DenseCNN: A Densely Connected CNN Model for Alzheimer's Disease Classification Based on Hippocampus MRI Data

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Biomarker Discovery

Background:

• Alzheimer's Disease (AD) is a prevalent dementia impacting cognition and behavior.
• The hippocampus is a key biomarker for AD diagnosis, with prior research focusing on its volume, texture, and shape.
• Current 3D Convolutional Neural Network (CNN) models for AD classification using hippocampus MRI are often complex and data-intensive.

Purpose of the Study:

• To develop an accurate and lightweight 3D CNN model for Alzheimer's Disease classification.
• To leverage hippocampus segments for improved AD diagnosis.
• To overcome the limitations of existing complex and data-hungry CNN models.

Main Methods:

• A Densely Connected 3D Convolutional Neural Network (DenseCNN) was designed for AD classification.
• The DenseCNN model was trained and tested using hippocampus MRI data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.
• The model utilized hippocampus segments for feature extraction and classification.

Main Results:

• The DenseCNN model achieved an average accuracy of 0.898.
• High sensitivity (0.985) and specificity (0.852) were recorded.
• A significant Area Under the Curve (AUC) of 0.979 was obtained, outperforming or matching state-of-the-art methods.

Conclusions:

• The proposed DenseCNN is an effective and efficient tool for Alzheimer's Disease classification based on hippocampus MRI.
• This lightweight model demonstrates superior or comparable performance to existing approaches.
• DenseCNN offers a promising avenue for improved, data-efficient AD diagnosis using neuroimaging biomarkers.