Brain-Shapelet: A Framework for Capturing Instantaneous Abnormalities in Brain Activity for Autism Spectrum Disorder...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Data Science

Background:

• Autism Spectrum Disorder (ASD) symptoms like anxiety and depression are often intermittent, complicating biomarker identification.
• Functional Connectivity Networks (FCNs) produce high-dimensional data, hindering the capture of instantaneous neurological abnormalities.

Purpose of the Study:

• To propose a novel framework, Brain-Shapelet, for extracting discriminative subsequences (Shapelets) from functional magnetic resonance imaging (fMRI) data.
• To capture instantaneous abnormalities in brain activity for improved Autism Spectrum Disorder (ASD) diagnosis.

Main Methods:

• Utilized a random walk algorithm on group-representative brain networks to identify brain region sets.
• Aggregated blood oxygen level-dependent (BOLD) signals to extract Shapelets reflecting real-time inter-regional brain associations.
• Developed a feature selection strategy to reduce Shapelet redundancy and optimize classification.

Main Results:

• Brain-Shapelet demonstrated superior performance in capturing short-term brain activity alterations.
• Achieved an 82.8% classification accuracy on the ABIDE dataset, outperforming traditional brain network modeling methods.
• Introduced novel metrics (co-occurrence rate, occurrence frequency, Gini coefficient) to quantify brain region contributions from a Shapelet perspective.

Conclusions:

• Brain-Shapelet offers a promising approach for identifying instantaneous biomarkers in neurological disorders like ASD.
• The framework provides valuable insights into ASD diagnosis by analyzing short-term brain activity patterns.
• This method enhances the understanding of brain activity abnormalities through the lens of Shapelet analysis.