Currently, breast cancer screening protocols are based on a woman's age, but not on other risk factors or on the physical
characteristics of her breasts. One commonly cited risk factor is dense breast tissue. This study is part of an effort to
provide basic information needed to develop automatically, individualized screening protocols, by clarifying the
relationships between age, risk, breast composition, lesion conspicuity, and other factors. In this project, a database was
established that includes 227 cancer negative cases and 116 cancer positive cases across a wide range of age groups. In
the cancer positive cases, we included a subgroup in which the cancer had been missed in the previous exam. Using our
physics based model of breast density, we quantified percentage of breast parenchyma as an index of density. Density
distributions and changes over time were analyzed. The most significant finding within this data was a significantly
slower density decrease over the time in the cancer positive group than in the cancer negative group, with no overall
difference in the density distribution in those two groups. False negative cases were found to be significantly more dense
than true positive cases. In addition, our results showed a trend of density decrease with increasing age, which is in
agreement with others' widely reported results.
Using electrical impedance spectroscopy (EIS) technology to detect breast abnormalities in general and cancer
in particular has been attracting research interests for decades. Large clinical tests suggest that current EIS systems can
achieve high specificity (≥ 90%) at a relatively low sensitivity ranging from 15% to 35%. In this study, we explore a new
resonance frequency based electrical impedance spectroscopy (REIS) technology to measure breast tissue EIS signals in
vivo, which aims to be more sensitive to small tissue changes. Through collaboration between our imaging research
group and a commercial company, a unique prototype REIS system has been assembled and preliminary signal
acquisition has commenced. This REIS system has two detection probes mounted in the two ends of a Y-shape support
device with probe separation of 60 mm. During REIS measurement, one probe touches the nipple and the other touches
to an outer point of the breast. The electronic system continuously generates sweeps of multi-frequency electrical pulses
ranging from 100 to 4100 kHz. The maximum electric voltage and the current applied to the probes are 1.5V and 30mA,
respectively. Once a "record" command is entered, multi-frequency sweeps are recorded every 12 seconds until the
program receives a "stop recording" command. In our imaging center, we have collected REIS measurements from 150
women under an IRB approved protocol. The database includes 58 biopsy cases, 78 screening negative cases, and other
"recalled" cases (for additional imaging procedures). We measured eight signal features from the effective REIS sweep
of each breast. We applied a multi-feature based artificial neural network (ANN) to classify between "biopsy" and
normal "non-biopsy" breasts. The ANN performance is evaluated using a leave-one-out validation method and ROC
analysis. We conducted two experiments. The first experiment attempted to classify 58 "biopsy" breasts and 58 "non-biopsy"
breasts acquired on 58 women each having one breast recommended for biopsy. The second experiment
attempted to classify 58 "biopsy" breasts and 58 negative breasts from the set of screening negative cases. The areas
under ROC curves are 0.679 ± 0.033 and 0.606 ± 0.035 for the first and the second experiment, respectively. The
preliminary results demonstrate (1) even with this rudimentary system with only one paired probes there is a measurable
signal of changes in breast tissue demonstrating the feasibility of applying REIS technology for identifying at least some
women with highly suspicious breast abnormalities and (2) the electromagnetic asymmetry between two breasts may be
more sensitive in detecting changes in the abnormal breast. To further improve the REIS system performance, we are
currently designing a new REIS system with multiple electrical probes and a more sophisticated analysis scheme.
Xiao Hui Wang, Janet Durick, David Herbert, Amy Lu, Sarawathi Golla, Dilip Shinde, Samaia Piracha, Kristin Foley, Carl Fuhrman, Betty Shindel, J. Ken Leader, Walter Good
To improve radiologist's performance in lesion detection and diagnosis on 3D medical image dataset, we have conducted a pilot study to test viability and efficiency of the stereo display for lung nodule detection and classification. Using our previously developed stereo compositing methods, stereo image pairs were prestaged and precalculated from CT slices for real-time interactive display. Three display modes (i.e., stereoscopic 3D, orthogonal MIP and slice-by-slice) were compared for lung nodule detection and total of eight radiologists have participated this pilot study to interpret the images. The performance of lung nodule detection was analyzed and compared between the modes using FROC analysis. Subjective assessment indicates that stereo display was well accepted by the radiologists, despite some uncertainty of beneficial results due to the novelty of the display. The FROC analysis indicates a trend that, among the three display modes, stereo display resulted in the best performance of nodule detection followed by slice-based display, although no statistically significant difference was shown between the three modes. The stereo display of a stack of thin CT slices has the potential to clarify three-dimensional structures, while avoiding ambiguities due to tissue superposition. Few studies, however, have addressed actual utility of stereo display for medical diagnosis. Our preliminary results suggest a potential role of stereo display for improving radiologists' performance in medical detection and diagnosis, and also indicate some factors likely affect the performance with new display, such as novelty of the display, training effect from projected radiography interpretation and confidence with the new technology.
