Considerations in compiling a database of clinical test images.
Explore the source record for details and available documents.
Hospitals↗
Biomedical subjects
G W Seeley
Publications and source records attributed to G W Seeley.
Ideal versus human observer for long-tailed point spread functions: does deconvolution help?
The ideal observer represents a Bayesian approach to performing detection tasks. Since such tasks are frequently used as a prototype tasks for radiological imaging systems, the detectability measured at the output of an ideal detector can be used as a figure of merit to characterize the imaging system. For the detectability achieved by the ideal observer to be a good figure of merit, it should predict the ability of the human observer to perform the same detection task. Of great general interest, especially to the medical community, are imaging devices with long-tailed point spread functions (PSFs). Such PSFs may occur due to septal penetration in collimators, veiling glare in image intensifiers or scattered radiation in the body. We have investigated the effect that this type of PSF has on human visual signal detection and whether any improvement in performance can be gained by deconvolving the tails of the PSF. For the ideal observer, it is straightforward to show that the performance is independent of any linear, invertible deconvolution filter. Our psychophysical studies show, however, that performance of the human observer is indeed improved by deconvolution. The ideal observer is, therefore, not a good predictor of human observer performance for detection of a signal imaged through a long-tailed PSF. We offer some explanations for this discrepancy by using some characteristics of the visual process and suggest a standard of comparison for the human observer that takes into account these characteristics. A look at the performance of the non-prewhitening (npw) ideal observer, before and after deconvolution, also brings some good insight into this study.
Bayes Theorem↗
Evaluation of a new method for verifying truth of diagnosis in clinical research.
Having a good methodology for establishing truth of diagnosis would be invaluable to researchers in radiology. A protocol to verify the truth of diagnosis of clinical images to be used in radiologic research has been proposed that specifies six classes, depending on how the diagnosis is verified. In this paper it was evaluated. The protocol was applied to sixty-nine liver-spleen scans. Using five of the possible six classes of the protocol, 42% of the scans were verifiable. Twenty-eight per cent of the cases could be verified by non-physicians using the first four classes of the protocol. In 96% of a sample of these cases, their diagnosis matched that of the physicians. Radiologists are needed for the last two classes. However, using three radiologists to investigate Class V protocol procedures only eight (16%) more cases were clearly classifiable. Class VI was not investigated. The protocol met its objectives of reducing physician involvement and developing a detailed methodology. The difficulties inherent in verification of diagnosis are also discussed, and suggestions for improving the protocol are proposed.
Diagnosis↗
Computer-simulated lung nodules in digital chest radiographs for detection studies.
Computer simulations of lung nodules overcome many shortcomings of creating radiographs using anthropomorphic nodule phantoms for lung nodule detection studies, but these algorithms can be cumbersome and involved. A simple, fast, and flexible computer program to simulate lung nodules in digital chest radiographs for detection studies is reported. To verify the realism of the simulated nodules, a psychophysical study and a statistical study were conducted. In the psychophysical study, six radiologists and four nonradiologists were asked to distinguish between 17 real lung nodules and 17 computer-simulated lung nodules shown in eight radiographs. The results show that the computer-simulated lung nodules are indistinguishable visually from real lung nodules. Using parameters from the Rose model of vision, results show that the simulated and real nodules are the same statistically. Thus, besides visual validity, statistical analysis in confirming the validity of the simulated lung nodules is included.
Computer Simulation↗
Detection of breast abnormalities on teleradiology transmitted mammograms.
The authors conducted an observer performance study to compare breast lesion detection of conventional mammography (CM) with teleradiology (TE) transmitted mammograms. One hundred four abnormal, single-projection mammographic images were transmitted by teleradiology. Abnormalities included 11 cases with skin thickening or retraction, 48 cases with microcalcifications, and 52 with parenchymal masses. The CM and TE images were reviewed by four mammographers who indicated the type and location of abnormalities and a level of confidence for their diagnostic decisions. For each of three detection tasks--skin and nipple abnormalities, microcalcifications, and masses--receiver operating characteristic curve analysis was performed for individual readers and all readers as a group. For detecting skin and nipple abnormalities, readers performed significantly better with CM than with TE (z = 2.05, P = less than 0.04). However, no significant differences were found among readers for detection of either microcalcifications or masses. Further improvements in hardware and imaging parameters may improve detection of soft tissue abnormalities. Further evaluation is necessary to determine whether teleradiology might be applicable to breast cancer screening.
