We present a computationally efficient method for analyzing H&E stained digital pathology slides with the objective of
discriminating diagnostically relevant vs. irrelevant regions. Such technology is useful for several applications: (1) It can
speed up computer aided diagnosis (CAD) for histopathology based cancer detection and grading by an order of magnitude
through a triage-like preprocessing and pruning. (2) It can improve the response time for an interactive digital pathology
workstation (which is usually dealing with several GByte digital pathology slides), e.g., through controlling adaptive
compression or prioritization algorithms. (3) It can support the detection and grading workflow for expert pathologists in a
semi-automated diagnosis, hereby increasing throughput and accuracy. At the core of the presented method is the statistical
characterization of tissue components that are indicative for the pathologist's decision about malignancy vs. benignity,
such as, nuclei, tubules, cytoplasm, etc. In order to allow for effective yet computationally efficient processing, we propose
visual descriptors that capture the distribution of color intensities observed for nuclei and cytoplasm. Discrimination
between statistics of relevant vs. irrelevant regions is learned from annotated data, and inference is performed via linear
classification. We validate the proposed method both qualitatively and quantitatively. Experiments show a cross validation
error rate of 1.4%. We further show that the proposed method can prune ≈90% of the area of pathological slides while
maintaining 100% of all relevant information, which allows for a speedup of a factor of 10 for CAD systems.
Detection of malignancy from histopathological images of breast cancer is a labor-intensive and error-prone
process. To streamline this process, we present an efficient Computer Aided Diagnostic system that can differentiate
between cancerous and non-cancerous H&E (hemotoxylin&eosin) biopsy samples. Our system uses novel
textural, topological and morphometric features taking advantage of the special patterns of the nuclei cells in
breast cancer histopathological images. We use a Support Vector Machine classifier on these features to diagnose
malignancy. In conjunction with the maximum relevance - minimum redundancy feature selection technique, we
obtain high sensitivity and specificity. We have also investigated the effect of image compression on classification
performance.
We aim to improve telepathology images for diagnoses using compression based on information about human visual
system. Underlying goal is to demonstrate utility of a visual discrimination model (VDM) for predicting observer
performance. 100 ROIs from breast biopsy virtual slides at 5 levels of compression (uncompressed, 8:1, 16:1, 32:1, 64:1,
128:1) were shown to 6 pathologists to determine benign vs malignant. There was a decrease in performance as a
function of compression (F = 14.58, p< 0.0001). The visibility of compression artifacts in the test images was predicted
using a VDM. JND metrics were computed for each image including mean, median, ≥90th percentiles, and maximum.
For comparison PSNR and SSIM were also computed. Image distortion metrics were computed as a function of
compression ratio and averaged across test images. All of the JND metrics were found to be highly correlated and
differed primarily in magnitude. Both PSNR and SSIM decreased with bit rate, correctly reflecting a loss of image
fidelity with increasing compression. The correlation of observer performance in the ROC experiment with image
distortion metrics is shown in Figures 3 and 4. Observer performance (Az) was nearly constant up to a compression ratio
of 32:1, then decreased significantly for 64:1 and 128:1 compression. The initial decline in Az occurred around a mean
JND of 3, Minkowski JND of 4, and 99th percentile JND of 6.5. Virtual pathology may be compressible to relatively high
levels before impacting diagnostic accuracy and the VDM accurately predicts human performance.
LCD monitors used for radiologic diagnosis exhibit a random spatial gain-fluctuation pattern that can be characterized as
multiplicative fixed-pattern noise. Our goal is to use a monitor-particular map of this noise to pre-process mammograms
for display. This noise map is obtained by processing a captured image of the monitor screen. In previous work we have
demonstrated a method that produces a measurable reduction in the noise level. In this paper we describe the results of
our first observer studies and of computer analysis using visual a discrimination model. The result proved to be negative,
i.e. the pre-processing provided an insufficient degree of improvement to be judged diagnostically effective. In light of
these first results, we have been examining our assumptions and processing methods.
We begin with a brief introduction then describe the construction of the noise map and its use in pre-processing. The two
key advances follow: the observer-study and the computer-modeling results. Finally, we describe our efforts to
understand the relation of our measurement, modeling, and observer results.
