An fNIRS-Based Feature Learning and Classification Framework to Distinguish Hemodynamic Patterns in Children Who...

Area of Science:

• Neuroscience
• Speech-language pathology
• Biomedical engineering

Background:

• Stuttering affects 1% of the population, with most preschool children recovering.
• Understanding why some children persist in stuttering is crucial but poorly understood.
• Identifying neurophysiological biomarkers for stuttering persistence is a key research objective.

Purpose of the Study:

• To detect neurophysiological biomarkers of stuttering persistence.
• To differentiate neural activation patterns in children who stutter, do not stutter, and have recovered from stuttering.
• To develop a data-driven diagnostic approach for stuttering.

Main Methods:

• Utilized functional near-infrared spectroscopy (fNIRS) brain imaging data from 46 children.
• Employed a novel supervised sparse feature learning approach.
• Analyzed cerebral hemodynamics during a speech production task.

Main Results:

• Identified a small set of discriminative cerebral hemodynamics features.
• Achieved 87.5% accuracy in differentiating neural patterns between children who stutter and those who do not.
• The selected features show promise as biomarkers for stuttering persistence.

Conclusions:

• The study presents promising biomarkers for understanding stuttering neurophysiology.
• These biomarkers can aid in differentiating persistent stuttering from recovery.
• Facilitates future data-driven diagnostic tools for children's stuttering.