Multivariate Multi-step Deep Learning Time Series Approach in Forecasting Parkinson's Disease Future Severity Progression

Parkinson's disease is a neurodegenerative disorder that affects the dopamine neurons production in the middle part of the brain. It is also recognized as the second most common degenerative nerve disorder in the United States after Alzheimer's disease. About 1% of the world population which estimated 7 to 10 million people with an average age of 62 are PD sufferers. Every year, approximately 60,000 Americans are diagnosed with PD, and the researchers believe this number will continue to grow. By providing a computational prognosis tool for PD, using patients' dataset containing clinical PD rating scale based on speech features could alleviate the PD progression. It can help a PD patient in monitoring the progress of unusual symptoms that they are currently facing based on previous and current recorded speech. This paper proposes a multi-step time series approach to forecasting the PD symptoms progression model using a deep neural network method, multichannel convolutional neural network (CNN). The experimental results show that our model could remarkably help in the forecasting of PD progression in the coming week/s.