Stereo vision-based co-registration of OPM-MEG and MRI: Achieving real-time sensor array localization and tracking

The advent of low-cost, wearable optically pumped magnetometers (OPMs) has significantly expanded the applications of magnetoencephalography (MEG). Magnetic source imaging (MSI) is a critical technique in MEG that enables the dynamic spatial imaging of neural activity. Accurate MSI relies on the precise co-registration of OPM sensors with magnetic resonance imaging (MRI). As OPM-MEG adoption grows, the demand for fast, accurate, and fully automated registration methods is increasing. However, existing methods often rely on manual scanning or digitization, leading to cumbersome procedures. In this study, we achieved automatic co-registration of OPM and MRI using a stereo vision approach to detect helmet markers and reconstruct the 3D facial structure of the subject. Experiments with a reference phantom showed that the proposed method achieved co-registration accuracy with position errors of below 1 mm and orientation errors below 0.6°. Simulations further confirmed that the MSI localization error was better than 1 mm at this accuracy level. Moreover, our method enables real-time localization and tracking of the sensor array with spatial localization accuracies exceeding 0.3 mm. The dynamic pose information acquired from the sensor array is expected to significantly enhance background magnetic field modeling, thereby suppressing motion artifacts.