A pairwise functional connectivity similarity measure method based on few-shot learning for early MCI detection

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Alzheimer's disease is irreversible, making early diagnosis critical for treatment.
• Early mild cognitive impairment (MCI) is a precursor stage requiring timely identification.
• Current diagnostic methods may not be sensitive enough for early-stage detection.

Purpose of the Study:

• To develop an automated method for detecting early mild cognitive impairment (MCI).
• To utilize functional connectivity (FC) patterns from resting-state functional magnetic resonance imaging (rs-fMRI) for diagnosis.
• To improve the accuracy of MCI detection by weighting the contribution of different brain region pairs.

Main Methods:

• A few-shot learning approach was used to measure pairwise functional connectivity (FC) similarity.
• Dynamic functional connectivity networks (FCNs) were generated using a sliding window strategy on rs-fMRI data.
• A self-attention mechanism was employed to weight FC features from different region pairs for enhanced classification.

Main Results:

• The proposed method successfully distinguished between normal controls (NCs) and early MCI patients.
• The use of a self-attention mechanism improved the classification accuracy by adaptively weighting FC features.
• Validation on the ADNI database demonstrated the approach's viability for early MCI detection.

Conclusions:

• The developed few-shot learning and self-attention based method is effective for early MCI detection.
• rs-fMRI coupled with advanced machine learning techniques shows promise for Alzheimer's disease diagnostics.
• This approach offers a potential tool for identifying individuals at the earliest stages of cognitive decline.