ABSTRACT
Parkinson's disease (PD) is one of the major health problems all over the world that affects human life. It occurs when there is a loss of dopamine in the nerve cells. Dopamine is a chemical substance present in the neuron which sends a signal to other nerve cells. There is a unique presentation of motor and non-motor features in the PD subjects when enough dopamine is not produced. If early prediction is possible then people can receive treatment at right time. Hence, there is a need to develop new techniques that predict PD in the early stages. In this paper, some motor features (MDS-UPDRS, Hoehn & Yahr, and Modified Schwab & England ADL), non-motor features (University of Pennsylvania Smell Identification Test (UPSIT), Montreal Cognitive Assessment (MoCA) Score, Geriatric Depression Scale Score (GDS), Scales for Outcomes in Parkinson's Disease - Autonomic Dysfunction (SCOPA-AUT)) and Neuroimaging markers (Single-photon Emission Computed Tomography (SPECT) Striatum binding ratios (SBR) – Left, Right caudate and Left, Right putamen) are used for simulation. Machine Learning models such as KNN, Naive Bayes (NB), Support Vector Machine(SVM), Decision Tree(DT) and Random Forest(RF) are employed for early prediction of PD. It is observed that SVM and RF gave better accuracy results as compared to other models. SVM achieves 99.45% accuracy and RF achieves 99.63%, with 10-fold cross validation.
