Diagnostic algorithms and clinical diagnostic thinking

Abstract

Diagnostic algorithms are evaluated as a model of medical reasoning in the differentiation of the two syndromes or diseases with overlapping clinical manifestations. Linear-discriminant analysis (LDA), nonlinear discriminant analysis (NLDA), and sequential-statistical analysis (SSA) are compared with the decision-making practices of physicians, as they differentiated 175 cases with anterior or posterior aphasia, 789 cases with cerebral infarction or intracerebral hemorrhage, and 200 cases of schizophrenia or schizophreniform organic psychosis. Comparisons showed that the clinical decision-making process used by physicians can be approximated by neural sets or parallel distributed processing. Implications for the implementation of computer diagnostic programs in clinical practice are discussed.<>