| Abstract | The differential diagnosis of human malformation patterns is difficult because of their great number and low occurrence. To compensate for deficiencies in experience, the authors have developed a microcomputer based differential diagnosis for dysmorphic syndromes. Written in PL/I-80 and using a MDBS III data base containing more than 700 signs and 300 syndromes, the algorithm of SYNDROC, the program's name, is built on three different approaches. An initial differential diagnosis is given by a pseudo-Bayesian method. Its refinement is possible through a heuristic approach, and undiagnosed cases are stored for subsequent Boolean retrieval, thus permitting the recognition of new dysmorphic disorders. The reliability of the program was tested in a retrospective study based on 100 dysmorphic cases. The results showed a 95% agreement between the first two diagnoses suggested by SYNDROC and the diagnosis made by experienced geneticists. |