2026-01-192026-01-192025-11-26https://repositorio.ifal.edu.br/handle/123456789/1376Parkinson’s disease (PD) is characterized as a neurodegenerative and progressive disorder that affects a significant portion of the elderly population worldwide. The most common symptoms in PD include cognitive problems, which predominantly begin to appear in more advanced stages, unlike the motor symptoms that manifest at the onset of the disease. The Movement Disorder Society Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) is one of the assessments that aid in diagnosing and monitoring the levels of Parkinson’s disease in patients. Within the MDS-UPDRS III, which assesses the motor aspects of patients, there is the Finger Tapping Test (FTT), used to assess bradykinesia and rigidity of movement. FTT is typically administered by a healthcare professional who observes the patient performing the test. However, the use of computer vision can provide a more effective alternative by accurately capturing the movement’s characteristics. The aim of this study was to develop a computer vision system capable of acquiring data from fine finger movements during the FTT and subsequently classify Parkinson’s disease levels according to the MDS-UPDRS. For this study, a computer vision algorithm was created to record data resulting from the abduction and adduction of hand movements during the FTT, calculating the amplitude generated by these movements. A dataset comprising 532 videos of both Parkinson’s and non-Parkinson’s patients, classified into four levels according to the MDS-UPDRS, was used. Subsequently, biomechanical signal processing procedures were conducted. The results obtained indicate that the use of computer vision in assisting healthcare professionals is promising. Data collection from movements allows for the creation of a database that can be used in machine learning models to predict the patient’s disease level based on the MDS-UPDRS. Additionally, a binary classifier was developed to determine the level at which each patient falls, using a Support Vector Machine (SVM) along with the leave-one-out cross-validation technique. This resulted in an accuracy of 73.3% when using the entire available dataset and 76.6% when using only the right-hand FTT.ptAttribution-ShareAlike 3.0 BrazilSistemas de InformaçãoVisão computacional – SistemaDoença de ParkinsonTeste de Toque de DedosFinger Tapping Test (FTT)Trabalho de Conclusão de CursoCIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::METODOLOGIA E TECNICAS DA COMPUTACAO::SISTEMAS DE INFORMACAO