A multi-model deep convolutional neural network for automatic hippocampus segmentation and classification in...

Area of Science:

• Neuroimaging and computational neuroscience
• Artificial intelligence in medical diagnostics

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder with no cure; early diagnosis is crucial for intervention.
• Mild cognitive impairment (MCI) often precedes AD, making its accurate identification vital.
• Traditional methods using hippocampus shape/volume from MRI have limitations, including segmentation errors and suboptimal feature extraction.

Purpose of the Study:

• To develop a novel deep learning framework for joint automatic hippocampal segmentation and AD classification using structural MRI.
• To improve the accuracy and robustness of early AD and MCI detection compared to existing methods.

Main Methods:

• A multi-task deep convolutional neural network (CNN) was designed for simultaneous hippocampal segmentation and disease classification.
• A 3D DenseNet model was employed to learn features from segmented hippocampal regions for classification.
• Features from both models were combined for final disease status classification (AD vs. Normal Control (NC), MCI vs. NC).

Main Results:

• The framework achieved 87.0% Dice similarity coefficient for hippocampal segmentation.
• Classification accuracy reached 88.9% (AUC 92.5%) for AD vs. NC and 76.2% (AUC 77.5%) for MCI vs. NC.
• The multi-model approach demonstrated superior performance over single-model and other competing methods.

Conclusions:

• The proposed deep learning framework effectively performs joint hippocampal segmentation and classification for AD and MCI detection.
• This integrated approach offers a promising tool for accurate and early diagnosis of Alzheimer's disease and its precursor.
• The findings highlight the potential of advanced AI models in neurodegenerative disease diagnostics using MRI data.