A Novel Unit-Based Personalized Fingerprint Feature Selection Strategy for Dynamic Functional Connectivity Networks.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biomedical Engineering

Background:

• Resting-state functional Magnetic Resonance Imaging (rs-fMRI) and sliding-window-based dynamic functional connectivity networks (SW-D-FCN) are crucial for diagnosing neurodegenerative diseases.
• Extracting and selecting discriminative features from SW-D-FCN remains a significant challenge.
• Conventional methods often use reductionist strategies, potentially missing personalized discriminative characteristics within functional connectivity (FC) sequences.

Purpose of the Study:

• To propose a novel unit-based personalized fingerprint feature selection (UPFFS) strategy.
• To enhance the capture of discriminative features associated with specific diseases from SW-D-FCN.
• To improve the utilization of personalized fingerprint features in classification tasks.

Main Methods:

• The proposed UPFFS strategy treats the FC sequence between each pair of brain regions of interest (ROIs) as a unit.
• For each unit, the most discriminative feature is identified using a specific feature evaluation method.
• All identified discriminative features are concatenated to form a comprehensive feature set for classification.

Main Results:

• The UPFFS strategy successfully identified relevant discriminative features from SW-D-FCN data.
• Experiments distinguishing individuals with autism spectrum disorder (ASD) from controls demonstrated the strategy's effectiveness.
• The proposed method achieved superior performance compared to existing benchmark techniques.

Conclusions:

• The UPFFS strategy offers a more effective approach to feature selection from SW-D-FCN.
• This method fully mines and utilizes personalized fingerprint features for improved classification accuracy.
• The findings suggest UPFFS holds significant potential for advancing diagnostic tools in neurodegenerative disease research.