A computer based """"""""intelligent tutor"""""""" will be developed to teach the interpretation of radiological images in clinical medicine. The system's general design will be guided by previous studies of image interpretation, which indicate that the process is effectively regarded as one of assigning status values to various features of the image an combining chose values with appropriate weights into a diagnostic decision. The system's specific design will be guided by new analyses of how image interpretation is done and taught by experts. The design of the tutorial system will draw also on recent work in the field of intelligent computer-assisted instruction (ICAI). This work suggests that the system should include, in addition to a model of the radiologic domain, a frequently-updated model of the student's knowledge, and a model of appropriate pedagogical strategy. The system will provide for a mixed-initiative dialogue, guided practice in interpretation of appropriate test images, feedback of the status values assigned the various features by experts, and graphical comparisons of the feature weights used by a student and the optimal weights as determined by discriminant analyses. The radiologic domain will be the interpretation of mammograms in the diagnosis of breast cancer. Instructional experiments will be carried out at the Harvard Medical School, both to improve and to evaluate the system's effectiveness and efficacy. A partially successful system would provide important information about the feasibility of using ICAI in radiology training and about the value of the feature-combining model of image interpretation. A fully successful system would provide a useful training tool that can be used in clinical settings to improve the interpretation of mammographic examinations. Enhancements of accuracy can have substantial effects in this particular imaging modality: the volume of examinations is now increasing very rapidly, and can be expected to increase primarily in community-based facilities in which, proficiency in any given modality is likely to fall short of that of specialists in referral centers. The proposed training tool should achieve an acceptability and continued value that is lacking in less flexible methods.
| Swets, J A (1992) The science of choosing the right decision threshold in high-stakes diagnostics. Am Psychol 47:522-32 |
| Maziarz, R T; Arceci, R J; Bernstein, S C et al. (1992) A gamma delta+ T-cell leukemia bearing a novel t(8;14)(q24;q11) translocation demonstrates spontaneous in vitro natural killer-like activity. Blood 79:1523-31 |
| Greenes, R A (1992) Medical education and decision support using network-based multimedia information resources. Ann N Y Acad Sci 670:244-56 |
| Greenes, R A (1991) A ""building block"" approach to application development for education and decision support in radiology: implications for integrated clinical information systems environments. J Digit Imaging 4:213-25 |
| Stein, L D; Snydr-Michal, J; Greenes, R A (1991) Realistic viewing and manipulation of radiographic images on a personal computer--a user interface for educational applications. J Digit Imaging 4:169-76 |
