spin density differences caused by the extreme rf magnetic field nonuniformity has been developed and tested with phantoms and images of the human back.