A0590
Title: Machine learning classification of functional brain imaging for Parkinsons disease stage prediction
Authors: Guan-Hua Huang - National Chiao Tung University (Taiwan) [presenting]
Abstract: The aim is to analyze a dataset containing functional brain imaging from 6 normal healthy controls and 196 patients with Parkinson's disease (PD), which can be divided into 5 stages according to the severity of illness. The goal is to predict patients' PD illness stages via their functional brain images. Used approaches include multivariate statistical methods, ensemble learning models, and deep convolutional neural network (CNN). For statistical and ensemble models, PCA is performed to extract features, and the best combination of parameters is found by grid search. For CNN modeling, we use the technique of image augmentation to increase data size and build the model by the architecture of VGG16. It is found that the deep learning VGG16 model outperforms other approaches, which can capture significant features from imaging and reach higher classification accuracy.
