Detecting and classifying global dermoscopic patterns are crucial steps for detecting melanocytic lesions from
non-melanocytic ones. An important stage of melanoma diagnosis uses pattern analysis methods such as 7-point
check list, Menzies method etc. In this paper, we present a novel approach to investigate texture analysis and
classification of 5 classes of global lesion patterns (reticular, globular, cobblestone, homogeneous, and parallel
pattern) in dermoscopic images. Our statistical approach models the texture by the joint probability distribution
of filter responses using a comprehensive set of the state of the art filter banks. This distribution is represented
by the frequency histogram of filter response cluster centers called textons. We have also examined other two
methods: Joint Distribution of Intensities (JDI) and Convolutional Restricted Boltzmann Machine (CRBM) to
learn the pattern specific features to be used for textons. The classification performance is compared over the
Leung and Malik filters (LM), Root Filter Set (RFS), Maximum Response Filters (MR8), Schmid, Laws and
our proposed filter set as well as CRBM and JDI. We analyzed 375 images of the 5 classes of the patterns. Our
experiments show that the joint distribution of color (JDC) in the L*a*b* color space outperforms the other
color spaces with a correct classification rate of 86.8%.
This paper presents a novel approach in computer aided skin lesion segmentation of dermoscopic images. We
apply spatial and color features in order to model the lesion growth pattern. The decomposition is done by
repeatedly clustering pixels into dark and light sub-clusters. A novel tree structure based representation of the
lesion growth pattern is constructed by matching every pixel sub-cluster with a node in the tree structure. This
model provides a powerful framework to extract features and to train models for lesion segmentation. The model
employed allows features to be extracted at multiple layers of the tree structure, enabling a more descriptive
feature set. Additionally, there is no need for preprocessing such as color calibration or artifact disocclusion.
Preliminary features (mean over RGB color channels) are extracted for every pixel over four layers of the growth
pattern model and are used in association with radial distance as a spatial feature to segment the lesion. The
resulting per pixel feature vectors of length 13 are used in a supervised learning model for estimating parameters
and segmenting the lesion. A dataset containing 116 challenging images from dermoscopic atlases is used to
validate the method via a 10-fold cross validation procedure. Results of segmentation are compared with six
other skin lesion segmentation methods. Our method outperforms ve other methods and performs competitively
with another method. We achieve a per-pixel sensitivity/specicity of 0.890 and 0.901 respectively.
A second order tensor is usually used to describe the diffusion of water for each voxel within Diffusion Tensor
Magnetic Resonance (DT-MR) images. However, a second order tensor approximation fails to accurately
represent complex local tissue structures such as crossing fibers. Therefore, higher order tensors are used to
represent more complex diffusivity profiles. In this work we examine and compare segmentations of both second
order and fourth order DT-MR images using the Random Walker segmentation algorithm with the emphasis of
pointing-out the shortcomings of second order tensor model in segmenting regions with complex fiber structures.
We first adopt the Random Walker algorithm for segmenting diffusion tensor data by using appropriate tensor
distance metrics and then demonstrate the advantages of performing segmentation on higher order DT-MR
data. The approach proposed takes advantage of all the information provided by the tensors by using suitable
tensor distance metrics. The distance metrics used are: the Log-Euclidean for the second order tensors and
the normalized L2 distance for the fourth order tensors. The segmentation is carried out on a weighted graph
that represents the image, where the tensors are the nodes and the edge weights are computed using the tensor
distance metrics. Applying the approach to both synthetic and real DT-MRI data yields segmentations that are
both robust and qualitatively accurate.
Fractional anisotropy, defined as the distance of a diffusion tensor from its closest isotropic tensor, has been
extensively studied as quantitative anisotropy measure for diffusion tensor magnetic resonance images (DT-MRI).
It has been used to reveal the white matter profile of brain images, as guiding feature for seeding and
stopping in fiber tractography and for the diagnosis and assessment of degenerative brain diseases. Despite
its extensive use in DT-MRI community, however, not much attention has been given to the mathematical
correctness of its derivation from diffusion tensors which is achieved using Euclidean dot product in 9D space.
But, recent progress in DT-MRI has shown that the space of diffusion tensors does not form a Euclidean vector
space and thus Euclidean dot product is not appropriate for tensors. In this paper, we propose a novel and
robust rotationally invariant diffusion anisotropy measure derived using the recently proposed Log-Euclidean
and J-divergence tensor distance measures. An interesting finding of our work is that given a diffusion tensor,
its closest isotropic tensor is different for different tensor distance metrics used. We demonstrate qualitatively
that our new anisotropy measure reveals superior white matter profile of DT-MR brain images and analytically
show that it has a higher signal to noise ratio than fractional anisotropy.
