Classification of Neurodegenerative Diseases Based on VGG 19 Deep Transfer Learning Architecture: A Deep Learning Approach
Classification of Neurodegenerative Diseases Based on VGG 19 Deep Transfer Learning Architecture: A Deep Learning Approach
K. Bhatele
TLDR
The proposed system based on VGG 19 outperforms the other three popular deep transfer learning models and delivers an average accuracy of 90% with 70/30 (training/validation split) and 93% with 80/20 ( training/ validation split) for the multiclass classification as Alzheimer disease, Healthy Control ADNI, Healthy control PPMI and Parkinson disease.
Abstract
This paper presents an automated system for the classification of two commonly diagnosed neurodegenerative diseases i.e. Alzheimer and Parkinsons, with the aid of deep transfer learning model Visual Geometry Group 19 (VGG19). As deep learning architectures are already very much used for the classification of various brain tumors and other types of cancers but not used much for the classification of Neurodegenerative diseases. There is an imperative need to have an automated system that can perform the classification among the Neurodegenerative diseases as the number of patients suffering from these disease are increasing. This system can even help the radiologists to enhance their diagnostic efficiency and accuracy of these Neurodegenerative diseases. In this paper, VGG19 deep transfer learning model is implemented, hyper tuned and trained over the Alzheimer and Parkinson disease (ADPP) dataset, which is developed using Alzheimer’s disease Neuroimaging Initiative (ADNI) and Parkinson's Progression Markers Initiative (PPMI) databases for the multiclass classification of Alzheimer, healthy control and Parkinson disease. As there is no such system proposed yet to perform this degree of multiclass classification. So other popular deep transfer learning models like ResNet 50, Inception Net and VGG 16 are also implemented over the same ADPP dataset and their performance are compared with the proposed system based on VGG19 architecture. The proposed system based on VGG 19 outperforms the other three popular deep transfer learning models and delivers an average accuracy of 90% with 70/30 (training/validation split) and 93% with 80/20 (training/validation split) for the multiclass classification as Alzheimer disease, Healthy Control ADNI, Healthy Control PPMI and Parkinson disease. This work can be extended in future in order to propose an automated system for the multiclass brain structural disorders classification.