Classification of Alzheimer's Disease Using Maximal Information Coefficient-Based Functional Connectivity with an...

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder causing cognitive decline and dementia.
• Accurate and early diagnosis of AD is crucial for effective management and treatment.
• Functional magnetic resonance imaging (fMRI) and deep learning offer potential for AD classification.

Purpose of the Study:

• To investigate the efficacy of fMRI-based functional connectivity (FC) measures combined with extreme learning machines (ELM) for classifying Alzheimer's disease (AD), mild cognitive impairment (MCI), and cognitively normal (CN) individuals.
• To compare the performance of linear (Pearson correlation coefficient - PCC) and non-linear (maximal information coefficient - MIC, extended maximal information coefficient - eMIC) FC measures in AD classification.

Main Methods:

• Utilized fMRI data to extract functional connectivity (FC) features using PCC, MIC, and eMIC.
• Employed extreme learning machines (ELM) as a deep learning model for classification tasks.
• Evaluated classification performance across three groups: CN vs. MCI, MCI vs. AD, and CN vs. AD.

Main Results:

• Non-linear techniques (MIC and eMIC) outperformed linear PCC for AD classification.
• eMIC achieved 94% accuracy in classifying CN vs. MCI, and 95% in CN vs. AD.
• MIC demonstrated higher accuracy (81%) than PCC (58%) and eMIC (78%) for MCI vs. AD classification.

Conclusions:

• fMRI-derived features from non-linear methods (MIC, eMIC) are effective for differentiating AD and MCI from CN individuals.
• These findings highlight the potential of advanced neuroimaging and machine learning for improving AD diagnosis.
• Non-linear connectivity analysis offers a promising avenue for enhancing the accuracy of AD classification systems.