| Research Abstract |
I developed a decision support system for diagnosis using essential laboratory tests according to the Japan Society of Clinical Pathology. On the basis of diagnostic logic, diagnostic categories were set up, such as inflammation, muscular or myocardial disease, anemia, myeloproliferative disease, renourinary disease, hepatobiliary disease, diabetes mellitsu, gastrointestinal disease, bone disease, hyperlipidemia, hyperuricemia and hypolipidemia. Further, comments are provided for each disease category, physiological variance, possible diagnosis and other items needed for diagnosis are shown. The present system was used for admission laboratory data on 211 patients whose diagnosis had been already confirmed in order to evaluate its diagnostic ability. In each diagnostic category, 11/26 cases of infectious disease, 34/90 cases of malignant tumor, 2/17 cases of muscular disease, 15/15 cases of anemia, 2/16 cases of myeloproliferative disease, 42/51 cases of reno-urinary disease, 8/14 case
… More
s of diabetes mellitus, 9/20 cases of gastrointestinal disease, 6/26 cases of bone disease and 13/13 cases of hyperlipidemia were correctly categorized. The present system was also used for 131 outpatients upon their initial clinic visit. Its diagnostic ability was comparatively good, and asymptomatic disease were detected (38/131 cases of hyperlipidemia, 29/131 cases of hepatobiliary disease, 21/131 cases of amemia) ; The system also proved effective for pinpointing latent disease. Upon initial examination the new system makes it possible to select the most effective and optimal test items. Besides, I developed a diagnostic system for each categorized case by means of a new type of computer technology. In 165 cases of chronic liver disease confirmed by biopsy or operation, the neural network showed an overall diagnostic accuracy of 64.2%, which was above mean diagnostic accuracy, 53.7% by hepatologists. This system is useful as an expert system substitute for specialists. With the use of fuzzy logic, high diagnostic accuracy was noted for acute hepatitis, alcoholic liver damage and cholestasis with characteristic laboratory data, in spite of diagnostic overlapping for chronic persistent hepatitis, chronic active hepatitis and fatty liver ; misdiagnosis was rare for healthy and non-hepatopathic cases. Accordingly, this new diagnostic system using fuzzy logic is useful as a screening system. Less
|
