QuSpin’s optically pumped magnetometers (OPMs) have transformed non-invasive neuroimaging by enabling high-resolution, on-scalp measurements with flexible sensor arrays. But with flexibility comes complexity: determining the precise position and orientation of each sensor is critical for accurate source localization. To streamline this process, QuSpin developed a complete analysis pipeline that begins with the HALO system for collecting sensor localization data and continues through LabVIEW– and Python-based software tools for extracting and formatting that data. This new tutorial completes the workflow by leveraging the powerful, open-source MNE-Python ecosystem—guiding users through every step from raw Neuro-1 MEG data to interactive 3D visualizations of localized phantom data.
Who Should Use This Tutorial?
This workflow is ideal for neuroscientists, neuroengineers, and technical staff working with the QuSpin Neuro-1 system. Whether you’re deploying flexible MEG arrays for cognitive studies, validating source models in a lab environment, or developing new biomagnetic instrumentation, this tutorial equips you with the practical tools to work with Neuro-1 data.
If you’re already familiar with Python and MNE-Python, this guide will fast-track your ability to analyze Neuro-1 data MEG data. If you’re new to source modeling but comfortable with scientific computing, the step-by-step layout provides an approachable entry point.
Click me to download the tutorial
What You’ll Learn
The tutorial uses a dataset recorded from a 63-sensor array placed on a grid, localized using QuSpin’s HALO system. It models a phantom dipole located ~6.5 cm above the HALO center—ideal for testing the accuracy of different inverse methods.
The workflow includes:
-
Sensor Localization – A brief video demo shows how to use QuSpin’s Sensor Localization Software to extract positions and orientations from a HALO scan. This spatial calibration is the backbone of source analysis.
-
Data Loading & Inspection – Raw MEG data are loaded and automatically scanned for bad channels using both geometric checks (orthonormality) and signal quality thresholds.
-
Preprocessing – The tutorial filters the data around the 40 Hz test signal and removes 60 Hz power line noise, preparing the data for event detection.
-
Sensor-Level Analysis – Stimulus triggers are used to define epochs and create an averaged evoked response. These steps help characterize the signal prior to source modeling.
-
Forward Modeling – A spherical head model is created to match the phantom, and a volumetric grid defines potential source locations.
-
Inverse Solution – Three methods—MNE, dSPM, and sLORETA—are applied to estimate where the signal originated, enabling side-by-side comparison of results.
-
Interactive Visualization – A 3D Plotly viewer presents sensor positions and localized sources, including inter-method distance metrics, for easy comparison.
Ready for the Next Step
Although this tutorial is designed for phantom data, the workflow translates directly to human studies. Researchers can substitute realistic BEM head models, MRI-derived source spaces, and subject-specific coregistration steps without changing the overall structure.
Get Started
If you’re working with OPM MEG systems—especially in flexible or portable formats—and need a proven path from sensor localization to source estimation, this tutorial is for you. It’s open-source, reproducible, and designed to work seamlessly with QuSpin’s hardware.
Contact Dave Bobela at dbobela@quspin.com with any questions or comments.