Predict Alzheimer's disease using hippocampus MRI data: a lightweight 3D deep convolutional network model with visual...

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder.
• Hippocampal atrophy is a key biomarker for AD diagnosis.
• Previous models like DenseCNN used hippocampus MRI segments for AD classification.

Purpose of the Study:

• To develop an improved deep learning model for AD classification.
• To incorporate global shape representations of the hippocampus alongside visual features.
• To enhance diagnostic accuracy for Alzheimer's disease.

Main Methods:

• Proposed DenseCNN2, a 3D deep convolutional network.
• Utilized T1-weighted structural MRI data from the Alzheimer's Disease Neuroimaging Initiative (ADNI).
• Trained and evaluated DenseCNN2 on 326 AD and 607 cognitively normal (CN) hippocampus MRI using 5-fold cross-validation.

Main Results:

• DenseCNN2 with combined features outperformed models using single feature types.
• Achieved high performance metrics: 0.925 accuracy, 0.882 sensitivity, 0.949 specificity, and 0.978 AUC.
• UMAP visualization confirmed improved class discrimination between AD and normal subjects with global shape features.

Conclusions:

• DenseCNN2 effectively classifies Alzheimer's disease using combined hippocampus segmentations and global shape features.
• The model demonstrates potential as an efficient diagnostic tool for AD.
• Integrating global shape features enhances the performance of deep learning models for AD detection.