A Transfer Learning Approach for Predicting Alzheimer's Disease
Sreeja Sasidharan Rajeswari, Manjusha Nair
Abstract
Alzheimer”s disease(AD) is a human brain disorder that gradually damages the memory and cognitive skills of an individual. The prediction of this condition was often laborious and time consuming. In order to reduce these constraints, different deep learning algorithms were investigated to automate the AD detection and prediction. In this study, the potency of transfer learning approach was analyzed in detail by fine tuning the deeper layers of Transfer Learning models like VGG-19, VGG-16, Resnet-50 and Xception. In prior work, VGG-16 model was experimented on the ADNI datasets to get an accuracy of 97% and a precision of 96%. From this study, it is analyzed that the overall classification accuracy and precision of VGG- 19 is exceptionally high (98 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> ) when compared to other models. This shows that the automated method can be a true guide in Alzheimer”s detection and prediction, especially when an early stage diagnosis may help to get the benefit of treatment.