Diagnosis of Parkinson’s Disease using Computer Aided Tool based on EEG

Parkinson’s Disease (PD) is a disorder affecting cortical and sub-cortical regions of the brain. The EEG recorded from these regions show variations in their patterns in patients with PD compared to that of healthy controls. Initially, Kolmogorov Complexity, Sample Entropy and Approximate Entropy are extracted from the epochs of 20 EEG electrodes collected from 18 subjects. These features are fed to the Back-Propagation Multi-Layer Perceptron (BP MLP) Neural Network and trained to perform diagnosis which results in 90% Accuracy, 100% Precision, 83% Sensitivity and 100% Specificity. Secondly, the above features from a set of EEG electrodes Fz, F1 and F2 that are placed close to cortical regions are considered and are fed to the BP MLP Neural Network, resulting in 97% Accuracy, 100% Precision, 96% Sensitivity and 100% Specificity. Therefore, PD can be effectively diagnosed by EEG readings recorded close to cortical regions thereby reducing the storage space and computation time.