Deep learning based prediction of Alzheimer's disease from magnetic resonance images

TL;DR

This paper proposes a deep learning approach using transfer learning with ResNet architectures to predict Alzheimer's disease stages from MRI images, achieving improved accuracy on a Kaggle dataset.

Contribution

It introduces a ResNet-based deep neural network model with transfer learning for AD prediction from MRI images, demonstrating superior performance over other models.

Findings

ResNet architecture outperforms VGG and other networks.

Transfer learning enhances model accuracy.

Best results achieved with deep ResNet models.

Abstract

Alzheimer's disease (AD) is an irreversible, progressive neuro degenerative disorder that slowly destroys memory and thinking skills and eventually, the ability to carry out the simplest tasks. In this paper, a deep neural network based prediction of AD from magnetic resonance images (MRI) is proposed. The state of the art image classification networks like VGG, residual networks (ResNet) etc. with transfer learning shows promising results. Performance of pre-trained versions of these networks are improved by transfer learning. ResNet based architecture with large number of layers is found to give the best result in terms of predicting different stages of the disease. The experiments are conducted on Kaggle dataset.