A major issue in telepathology is the extremely large and growing size of digitized "virtual" slides, which can require
several gigabytes of storage and cause significant delays in data transmission for remote image interpretation and
interactive visualization by pathologists. Compression can reduce this massive amount of virtual slide data, but
reversible (lossless) methods limit data reduction to less than 50%, while lossy compression can degrade image quality
and diagnostic accuracy. "Visually lossless" compression offers the potential for using higher compression levels
without noticeable artifacts, but requires a rate-control strategy that adapts to image content and loss visibility. We
investigated the utility of a visual discrimination model (VDM) and other distortion metrics for predicting JPEG 2000 bit
rates corresponding to visually lossless compression of virtual slides for breast biopsy specimens. Threshold bit rates
were determined experimentally with human observers for a variety of tissue regions cropped from virtual slides. For test
images compressed to their visually lossless thresholds, just-noticeable difference (JND) metrics computed by the VDM
were nearly constant at the 95th percentile level or higher, and were significantly less variable than peak signal-to-noise
ratio (PSNR) and Structural Similarity (SSIM) metrics. Our results suggest that VDM metrics could be used to guide the
compression of virtual slides to achieve visually lossless compression while providing 5 to 12 times the data reduction of
reversible methods.
KEYWORDS: LCDs, Mammography, Signal to noise ratio, Breast cancer, Quantization, CCD cameras, Visual process modeling, Statistical analysis, Cancer, Denoising
This presentation describes work in progress that is the result of an NIH SBIR Phase 1 project that
addresses the wide- spread concern for the large number of breast-cancers and cancer victims [1,2].
The primary goal of the project is to increase the detection rate of microcalcifications as a result of
the decrease of spatial noise of the LCDs used to display the mammograms [3,4]. Noise reduction is
to be accomplished with the aid of a high performance CCD camera and subsequent application of
local-mean equalization and error diffusion [5,6]. A second goal of the project is the actual detection
of breast cancer. Contrary to the approach to mammography, where the mammograms typically have
a pixel matrix of approximately 1900 x 2300 pixels, otherwise known as FFDM or Full-Field Digital
Mammograms, we will only use sections of mammograms with a pixel matrix of 256 x 256 pixels.
This is because at this time, reduction of spatial noise on an LCD can only be done on relatively
small areas like 256 x 256 pixels. In addition, judging the efficacy for detection of breast cancer will
be done using two methods: One is a conventional ROC study [7], the other is a vision model
developed over several years starting at the Sarnoff Research Center and continuing at the Siemens
Corporate Research in Princeton NJ [8].
In our work with LCD monitors for medical images, we have created a computer-program simulation-suite that mimics
the appearance of an LCD screen. It uses high-magnification digital-camera capture of individual monitor pixels to
compose realistic the sub-pixel patterns used in the simulations. These patterns are then weighted by digital driving
levels, DDL's, that correspond to the image being displayed and inserted into a digital monitor field so as to compose an
image of pixels that correspond to those of a monitor. The program suite also simulates the area-capture of a screenimage
by a digital camera at a selectable magnification. The research project to which we are currently applying this
simulation is the reduction of near-pixel-sized fixed-pattern noise. In the actual experiment a camera is used to capture a
magnified portion of the monitor. Typical magnifications are 4:1 and 8:1 CCD to LCD pixels. From this captured
image, a fixed-pattern multiplicative-noise gain map is generated that is used to adjust DDL's in order to pre-compensate
for that noise. In addition to the spatial characteristics of the LCD monitor and CCD camera sensor, our simulation
addresses nonlinearities found in the display and capture processes. The nonlinearities become important because the
captured CCD digital values, or DSL's for digital sensor levels, are converted to luminance. This conversion is necessary
because we employ a subsequent local-area processing step that relies on linearity of image-spread being in energy fluxdensity.
This presentation focuses specifically on the comparison of the simulation results to physical experiments.
A major issue in telepathology is the extreme size of digitized slides, which require several gigabytes of storage and
cause significant delays in image delivery to pathologists. We investigated the utility of a visual discrimination model
(VDM) to predict bit rates for visually lossless JPEG2000 compression of breast biopsy virtual slides. Visually lossless
bit rates were determined experimentally with human observers. VDM metrics computed for those bit rates were nearly
constant, suggesting that VDMs could be used to achieve visually lossless image quality while providing about four
times the data reduction of reversible compression.
