ANALYSIS OF CHEST X-RAY IMAGES USING SUPPORT VECTOR MACHINE

Modern medicine is able to effectively combat a wide range of diseases. This development has become possible due to the constant improvement of therapeutic methods. However, the success of the latter significantly depends on the quality and timeliness of diagnostics, which in severe cases can critically reduce the effectiveness of treatment. Traditional diagnostics based on expert assessment is susceptible to subjective factors. A promising alternative is computer analysis methods, such as radiography, which help minimize this negative impact. Machine learning algorithms provide the ability to process X-ray images, but such approaches require significant computing resources, giving rise to the problem of balancing the accuracy of the analysis and hardware requirements. The purpose of this study was to test two hypotheses: 1) the possibility of using the «weak» support vector machine method to predict diseases based on lung radiographs and 2) the feasibility of using «masks» instead of original images to train mathematical models. As a result, the truth of the first hypothesis was confirmed; at the same time, the second, suggesting the advantage of modified images over the original ones, turned out to be false.