Image segmentation is a method of separating an image into regions of interest, such as separating an object
from the background. The random walker image segmentation technique has been applied extensively to scalar
images and has demonstrated robust results. In this paper we propose a novel method to apply the random walker
method to segmenting non-scalar diffusion tensor magnetic resonance imaging (DT-MRI) data. Moreover, we
used a non-parametric probability density model to provide estimates of the regional distributions enabling the
random walker method to successfully segment disconnected objects. Our approach utilizes all the information
provided by the tensors by using suitable dissimilarity tensor distance metrics. The method uses hard constraints
for the segmentation provided interactively by the user, such that certain tensors are labeled as object or
background. Then, a graph structure is created with the tensors representing the nodes and edge weights
computed using the dissimilarity tensor distance metrics. The distance metrics used are the Log-Euclidean and
the J-divergence. The results of the segmentations using these two different dissimilarity metrics are compared
and evaluated. Applying the approach to both synthetic and real DT-MRI data yields segmentations that are
both robust and qualitatively accurate.
The main challenge in an automated diagnostic system for the early diagnosis of melanoma is the correct segmentation
and classification of moles, often occluded by hair in images obtained with a dermoscope. Hair occlusion causes
segmentation algorithms to fail to identify the correct nevus border, and can cause errors in estimating texture measures.
We present a new method to identify hair in dermoscopic images using a universal approach, which can segment both
dark and light hair without prior knowledge of the hair type. First, the hair is amplified using a universal matched
filtering kernel, which generates strong responses for both dark and light hair without prejudice. Then we apply local
entropy thresholding on the response to get a raw binary hair mask. This hair mask is then refined and verified by a
model checker. The model checker includes a combination of image processing (morphological thinning and label
propagation) and mathematical (Gaussian curve fitting) techniques. The result is a clean hair mask which can be used to
segment and disocclude the hair in the image, preparing it for further segmentation and analysis. Application on real
dermoscopic images yields good results for thick hair of varying colours, from light to dark. The algorithm also performs
well on skin images with a mixture of both dark and light hair, which was not previously possible with previous hair
segmentation algorithms.
Detecting pigmented network is a crucial step for melanoma diagnosis. In this paper, we present a novel graphbased
pigment network detection method that can find and visualize round structures belonging to the pigment
network. After finding sharp changes of the luminance image by an edge detection function, the resulting binary
image is converted to a graph, and then all cyclic sub-graphs are detected. Theses cycles represent meshes that
belong to the pigment network. Then, we create a new graph of the cyclic structures based on their distance.
According to the density ratio of the new graph of the pigment network, the image is classified as "Absent" or
"Present". Being Present means that a pigment network is detected in the skin lesion. Using this approach, we
achieved an accuracy of 92.6% on five hundred unseen images.
KEYWORDS: Radiology, Visualization, Medical imaging, Magnetic resonance imaging, 3D acquisition, Windows XP, Computer aided design, Computer aided diagnosis and therapy, Computed tomography, 3D displays
Radiologists make their main analysis and diagnosis based on careful observation of medical images, although
there are all kinds of automatic methods under development. Radiologists typically use a scroll mouse to click on
an image when they find something interesting, and they also use the mouse to navigate through the image slices
in volumetric scans. Thus they perform many thousands of mouse clicks every day, causing wrist fatigue. This
paper presents a method of improving the mouse pointing performance by reducing the time taken to move the
mouse to a target. We use a dynamic Control-to-Display (C-D) ratio of the mouse, by adjusting the C-D ratio
according to the current distance to the target. In theory this reduces the difficulty of the target selection, and
also reduces the movement time. The result of preliminary study demonstrates that the speed of pointing can
be improved under certain conditions, particularly for small targets and for long distances to move. In addition,
all participants claim that this mouse speed change reduces the difficulty of selecting a small target.
This paper explores a novel approach to interactive user-guided image segmentation, using eyegaze information
as an input. The method includes three steps: 1) eyegaze tracking for providing user input, such as setting
object and background seed pixel selection; 2) an optimization method for image labeling that is constrained
or affected by user input; and 3) linking the two previous steps via a graphical user interface for displaying the
images and other controls to the user and for providing real-time visual feedback of eyegaze and seed locations,
thus enabling the interactive segmentation procedure. We developed a new graphical user interface supported
by an eyegaze tracking monitor to capture the user's eyegaze movement and fixations (as opposed to traditional
mouse moving and clicking). The user simply looks at different parts of the screen to select which image to
segment, to perform foreground and background seed placement and to set optional segmentation parameters.