We deal with monochrome high-resolution LCD monitors used for displaying medical images. We discuss reducing
near-pixel-sized components of fixed-pattern noise. This noise is composed of irregularities in the LCD pixel structure
and fine background structures between pixels. We display a series of test images on the monitor with controlled LCD
digital driving levels or DDL's. A calibrated CCD camera is used to magnify and capture a portion of the monitor for
each image. The captured CCD digital values, or DSL's for digital sensor levels, are converted to luminance. This
conversion is necessary because we employ a subsequent local-area processing step that relies on linearity of image-spread
being in energy flux-density. Because we are working with two digital systems, the CCD camera and the LCD
display, there is no continuous map between the CCD DSL's and LCD DDL's. We map the discrete LCD DDL's to a
quantized luminance space. Once we have determined the target luminance values, we use an error-diffusion algorithm
to select luminance values from the discrete addressable set for each pixel and then map those values to LCD DDL's.
The result is a set of adjusted LCD DDL's that reduces the fixed-pattern noise in a locally averaged sense.
Breast tomosynthesis is currently an investigational imaging technique requiring optimization of its many combinations
of data acquisition and image reconstruction parameters for optimum clinical use. In this study, the effects of several
acquisition parameters on the visual conspicuity of diagnostic features were evaluated for three breast specimens using a
visual discrimination model (VDM). Acquisition parameters included total exposure, number of views, full resolution
and binning modes, and lag correction. The diagnostic features considered in these specimens were mass margins,
microcalcifications, and mass spicules. Metrics of feature contrast were computed for each image by defining two
regions containing the selected feature (Signal) and surrounding background (Noise), and then computing the difference
in VDM channel metrics between Signal and Noise regions in units of just-noticeable differences (JNDs). Scans with
25 views and exposure levels comparable to a standard two-view mammography exam produced higher levels of feature
contrast. The effects of binning and lag correction on feature contrast were found to be generally small and isolated,
consistent with our visual assessments of the images. Binning produced a slight loss of spatial resolution which could
be compensated in the reconstruction filter. These results suggest that good image quality can be achieved with the
faster and therefore more clinically practical 25-view scans with binning, which can be performed in as little as 12.5
seconds. Further work will investigate other specimens as well as alternate figures of merit in order to help determine
optimal acquisition and reconstruction parameters for clinical trials.
We evaluated human observer and model (JNDmetrix) performance to assess whether the veiling glare of a digital display influences performance in softcopy interpretation of mammographic images. 160 mammographic images, half with a single mass, were processed to simulate four levels of veiling glare: none, comparable to a typical cathode ray tube (CRT) display, double a CRT and quadruple a CRT. Six radiologist observers were shown the images in a randomized presentation order on a liquid crystal display (LCD) that had relatively no veiling glare. The JNDmetrix human visual system model also analyzed the images. Receiver Operating Characteristic (ROC) techniques showed that performance declined with increasing veiling glare (F = 6.884, p = 0.0035). Quadruple veiling glare yielded significantly lower performance than the lower veiling glare levels. The JNDmetrix model did not predict a reduction in performance with changes in veiling glare, and correlation with the human observer data was modest (0.588). Display veiling glare may influence observer performance, but only at very high levels.
The JPEG2000 compression standard is increasingly a preferred industry method for 2D image compression. Some vendors, however, continue to use proprietary discrete cosine transform (DCT) JPEG encoding. This study compares image quality in terms of just-noticeable differences (JNDs) and peak signal-to-noise ratios (PSNR) between DCT JPEG encoding and JPEG2000 encoding. Four computed tomography and 6 computed radiography studies were compressed using a proprietary DCT JPEG encoder and JPEG2000 standard compression. Image quality was measured in JNDs and PSNRs. The JNDmetrix computational visual discrimination model simulates known physiological mechanisms in the human visual system, including the luminance and contrast sensitivity of the eye and spatial frequency and orientation responses of the visual cortex. Higher JND values indicate that a human observer would be more likely to notice a significant difference between compared images. DCT JPEG compression showed consistently lower image distortions at lower compression ratios, whereas JPEG2000 compression showed benefit at higher compression ratios (>50:1). The crossover occurred at ratios that varied among the images. The magnitude of any advantage of DCT compression at low ratios was often small. Interestingly, this advantage of DCT JPEG compression at lower ratios was generally not observed when image quality was measured in PSNRs. These results suggest that DCT JPEG may outperform JPEG2000 for compression ratios generally used in medical imaging and that the differences between DCT and JPEG2000 could be visible to observers and thus clinically significant.
