Improving neuroimaging headgear placement robustness using facial-landmark-guided augmented reality

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Computer Science

Background:

• Accurate probe placement is critical for functional near-infrared spectroscopy (fNIRS) and electroencephalogram (EEG) studies.
• Operator experience and subject head shape variability impact placement accuracy, especially in longitudinal and group studies.

Purpose of the Study:

• To develop NeuroNavigatAR (NNAR), an augmented reality (AR) and machine learning-based software for real-time estimation and display of cranial landmarks.
• To guide consistent headgear placement in fNIRS and EEG experiments.

Main Methods:

• Utilized a facial recognition toolbox to track 3D facial landmarks from video input.
• Developed a transformation model between facial and cranial landmarks using a large head-model library (>1000 subjects).
• Rendered atlas-derived head landmarks onto the subject's camera stream in real-time.

Main Results:

• Achieved 15 frames per second processing speed with an open-source GUI on a laptop.
• Reported a median 10-20 position error of 1.52 cm, reduced to 0.75 cm with subject-specific head surfaces.
• Demonstrated consistent landmark prediction and no significant accuracy difference across age groups.

Conclusions:

• NNAR is an intuitive AR tool for monitoring headgear placement in neuroimaging.
• Expected to significantly improve consistency and decrease setup duration for fNIRS and EEG probe application.
• Facilitates more reliable data acquisition across diverse research applications.