Sistema inteligente para identificação de patologias laríngeas

Abstract

Every year, thousands of people are diagnosed with some type of laryngeal pathology, the most common being polyps, nodules and Reinke's edema. The examinations and diagnoses of these cases can often be inaccurate and difficult to perform during routine medical practice. The focus of this monograph was to propose an intelligent system capable of classifying patients with possible laryngeal diseases, including dysphonia and Reinke's edema, based on audio recordings of healthy and sick patients taken from a world-renowned database. The objective of this system is to assist health professionals in the analysis and diagnosis of these pathologies during their work routine. The focus of this work was also to use several different algorithms to perform this classification (Random Forest, Support Vector Machine, K-Nearest Neighbor and Multi-layer Perceptron), comparing their effectiveness metrics with the metrics of the same models used in intelligent systems proposed in other works in the scientific literature. The Multi-layer Perceptron model obtained the best precision and accuracy values ​​(0.836 and 0.700) of the models analyzed, showing promise for the analysis of voice recordings and subsequent identification of laryngeal pathologies. However, as an opportunity for improvement and future work, it may be interesting to work with a larger database, extract other features, or even use other classifiers. All this with the aim of complementing the experimentation carried out or obtaining even more convincing results.