The authors identify a fundamental disconnect between the ways in which industry and radiologists assess and even discuss product performance. What is needed is a quantitative methodology that can assess both subjective image quality and observer task performance. In this study, we propose and evaluate the use of a visual discrimination model (VDM) that assesses just-noticeable differences (JNDs) to serve this purpose. The study compares radiologists' subjective perceptions of image quality of computer tomography (CT) and computed radiography (CR) images with quantitative measures of peak signal-to-noise ratio (PSNR) and JNDs as measured by a VDM. The study included 4 CT and 6 CR studies with compression ratios ranging from lossless to 90:1 (total of 80 sets of images were generated [n = 1,200]). Eleven radiologists reviewed the images and rated them in terms of overall quality and readability and identified images not acceptable for interpretation. Normalized reader scores were correlated with compression, objective PSNR, and mean JND values. Results indicated a significantly higher correlation between observer performance and JND values than with PSNR methods. These results support the use of the VDM as a metric not only for the threshold discriminations for which it was calibrated, but also as a general image quality metric. This VDM is a highly promising, reproducible, and reliable adjunct or even alternative to human observer studies for research or to establish clinical guidelines for image compression, dose reductions, and evaluation of various display technologies.
The widespread use of multi-detector CT scanners has been associated with a remarkable increase in the number of CT slices as well as a substantial decrease in the average thickness of individual slices. This increased number of thinner slices has created a marked increase in archival and network bandwidth requirements associated with storage and transmission of these studies. We demonstrate that although compression can be used to decrease the size of these image files, thinner CT slices are less compressible than thicker slices when measured by either a visual discrimination model (VDM) or the more traditional peak signal to noise ratio. The former technique (VDM) suggests that the discrepancy in compressibility between thin and thick slices becomes greater at greater compression levels while the latter technique (PSNR), suggests that this is not the case. Previous studies that we and others have performed suggest that the VDM model probably corresponds more closely with human observers than does the PSNR model. Additionally we demonstrated that the poor relative compressibility of thin sections can be substantially negated by the use of JPEG 2000 3D compression which yields superior image quality at a given level of compression in comparison with 2D compression. Additionally, thin and thick sections are approximately equally compressible for 3D compression with little change with increasing levels of compression.
CRT displays are generally used for softcopy display in the digital reading room, but LCDs are being used more frequently. LCDs have many useful properties, but can suffer from significant degradation when viewed off-axis. We compared observer performance and human visual system model performance for on and off-axis CRT and LCD viewing. 400 mammographic regions of interest with different lesion contrasts were shown on and off-axis to radiologists on a CRT and LCD. Receiver Operating Characteristic (ROC) techniques were used to analyze observer performance and results were correlated with the predictions of the human vision model (JNDmetrix model). Both sets of performance metrics showed that LCD on-axis viewing was better than the CRT; and off-axis was significantly better with the CRT. Off-axis LCD viewing of radiographs can degrade observer performance compared to a CRT.
For diagnosis of breast cancer by mammography, the mammograms must be viewed by a radiologist. The purpose of this study was to determine the effect of display resolution on the specific clinical task of detection of breast lesions by a human observer. Using simulation techniques, this study proceeded through four stages. First, we inserted simulated masses and calcifications into raw digital mammograms. The resulting images were processed according to standard image processing techniques and appropriately windowed and leveled. The processed images were blurred according to MTFs measured from a clinical Cathode Ray Tube display. JNDMetrix, a Visual Discrimination Model, examined the images to estimate human detection. The model results suggested that detection of masses and calcifications decreased under standard CRT resolution. Future work will confirm these results with human observer studies. (This work was supported by grants NIH R21-CA95308 and USAMRMC W81XWH-04-1-0323.)
