Secure Comparative Evaluation of Alzheimer MRI Classification Models Using Blockchain

Abstract

Alzheimer’s disease (AD) is a chronic neurodegenerative disorder profoundly affecting memory and cognitive functions for which an early and precise diagnosis is essential to achieve timely intervention and disease management. Magnetic Resonance Imaging (MRI) is an important tool for detecting structural changes in the brain such as hippocampal shrinkage and ventricular enlargement, which can be correlated with Alzheimer’s disease’s stages of progression. In this work, we present a framework that couples deep learning-based Alzheimer’s MRI classification with blockchain-supported image authenticity verification. Our experimental setup compares five classification approaches, Xception, Long Short-Term Memory (LSTM) networks, ResNet50, Random Forest, and Gradient Boosting across different training durations. The best performing model is integrated to the local IPFS node and Ethereum smart contract through Ganache. This comparative investigation examines the balance of accuracy, efficiency, and training time in a variety of model designs. It also illustrates the viability of a secure, decentralized framework for both diagnostic accuracy and data integrity through blockchain.