Many diagnostic problems involve the assessment of vascular structures or bronchial trees depicted in volumetric
datasets, but previous algorithms for segmenting cylindrical structures are not sufficiently robust for them to be widely
applied clinically. Local geometric information that is of importance in segmentation consists of voxel values and their
first and second derivatives. First derivatives can be generalized to the gradient and more generally the structure tensor,
while the second derivatives can be represented by Hessian matrices. It is desirable to exploit both kinds of information,
at the same time, in any voxel classification process, but few segmentation algorithms have attempted to do this. This
project compares segmentation based on the structure tensor to that based on the Hessian matrix, and attempts to
determine whether some combination of the two can demonstrate better performance than either individually. To
compare performance in a situation where a gold standard exists, the methods were tested on simulated tree structures.
We generated 3D tree structures with varying amounts of added noise, and processed them with algorithms based on the
structure tensor, the Hessian matrix, and a combination of the two. We applied an orientation-sensitive filter to smooth
the tensor fields. The results suggest that the structure tensor by itself is more effective in detecting cylindrical structures
than the Hessian tensor, and the combined tensor is better than either of the other tensors.
A workstation for testing the efficacy of stereographic displays for applications in radiology has been developed, and is currently being tested on lung CT exams acquired for lung cancer screening. The system exploits pre-staged rendering to achieve real-time dynamic display of slabs, where slab thickness, axial position, rendering method, brightness and contrast are interactively controlled by viewers. Stereo presentation is achieved by use of either frame-swapping images or cross-polarizing images. The system enables viewers to toggle between alternative renderings such as one using
distance-weighted ray casting by maximum-intensity-projection, which is optimal for detection of small features in many cases, and ray casting by distance-weighted averaging, for characterizing features once detected. A reporting mechanism is provided which allows viewers to use a stereo cursor to measure and mark the 3D locations of specific features of interest, after which a pop-up dialog box appears for entering findings. The system's impact on performance is being tested on chest CT exams for lung cancer screening. Radiologists' subjective assessments have been solicited for other
kinds of 3D exams (e.g., breast MRI) and their responses have been positive. Objective estimates of changes in performance and efficiency, however, must await the conclusion of our study.
Based on the need to increase the efficacy of chest CT for lung cancer screening, a stereoscopic display for viewing chest CT images has been developed. Stereo image pairs are generated from CT data by conventional stereo projection derived from a geometry that assumes the topmost slice being displayed is at the same distance as the screen of the physical display. Image grayscales are modified to make air transparent so that soft tissue structures of interest can be more easily seen. Because the process of combining multiple slices has a tendency to reduce the effective local contrast, we have included mechanisms to counteract this, such as linear and nonlinear local grayscale transforms. The physical display, which consists of a CRT viewed through shutter glasses, also provides for real-time adjustment of displayed thickness and axial position, as well as for changing brightness and contrast. While refinement of the stereo projection, contrast, and transparency models is ongoing, subjective evaluation of our current implementation indicates that the method has considerable potential for improving the efficiency of the detection of lung nodules. A more quantitative effort to assess its impact on performance, by ROC type methods, is underway.
The widespread adoption of chest CT for lung cancer screening will greatly increase the workload of chest radiologists. Contributing to this effort is the need for radiologists to differentiate between localized nodules and slices through linear structures such as blood vessels, in each of a large number of slices acquired for each subject. To increase efficiency and accuracy, thin slices can be combined to provide thicker slabs for presentation, but the resulting superposition of tissues can make it more difficult to detect and characterize smaller nodules. The stereo display of a stack of thin CT slices may be able to clarify three-dimensional structures, while avoiding the loss of resolution and ambiguities due to tissue superposition.
The current work focuses on the development and evaluation of stereo projection models that are appropriate for chest CT. As slices are combined into a three dimensional structure, maximum image intensity, which is limited by the display, must be preserved. But, compositing methods that effectively average slices together typically reduce contrast of subtle nodules. For monoscopic viewing, orthographic maximum-intensity projection (MIP), of thick slabs, has been employed to overcome this effect, but this method provides no information of depth or of the geometrical relationships between structures. Our comparison of various rendering options indicates that a stereographic perspective transformation, used in conjunction with a compositing model that combines maximum-intensity projection with an appropriate brightness weighting function, shows promise for this application. The main drawback uncovered was that, for the images used in this study, the lung volume was undersampled in the z-direction, resulting in certain unavoidable image artifacts.