Previous studies in which the JNDmetrix visual discrimination model (VDM) was applied to predict effects of image display and processing factors on lesion detectability have shown promising results for mammographic images with microcalcification clusters. In those studies, just-noticeable-difference (JND) metrics were computed for signal-present and signal-absent image pairs with the same background. When this "paired discriminability" method was applied to Gaussian signals in 1/f3 filtered noise, however, it was unable to predict detection thresholds measured in 2AFC trials for different backgrounds. We suggested previously (SPIE 2002) that a statistical model observer using channel responses from "single-ended" VDM simulations could predict detection performance with different backgrounds. The implementation and evaluation of that VDM-channelized model observer is described in this paper. Model performance was computed for sets of signal and noise images from two observer performance studies involving the detection of simulated or real breast masses. For the first study, the VDM-channelized model observer was able to predict the dependence of detection thresholds on signal size (contrast-detail slope) for 2AFC detection of Gaussian signals on different 1/f3 noise backgrounds. Variations in the detectability of masses in mammograms from the second study correlated well with model performance as a function of display type (LCD vs. CRT) and viewing angle (on-axis vs. 45° off-axis). The performance of the VDM-channelized model observer was superior to results obtained using either the VDM paired discriminability method or a conventional nonprewhitening model observer.
The JNDmetrix human visual system model developed by the Sarnoff Corporation is used to predict observer performance on visual discrimination tasks. It begins with two paired images as the initial input and ends with a JND map that shows the magnitude and spatial location of visible differences between the two input images. The goal of this experiment was to determine if the location and magnitude of JNDs identified by the model corresponded to visual search parameters of the human observer. Radiologists searched 20 mammograms with multiple masses and microcalcifications of different subtleties as their eye-position was recorded. The JNDmetrix model analyzed the same images and identified, with JNDs, discriminable areas on the images. Lesions with lower subtlety ratings were detected later in search than more obvious ones (FNs later than TPs). When the subtler lesions were detected (TP) dwell time was longer than more obvious lesions, but the FNs received shorter total dwell. The subtler lesions when detected (TP) received more total fixation clusters than more obvious ones, but the FNs received fewer. The correlation between the model JNDs and the eye-position parameters was high. Understanding the influence of lesion subtlety on search may help us better model and predict human observer performance.
A discrepancy exists between two studies that investigated psychophysical detection of simulated lesions (e.g. gaussians or designer nodules) embedded in filtered noise images (Johnson et al, 2002; Burgess et al, 2003). Johnson et al, 2002 identified a significant difference in the slope of the contrast detail plots (CD plots) as the presentation methodology in a 2AFC task was changed from the unlike background (unpaired) to identical backgrounds (paired). In comparable experiments, Burgess et al, 2003 challenged the results by finding no difference between the slopes (both positive) of the CD plots when using paired backgrounds or unpaired backgrounds. We found that a significant difference between the two studies, namely the presence of a circular fixation cue was responsible for the discrepancy. The detection noise due to positional uncertainty was sufficient to reduce subject's threshold for small target diameters. This effect was amplified in the paired background, switching the CD plot from a negative slope (without fixation) to a positive slope (with fixation). The effect was less dramatic with the unpaired backgrounds, however intra-observer variability seemed to be reduced with fixation cues. These results significantly reduce the discrepancies in C-D characteristics between the two studies.
Evaluation of images generated by new MR pulse sequences or reconstruction methods is traditionally done using subjective measures of image quality in a clinical study or by using quantitative measures such as local SNR or CNR to support the subjective findings. In order to accelerate evaluation of new candidate MR techniques, objective measures related to human perception and performance are desirable. Therefore, the goal of this study was to determine if the effects of parallel-imaging artifacts on subjective image quality could be predicted using the JNDmetrix vision model as a first step in developing objective measures to guide MR development. Single-shot fast spin echo images (HASTE) were acquired with increasing acceleration factors (0, 2, 3, and 4) and reconstructed with two algorithms, mSENSE and GRAPPA. Subjective quality ratings (0-10 scale) for these images were compared to spatial-frequency channel responses of the JNDmetrix model and to PSNR. Our results confirmed the anticipated degradation in quality for GRAPPA and mSENSE images with increasing acceleration factor. The mSENSE method yielded significantly lower quality ratings than GRAPPA for the higher acceleration factors (3 and 4). Full matrix images with no partial parallel acquisition (noppa) showed blurring due to longer shot time and T2 decay and were rated most comparable to the GRAPPA acceleration factor 4 images. There was a strong linear relationship between just-noticeable difference (JND) changes and observer ratings, while PSNR showed no correlation with observer ratings. The JNDmetrix results better reflected image degradation due to both blurring and noise. These results give confidence that the JNDmetrix approach may become a useful tool for the design and evaluation of MR pulse sequences and reconstruction methods.