There is an eyegaze-controlled "zoom" feature for difficult images containing objects with narrow parts, holes
or weak boundaries. The image is then segmented using the random walker image segmentation method. We
performed a pilot study with 7 subjects who segmented synthetic, natural and real medical images. Our results
show that getting used the new interface takes about only 5 minutes. Compared with traditional mouse-based
control, the new eyegaze approach provided a 18.6% speed improvement for more than 90% of images with high
object-background contrast. However, for low contrast and more difficult images it took longer to place seeds
using the eyegaze-based "zoom" to relax the required eyegaze accuracy of seed placement.
Inpainting, a technique originally used to restore film and photographs, is used to disocclude hair from dermascopic images of skin lesions. The technique is compared to the conventional software DullRazor, which uses linear interpolation to perform disocclusion. Comparison was performed by simulating occluding hair on a dermascopic image, applying DullRazor and inpainting and calculating the error induced. Inpainting is found to perform approximately 33% better than DullRazor's linear interpolation, and is more stable under heavy occlusion. The results are also compared to published results from two other alternatives: auto-regressive (AR) model signal extrapolation and band-limited (BL) signal interpolation.
Texture is known to predict atypicality in pigmented skin lesions. This paper describes an experiment that was
conducted to determine 1) if this textural information is present in the center of skin lesions, and 2) how color
affects the perception of this information. Images of pigmented skin lesions from three categories were shown
to subjects in such a way that only textural information could be perceived; other factors known to predict
atypicality were removed or held constant. These images were shown in both color and grayscale. Each subject
assigned a score of atypicality to each image.
The experiment was conducted on 5 subjects of varying backgrounds, including one expert. Each subject's
accuracy under each modality was measured by calculating the volume under a 3-way ROC surface. The
modalities were compared using the Dorfman-Berbaum-Metz (DBM) method of ROC analysis, giving a p-value
of 0.8611. Therefore the null hypothesis that there is no difference between the predictive power of the modalities
cannot be rejected. Also, a two one-sided test of equivalence (TOST) was performed giving a p-value pair of
< 0.01; strong evidence that the textural information is independent of color.
Additionally, the subjects' accuracies were compared to a set of random readers using the DBM and TOST
methods. This was done for accuracies under the color modality, the grayscale modality and both modalities
simultaneously. The results (all p-values < 0.001) confirm the existence of textural information predictive of
atypia in the center of pigmented skin lesions.
KEYWORDS: Magnetic resonance imaging, Wavelets, Reconstruction algorithms, Interference (communication), Signal to noise ratio, Detection and tracking algorithms, Image processing, Data modeling, Signal detection, Target detection
Some diagnostic tasks in MRI involve determining the presence of a faint feature (target) relative to a dark
background. In MR images produced by taking pixel magnitudes it is well known that the contrast between faint
features and dark backgrounds is reduced due to the Rician noise distribution. In an attempt to enhance detection
we implemented three different MRI reconstruction algorithms: the normal magnitude, phase-corrected real, and
a wavelet thresholding algorithm designed particularly for MRI noise suppression and contrast enhancement.
To compare these reconstructions, we had volunteers perform a two-alternative forced choice (2AFC) signal
detection task. The stimuli were produced from high-field head MRI images with synthetic thermal noise added
to ensure realistic backgrounds. Circular targets were located in regions of the image that were dark, but
next to bright anatomy. Images were processed using one of the three reconstruction techniques. In addition
we compared a channelized Hotelling observer (CHO) to the human observers in this task. We measured the
percentage correct in both the human and model observer experiments.
Our results showed better performance with the use of magnitude or phase-corrected real images compared
to the use of the wavelet algorithm. In particular, artifacts induced by the wavelet algorithm seem to distract
some users and produce significant inter-subject variability. This contradicts predictions based only on SNR.
The CHO matched the mean human results quite closely, demonstrating that this model observer may be used
to simulate human response in MRI target detection tasks.
KEYWORDS: Medical imaging, Radiology, 3D image processing, Human-machine interfaces, Magnetic resonance imaging, 3D acquisition, Human-computer interaction, Eye, Image enhancement, Medical research
The goal of this research was to evaluate two different stack mode layouts for 3D
medical images - a regular stack mode layout where just the topmost image was visible,
and a new stack mode layout, which included the images just before and after the main
image. We developed stripped down user interfaces to test the techniques, and designed a
look-alike radiology task using 3D artificial target stimuli implanted in the slices of
medical image volumes. The task required searching for targets and identifying the range
of slices containing the targets.
Eight naive students participated, using a within-subjects design. We measured the
response time and accuracy of subjects using the two layouts and tracked the eyegaze of
several subjects while they performed the task. Eyegaze data was divided into fixations
and saccades
Subjects were 19% slower with the new stack layout than the standard stack layout,
but 5 of the 8 subjects preferred the new layout. Analysis of the eyegaze data showed that
in the new technique, the context images on both sides were fixated once the target was
found in the topmost image. We believe that the extra time was caused by the difficulty in controlling the rate of scrolling, causing overshooting. We surmise that providing some contextual detail such as adjacent slices in the new stack mode layout is helpful to reduce cognitive load for this radiology look-alike task.
