An analytical method for measuring the Parkinson's disease progression: A case on a Parkinson's telemonitoring dataset. (March 2019)

• Record Type: Journal Article
• Authors: Nilashi, Mehrbakhsh; Ibrahim, Othman; Samad, Sarminah; Ahmadi, Hossein; Shahmoradi, Leila; Akbari, Elnaz
• Is Part Of: Measurement. Volume 136(2019)
• Page Start: 545
• Page End: 557
• Publication Date: 2019-03
• Subjects: Big data -- Prediction -- Data mining -- Parkinson -- Diseases diagnosis -- SVD -- ANFIS -- Ensemble; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2019.01.014
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A hybrid method is proposed for PD disease diagnosis. SOM and EM are used for the clustering the data. ANFIS ensemble is used for Total and Motor-UPDRS. SVD is used for dimensionality reduction for computation time reduction. ANFIS ensemble is used for predicting Total-UPDRS and Motor-UPDRS. Abstract: The use of machine learning techniques for early diseases diagnosis has attracted the attention of scholars worldwide. Parkinson's Disease (PD) is one of the most common neurological and complicated diseases affecting the central nervous system. Unified Parkinson's Disease Rating Scale (UPDRS) is widely used for tracking PD symptom progression. Motor- and Total-UPDRS are two important clinical scales of PD. The aim of this study is to predict UPDRS scores through analyzing the speech signal properties which is important in PD diagnosis. We take the advantages of ensemble learning and dimensionality reduction techniques and develop a new hybrid method to predict Total- and Motor-UPDRS. We accordingly improve the time complexity and accuracy of the PD diagnosis systems, respectively, by using Singular Value Decomposition (SVD) and ensembles of Adaptive Neuro-Fuzzy Inference System (ANFIS). We evaluate our method on a large PD dataset and present the results. The results showed that the proposed method is effective in predicting PD progression by improving the accuracy and computation time of the disease diagnosis. The method can be implemented as a medical decision support system for real-time PD diagnosis when big data from the patients is available in the medical datasets.