The goal of this project was to evaluate a human visual system model (JNDmetrix) based on JND and frequency-channel vision-modeling principles to predict the effects of monitor veiling glare on observer performance in interpreting radiographic images. The veiling glare of a high-performance CRT and an LCD display was measured. A series of mammographic images with masses of different contrast levels was generated. Six radiologists viewed the sets of images on both monitors and reported their decision confidence about the presence of a mass. The images were also run through the JNDmetrix model. Veiling glare affected observer performance (ROC Az). Performance was better on the LCD display with lower veiling glare compared to the CRT with higher veiling glare. The JNDmetrix model predicted the same pattern of results and the correlation between human and computer observers was high. Veiling glare can affect significantly observer performance in diagnostic radiology. A possible confound exists in that two different monitors were used and other physical parameters may contribute to the differences observed. A new set of studies is underway to remove that confound.
KEYWORDS: Visual process modeling, Performance modeling, Data modeling, Eye models, CRTs, CCD cameras, Visual system, Signal to noise ratio, Human vision and color perception, Visualization
The goal was to develop an efficient method of optimizing CRT monitor performance for digital mammography. The Sarnoff JNDmetrix vision model is based on just-noticeable difference measurement and frequency-channel vision-modeling principles. Given 2 images as input the model returns accurate, robust estimates of discriminability. Model predictions are compared with human performance. Mammographic images with microcalcifications were viewed by six radiologists, once on a monitor with P45 and once on one with P104 phosphor. Results were compared with output of the model used to predict differences in perceptibility of calcifications using luminance data measured with a high-resolution CCD camera. Human performance was best with high contrast clusters and got worse with each decrease in contrast. Performance was better with the P45 than the P104 for targets at all contrast levels. The JNDmetrix model predicted the same pattern of results. Correlation between human and model observer performance was very high. We have demonstrated the utility of using a vision model to accurately predict human detection performance. The type of phosphor in a monitor influences observer performance at least for the detection of microcalcifications. The main reason is that the P104 has a higher luminance, but the P45 has a higher signal-to-noise ratio.
The goal of this project was to develop an efficient method of optimizing CRT monitor performance for digital mammography. In this study we examined the effects on performance of processing images to compensate for limitations in the MTF of the CRT monitor. The Sarnoff JNDmetrix vision model is based on just-noticeable difference measurement and frequency-channel vision-modeling principles. Given two images as input the model returns accurate, robust estimates of their discriminability. Model predictions are then compared with human performance. Mammographic images (n = 250) with microcalcifications were viewed by six radiologists. The images were viewed once in original unprocessed form and once after processing. Results were compared with output of the model that was used to predict differences in perceptibility of calcifications using luminance data measured with a high-resolution CCD camera. Human performance was better with the MTF compensated images at all contrast levels. The JNDmetrix model predicted the same pattern of results. Correlation between human and model observer performance was very high. Using image processing methods to compensate for limitations in the MTF of CRT monitors can improve the detection performance of radiologists searching for microcalcifications.
The Sarnoff JNDmetrix visual discrimination model (VDM) was applied to predict the visibility of compression artifacts in mammographic images. Sections of digitized mammograms were subjected to irreversible (lossy) JPEG and JPEG 2000 compression. The detectability of compressed images was measured experimentally and compared with VDM metrics and PSNR for the same images. Artifacts produced by JPEG 2000 compression were generally easier for observers to detect than those produced by JPEG encoding at the same compression ratio. Detection thresholds occurred at JPEG 2000 compression ratios from 6:1 to 10:1, significantly higher than the average 2:1 ratio obtained for reversible (lossless) compression. VDM predictions of artifact visibility were highly correlated with observer performance for both encoding techniques. Performance was less correlated with encoder bit rate and PSNR, which was a relatively poor predictor of threshold bit rate across images. Our results indicate that the VDM can be used to predict the visibility of compression artifacts and guide the selection of encoder bit rate for individual images to maintain artifact visibility below a specified threshold.
