Volume 19, No. 4, 2022
Machine Learning Based Idiopathic Parkinson’s Disease Detection Using Speech Data
Mafas Raheem
Abstract
The World Health Organization (WHO) has identified neurodegenerative disorders as one of the critical threats to public health. Currently, it is estimated that 16 out of 60 people suffer from neurological related diseases. Parkinson's disease is one of the most serious neurological diseases since there is no exact cure as yet. It is also referred to as Idiopathic Parkinson’s disease since there is no reliable cause for the disease. Speech impairments analysis has been used as an efficient tool for the early detection of Parkinson's disease. Voice or speech can be a reliable biomarker for Parkinson’s disease since 90% of Parkinson’s disease patients normally experience hypokinetic dysarthria. There are several problems associated with Parkinson’s disease such as there are obstacles to early diagnosis of Parkinson’s disease, lower efficiency of the existing diagnosis method and expensive cost for the existing diagnosis. Machine learning has become a solution for detecting many types of diseases, and it can potentially be applied for the detection of Parkinson’s disease as well. In this line, several commonly used machine learning predictive models were built and used speech data that can detect Parkinson’s disease. Among the models, with the support of proper preprocessing and model tuning with cross-validation, the Deep Neural Network outperformed by obtaining 87.17% of accuracy with 0.88 of precision.
Pages: 308-325
Keywords: Parkinson’s disease, Speech data, Machine Learning, Classification, Deep Neural Network.