Использование нейронной сети для совершенствования дифференциальной диагностики хронического аллергического и хронического вазомоторного ринита

Abstract

The clinical symptoms of chronic allergic rhinitis (A-rhinitis) and chronic vasomotor rhinitis (V-rhinitis) are so similar that it results in subjective diagnosis. There have been attempts to clarify this problem with the help of statistics. However, in the process of using statistical methods (which are based on the selective method), there arise some difficulties due to a number of objective reasons. They are the high variability of the studied characteristics (symptoms) due to the effect of a great number of uncontrolled factors; the problems of processing of small-volume samplings. Moreover, the samples obtained by using the actual population of patients in a particular hospital or clinic are, in general, neither random nor homogeneous. Various factors (the subjectivism in complaints, anamnesis, in the examination of patient by a doctor), affecting the selection of patients, resulting in the formation of a certain group of patients who are not the representatives of the general population. The constant pathomorphism of the diseases makes the search for other methods and technologies that allow considering the effect of various factors on the course on the pathological process especially relevant. The developing techniques of artificial neural networks have been widely used these days. The authors have developed a multilayer artificial feedforward neural network that provided a classification of patients with chronic allergic and chronic vasomotor rhinitis. The results indicate the complex interrelations between the symptoms of these two forms of rhinitis. The authors have studied the following categories of allergic and vasomotor rhinitis symptoms: “Complaints”, “Anamnesis”, “Objective Status” from the viewpoint of effect of the degree of subjectivism (“high”, “medium”, “low”) on the accuracy of the decision as to the differential diagnosis of the pathology. It has been established that the minimum degree of subjectivism (“low”, 2.27%) refers to the category of “Objective Status” symptoms. The application of logical filtration operation to the output results of the neural network made it possible to identify another form in the groups of patients with allergic and vasomotor rhinitis – the mixed rhinitis. The suggested technology reduces subjectivity in differential diagnostics of the forms of rhinitis.