A PARALLEL MEDICAL THERMOGRAPHIC IMAGE ANALYSIS USING A SUPPORT VECTOR MACHINE AND A NAIVE BAYESIAN CLASSIFIER

Abstract

The system for automated analysis of medical thermograms has a wide range of applications, covering such fields as medicine, sports, military technology, and industry. In medicine, thermography is an important tool for diagnosing a variety of diseases, including breast, joint, and other organ cancers, due to its ability to display temperature differences that may indicate pathological processes. In the sports industry, thermography is used to monitor and predict injuries, helping to reduce the risk of injury and improve training efficiency. In military technology, thermography is used to detect terrorists or suspicious persons on the ground, as well as to localize the wounded in a combat zone, ensuring a quick and accurate response. In this paper, we propose a parallel approach to automatically process thermographic images and detect possible pathologies using an ensemble of classification methods. This ensemble consists of a support vector machine (SVM) and a naive Bayesian classifier (NB), and a voting method is used to make the final decision. A high F1-Score of 0.93 was obtained, which indicates the effectiveness of the developed system. To speed up the training time of the model, parallelization was used using the joblib library and its parallel_backend function, which led to an acceleration factor close to two. This significantly improves the system's performance and ensures fast processing of large amounts of data. Further recommendations and goals for the next steps were also discussed, including expanding the functionality of the system by training the model on different types of diseases, improving data processing algorithms, and using more complex machine learning models to obtain more accurate results.