DFC-Igloo: A dynamic functional connectome learning framework for identifying neurodevelopmental biomarkers in very...

Area of Science:

• Neuroscience
• Developmental Biology
• Medical Imaging

Background:

• Very preterm infants face risks of neurodevelopmental impairments.
• Early detection of prognostic biomarkers is crucial for timely intervention.
• Previous studies were limited by small sample sizes and inefficient functional connectome (FC) construction algorithms for noisy neonatal data.

Purpose of the Study:

• To explore functional biomarkers for very preterm infants' future cognitive and motor development using resting-state fMRI.
• To propose an enhanced functional connectome construction algorithm to address limitations in prior research.
• To apply the new algorithm to a large cohort of very preterm infants to identify multi-level neurodevelopmental biomarkers.

Main Methods:

• Proposed an advanced dynamic FC (dFC) learning model, dFC-Igloo, leveraging structural connectome (SC) information.
• Applied graph signal smoothness to iteratively refine dFC estimations using both FC and SC.
• Evaluated the model on artificial graphs and simulated graph signals before application to a human cohort.

Main Results:

• The dFC-Igloo model demonstrated robust recovery of ground truth graphs across various noise levels.
• Applied to a cohort of very preterm infants, the model generated enhanced dFC for each infant.
• Identified neurodevelopmental biomarkers at connectome-wide, regional, and subnetwork levels based on improved dFC.

Conclusions:

• The identified biomarkers correlate with cognitive and motor developmental outcomes in very preterm infants.
• These findings provide insights into early brain development and potential neurodevelopmental challenges.
• The enhanced dFC construction algorithm offers a promising tool for neonatal neuroimaging research.