Neural Network Prediction of the Effectiveness of Pharmacotherapy in Hepatitis

Abstract

Despite the standardization of medical diagnostics, the treatment remains and is regarded as an art, since the experience of a clinical specialist is required to select effective tactics. And the task of determining the effectiveness itself involves dealing with great uncertainty, the work with which is not fully available when using standard algorithms in modern computers. Using the methodology of forming topology and training neural networks for the task of determining pharmacological effectiveness is a new concept, the study in which will bring some degree of objectivity to the study. In the proposed publication, an analysis and adjustment of the methodology of forming the topology of neural networks to evaluate the effectiveness of hepatitis pharmacotherapy is carried out, it is proposed to evaluate the effectiveness by dynamically assessing the severity of patients. At the initial stage, a selection of important clinical factors is carried out, which mainly characterize the state of the disease; a scale is selected to assess the condition of the patient with hepatitis disease. For the proposed state assessment, the use of neural networks is considered, an effective algorithm for evaluating and forming the neural network topology for the problem under consideration is selected. A comparison is made of the selected methodology for modular construction of the neural network topology with analogs that are used for medical purposes. In the final part of the study, using the decision support system, we analyze the effectiveness of seven drugs for hepatitis pharmacotherapy.