Detecção de lesões de joelho por técnicas de vibroartrografia e machine learning

Abstract

The diagnosis of knee injuries is currently performed using high-cost and invasive techniques, which demand high financial investments and a series of car in the post-procedure period. Among the techniques used today are magnetic resonance imaging and computed tomography, common and expensive procedures, although not invasive, and also arthroscopy, which is not only expensive but also invasive for the patient. Given this scenario, the present work seeks to develop an alternative method for the diagnosis of knee injuries based on the technique known as vibroatrography, with low cost and non-invasive. Diagnostic vibroarthrography uses sensors to measure the acceleration and angular velocity in the patient’s knee. These signals are mathematically processed by Fourier coding and then are analyzed by different Machine Learning methods. The present work used information from 43 patients with different knee weaknesses as data for training the equipment, in the period between April 28, 2022 and August 4 of the same year, which were collected in partnership with Hospital do Trabalhador, Curitiba – PR . The project completed the diagnosis of knee problems using the vibroarthrography technique and obtained 99.39% of Precision, 100% of Sensitivity, 99.03% of Accuracy and 100% of Specificity for the set of sensors measuring the medial y-axis combined with the lateral angular velocity of the x-axis and the Machine Learning Decision Tree algorithm, which indicates that the method has good results in detecting knee injuries.