Computer-aided diagnosis of the character of bowel obstruction

A computer programme for the differential diagnosis of bowel obstruction was created and put into clinical practice to accelerate and simplify the diagnostic process. The training material was 503 cases of bowel obstruction admitted to the Surgical Clinic of Kaunas Medical University during 1990-1996. Based on 36 statistically significant anamnestical, clinical, laboratory investigations and plain abdominal X-ray findings, the computer programme was built up using a Bayesian formula. Retrospectively, the prognostic diagnosis was compared with the final clinical diagnosis based on instrumental or operative findings. Then, in a control group of 136 patients, the prospective prognostic diagnosis was obtained. The accuracy of the prognostic diagnosis in the control group of patients with complete small bowel obstruction amounted to 88.7% and for the patients with complete large bowel ileus 95.8%. Prognostic accuracy for partial small bowel obstruction was the most precise (96.1%) and for partial large bowel ileus (87.5%). The overall diagnostic accuracy of the computer algorithm was 92.6%. All cases were classified. This computer algorithmic programme for the differential diagnosis of the character of mechanical bowel obstruction has markedly shortened and facilitated the process of diagnosis of ileus.