We developed and tested an automated scheme to segment lung areas depicted in CT images. The scheme includes a series of six steps. 1) Filtering and removing pixels outside the scanned anatomic structures. 2) Segmenting the potential lung areas using an adaptive threshold based on pixel value distribution in each CT slice. 3) Labeling all selected pixels ingo segmented regions and deleting isolated regions in non-lung area. 4) Labeling and filling interior cavities (e.g., pleural nodules, airway wall, and major blood vessels) inside lung areas. 5) Detecting and deleting the main airways (e.g., trachea and central bronchi) connected to the segmented lung areas. 6) Detecting and separating possible anterior or posterior junctions between the lungs. Five lung CT cases (7-10 mm in slice thickness) with variety of disease patterns were used to train or set up the classification rules in the scheme. Fifty examinations of emphysema patients were then used to test the scheme. The results were compared with the results generated from a semi-automated method with manual interaction by an expert observer. The experimental results showed that the average difference in estimated lung volumes between the automated scheme and manually corrected approach was 2.91%±0.88%. Visual examination of segmentation results indicated that the difference of the two methods was larger in the areas near the apices and the diaphragm. This preliminary study demonstrated that a simple multi-stage scheme had potential of eliminating the need for manual interaction during lunch segmentation. Hence, it can ultimately be integrated into computer schemes for quantitative analysis and diagnosis of lung diseases.
A novel technique for assessing local and global differences between mammographic images was developed. This method uses correlations between abstract features extracted from corresponding views to compare image properties without resorting to processes that depend on exact geometrical congruence, such as image subtraction, which have a tendency to produce excessive artifact. The method begins by normalizing both digitized mammograms, after which a series of global and local feature filters are applied to each image. Each filter calculates values characterizing a particular property of the given image, and these values, for each property of interest are arranged in a feature vector. Corresponding elements in the two feature vectors are combined to produce a difference vector that indicates the change in the particular properties between images. Features are selected which are expected to be relatively invariant with respect to breast compression.
Breast tissue density is one of the most cited risk factors in breast cancer development. Nevertheless, estimates of the magnitude of breast cancer risk associated with density vary substantially because of the inadequacy of methods used in tissue density assessment (e.g., subjective and/or qualitative assessment) and lack of a reliable gold standard. We have developed automated algorithms for quantitatively measuring breast composition from digitized mammograms. The results were compared to objective truth as determined by quantitative measures from breast MR images, as well as to subjective truth as determined by radiologists' readings from digitized mammograms using BI-RAD standards. Higher linear correlation between estimates calculated from mammograms using the methods developed herein and estimates derived from breast MR images demonstrates that the mammography-based methods will likely improve our ability to accurately determine the breast cancer risk associated with breast density. By using volumetric measures from breast MR images as a gold standard, we are able to estimate the adequacy and accuracy of our algorithms. The results can be used for providing a calibrated method for estimating breast composition from mammograms.
The purpose was to evaluate the effect of incorporating negative but suspicious regions into a knowledge-based computer-aided detection (CAD) scheme of masses depicted in mammograms. To determine if a suspicious region is positive for a mass, the region was compared not only with actually positive regions (masses), but also with known negative regions. A set of quantitative measures (i.e., a positive, a negative, and a combined likelihood measure) was computed. In addition, a process was developed to integrate two likelihood measures that were derived using two selected features. An initial evaluation with 300 positive and 300 negative regions was performed to determine the parameters associated with the likelihood measures. Then, an independent set of 500 positive and 500 negative regions was used to test the performance of the CAD scheme. During the training phase, the performance was improved from Az=0.83 to 0.87 with the incorporation of negative regions and the integration process. During the independent test, the performance was improved from Az=0.80 to 0.83. The incorporation of negative regions and the integration process was found to add information to the scheme. Hence, it may offer a relatively robust solution to differentiate masses from normal tissue in mammograms.
We present a simple algorithm for determining the fat fraction in magnetic resonance images of the breast. These computed values are intended to help train neural networks for determining breast composition from x-ray mammograms. The method relies on simple intensity thresholding to form a binary mask followed by morphological dilations and erosions, automated region selection and clustering the tissues within the mask into fat and parenchymal components. Correcting the image intensity nonuniformity due to the spatial sensitivity profile of the breast coil was found to be essential and easily accomplished with homologous filtering. In the absence of large artifacts, the algorithm was able to accurately calculate breast fat fractions.