The Sarnoff JNDmetrix visual discrimination model (VDM) was applied to predict human psychophysical performance in the detection of simulated mammographic lesions. Contrast thresholds for the detection of synthetic Gaussian masses on mean backgrounds and simulated mammographic backgrounds were measured in two-alternative, forced-choice (2AFC) trials. Experimental thresholds for 2-D Gaussian signal detection decreased with increasing signal size on mean backgrounds and on 1/f3 filtered noise images presented with identical (paired) backgrounds. For 2AFC presentations of different (unpaired) filtered noise backgrounds, detection thresholds increased with increasing signal diameter, consistent with a decreasing signal-to-noise ratio. Thresholds for mean and paired filtered noise backgrounds were used to calibrate a new low-pass, spatial-frequency channel in the VDM. The calibrated VDM was able to predict accurate detection thresholds for Gaussian signals on mean and paired 1/f3 filtered noise backgrounds. To simulate noise-limited detection thresholds for unpaired backgrounds, an approach is outlined for the development of a VDM-based model observer based on statistical decision theory.
The purpose of this study was to evaluate a method of creating synthetic normal and abnormal mammograms. Images consisting of 1024 x 1024 regions were extracted from digitized mammograms. Twenty-five regions included a single microcalcification cluster. A second set of twenty-five regions without calcifications was also selected. Calcifications were digitally removed by application of a median filter to form a third set of images. Finally, extracted calcifications were superposed on normal images to create a fourth set. Three mammographers evaluated the quality of the simulations. Their task was to classify the images according to real or simulated status using a 10-point rating scale. The classification accuracy was calculated by Receiver Operating Characteristic (ROC) analysis. Two other radiologists performed a paired image task on a subset of the images. They attempted to discriminate between real and simulated images that were simultaneously displayed, which was analyzed by a forced-choice method. In either case it was found that the probability of correct classification was insignificantly different from the chance level. We conclude that the simulation methodology employed was satisfactory. The ability to create synthetic images, that are indistinguishable from real images, is expected to facilitate modality evaluation studies in mammography.
KEYWORDS: Visual process modeling, Mammography, CRTs, Data modeling, Performance modeling, Visualization, Digital mammography, CCD cameras, Breast cancer, LCDs
The goal of the project was to develop an efficient method of optimizing CRT performance for digital mammography. The paradigm measures radiologist performance for various display characteristics and uses these results to validate a model of human visual performance. The Sarnoff JNDmetrix vision model is based on psychophysical just-noticeable difference measurement and frequency-channel vision-modeling principles. Given 2 images as input the model returns accurate, robust estimates of their discriminability. Model predictions are compared with human performance. Mammographic images with microcalcifications were viewed by radiologists. Results were analyzed using ROC techniques. The images were viewed once on a monitor with P45 and once on a monitor with P104 phosphor. Results were compared with output of the model that was used to predict differences in perceptibility of calcifications using luminance data measured with a high-resolution CCD camera. Early results suggest that human performance is best with high contrast clusters and progressively gets worse with each decrease in contrast. Performance so far is better with the P45 than the P104 for targets at all contrast levels. The JNDmetrix model should predict the same pattern of results. The type of phosphor in a CRT monitor seems to influence observer performance.
Numerous studies have been conducted to determine experimentally the effects of image processing and display parameters on the diagnostic performance of radiologists. Comprehensive optimization of imaging systems for digital mammography based solely on measurements of reader performance is impractical, however, due to the large number of interdependent variables to be tested. A reliable, efficient alternative is needed to improve the evaluation and optimization of new imaging technologies. The Sarnoff JNDmetrixTM Visual Discrimination Model (VDM) is a computational, just-noticeable difference model of human vision that has been applied successfully to predict performance in various nonmedical detection and rating tasks. To test the applicability of the VDM to specific detection tasks in digital mammography, two observer performance studies were conducted. In the first study, effects of display tone scale and peak luminance on the detectability of microcalcifications were evaluated. The VDM successfully predicted improvements in reader performance for perceptually linearized tone scales and higher display luminances. In the second study, the detectability of JPEG and wavelet compression artifacts was evaluated, and performance ratings were again found to be highly correlated with VDM predictions. These results suggest that the VDM would be useful in the assessment and optimization of new imaging and compression technologies for digital mammography.
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.