Motion correction for infant functional near-infrared spectroscopy with an application to live interaction data

Area of Science:

• Neuroscience
• Biomedical Engineering
• Developmental Psychology

Background:

• Motion artifacts are a significant challenge in infant functional near-infrared spectroscopy (fNIRS) studies.
• Effective motion correction is essential for reliable neuroimaging data in infants.

Purpose of the Study:

• To evaluate conventional motion correction methods for infant fNIRS data.
• To compare motion and data quality across different infant ages and stimulus presentation methods.
• To assess the performance of wavelet filtering and targeted principal component analysis (tPCA).

Main Methods:

• Collected fNIRS data from infants at 5, 7, and 12 months.
• Utilized video and live stimulus presentation methods.
• Applied wavelet filtering and tPCA for motion correction.
• Analyzed motion metrics, data quality, and hemodynamic response recovery.

Main Results:

• Infant head speed varied by age, but data quality and hemodynamic recovery were consistent across ages (5, 7, 12 months).
• Video and live stimulus presentation yielded similar data quality.
• Wavelet filtering and tPCA demonstrated good performance with infant-specific parameters, requiring no fine-tuning for age or stimulus type.
• Trial rejection alone did not improve hemodynamic response recovery.

Conclusions:

• Wavelet filtering is recommended for infant fNIRS motion correction, with specific parameters suggested but flexibility noted.
• Motion correction methods, particularly wavelet filtering, perform reliably across infant ages and stimulus types.
• Data quality metrics from uncorrected data can predict motion correction efficacy, potentially reducing the need for simulation studies.