A novel figure-of-merit (FOM) for automatically quantifying the types of artifacts that appear in compressed images was investigated. This FOM is based on task specific linear combinations of magnitude, frequency and 'localized' structure information derived from difference images. For each elemental diagnostic task (e.g., detection of microcalcifications) a value is calculated as the weighted linear combination of the output of an array of filters, and the FOM is defined to be the maximum of these values, taken over all relevant diagnostic tasks. This FOM was tested by applying it to a previously assembled set of 60 mammograms that had been digitized and compressed at five different compression levels using our version of the original JPEG algorithm. The FOM results were compared to subjective assessments of image quality provided by nine radiologists. A subset consisting of 25 images was also processed with the JPEG 2000 algorithm and evaluated by the FOM. A significant correlation existed between readers' subjective ratings and FOMs for JPEG compressed images. A comparison between the results of the two compression algorithms reveals that, to achieve a comparable FOM level, the JPEG 2000 images were compressed at a bitrate that was typically 15% lower than that of images compressed with the original JPEG algorithm.
In this study, we test a new method to automatically search for matched regions in bilateral digitized mammograms and to compute differences in region conspicuities in pairs of matched regions. One hundred pairs of bilateral images of the same view were selected for the experiment. Each pair of images depicted one verified mass. These 100 mass regions, along with 356 suspicious but actually negative mass regions, were first detected by a single-image-based CAD scheme. To find the matched regions in the corresponding bilateral images, a Procrustean-type technique was used to register the two images, which corrects the deformation of tissue structure between images by guaranteeing the registration of nipples, skin lines, and chest walls. Then, a region growth algorithm was applied to generate a growth region in the matched area, which has the same effective size as the suspicious region in the abnormal image. The conspicuities in the two matched regions, as well as their differences, were computed. Using the conspicuity in the original mass regions and the difference of conspicuities in the two matched regions as two identification indices to classify this set of 456 suspicious regions, the computed areas under the ROC curves (Az) were 0.77 and 0.75, respectively. This preliminary study indicates that by comparing the difference of conspicuities in two matched regions that a very useful feature for the CAD schemes can be extracted.
In this paper, a novel method is used for computerized lesion detection and analysis in three-dimensional(3D) contrast enhanced MR breast images. The automatic analysis involves three steps: 1) alignment between series; 2) extraction of suspicious regions; and 3) application of feature classification to each region. Assuming that there are only small geometric deformations after global registration, we adopted a 3D thin-plate spline based registration method, in which the control points are determined using 3D gradient and local correlation. Experiments show superior correlation between neighboring slices with 3D alignment as compared to a previous two-dimensional(2D) method. After registration, a new series named enhancement rate images(ERIs) are created. Suspicious volumes-of-interest(VOIs) are identified by 3D region labeling after thresholding the ERIs. Since carcinomas can typically be characterized by irregular borders and rapid and high uptake of contrast followed by a washout, a set of morphological features(irregularity, spiculation index, etc) and enhancement features(small volume enhancement rate, slope of average rate, etc) are calculated for selected VOIs and evaluated in a rule-based classifier to identify malignant lesions from benign lesions or normal tissues.
The wide dynamic range present in digitized mammographic data, partially resulting from the non-uniform thickness of tissue during breast compression, makes it difficult to find window and level values that are appropriate to display the entire image. Further, this factor combined with the non- linearity of the relationship between density and log exposure, confound attempts to automatically derive tissue composition information directly from uncorrected data. This project attempts to address these issues by making appropriate local image corrections based on the characteristic curves of film and digitizer, as well as on the variations in tissue thickness during breast compression. Subjective comparisons of the display techniques developed in this project, to mammography displays based on local histogram equalization methods to reduce image dynamic range, clearly demonstrate superior performance of the methods presented in this paper. In addition to this subjective observation about image display, we also investigated the possibility of using corrected data to improve the performance of tissue composition measurements. A neural network classifier was developed to use features derived from the volume-corrected histogram of the corrected mammographic data to estimate tissue composition. Results indicate that tissue composition measurements are more highly correlated to radiologists' estimates, when they are derived from corrected images.
