Welcome to Magnivel International
Robust Imputation and Classification of Parkinson’s Disease Using a Questionnaire Dataset: A Novel Variational Autoencoder Approach
Abstract
This study introduces a novel Fully Connected Variational Autoencoder (VICPute) to effectively address the challenges of missing data imputation and classification of Parkinson’s disease (PD) using the Parkinson’s Progression Markers Initiative (PPMI) dataset. The VICPute model, which integrates an encoder with dual decoders, refines imputation and classification processes through a pioneering architecture that employs computational blocks for enhanced feature extraction. This model distinguishes between healthy individuals and those with PD and robustly handles significant missing data issues within a complex dataset that includes 24,182 samples from 2,427 patients. Our approach optimizes training by applying advanced preprocessing techniques such as Multivariate Imputation by Chained Equations (MICE) and generating artificial missing data. The model’s efficiency is underscored by a high f1-score of 0.87 in testing, highlighting the potential of deep learning techniques to enhance diagnostic accuracy and contribute to the broader understanding of PD progression.
Keywords: Classification, Imputation, Parkinson’s disease, Progression of disease, Autoencoder, PPMI dataset
Magnivel International is an Open Access publisher and international scientific conferences and expo Organizer. We own and operate peer-reviewed clinical, medical, life sciences, engineering, and management journals and hosts scholarly conferences per year in the fields of clinical, medical, pharmaceutical, life sciences, business, engineering and technology.