KEYWORDS: Signal to noise ratio, Commercial off the shelf technology, Magnetic resonance imaging, Phase shifts, Error analysis, Tellurium, Interference (communication), Image restoration, Image analysis, Data acquisition
Signal from fat is normally removed from MR images either by fat separation techniques that distinguish water from fat signal after the data has been received, or by fat suppression techniques that prevent the fat signal from being received. Most approaches to fat separation are variations on Dixon imaging. The primary downside to Dixon imaging is the requirement for multiple images with stationary anatomy, often with specific TEs. An alternate approach is to take only one image, estimate phase errors to correct for inhomogeneity or other effects, and then separate the water and fat using the known phase shift. This has shown promise in previously published work, but the water and fat signals were always perpendicular, requiring a fixed TE. We consider the possibility of separation from a single, phase-corrected image with an arbitrary angle between water and fat signals. We note that a change of basis will separate water and fat signals into two images with additive zero-mean Gaussian noise. However, as the angle between water and fat nears pi or 0, the noise power in the separated images increases rapidly. We discuss techniques for reducing this noise magnification.
If the phase error at each pixel in a complex-valued MRI image is known the noise in the image can be reduced resulting in improved detection of medically significant details. However, given a complex-valued MRI image, estimating the phase error at each pixel is a difficult problem. Several approaches have previously been suggested including non-linear least squares fitting and smoothing filters. We propose a new scheme based on iteratively applying a series of non-linear filters, each used to modify the estimate into greater agreement with one piece of knowledge about the problem, until the output converges to a stable estimate. We compare our results with other phase estimation and MRI denoising schemes using synthetic data.
A novel automatic method of segmenting nevi is explained and analyzed in this paper. The first step in nevi segmentation is to iteratively apply an adaptive mean shift filter to form clusters in the image and to remove noise. The goal of this step is to remove differences in skin intensity and hairs from the image, while still preserving the shape of nevi present on the skin. Each iteration of the mean shift filter changes pixel values to be a weighted average of pixels in its neighborhood. Some new extensions to the mean shift filter are proposed to allow for better segmentation of nevi from the skin. The kernel, that describes how the pixels in its neighborhood will be averaged, is adaptive; the shape of the kernel is a function of the local histogram. After initial clustering, a simple merging of clusters is done. Finally, clusters that are local minima are found and analyzed to determine which clusters are nevi. When this algorithm was compared to an assessment by an expert dermatologist, it showed a sensitivity rate and diagnostic accuracy of over 95% on the test set, for nevi larger than 1.5mm.
As radiologists progress from reading images presented on film to modern computer systems with images presented on high-resolution displays, many new problems arise. Although the digital medium has many advantages, the radiologist’s job becomes cluttered with many new tasks related to image manipulation. This paper presents our solution for supporting radiologists’ interpretation of digital images by automating image presentation during sequential interpretation steps. Our method supports scenario based interpretation, which group data temporally, according to the mental paradigm of the physician. We extended current hanging protocols with support for “stages”. A stage reflects the presentation of digital information required to complete a single step within a complex task. We demonstrated the benefits of staging in a user study with 20 lay subjects involved in a visual conjunctive search for targets, similar to a radiology task of identifying anatomical abnormalities. We designed a task and a set of stimuli which allowed us to simulate the interpretation workflow from a typical radiology scenario - reading a chest computed radiography exam when a prior study is also available. The simulation was possible by abstracting the radiologist’s task and the basic workstation navigation functionality. We introduced “Stages,” an interaction technique attuned to the radiologist’s interpretation task. Compared to the traditional user interface, Stages generated a 14% reduction in the average interpretation.
It has been shown that the presence of a blood oxygen level dependent (BOLD) signal in high-field (3T and higher) fMRI datasets can cause stimulus-correlated registration errors, especially when using a least-squares registration method. These errors can result in systematic inaccuracies in activation detection. The authors have recently proposed a new method to solve both the registration and activation detection least-squares problems simultaneously. This paper gives an outline of the new method, and demonstrates its robustness on simulated fMRI datasets containing various combinations of motion and activation. In addition to a discussion of the merits of the method and details on how it can be efficiently implemented, it is shown that, compared to the standard approach, the new method
consistently reduces false-positive activations by two thirds and reduces false-negative activations by one third.