Registration of mammograms is frequently used in computer- aided-detection algorithms, and has been considered for use in the analysis of temporal sequences of screening exams. Previous image registration methods, employing affine transformations or Procrustean transforms based ona small number of fiducial points, have not proven to be entirely adequate. A significantly improved method to facilitate the display and analysis of temporal sequences of mammograms by optimizing image registration and grayscales, has been developed. This involves a fully automatic nonlinear geometric transformation, which puts corresponding skin lines, nipples and chest walls in registration and locally corrects pixel values based on the Jacobian of the transformation. Linear regression is applied between pairs of corresponding pixels after registration, and the derived regression equation is used to equalize grayscales. Although the geometric transformation is not able to correct interior tissue patterns for gross differences in the angle of view or differences resulting from skewing of the breast tissue parallel to the detector, sequences studied have been sufficiently consistent that typically only about 30 percent of images in a sequence are considered to be seriously incompatible with the remaining images. These methods clearly demonstrate a significant benefit for the display and analysis of sequences of digital mammograms.
A constrained ROC formulation from probability summation is proposed for measuring observer performance in detecting abnormal findings on medical images. This assumes the observer's detection or rating decision on each image is determined by a latent variable that characterizes the specific finding (type and location) considered most likely to be a target abnormality. For positive cases, this 'maximum- suspicion' variable is assumed to be either the value for the actual target or for the most suspicious non-target finding, whichever is the greater (more suspicious). Unlike the usual ROC formulation, this constrained formulation guarantees a 'well-behaved' ROC curve that always equals or exceeds chance- level decisions and cannot exhibit an upward 'hook.' Its estimated parameters specify the accuracy for separating positive from negative cases, and they also predict accuracy in locating or identifying the actual abnormal findings. The present maximum-likelihood procedure (runs on PC with Windows 95 or NT) fits this constrained formulation to rating-ROC data using normal distributions with two free parameters. Fits of the conventional and constrained ROC formulations are compared for continuous and discrete-scale ratings of chest films in a variety of detection problems, both for localized lesions (nodules, rib fractures) and for diffuse abnormalities (interstitial disease, infiltrates or pnumothorax). The two fitted ROC curves are nearly identical unless the conventional ROC has an ill behaved 'hook,' below the constrained ROC.
Six radiologists used continuous scales to rate 529 chest-film cases for likelihood of five separate types of abnormalities (interstitial disease, nodules, pneumothorax, alveolar infiltrates and rib fractures) in each of six replicated readings, yielding 36 separate ratings of each case for the five abnormalities. Analyses for each type of abnormality estimated the relative gains in accuracy (area below the ROC curve) obtained by averaging the case-ratings across: (1) six independent replications by each reader (30% gain), (2) six different readers within each replication (39% gain) or (3) all 36 readings (58% gain). Although accuracy differed among both readers and abnormalities, ROC curves for the median ratings showed similar relative gains in accuracy. From a latent-variable model for these gains, we estimate that about 51% of a reader's total decision variance consisted of random (within-reader) errors that were uncorrelated between replications, another 14% came from that reader's consistent (but idiosyncratic) responses to different cases, and only about 35% could be attributed to systematic variations among the sampled cases that were consistent across different readers.
The JPEG compression algorithm was tested on a set of 529 chest radiographs that had been digitized at a spatial resolution of 100 micrometer and contrast sensitivity of 12 bits. Images were compressed using five fixed 'psychovisual' quantization tables which produced average compression ratios in the range 15:1 to 61:1, and were then printed onto film. Six experienced radiologists read all cases from the laser printed film, in each of the five compressed modes as well as in the non-compressed mode. For comparison purposes, observers also read the same cases with reduced pixel resolutions of 200 micrometer and 400 micrometer. The specific task involved detecting masses, pneumothoraces, interstitial disease, alveolar infiltrates and rib fractures. Over the range of compression ratios tested, for images digitized at 100 micrometer, we were unable to demonstrate any statistically significant decrease (p greater than 0.05) in observer performance as measured by ROC techniques. However, the observers' subjective assessments of image quality did decrease significantly as image resolution was reduced and suggested a decreasing, but nonsignificant, trend as the compression ratio was increased. The seeming discrepancy between our failure to detect a reduction in observer performance, and other published studies, is likely due to: (1) the higher resolution at which we digitized our images; (2) the higher signal-to-noise ratio of our digitized films versus typical CR images; and (3) our particular choice of an optimized quantization scheme.
