Banadyga A. Prediction of acute pancreatitis and differentiated choice of treatment tactics

The dissertation theoretically generalizes and solves in a new way the scientific task – to improve the results of treatment of patients with acute pancreatitis by using mathematical cluster systems and creating a neural network model for predicting the severity of clinical course and complications, introduction of differentiated treatment tactics. For the first time, a neural network model for predicting the course of acute pancreatitis was created, which improved the results of patient management. The group of prognostic indicators (procalcitonin> 3.08 ng/ml, C-reactive protein >0.5 mg/ml, glycemia >10 mmol/l, abdominal compartment syndrome of the III degree) which in combination with clinical criteria of severity define the approach to treatment is established by a method of the cluster analysis. The introduced use of the algorithm of differentiated choice of treatment tactics has reduced the number of complications. Thus, in acute pancreatitis of moderate severity, the percentage of complications decreased more than twice: from 28.48 % to 12.33 %, in severe: from 37.51 % to 23.12 %, and in critical from 45.80 % to 34, 04 %. The use of the created diagnostic algorithm in acute necrotic pancreatitis provided a reduction in the number of postoperative complications by 1.5 times (p<0.01), overall mortality by 9.21 % (p<0.05), the duration of treatment of the patient in the hospital in 1,9 times (p<0.05).