The goal is to provide a smooth, efficient and automatic display for interpretation of medical images by using a new generation of hanging protocols (HPs). HPs refer to a set of rules defining the way images are arranged on the computer screen immediately after opening a case. HPs usually include information regarding placement of the sequences, viewing mode, layout, window width and level (W/L) settings, zoom and pan. We present the results of a survey of 8 radiologists on (1) the necessity of using HPs, (2) the applicability of a hierarchical organization of HPs and (3) the number of HPs required for interpretation. We discuss some limitations and challenges associated with the HP including automatic placement of the series on the screen despite non-standard series labeling, generation of pseudo-series, creation of the 'study context' and identification of relevant priors, and image display standardization with automatic orientation and shuttering. The paper also addresses the HP selection based on the workstation's hardware such as number and type of monitors, size of the study, and presence of image processing routines tailored to the information needs and level of expertise of particular users. Our 'heads-up' approach is meant to free the user's conscious processing for reasoning such as detection of patterns so allowing for the execution of the tasks in an efficient, yet highly adaptive manner, sensitive to shifting concepts. Automation of routine tasks is maximized through the creation of shortcuts and macros embedded in features like multi-stage HP.
In the radiology workstation design, the race for adding more features is now morphing into an iterative user centric design with the focus on ergonomics and usability. The extent of the list of features for the radiology workstation used to be one of the most significant factors for a Picture Archiving and Communication System (PACS) vendor's ability to sell the radiology workstation. Not anymore is now very much the same between the major players in the PACS market. How these features work together distinguishes different radiology workstations. Integration (with the PACS/Radiology Information System (RIS) systems, with the 3D tool, Reporting Tool etc.), usability (user specific preferences, advanced display protocols, smart activation of tools etc.) and efficiency (what is the output a radiologist can generate with the workstation) are now core factors for selecting a workstation. This paper discusses these new trends in radiology workstation design. We demonstrate the importance of the interaction between the PACS vendor (software engineers) and the customer (radiologists) during the radiology workstation design. We focus on iterative aspects of the workstation development, such as the presentation of early prototypes to as many representative users as possible during the software development cycle and present the results of a survey of 8 radiologists on designing a radiology workstation.
Other researchers have proposed that the brain parenchymal fraction (or brain atrophy) may be a good surrogate measure for disease progression in patients with Multiple Sclerosis. This paper considers various factors influencing the measure of the brain parenchymal fraction obtained from dual spin-echo PD and T2-weighted head MRI scans. We investigate the robustness of the brain parenchymal fraction with respect to two factors: brain-mask border placement which determines the brain intra-dural volume, and brain scan incompleteness. We show that an automatic method for brain segmentation produces an atrophy measure which is fairly sensitive to the brain-mask placement. We also show that a robust, reproducible brain atrophy measure can be obtained from incomplete brain scans, using data in a centrally placed subvolume of the brain.
KEYWORDS: Brain, Magnetic resonance imaging, Image segmentation, Neuroimaging, Head, Data modeling, Evolutionary algorithms, Data centers, Data corrections, Internet
This paper looks at the difficulties that can confound published T1-weighted Magnetic Resonance Imaging (MRI) brain segmentation methods, and compares their strengths and weaknesses. Using data from the Internet Brain Segmentation Repository (IBSR) as a gold standard, we ran three different segmentation methods with and without correcting for intensity inhomogeneity. We then calculated the similarity index between the brain masks produced by the segmentation methods and the mask provided by the IBSR. The intensity histograms under the segmented masks were also analyzed to see if a Bi-Gaussian model could be fit onto T1 brain data. Contrary to our initial beliefs, our study found that intensity based T1-weighted segmentation methods were comparable or even superior to, methods utilizing spatial information. All methods appear to have parameters that need adjustment depending on the data set used. Furthermore, it seems that the methods we tested for intensity inhomogeneity did not improve the segmentations due to the nature of the IBSR data set.
Conventional methods to diagnose and follow treatment of Multiple Sclerosis require radiologists and technicians to compare current images with older images of a particular patient, on a slic-by-slice basis. Although there has been progress in creating 3D displays of medical images, little attempt has been made to design visual tools that emphasize change over time. We implemented several ideas that attempt to address this deficiency. In one approach, isosurfaces of segmented lesions at each time step were displayed either on the same image (each time step in a different color), or consecutively in an animation. In a second approach, voxel- wise differences between time steps were calculated and displayed statically using ray casting. Animation was used to show cumulative changes over time. Finally, in a method borrowed from computational fluid dynamics (CFD), glyphs (small arrow-like objects) were rendered with a surface model of the lesions to indicate changes at localized points.
