Parkinson's Disease Screening using Spiral Handsketch with Lightweight Deep Learning
V. Rajinikanth, Mathiyazhagan Narayanan
Abstract
Application of the artificial-Intelligence methods are widely considered in variety of data evaluation procedures due to its merit and adaptability on various image databases. The Lightweight Deeplearning (LD) schemes are considered to be simple and effective technique for analyzing RGB/grey scale images. This work proposes a LD-model based examination of Parkinson's disease (PD) using hand-sketch collected from individuals. The various phases of the developed LD-scheme include (i) labeled data collection and processing, (ii) feature extraction with chosen LD-model, (iii) classification with SoftMax and identification of best two models and generating fused-features (FF) based on 50% features reduction and serial concatenation, and (iv) detection and 5-fold cross validation. This study used 800 images per class and attained results verifies the merit of FF (accuracy >94%) with SoftMax.