Using convolutional dictionary learning to detect task-related neuromagnetic transients and ageing trends in a large...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biophysics

Background:

• Human neuromagnetic activity comprises complex transient bursts with dynamic spatial and temporal features.
• These burst characteristics evolve with task engagement and aging, providing insights into cortical activity.
• Existing burst detection methods often lack scalability for large datasets and group-level analysis.

Purpose of the Study:

• To apply a data-driven convolutional dictionary learning (CDL) approach for detecting neuromagnetic transient bursts in a large cohort.
• To analyze age-related trends in the spatiotemporal characteristics of identified neural bursts.
• To validate CDL's efficacy for large-scale neuromagnetic data analysis.

Main Methods:

• Utilized a data-driven convolutional dictionary learning (CDL) method on magnetoencephalography (MEG) data from the Cam-CAN dataset (538 participants, ages 18-88).
• Extracted repeating spatiotemporal motifs during a sensorimotor task.
• Clustered motifs across participants and analyzed age-related trends in task-related clusters.

Main Results:

• Identified seven task-related motifs, including beta, mu, and alpha bursts.
• All identified burst types exhibited increased activation levels with age, primarily due to a higher burst rate.
• Demonstrated positive age-related trends in the spatiotemporal characteristics of neural bursts.

Conclusions:

• Validated the convolutional dictionary learning (CDL) approach for robust transient burst detection in large-scale neuromagnetic datasets.
• Revealed significant age-related changes in human brain signal characteristics, specifically in burst rates.
• Provided a data-driven framework for understanding neural dynamics and their modulation by aging.