Neural Network Approach to Medical Express Diagnostics

Abstract

Modern computerized medical diagnostics systems effectively solve complex and important problems in the field of medicine, including diagnosing diseases, monitoring patients, predicting treatment outcomes, supporting decision-making about diagnosing and treating patients. This is due to their ability to instantly analyze and summarize many factors in the process of diagnosing biomedical data. The use of neural network technologies in medical expert systems, in particular for express diagnostics, can significantly improve this process. Neural networks enable data mining, information retrieval, recognition (classification) of objects (symptoms) and visualization of the obtained results. In this work, the features of the neural network approach to medical express diagnostics are analyzed. The analysis of biomedical diagnostics methods and means showed the relevance and perspective of the use of neural network technologies. The neural network classifier based on the advanced Hamming network with the formation of discriminant functions is proposed. It allows to perform express diagnostics on a set of defined symptoms with the use of the generated in the learning the memory of weights for specific diseases. A single y_k signal indicates the disease at k-number when the output layer of the neural network classifier is triggered. Therefore, the formation of the binary output signal Y = {y_i} in the proposed neural network classifier provides the ability to visualize the result of diagnosis with using of LEDs. The hardware implementation of the proposed neural network classifier, along with software support, will significantly speed up the process of diagnosing biomedical data using neural network express diagnostics systems.