Expert system for pathology learning

Abstract

Context and setting Expert system programs follow the same rational steps as a human expert to solve a specific problem. Their use in teaching may be effective, as users can learn by observing the behaviour of the system. We developed an expert system for lymphoid pathology diagnosis, based on pattern recognition, the key to diagnosis in pathology. Why the idea was necessary A recent call for novel computer-based approaches to the teaching of pathology in European schools of medicine prompted our group to adapt and test our expert system for this purpose. What was done The modified expert system for lymphoid pathology knowledge (SEACOPALI) shows images selected for their similarity to a problem case. According to the pattern selected, further patterns and descriptive and diagnostic data are presented, requiring the user to focus on specific characteristics and offering differential diagnoses. A study was performed with 3 groups of 3rd-year medical students (n = 101 students) who used either the SEACOPALI system, or SEACOPALI documentation or lymphoid pathology text plus atlas. Students were tested on their existing knowledge (pretest), after which they completed a questionnaire on 4 clinical cases using their respective learning resources (computer system or documentation or traditional materials). This was immediately followed by a second test of their knowledge (post-test). The expert system group also completed a questionnaire on the usability of the system. Evaluation of results and impact Performance of the test itself impacted positively on knowledge acquisition (correct results: 27·5% pretest versus 56·6% post-test), as reported in previous studies of problem-based learning strategies. The expert system group achieved the highest level of learning in the histological diagnosis of low- and high-grade lymphomas (P < 0·05). The groups did not differ in their ability to diagnose Hodgkin's lymphoma, but the expert system and the expert system documentation groups showed greater skill in pattern recognition than the traditional learners (P < 0·001); 68·6% of the expert system group said they would recommend its use to their peers and 42·9% indicated they would like to use the system daily. Results were validated by using the Kruskal–Wallis non-parametric test. These preliminary results indicate that use of the expert system assists acquisition of practical knowledge and pattern recognition and that the system would be optimally used as a complement to traditional teaching. The need for modifications to the system was revealed by the study, leading to the development of a new expert system prototype that incorporates a screen to display problem cases and can be used interactively by the student.