Apresentação e avaliação de um programa de computador de auxílio ao diagnóstico médico: uma contribuição ao ensino da lógica diagnóstica

• Ano de defesa: 1995
• Autor(a) principal: Andrade, Pedro José Negreiros de
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Diagnóstico por Computador; Cardiologia; Medicina Interna
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/67250

The author reviews the basis of medicai reasoning in relation to a software to suppdrt the elaboration of diferential diagnosis in Internai Medicine. The program runs on IBM-corripatible computers, requiring 1 megabyte of RAM and 3.2 megabytes of hard disk space. The system simulates medicai reasoning using a Bayesian approach. In order to measure the diagnostic accuracy of lhe system, the program was tested comparing its performance in the diagnosis of 30 cases taken ffom Clinical- Pathological Conferences of the University Hospital of Federal University of Ceará (UFC) with the performance of theTaculty case discussants. The program listed the correct. diagnosis among the six most probable in 23 cases, versus 2.5 by the Case discussants (PX1.05) and listed the correct diagnosis as the first choise in 16 cases, equal to the numbers of correct citations by the case discussants (P=l). The software was also tested on Cardiology cases, comparing its performance with that of the case discussants in 91 Clinical-Pathological Conferences (taken from the New England Journal of Medicine, The Washington University School of Medicine and froin UFC). The program listed the correct diagnosis among the six most probable in 83 cases, versus 85 correct citations by the case discussants, and listed lhe correct diagnosis as the first. choise in 66 cases, compared to 80 correct diagnosis by the case discussants. Considering only the 73 Clinical-Pathological Conferences ofthe New England Journal of Medicine the program listed the correct diagnosis as the first choise in 54 cases, compared to 39 correct diagnosis by the Massachusets General Hospital Clinicians. The evaluacion of the system by interns and residents has been quite favorable, particulady in regard to its ease of use and its capacity to generate lists of dirferentia! diagnosis correctly.