Neural network model of treatment policy decision in patients with ulcer disease

: Based on neuronetwork computer technologies expert neuronetwork complex Axon was developed providing individual prognosis of ulcer disease and recommendations. The base of this expert system is information data base consisting of two blocks. The first block consists of detailed information about 634 patients with gastric ulcers (GU), the second block - about 417 patients with duodenal ulcers (DU). Control evaluation of the system demonstrated that the coefficient of sensitivity in patients with GU was 0,706 - 0,724, coefficient of specificity - 0,701-0,790. Coefficient of sensitivity in patients with DU was 0,765-0,817, coefficient of specificity - 1,000-0,877. The best neuronetwork evaluation was carried out in 57 new hospitalized patients. In 7 of 10 operated patients indications for surgery were based on combination of prognostic and clinical data. Only 2 (4,3%) of 47 patients demonstrated disagreement of prognostic and clinical data. In comparison of results of complications prognosis with real clinical results in surgical group 7 (70,0%) of 10 patients demonstrated agreement of these data. In 32 (74,4%) of 43 patients treated conservatively prognosis of complications corresponded to real clinical data.