Mathematical model of diagnostics of malignant pancreas pathology

Abstract

Annotation. Chronic pancreatitis (CP) is a common recurrent pathology of the pancreas. The long course of the inflammatory process, accompanied by chronicity, is the cause of concomitant complications of the gastrointestinal tract, and is also often attributed to the causes that lead to the development of a pathology with an extremely unfavorable prognosis – pancreatic cancer (PC). The purpose of the work is to build a mathematical model for the diagnosis of malignant pathology of the pancreas based on the available group of real statistical data in the form of symptoms of the type of presence (presence). 45 patients who were operated on between 2018 and 2022 were analyzed. A problem of automatization of diagnosing malignant pancreas pathology or its absence is considered. The goal is to build a mathematical model of diagnosing the malignant pathology based on an available group of statistical data in the form of 0 and 1, which indicate the absence and presence of the definite symptom. Based on the selected symptoms of the pathology development likelihood increase, a mathematical model in the form of binary classification is built by using probabilistic neural networks. A set of the selected symptoms is divided into a group of 14 more influential symptoms and a group of 13 less influential symptoms. A set of statistical data of 20 patients with a correctly diagnosed presence of pathology and a set of statistical data of 25 patients with a correctly diagnosed absence of malignant pathology were formed for the construction and testing of a classifier. Thus, if the number of false symptoms did not exceed five in each of the sets, then the corresponding classifier, which is an average of 20,000 probabilistic networks, determines the diagnosis without error. It is noted that the slower 53130 classifier is more reliable.