This study investigates the degree to which the performance of Bayesian belief networks (BBNs), for computer-assisted diagnosis of breast cancer, can be improved by optimizing their input feature sets using a genetic algorithm (GA). 421 cases (all women) were used in this study, of which 92 were positive for breast cancer. Each case contained both non-image information and image information derived from mammograms by radiologists. A GA was used to select an optimal subset of features, from a total of 21, to use as the basis for a BBN classifier. The figure-of-merit used in the GA's evaluation of feature subsets was Az, the area under the ROC curve produced by the corresponding BBN classifier. For each feature subset evaluated by the GA, a BBN was developed to classify positive and negative cases. Overall performance of the BBNs was evaluated using a jackknife testing method to calculate Az, for their respective ROC curves. The Az value of the BBN incorporating all 21 features was 0.851 plus or minus 0.012. After a 93 generation search, the GA found an optimal feature set with four non-image and four mammographic features, which achieved an Az value of 0.927 plus or minus 0.009. This study suggests that GAs are a viable means to optimize feature sets, and optimizing feature sets can result in significant performance improvements.
On mammograms, certain kinds of features related to masses (e.g., location, texture, degree of spiculation, and integrated density difference) tend to be relatively invariant, or at last predictable, with respect to breast compression. Thus, ipsilateral pairs of mammograms may contain information not available from analyzing single views separately. To demonstrate the feasibility of incorporating multi-view features into CAD algorithm, `single-image' CAD was applied to each individual image in a set of 60 ipsilateral studies, after which all possible pairs of suspicious regions, consisting of one from each view, were formed. For these 402 pairs we defined and evaluated `multi-view' features such as: (1) relative position of centers of regions; (2) ratio of lengths of region projections parallel to nipple axis lines; (3) ratio of integrated contrast difference; (4) ratio of the sizes of the suspicious regions; and (5) measure of relative complexity of region boundaries. Each pair was identified as either a `true positive/true positive' (T) pair (i.e., two regions which are projections of the same actual mass), or as a falsely associated pair (F). Distributions for each feature were calculated. A Bayesian network was trained and tested to classify pairs of suspicious regions based exclusively on the multi-view features described above. Distributions for all features were significantly difference for T versus F pairs as indicated by likelihood ratios. Performance of the Bayesian network, which was measured by ROC analysis, indicates a significant ability to distinguish between T pairs and F pairs (Az equals 0.82 +/- 0.03), using information that is attributed to the multi-view content. This study is the first demonstration that there is a significant amount of spatial information that can be derived from ipsilateral pairs of mammograms.
The study is to investigate the use of a Bayesian belief network (BBN) in a computer-assisted diagnosis (CAD) scheme for mass detection in digitized mammograms. Two independent image sets were used in the experiments. After initial processing of image segmentation and adaptive topographic region growth in our CAD scheme, 288 true-positive mass regions and 2,204 false-positive regions were identified in the training image set. In the testing set, 304 true-positive and 1,586 false-positive regions were identified. Fifty features were computed for each region. After using a genetic algorithm search, a BBN was constructed based on 12 local and four global features in order to classify these regions as positive or negative for mass. The performance of the BBN was evaluated using an ROC methodology. The BBN achieved an area under the ROC curve of 0.873 plus or minus 0.009 in classifying the 304 positive and 1,586 negative regions in the testing set. This result was better than using an artificial neural network with the same set of input features. After incorporating the BBN into our CAD scheme as the last classification stage, we detected 80% of 189 positive mass cases (in 433 testing images) with an average detection rate of 0.76 false-positive regions per image. Therefore, this study demonstrated that a BBN approach could yield a comparable performance to that using other classifiers. Using a probabilistic learning concept and interpretable topology, the BBN provides a flexible approach to improving CAD schemes.
This project is a preliminary evaluation of two simple fully automatic nonlinear transformations which can map any mammographic image onto a reference image while guaranteeing registration of specific features. The first method automatically identifies skin lines, after which each pixel is given coordinates in the range [0,1] X [0,1], where the actual value of a coordinate is the fractional distance of the pixel between tissue boundaries in either the horizontal or vertical direction. This insures that skin lines are put in registration. The second method, which is the method of primary interest, automatically detects pectoral muscles, skin lines and nipple locations. For each image, a polar coordinate system is established with its origin at the intersection of the nipple axes line (NAL) and a line indicating the pectoral muscle. Points within a mammogram are identified by the angle of their position vector, relative to the NAL, and by their fractional distance between the origin and the skin line. This deforms mammograms in such a way that their pectoral lines, NALs and skin lines are all in registration. After images are deformed, their grayscales are adjusted by applying linear regression to pixel value pairs for corresponding tissue pixels. In a comparison of these methods to a previously reported 'translation/rotation' technique, evaluation of difference images clearly indicates that the polar coordinates method results in the most accurate registration of the transformations considered.
