Automated analysis of the American College of Radiology mammographic accreditation phantom images.

A significant metric in federal mammography quality standards is the phantom image quality assessment. The present work seeks to demonstrate that automated image analyses for American College of Radiology (ACR) mammographic accreditation phantom (MAP) images may be performed by a computer with objectivity, once a human acceptance level has been established. Twelve MAP images were generated with different x-ray techniques and digitized. Nineteen medical physicists in diagnostic roles (five of which were specially trained in mammography) viewed the original film images under similar conditions and provided individual scores for each test object (fibrils, microcalcifications, and nodules). Fourier domain template matching, used for low-level processing, combined with derivative filters, for intermediate-level processing, provided translation and rotation-independent localization of the test objects in the MAP images. The visibility classification decision was modeled by a Bayesian classifer using threshold contrast. The 50% visibility contrast threshold established by the trained observers' responses were: fibrils 1.010, microcalcifications 1.156, and nodules 1.016. Using these values as an estimate of human observer performance and given the automated localization of test objects, six images were graded with the computer algorithm. In all but one instance, the algorithm scored the images the same as the diagnostic physicists. In the case where it did not, the margin of disagreement was 10% due to the fact that the human scoring did not allow for half-visible fibrils (agreement occurred for the other test objects). The implication from this is that an operator-independent, machine-based scoring of MAP images is feasible and could be used as a tool to help eliminate the effect of observer variability within the current system, given proper, consistent digitization is performed.

A new approach to film dosimetry for high energy photon beams: lateral scatter filtering.

A method of film dosimetry for high energy photon beams is proposed which reduces the required film calibration exposures to a set of films obtained for a small radiation field size and shallow depth (6 cm x 6 cm at 5 cm depth). It involves modification of a compression type polystyrene film phantom to include thin lead foils parallel to the vertical film plane at approximately 1 cm from both sides of the film emulsion. The foils act as high atomic number filters which remove low energy Compton scatter photons that otherwise would cause the film sensitivity to change with field size and depth. The proposed method is best described as "lateral scatter filtering." To validate the proposed method, central axis depth doses and isodose curves for a 4 MV photon beam were determined from films exposed within the modified phantom and the results compared with ionization chamber measurements. When no lateral filtering was used, for field sizes of 6 cm x 6 cm and 25 cm x 25 cm, this comparison demonstrated up to a 65% difference between film and ionization chamber central axis depth dose measurements. When using the lateral scatter filtering technique, less than a 4% difference was observed for these field sizes.

Computer-based radiology information system: from floppy disk to CD-ROM.

The authors have developed a comprehensive computer-based radiology information system known as "Radiology Resource and Review" (R3). The content is divided into 10 radiology information categories spanning the entire human body and presently includes more than 4 Mbytes of text, 9,000 topics, and 6,000 images. The R3 software and the information content are stored on compact disk, read-only memory (CD-ROM) media. The images are compressed by using a standard compression algorithm. Images and text are cross-indexed with more than 13,000 key words, which can be linked together in searches by using Boolean logic. Four different retrieval interfaces support browsing of text and image information, diagnosis decision making, self-study, and teaching file preparation. The 10-year university-funded evolution of R3 is an example of the transition in storage media from floppy disk to CD-ROM.

Computerized case-based instructional system for computed tomography and magnetic resonance imaging of brain tumors.

RATIONALE AND OBJECTIVES: A computerized system that applies the case-based approach to training radiologists to diagnose brain tumors was designed. The authors attempted to provide residents a tool that supports their visual memory and inducts case-based reasoning. METHODS: A relational database with a digital image library was implemented and incorporated into a computer aided instruction environment based on case presentation. An indexing system was structured around case features (case history and radiologic findings). "If-then" type rules were used to control the search direction within the case library and to generate lists of diagnostic hypotheses. The indexing system was evaluated against cases "known" to the system. RESULTS: The current case library consists of 122 cases with 640 digitized images (computed tomography and magnetic resonance imaging). The accuracy of retrieval for "known" to the system cases was 80.7%. CONCLUSIONS: A case library stored on a personal computer can be efficiently searched for a combination of radiologic findings and can offer quality images for comparison to the case in question. A case library is a source of the information that may be used by different teaching applications.

