| Title | Modeling Techniques for Medical Diagnosis: II. Differential Diagnosis of Neonatal Hepatitis and Biliary Atresia |
| Author(s) | Charles E. Mize; William M. Lively; Stephen A. Szygenda |
| Source | Computers and Biomedical Research, Vol. 9, Pages 239-245 |
| Publication Date | 1976 |
| Abstract | Heuristic liver modeling and learning program techniques have been utilized to develop a method for augmenting presurgical guidance in anatomic biliary tree obstruction and hepatitis in neonates. A liver model programmed for evaluation of clinical data leads to one of four categories: (1) hepatitis, (2) atresia, (3) normal, (4) no diagnosis possible. Using a simple set of data (serum glutamate-oxalacetate transaminase and direct Van den Bergh bilirubin, a defined disease duration, and Rose Bengal fecal excretion), 36 test patients whose diagnoses were unknown were tested. Comparison with actual subsequent diagnoses in these patients confirmed the method as a reliable aid in medical diagnosis and management decisions. |