We have implemented a prototype system consisting of a Java- based image viewer and a web server extension component for transmitting Magnetic Resonance Images (MRI) to an image viewer, to test the performance of different image retrieval techniques. We used full-resolution images, and images compressed/decompressed using the Set Partitioning in Hierarchical Trees (SPIHT) image compression algorithm. We examined the SPIHT decompression algorithm using both non- progressive and progressive transmission, focusing on the running times of the algorithm, client memory usage and garbage collection. We also compared the Java implementation with a native C++ implementation of the non- progressive SPIHT decompression variant. Our performance measurements showed that for uncompressed image retrieval using a 10Mbps Ethernet, a film of 16 MR images can be retrieved and displayed almost within interactive times. The native C++ code implementation of the client-side decoder is twice as fast as the Java decoder. If the network bandwidth is low, the high communication time for retrieving uncompressed images may be reduced by use of SPIHT-compressed images, although the image quality is then degraded. To provide diagnostic quality images, we also investigated the retrieval of up to 3 images on a MR film at full-resolution, using progressive SPIHT decompression. The Java-based implementation of progressive decompression performed badly, mainly due to the memory requirements for maintaining the image states, and the high cost of execution of the Java garbage collector. Hence, in systems where the bandwidth is high, such as found in a hospital intranet, SPIHT image compression does not provide advantages for image retrieval performance.
Medical images are increasingly being examined on computer monitors. In contrast to the traditional film viewbox, the use of computer displays often involves a trade-off between the number and size of images shown and the available screen space. This paper focuses on two solutions to this problem: the thumbnail technique and the detail-in-context technique. The thumbnail technique, implemented in many current commercial medical imaging systems, presents an overview of the images in a thumbnail bar while selected images are magnified in a separate window. Our earlier work suggested the use of a detail-in-context technique which displays all images in one window utilizing multiple magnification levels. We conducted a controlled experiment to evaluate both techniques. No significant difference was found for performance and preference. However, differences were found in the interaction patterns and comments provided by the participants. The detail-in-context technique accommodated many individual strategies and offered good capabilities for comparing different images whereas the thumbnail technique strongly encouraged sequential examination of the images and allowed for high magnification factors. Given the results of this study, our research suggests new alternatives to the presentation of medical images and provides an increased understanding of the usability of existing medical image viewing methods.
Recognizing that conspicuous multiple sclerosis (MS) lesions have high intensities in both dual-echo T2 and PD-weighted MR brain images, we show that it is possible to automatically determine a thresholding mechanism to locate conspicuous lesion pixels and also to identify pixels that suffer from reduced intensity due to partial volume effects. To do so, we first transform a T2-PD feature space via a log(T2)- log(T2+PD) remapping. In the feature space, we note that each MR slice, and in fact the whole brain, is approximately transformed into a line structure. Pixels high in both T2 and PD, corresponding to candidate conspicuous lesion pixels, also fall near this line. Therefore we first preprocess images to achieve RF-correction, isolation of the brain, and rescaling of image pixels into the range 0 - 255. Then, following remapping to log space, we find the main linear structure in feature space using a robust estimator that discounts outliers. We first extract the larger conspicuous lesions which do not show partial volume effects by performing a second robust regression for 1D distances along the line. The robust estimator concomitantly produces a threshold for outliers, which we identify with conspicuous lesion pixels in the high region. Finally, we perform a third regression on the conspicuous lesion pixels alone, producing a 2D conspicuous lesion line and confidence interval band. This band can be projected back into the adjacent, non-conspicuous, region to identify tissue pixels which have been subjected to the partial volume effect.
The generation of magnitude magnetic resonance images comprises a sequence of data encodings or transformations, from detection of an analog electrical signal to a digital phase/frequency k-space to a complex image space via an inverse Fourier transform and finally to a magnitude image space via a magnitude transformation and rescaling. Noise present in the original signal is transformed at each step of this sequence. Denoising MR images from low field strength scanners is important because such images exhibit low signal to noise ratio. Algorithms that perform denoising of magnetic resonance images may be usefully classified according to the data domain on which they operate (i.e. at which step of the sequence of transformations they are applied) and the underlying statistical distribution of the noise they assume. This latter dimension is important because the noise distribution for low SNR images may be decidedly non-Gaussian. Examples of denoising algorithms include 2D wavelet thresholding (operates on the wavelet transform of the magnitude image; assumes Gaussian noise), Nowak's 2D wavelet filter (operates on the squared wavelet transform of the magnitude image; assumes Rician noise), Alexander et. al.'s complex 2D filters (operates on the wavelet transform of the complex image space; assumes Gaussian noise), wavelet packet denoising (wavelet packet transformation of magnitude image; assumes Rician noise) and anisotropic diffusion filtering (operates directly on magnitude image; no assumptions on noise distribution). Effective denoising of MR images must take into account both the availability of the underlying data, and the distribution of the noise to be removed. We classify a number of recently published denoising algorithms and compare their performance on images from a 0.35T permanent magnet MR scanner.
