Computer-assisted Parkinson's disease diagnosis using fuzzy optimum- path forest and Restricted Boltzmann Machines

• Published in: Computers in biology and medicine Vol. 131; p. 104260
• Main Authors: de Souza, Renato W.R., Silva, Daniel S., Passos, Leandro A., Roder, Mateus, Santana, Marcos C., Pinheiro, Plácido R., de Albuquerque, Victor Hugo C.
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.04.2021; Elsevier Limited
• Subjects: Algorithms; Diagnosis; Feature extraction; Fuzzy logic; Fuzzy optimum-path forest; Internal Medicine; Machine learning; Medical diagnosis; Motor skill; Movement disorders; Neural networks; Neurodegenerative diseases; Other; Parkinson's disease; Reliability aspects; Signs and symptoms; Support vector machines; Parkinson's disease; Fuzzy optimum-path forest; Machine learning
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2021.104260

Parkinson's disease (PD) is a progressive neurodegenerative illness associated with motor skill disorders, affecting thousands of people, mainly elderly, worldwide. Since its symptoms are not clear and commonly confused with other diseases, providing early diagnosis is a challenging task for traditional methods. In this context, computer-aided assistance is an alternative method for a fast and automatic diagnosis, accelerating the treatment and alleviating an excessive effort from professionals. Moreover, the most recent studies proposing a solution to this problem lack in computational efficiency, prediction power, reliability among other factors. Therefore, this work proposes a Fuzzy Optimum Path Forest for automated PD identification, which is based on fuzzy logic and graph-based framework theory. Experiments consider a dataset composed of features extracted from hand-drawn images using Restricted Boltzmann Machines, and results are compared with baseline models such as Support Vector Machines, KNN, and the standard OPF classifier. Results show that the proposed model outperforms the baselines in most cases, suggesting the Fuzzy OPF as a viable alternative to deal with PD detection problems.