Classification of Alzheimer's Progression Using fMRI Data

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Functional magnetic resonance imaging (fMRI) has advanced brain research over three decades.
• Deep learning models show promise in various complex data analysis tasks.

Purpose of the Study:

• To develop and evaluate a deep learning model for diagnosing Alzheimer's disease (AD) progression.
• To classify individuals into categories: condition normal (CN), early mild cognitive impairment (EMCI), late mild cognitive impairment (LMCI), and AD.

Main Methods:

• A 4D fMRI dataset was pre-processed in four steps to remove noise.
• A 3D-CNN-LSTM classification model was employed, utilizing U-Net for spatial feature extraction and LSTM for temporal feature extraction.
• Comparative experiments involved training three models with adjusted time dimensions.

Main Results:

• The proposed 3D-CNN-LSTM model achieved an average accuracy of 96.4% using five-fold cross-validation.
• The model effectively extracted both spatial and temporal features from fMRI data.
• The method demonstrated high potential in identifying Alzheimer's disease progression.

Conclusions:

• The developed 3D-CNN-LSTM model shows significant potential for the early identification and diagnosis of Alzheimer's disease progression using 4D fMRI data.
• This approach offers a non-invasive method for monitoring cognitive decline and disease staging.