One of the important clinical features to differentiate benign melanocytic nevi from malignant melanomas is the irregularity of the lesion border. A careful examination of a lesion border reveals two types of irregularity: texture irregularity and structure irregularity. Texture irregularities are the small variations along the border, while structure irregularities are the global indentations and protrusions, which may suggest excess of cell growth or regression of a melanoma. Therefore, measuring border irregularity by structural indentations and protrusions may detect the malignancy of the lesion. The common shape descriptors such as compactness index and fractal dimension are more sensitive to texture irregularities than structure irregularities. They do not provide an accurate estimation for the structure irregularity. Therefore, we have designed a new measurement for border irregularity. The proposed method first locates all indentations and protrusions along the lesion border. Then a new area-based index, called irregularity index, is computed for each indentation and protrusion. The overall border irregularity is estimated by the sum of all individual indices. In addition, the new method offers an extra feature: localization of the significant indentations and protrusions. As the result, the new measure is sensitive to structure irregularities and may be useful for diagnosing melanomas.
This paper describes the compression of grayscale medical ultrasound images using a new compression technique, space- frequency segmentation. This method finds the rate- distortion optimal representation of an image from a large set of possible space-frequency partitions and quantizer combinations. The method is especially effective when the images to code are statistically inhomogeneous, which is the case for medical ultrasound images. We implemented a real compression algorithm based on this method, and applied the resulting algorithm to representation ultrasound images. The result is an effective technique that performs significantly better than a current leading wavelet transform coding algorithm, Set Partitioning In Hierarchical Trees (SPIHT), using the standard objective PSNR distortion measure. The performance of our space-frequency codec is illustrated, and the space-frequency partitions described. To obtain a qualitative measure of our method's performance, we describe an expert viewer study, where images compressed using both space-frequency compression and SPIHT were presented to ultrasound radiologists to obtain expert viewer assessment of the differences in quality between images from the two different methods. The expert viewer study showed the improved quality of space-frequency compressed images compared to SPIHT compressed images.
A new numerical wavelet transform, the discrete torus wavelet transform, is described and an application is given to the denoising of abdominal magnetic resonance imaging (MRI) data. The discrete tori wavelet transform is an undecimated wavelet transform which is computed using a discrete Fourier transform and multiplication instead of by direct convolution in the image domain. This approach leads to a decomposition of the image onto frames in the space of square summable functions on the discrete torus, l2(T2). The new transform was compared to the traditional decimated wavelet transform in its ability to denoise MRI data. By using denoised images as the basis for the computation of a nuclear magnetic resonance spin-spin relaxation-time map through least squares curve fitting, an error map was generated that was used to assess the performance of the denoising algorithms. The discrete torus wavelet transform outperformed the traditional wavelet transform in 88% of the T2 error map denoising tests with phantoms and gynecologic MRI images.
This paper examines the presentation of MRI on a computer screen. In order to understand the issues involved with the diagnostic-viewing task performed by the radiologist, field observations were obtained in the traditional light screen environment. Requirement issues uncovered included: user control over grouping, size and position of images; navigation of imags and image groups; and provision of both presentation detail and presentation context. Existing presentation techniques and variations were explored in order to obtain an initial design direction to address these issues.
This paper presents an automatic computer system for analyzing the structural shape of cutaneous melanocytic lesion borders. The computer system consists of two steps: pre-preprocessing the skin lesion images and lesion border shape analysis. In the preprocessing step, the lesion border is extracted from the skin images after the dark thick hairs are removed by a program called DullRazor. The second step analyzes the structural shape of the lesion border using a new measure called sigma-ratio. The new measure is derived from scale- space filtering technique with an extended scale-space image. When comparing the new measure with other common shape descriptors, such as compactness index and fractal dimensional, sigma-ratio is more sensitive to the structural protrusions and indentations. In addition, the extended scale- space image can be used to pinpoint the locations of the structural indentations and protrusions, the potential problem areas of the lesion.
This paper describes an intensity-based method for the segmentation of multiple sclerosis lesions in dual-echo PD and T2-weighted magnetic resonance brain images. The method consists of two stages: feature extraction and image analysis. For feature extraction, we use a ratio filter transformation on the proton density (PD) and spin-spin (T2) data sequences to extract the white matter, cerebrospinal fluid and the lesion features. The one and two dimensional histograms of the features are then analyzed to obtain different parameters, which provide the basis for subsequent image analysis operations to detect the multiple sclerosis lesions. In the image analysis stage, the PD images of the volume are first pre-processed to enhance the lesion tissue areas. White matter and cerebrospinal fluid masks are then generated and applied on the enhanced volume to remove non- lesion areas. Segmentation of lesions is performed in two steps: conspicuous lesions are extracted in the first step, followed by the extraction of the subtle lesions.
