Method for predicting surgical complications

Abstract

he problem of predicting the risk of purulent-inflammatory complications after surgery in patients with purulent-destructive lung diseases is still unsolved. When analyzing a sample of 543 patients with purulent-destructive lung diseases in the Penza Regional Clinical Hospital, 45 (8.3 %) had purulent-inflammatory complications. The aim of the study is to create a neural network system for predicting the risk of surgical complications in patients with purulent-destructive lung diseases. As a result of this study, the technology of constructing neural network models for predicting complications in thoracic surgery was developed. In particular: methods of selection and transformation of features have been developed and the neural network system «Neuropredictor» has been developed, which has demonstrated high accuracy rates.