A Multi-Stream Convolutional Neural Network for Classification of Progressive MCI in Alzheimer's Disease Using...

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Neurology

Background:

• Early diagnosis of Alzheimer's disease (AD) and mild cognitive impairment (MCI) is crucial for timely intervention.
• Distinguishing between stable MCI and progressive MCI is essential for predicting AD development.

Purpose of the Study:

• To develop and evaluate a multi-stream deep convolutional neural network (CNN) for classifying stable MCI and progressive MCI.
• To leverage patch-based MRI data and anatomical landmarks for improved MCI classification.

Main Methods:

• Utilized a multi-stream deep CNN architecture fed with patch-based MRI data.
• Identified distinct anatomical landmarks by comparing AD and cognitively normal MRI scans using multivariate statistical tests.
• Employed a transfer learning approach, pre-training the model on AD and normal cognition data before fine-tuning on MCI data.

Main Results:

• The proposed method achieved an F1-score of 85.96% on the ADNI-1 dataset for MCI classification.
• Outperformed existing methods in differentiating between stable and progressive MCI.
• Demonstrated the effectiveness of patch-based MRI analysis and transfer learning in MCI diagnosis.

Conclusions:

• The developed multi-stream deep CNN shows significant promise for accurate and early detection of progressive MCI.
• This AI-driven approach can aid clinicians in identifying individuals at higher risk of developing Alzheimer's disease.
• The findings support the use of advanced neuroimaging analysis techniques for neurodegenerative disease diagnostics.