In this paper, a novel image compression scheme is presented, which is specially suited for image transmission over a narrow-band network typically required for telemedicine to remote regions. A wavelet compression algorithm is enhanced with the feature of dynamically compressing different regions of the image. This feature is provided while keeping the algorithm's embedding ability, which leads to an 'importance' embedding rather than the traditional 'energy' based embedding. To incorporate regions in a wavelet-based compression algorithm the region edges were carefully tuned to eliminate the negative influence that the wavelet transform has on the region algorithm. Test of this new algorithm on standard test images and ultrasound images showed that both the dynamic and region-based features could be incorporated into the wavelet algorithm with only a small overhead.
A new task-oriented image quality metric is used to quantify the effects of distortion introduced into magnetic resonance images by lossy compression. This metric measures the similarity between a radiologist's manual segmentation of pathological features in the original images and the automated segmentations performed on the original and compressed images. The images are compressed using a general wavelet-based lossy image compression technique, embedded zerotree coding, and segmented using a three-dimensional stochastic model-based tissue segmentation algorithm. The performance of the compression system is then enhanced by compressing different regions of the image volume at different bit rates, guided by prior knowledge about the location of important anatomical regions in the image. Application of the new system to magnetic resonance images is shown to produce compression results superior to the conventional methods, both subjectively and with respect to the segmentation similarity metric.
Imaginer is a graphical user interface currently being developed for automated analysis of emission images. It will be the first application to implement a new feature extraction method of contiguous volume analysis on an unlimited number of image formats. Its development was prompted by the desire to simplify the steps involved in that analysis and to improve visualization and interpretation of results through a graphical user interface. This paper discusses difficulties that have arisen in generalizing the method of contiguous volume analysis to work with an unlimited number of image formats, as well as in abstracting the visualization techniques to effectively represent all types of data used during analysis. Issues in creating a flexible, intuitive, and extensible user interface for scientific investigation and clinical use are discussed, along with several usability issues that have arisen during development. Prototypes of Imaginer and its software components are described. Designed for ease of use, flexibility, extensibility and portability, Imaginer will enable users to assess the appropriateness of this method of feature extraction for various clinical and research purposes, and to use it in contexts for which it is found to be appropriate.
This paper describes current iterative surface matching methods for registration, and our new extensions. Surface matching methods use two segmented surfaces as features (one dynamic and one static) and iteratively search parameter space for an optimal correlation. To compare the surfaces we use an anisotropic Euclidean chamfer distance transform, based on the static surface. This type of DT was analyzed to quantify the errors associated with it. Hierarchical levels are attained by sampling the dynamic surface at various rates. In using the reduced amount of data provided by the surface segmentation each hierarchical level is formed quickly and easily and only a single distance transform is needed, thus increasing efficiency. Our registrations were performed in a data-flow environment created for multipurpose image processing. The new modifications were tested on a large number of simulations, over a wide range of rigid body transformations and distortions. Multimodality, and multipatient registration tests were also completed. A thorough examination of these modifications in conjunction with various minimization methods was then performed. Our new approaches provide accuracy and robustness, while requiring less time and effort than conventional methods.
An important first step in diagnosis and treatment planning using tomographic imaging is differentiating and quantifying diseased as well as healthy tissue. One of the difficulties encountered in solving this problem to date has been distinguishing the partial volume constituents of each voxel in the image volume. Most proposed solutions to this problem involve analysis of planar images, in sequence, in two dimensions only. We have extended a model-based method of image segmentation which applies the technique of iterated conditional modes in three dimensions. A minimum of user intervention is required to train the algorithm. Partial volume estimates for each voxel in the image are obtained yielding fractional compositions of multiple tissue types for individual voxels. A multispectral approach is applied, where spatially registered data sets are available. The algorithm is simple and has been parallelized using a dataflow programming environment to reduce the computational burden. The algorithm has been used to segment dual echo MRI data sets of multiple sclerosis patients using lesions, gray matter, white matter, and cerebrospinal fluid as the partial volume constituents. The results of the application of the algorithm to these datasets is presented and compared to the manual lesion segmentation of the same data.
An algorithm has been developed which uses stochastic relaxation in three dimensions to segment brain tissues from images acquired using multiple echo sequences from magnetic resonance imaging (MRI). The initial volume data is assumed to represent a locally dependent Markov random field. Partial volume estimates for each voxel are obtained yielding fractional composition of multiple tissue types for individual voxels. A minimum of user intervention is required to train the algorithm by requiring the manual outlining of regions of interest in a sample image from the volume. Segmentations obtained from multiple echo sequences are determined independently and then combined by forming the product of the probabilities for each tissues type. The implementation has been parallelized using a dataflow programming environment to reduce the computational burden. The algorithm has been used to segment 3D MRI data sets using multiple sclerosis lesions, gray matter, white matter, and cerebrospinal fluid as the partial volumes. Results correspond well with manual segmentations of the same data.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.