The purpose of this study is to explore the potential application of region conspicuity as an index of difficulty for mass detection using computer-assisted diagnosis (CAD) schemes on mammograms and to assess the performance improvement of our own CAD scheme by incorporation of conspicuity as well as other features related to tissue background.
The current trend of using ROC style observer performance studies to evaluate image compression schemes is inefficient and has greatly limited our ability to apply image compression to radiographic images. Because of this, we are developing more efficient automated techniques for assessing the loss of image quality due to compression. The figure- of-merit (FOM) presented here is based on plausible hypotheses about what factors are important in observer performance and evaluates not only the magnitude of difference images but also their structure. This FOM should avoid the limitations of simpler measures such as mean-square-error as well as the limitations and complexity of measures based on psychovisual theory. We applied our FOM to portions of compressed mammograms that had been previously evaluated in a just-noticeable-difference study, and to CT images that had been degraded by compression at various compression ratios with the JPEG algorithm or through other processes. In general, we found that this FOM offers many advantages over other common FOMs and demonstrates the feasibility of developing efficient measures. Although no one-dimensional measure can be expected to be highly correlated to observer performance, for small levels of image degradation we expect this measure to be useful in placing limits on the loss of observer performance.
This investigation is a preliminary attempt to derive a figure-of-merit (FOM) characterizing the image degradation that can occur as a result of image compression. The usefulness of a visible difference predictor (VDP), which is based on previously published psychophysical models, is assessed. Specifically, this is a preliminary attempt to relate physical differences between an original image and a compressed version of the original to visible differences and then to calculate a FOM indicating the degree to which images have been degraded based on this visible difference. The FOM was applied to phantom images and to CT images that had been compressed at various ratios using either the JPEG algorithm or a wavelet compression algorithm. Comparisons between the VDP based measures and more traditional measures of image degradation suggest that, while this FOM may overcome some of the shortcomings of simpler measures such as root-mean-square (RMSE), the use of a VDP has limitations of its own.
To assess the type and frequency of data entry errors, we implemented an interface between a radiology information system and a Kodak image-management system. During a patient verification process that included 2,550 cases transmitted from 15 acquisition devices and archived on our system, 195 cases (7.6%) had mismatches that could not be corrected without manual intervention. These included multiple misspellings, errors in the patient identification numbers, and errors due to the incomplete entry of information from the emergency room (trauma cases). Our results clearly demonstrate the need for a comprehensive patient identification and verification function prior to permanent archiving of imaging information.
A fast, easy-to-use, manufacturer-non-specific archival display and filming system was assembled and tested in the clinical environment. In this system, character recognition software decodes patient identification information and the type of examination for the archival data base in a manner that is transparent to the operator. Images are stored on an optical disk jukebox and can be retrieved to soft display for review and/or reprinting. Preliminary clinical evaluations with the system connected to one and/or two CT scanners clearly indicate that such a concept can successfully replace and/or serve as a backup to conventional film libraries. It can easily be attached to multiple devices (currently three) that are manufactured by the same or different vendors. Technologist and physician responses to the system have been favorable.
The Joint Photographic Expert Group (JPEG) compression standard specifies a quantization procedure but does not specify a particular quantization table. In addition, there are quantization procedures which are effectively compatible with the standard but do not adhere to the simple quantization scheme described therein. These are important considerations, since it is the quantization procedure that primarily determines the compression ratio as well as the kind of information lost or artifacts introduced. A study has been conducted of issues related to the design of quantization techniques tailored for the compression of 12-bit chest images in radiology. Psycho-physical based quantization alone may not be optimal for images that are to be compressed and then used for primary diagnosis. Two specific examples of auxiliary techniques which can be used in conjunction with JPEG compression are presented here. In particular, preprocessing of the source image is shown to be advantageous under certain circumstances. In contrast, a proposed quantization technique in which isolated nonzero coefficients are removed has been shown to be generally detrimental. Image quality here is primarily measured by mean square error (MSE), although this study is in anticipation of more relevant reader performance studies of compression.
The display of a large number of projection radiographs (e.g., AP chest images) for comparison purposes poses potential problems for any electronic environment. In an attempt to assess the concept of rapid sequential viewing, 10 series of AP chest images were each reviewed on a high-resolution workstation under two conditions: (1) simultaneous display of each series in a mosaic configuration; and (2) separate image display in which each image was viewed individually in a rapid sequential mode. In our study, the sequential display was believed subjectively to be of comparable or higher quality by four of six readers. Diagnostic performance (patient improved; no change; patient condition worsened) was comparable for both display modes. Readers were somewhat more comfortable with the simultaneous (mosaic) configuration. Our preliminary results indicate that after minimal training, rapid sequential viewing of AP-chest images may be a reasonable alternative for the display of a series of AP chest images in the ICU.