Breast Neoplasms↗
Design and testing of artifact-suppressed adaptive histogram equalization: a contrast-enhancement technique for display of digital chest radiographs.
One of the goals of our research in the field of digital radiography has been to develop contrast-enhancement algorithms for eventual use in the display of chest images on video devices with the aim of preserving the diagnostic information presently available with film, some of which would normally be lost because of the smaller dynamic range of video monitors. The ASAHE algorithm discussed in this article has been tested by investigating observer performance in a difficult detection task involving phantoms and simulated lung nodules, using film as the output medium. The results of the experiment showed that the algorithm is successful in providing contrast-enhanced, natural-looking chest images while maintaining diagnostic information. The algorithm did not effect an increase in nodule detectability, but this was not unexpected because film is a medium capable of displaying a wide range of gray levels. It is sufficient at this stage to show that there is no degradation in observer performance. Future tests will evaluate the performance of the ASAHE algorithm in preparing chest images for video display.
Algorithms↗
Comparison of conventional fluorescein angiography film images with a cathode ray tube display.
We evaluated the effectiveness of cathode ray tube (CRT) displays in the transfer of diagnostic information to the ophthalmologist. Retinal fluorescein angiography images were used and data were taken using the receiver operating characteristic experimental paradigm. Three ophthalmologists with experience ranging from two to 12 years participated in the study. Sixty-eight images were selected from proof sheets of 800 original images. Next, these images were digitized, then each set was shown on its respective display (slide projection for the proof sheet images and CRT for the digital images). These 68 images were from a variety of patients and represented a broad range of normal and abnormal retinas (including diabetic retinopathy, hypertension independent of diabetes, and sickle-cell disease). Results show (1) that there is no difference between the systems in terms of diagnostic accuracy; (2) that certainty of diagnosis was not a function of the system; and (3) that although there was an effect due to experience of the ophthalmologist, it did not affect diagnosis accuracy. We conclude that CRT displays at conventional video resolutions can deliver the needed diagnostic information as well as film. We also hypothesize that digital enhancement techniques can increase the available diagnostic information beyond that of film.
Computer Systems↗
Computer-assisted analysis of mammographic clustered calcifications.
A total of 127 biopsy-proven clustered mammographic calcifications were diagnosed and evaluated using multiple parameters. The results were subjected to computer analysis. The number of calcifications/cm2 was the most important parameter. Less than 10 calcifications/cm2 resulted in an 82% chance of being benign while 10 or more calcifications/cm2 had a 44% chance of being malignant. The average distance between the calcifications in the cluster was also significant, with a 92% chance of being benign if greater than 1 mm and a 52% chance of being malignant if less than 1 mm. The remaining parameters did not yield statistically significant results, but there were trends which suggest that less than 10 clustered calcifications that are regular in size and shape and orderly in appearance tend to be benign while disorderly clusters tend to be malignant. In order to decrease the number of false positive biopsies and improve the cost-effectiveness of screening mammography, radiologists should consider 6-month follow-up mammography as a reasonable alternative for clustered calcifications that appear benign.
Breast↗
Intraoperative arteriography: comparison of conventional screen-film with photostimulable imaging plate radiographs.
This prospective study compared images obtained with a photostimulable imaging plate with matched images obtained with a conventional screen-film combination in 26 patients undergoing intraoperative arteriography. Diagnostic accuracy of the two techniques was assessed objectively, and image quality was assessed subjectively. In 16 patients (62%), the radiation exposure was reduced by 50% for the imaging plate technique by decreasing the mAs level generally used for the screen-film combination. Because of the dynamic range of the imaging plate system, no repeat examinations were necessary, while 12% of the screen-film studies had to be repeated because of over- or under-penetration. Imaging plate studies required 6% more time for processing than screen-film studies. Receiver-operating-characteristic analysis indicated no difference in diagnostic accuracy between the two imaging techniques. Subjective evaluation also revealed no difference in observer preference for imaging plate or screen-film studies. The imaging plate technique is an excellent alternative to screen-film studies in the operating room.
Angiography↗
Spinal cord compression due to metastatic disease: diagnosis with MR imaging versus myelography.
