Parkinson’s Disease Prediction Using Artificial Neural Network Algorithms and A Studyon Effects of Data Attributes to Prediction Results



Parkinson’s disease is a chronic neurodegenerative impairment which causes movement impairment. Dopaminergic deficiency resulted from the loss of dopa-minergic neurons in the substantia nigra causes the disease. UPDRS (Unified Parkinson's disease rating scale) is an important scale for evaluation of clinical severity of Parkinson’s disease. Recent computational studies using in silico pre-diction methods show promising results in terms of their potential diagnostic relevance. In this study we aimed to demonstrate a prediction method for evalu-ation of clinical motor and total UPDRS using regression analysis with neural network. In addition, we investigated the importance of different attributes in our regression algorithm provided from telemonitoring for evaluation of their predictive relevance. The correlation between predicted motor UPDRS score and clinical motor UPDRS score was found as 97%. Exclusion of Jitter values did not directly affect the predictive power of the model whereas clinical UPDRS scoring proved its importance to achieve to generate more predictive models.