An increasing practical problem in the evaluation of the accuracy of new imaging systems, as well as the effect of modifications in the display of current imaging systems, is the effort entailed in performing the necessary readings. Although some constraints are dictated by the specific evaluation being conducted, some aspects of the experiment can be determined by the investigator. These include, but are not limited to, the method used to select the cases (selected, stratified, or random) and whether continuous variables that are being evaluated (i.e., pixel size, brightness, contrast) are grouped into discrete categories. The selection of the experimental design has an impact on the sample size required to answer the study question and thus impacts on the cost and effort required to do the study.
KEYWORDS: Diagnostics, Image resolution, Image processing, Chest, Radiography, Databases, Data storage, Medical imaging, Picture Archiving and Communication System, Imaging systems
We designed and implemented a high-resolution video workstation as the central
hardware component in a comprehensive multi-project program comparing the use of
digital and film modalities. The workstation utilizes a 1.8 GByte real-time disk
(RCI) capable of storing 400 full-resolution images and two Tektronix (GMA251) display
controllers with 19" monitors (GMA2O2). The display is configured in a portrait
format with a resolution of 1536 x 2048 x 8 bit, and operates at 75 Hz in a noninterlaced
mode. Transmission of data through a 12 to 8 bit lookup table into the
display controllers occurs at 20 MBytes/second (.35 seconds per image). The
workstation allows easy use of brightness (level) and contrast (window) to be
manipulated with a trackball, and various processing options can be selected using
push buttons. Display of any of the 400 images is also performed at 20MBytes/sec (.35
sec/image). A separate text display provides for the automatic display of patient
history data and for a scoring form through which readers can interact with the system
by means of a computer mouse. In addition, the workstation provides for the
randomization of cases and for the immediate entry of diagnostic responses into a
master database. Over the past year this workstation has been used for over 10,000
readings in diagnostic studies related to 1) image resolution; 2) film vs. soft
display; 3) incorporation of patient history data into the reading process; and 4)
usefulness of image processing.
In an ongoing, multi-reader, multi-project program dealing with the
interpretation of radiological images, we have examined several issues of methodology
which have not as yet been addressed that may impact on the determination of reader
performance as measured by receiver operating characteristic (ROC) analysis. Among
these are issues associated with the training of observers prior to such studies. We
employed three types of observer training that we found to be necessary for the
successful performance of such studies: 1) a general instructional session for
observers on the study protocol and system operation; 2) practice with an interactive
computerized feedback teaching file that demonstrates the imaging systems and
familiarized readers with an idea of the types of cases that were used in the study
along with their correct diagnoses; and 3) training sessions in which readers were
taught the manner in which to distribute answer ratings over an ordinal confidence
scale. The possible effect of such types of training on the performance and results
of ROC studies should be carefully considered prior to their commencement.
We are investigating a prototype x-ray imaging system in which a scintillating
fiberoptic glass plate and/or a fluorescent screen is fiberoptically coupled
to a 2048 x 2048 CCD array (Tektronix). The imaging system includes a fiberoptic
minifier to increase the imaging field of view to a clinically usable one. The
system also allows for cooling of the CCD to reduce the effect of dark noise on
image quality and the use of single-stage light amplification to act as a shutter
and to provide gain control. Images are software corrected for dark current,
individual pixel gain, and geometric distortion. Preliminary results indicate
that high quality x-ray imaging can be obtained using this methodology. This
paper describes design concepts and configuration of the system as well as characterizations
of the initial x-ray images acquired with the camera.
KEYWORDS: Charge-coupled devices, Fiber optics, Image resolution, Signal to noise ratio, Medical imaging, Image acquisition, Imaging systems, Modulation transfer functions, Digital x-ray imaging, Cameras
We are investigating the characteristics of a prototype digital radiography
imaging system in which six two-dimensional diode arrays (CCD) are directly coupled
through a bonded matrix (3 x 2) of fiberoptic minifiers to either a scintillating fiberoptic glass plate or to a fluorescent screen. Images are digitally
acquired at a rate of up to 30 frames/sec and software corrected for pixel gain,
dark current, and geometric distortion. This paper describes the concepts and
design configuration of this approach, as well as preliminary results from several
phantom and animal studies. Our results indicate that high resolution (> 4 lp/rnm)
and high signal-to-noise ratio images can be obtained with this method. However,
the complexity associated with this concept cannot be discounted.
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.