Wavelet coherence-based classifier: A resting-state functional MRI study on neurodynamics in adolescents with...

Area of Science:

• Neuroscience
• Biomarkers
• Medical Imaging

Background:

• Autism spectrum disorder (ASD) diagnosis is subjective and lengthy due to the absence of a reliable biomarker.
• Previous attempts using functional MRI (fMRI) for ASD classification yielded suboptimal accuracy (<80%) and lacked validation.
• Emerging evidence suggests temporal brain dynamics, not network topology, are crucial for understanding ASD.

Purpose of the Study:

• To develop a novel, objective MRI-based biomarker for ASD by analyzing temporal brain dynamics.
• To investigate the utility of wavelet coherence in identifying ASD-related neurodynamic changes.
• To assess the diagnostic performance and robustness of a novel ASD biomarker.

Main Methods:

• Resting-state fMRI data from two independent adolescent datasets were analyzed.
• Independent component analysis identified socio-executive resting-state networks (RSNs) and their time series.
• Wavelet coherence maps were used to calculate a novel metric: time of in-phase coherence, for ASD classification.

Main Results:

• Classifiers achieved 86.7% accuracy in distinguishing ASD from non-ASD adolescents across datasets.
• Sensitivity and specificity ranged from 83.3% to 100% and 66.7% to 88.9%, respectively.
• Cross-site validation demonstrated robust classification performance, with 80% accuracy when training on one dataset and testing on another.

Conclusions:

• Changes in temporal neurodynamic coherence serve as a reliable biomarker for ASD.
• Wavelet coherence-based classifiers provide robust and replicable diagnostic results.
• This approach holds potential as an objective diagnostic tool for ASD.