An Efficient Algorithm for Analysis of Handwriting Task for the Detection of Parkinson's disease

Abstract

Parkinson's disease results in tremors, stiffness, bradykinesia, and rigidity due to dopamine depletion. Several diagnostic tests were applied to detect Parkinson's disease such as MRI images and EEG signals. The lack of conventional methods is taking a relatively long time to produce results and These methods are mostly expensive. The new concept to detect Parkinson's disease is depending on the image processing for handwriting images. In previous studies, scientists found that the sketching rate becomes steady in healthy cases and instability appears among patient cases. In this paper, we used Radon Transform to enhance the training of the CNN and increase its ability to detect Parkinson’s. Also, we used various types of machine learning classifiers and the comparing the accuracy between the different selected classifiers and the previous techniques. The accuracy obtained by the proposed technique reached up to 92.45% with 70.38% specificity, 88.98% F1 score , and precision of 87.68% Consequently.