[Paper Review] Deep Learning for Early Alzheimer Disease Detection with MRI Scans

Mohammad Rafsan, Tamer Oraby+3 more|ArXiv.org|Jan 17, 2025

Brain Tumor Detection and Classification3 citations

TL;DR

This study compares CNN, Bayesian CNN, and U-Net models on the OASIS brain MRI dataset to detect early Alzheimer’s disease, with Bayesian CNN achieving accuracy above 95% and Grad-CAM for interpretability.

ABSTRACT

Alzheimer's Disease is a neurodegenerative condition characterized by dementia and impairment in neurological function. The study primarily focuses on the individuals above age 40, affecting their memory, behavior, and cognitive processes of the brain. Alzheimer's disease requires diagnosis by a detailed assessment of MRI scans and neuropsychological tests of the patients. This project compares existing deep learning models in the pursuit of enhancing the accuracy and efficiency of AD diagnosis, specifically focusing on the Convolutional Neural Network, Bayesian Convolutional Neural Network, and the U-net model with the Open Access Series of Imaging Studies brain MRI dataset. Besides, to ensure robustness and reliability in the model evaluations, we address the challenge of imbalance in data. We then perform rigorous evaluation to determine strengths and weaknesses for each model by considering sensitivity, specificity, and computational efficiency. This comparative analysis would shed light on the future role of AI in revolutionizing AD diagnostics but also paved ways for future innovation in medical imaging and the management of neurodegenerative diseases.

Motivation & Objective

• Motivate early detection of Alzheimer's disease using MRI and AI driven analysis.
• Evaluate and contrast three deep learning architectures (CNN, Bayesian CNN, U-Net) on a public MRI dataset.
• Address class imbalance in medical imaging through SMOTE-Tomek preprocessing.
• Assess model performance on accuracy, precision, recall, and F1-score.
• Provide interpretability of model predictions using Grad-CAM.

Proposed method

• Apply CNN, Bayesian CNN, and U-Net models to MRI slices from the OASIS dataset.
• Balance data with SMOTE-Tomek before training and evaluation.
• Evaluate models using accuracy, precision, recall, and F1-score.
• Use Grad-CAM to visualize brain regions contributing to predictions for interpretability.
• Describe Bayesian CNN via Bayes by Backprop and variational inference with ELBO objective.
• Implement and compare uncertainty-aware modeling in Bayesian CNN with stochastic gradient variational Bayes.

Experimental results

Research questions

• RQ1 Can Bayesian CNN outperform standard CNN and U-Net in early AD detection on MRI data?
• RQ2 How do these models perform in terms of accuracy, precision, recall, and F1 when detecting early stages of AD?
• RQ3 Does Grad-CAM provide meaningful localization of relevant brain regions for model predictions?
• RQ4 What is the impact of SMOTE-Tomek balancing on model performance?
• RQ5 Can uncertainty modeling in Bayesian CNN contribute to more reliable diagnostics?

Key findings

• Bayesian CNN achieved accuracy above 95% on the task.
• CNN and U-Net achieved the next highest performance among the evaluated models.
• Grad-CAM was used to visualize brain regions contributing to predictions, aiding interpretability.
• The study employed SMOTE-Tomek to balance the dataset before training and evaluation.
• The analysis discusses uncertainty quantification provided by Bayesian CNN and its potential clinical utility.