To determine the efficacy of magnetic resonance (MR) imaging and myelography for the diagnosis of spinal cord compression due to metastatic disease, the authors prospectively examined 70 patients who had known or suspected spinal involvement by malignancy. Most MR examinations consisted of T1-weighted sagittal imaging of the entire spine, with additional sequences as needed for clarification. Extradural masses were found in 46 patients, 25 of whom had cord compression. For extradural masses causing cord compression, the sensitivity and specificity of MR imaging was .92 and .90, respectively, compared with .95 and .88 for myelography. For extradural masses without cord compression the sensitivity and specificity of MR imaging was .73 and .90, versus .49 and .88 for myelography. MR imaging was much more sensitive for metastases to bone (.90 vs .49), as expected. MR imaging is an acceptable alternative to myelography for diagnosing spinal cord compression and is preferable as a first study because it is noninvasive and better tolerated.
Adult↗
A protocol for verifying truth of diagnosis. Preliminary report.
Explore the source record for details and available documents.
Clinical Protocols↗
Displays in radiology: past, present, and future.
Explore the source record for details and available documents.
Forecasting↗
Total digital radiology department: spatial resolution requirements.
The minimum spatial resolution required for a total digital radiology department has yet to be defined. A pilot study designed to provide this information was performed. Abnormal and normal radiographic images of children were digitized and redisplayed on film at spatial resolutions of 5.0, 2.5, 1.25, and 0.625 lp/mm. These resolutions are comparable to a digital display of a 14 X 14 in. chest image having pixel elements of 4096 X 4096, 2048 X 2048, 1024 X 1024, and 512 X 512, respectively. Contrast resolution was maintained at 12 bits or 4096 gray levels. The three phases of data acquisition were (1) the standard analysis of receiver operating characteristics, (2) a checklist evaluation of the "seeability" of important structures, and (3) a comparison of all resolutions and a discernment of usability. Fifteen radiologists participated in the study. On the basis of the pediatric cases used, the results showed that the needed spatial resolution for a total digital radiology department may be around 2.5 lp/mm (2048 X 2048). Checklist data on seeability of structures and comparisons of all resolutions give information on specific changes that are occurring as the resolution is decreased, and, when included with the receiver-operating-characteristic data, they become a major component in developing a resolution standard. The finding that 2.5 lp/mm is the required spatial resolution makes construction of a total digital radiology department possible with present state-of-the-art technology.
Child↗
Radiologic total lung capacity measurement. Development and evaluation of a computer-based system.
The development of a computer-based radiologic total lung capacity (TLC) measurement system designed to be used by non-physician personnel is detailed. Four operators tested the reliability and validity of the system by measuring inspiratory PA and lateral pediatric chest radiographs with a Graf spark pen interfaced to a DEC VAX 11/780 computer. First results suggest that the ultimate goal of developing an accurate and easy to use TLC measurement system for non-physician personnel is attainable.
Adult↗
Effect of noise correlation on detectability of disk signals in medical imaging.
Pixel signal-to-noise ratio is one accepted measure of image quality for predicting observer performance in medical imaging. We have found, however, that images with equal pixel signal-to-noise ratio (SNRp) but different correlation properties give quite different observer-performance measures for a simple detection experiment. The SNR at the output of an ideal detector with the ability to prewhiten the noise is also a poor predictor of human performance for disk signals in high-pass noise. We have found constant observer efficiencies for humans relative to the performance of a nonprewhitening detector for this task.
Fourier Analysis↗
The total digital radiology department: an alternative view.
Explore the source record for details and available documents.
Computers↗
The use of psychophysical principles in the design of a total digital radiology department.
In this article, the current design questions in developing a total digital radiology department (TDRD) are discussed, and the psychophysical techniques being used to determine the necessary spatial resolution in such systems are described. In addition, a review of the problems of designing the system and preliminary design results from the authors' institution are presented.
Equipment Design↗
The digital radiology department of the future.
The boom in microelectronics, including cost-effectiveness, has now allowed us to consider the use of these objects to store digital images. There remains much research, development, and clinical evaluation to be done in receptor technology. Further improvements in image processing, optical laser disk storage, and optical transmission and further commercial development of display technology must take place. All of these developments are occurring simultaneously. Within 5 to 10 years, radiology departments will most likely be totally electronic, probably cost-effective, and, it is hoped, more diagnostically accurate.
Data Display↗