Subjective evaluations of mammographic accreditation phantom images by three observer groups.

RATIONALE AND OBJECTIVES: Mammography providers are under increasing pressure to become certified by the American College of Radiology (ACR). Successful accreditation is contingent on passing a phantom image quality test. This study was undertaken to evaluate viewer performance with the phantom image evaluation process and to determine any observer group differences. METHODS: A series of standard phantom images were viewed by 30 medical physicists, 30 diagnostic radiologists, and 30 inexperienced observers. From the responses, object detection rates and passing rates according to the ACR criteria were established. These responses were analyzed with standard nonparametric tests to assess the degree of variability, correlation, and agreement among different observer groups. RESULTS: Median passing scores were similar for the radiologists and untrained readers, but the medical physicists appear to differ from the other two groups. There was not sufficient evidence to indicate that special training among physicists made a significant difference in median passing scores or mass detection rates. However, such training appeared to significantly affect the detection of microcalcification and fibril test patterns among the physicists' subgroups. Agreement among observer groups was high for all groups, but tended to be lower for overall passing rates than for any of the individual test objects. Agreement among physicists was affected by their subspecialty, presumably caused by their levels of specific training for these visual tasks. CONCLUSIONS: The authors conclude that choosing medical physicists to evaluate mammographic phantom films appears to be a good choice among potential observer groups, and that special training for reading these images affects their detection abilities and consistency. However, because passing rates did not appear to be affected by special training and given the current rapid degree of change in this area, more testing of medical physicists is desirable to examine these effects over time, and to study the effect of developing standards for training.

Preparation of a 99mTc wipe standard for survey instrument calibration in nuclear medicine.

A method was developed for the preparation of a standard source to satisfy the U.S. Nuclear Regulatory Commission requirement for calibration of wipe-assay procedures used in nuclear medicine laboratories. An essential element of this standard was that a locally prepared calibration wipe could be used with thin-window Geiger-Muller detectors that detect 99mTc conversion electrons with high sensitivity. This proposed calibration wipe was evaluated with both a NaI(Tl) well counter and a thin-window Geiger-Muller pancake probe. Average detector sensitivities of 1,540 and 160 counts min-1 per 2000 disintegrations min-1, respectively, were found for 99mTc. An analysis of the errors for each step of the procedure showed most to be less than 3% (95% confidence level). The nuclear medicine dose calibrator, used to assay the activity from which the wipe was prepared, was estimated to contribute a 9% absolute calibration error. Use of the pancake probe contributed a 12% geometry-reproducibility error. The total error associated with the use of the calibration wipe standard to calibrate a typical 99mTc wipe-assay procedure, exclusive of the count-dependent error associated with the number of counts acquired, was found for the NaI(Tl) well counter to be given by a percentage of 2 SD = 10% and for the thin-window Geiger-Muller pancake probe, percentage 2 SD = 16%. Coating of the calibration wipe with an acrylic spray, to prevent it from contaminating instruments during use, was evaluated. This coating was effective while it reduced the detectability of the wipe activity by a negligible degree.

Corneal contact times of ophthalmic vehicles. Evaluation by microscintigraphy.

Lacrimal microscintigraphy, in conjunction with a recently developed computer system, was used to evaluate the corneal contact time of three ophthalmic vehicles in 18 humans. The percentage of a radioactively labeled vehicle remaining over the cornea after 90 seconds was 2.9% plus and minus 2.2% for saline, 4.3% plus and minus 2.4% for polyvinyl alcohol, and 8.8% plus and minus 4.1% for hydroxypropyl methylcellulose.