Fig 1.
Experimental setup of the SQUID and OPM-MEG systems.
(A) Schematic representation of a SQUID-MEG system with a gradiometer array inside a liquid helium Dewar. (B) Dual-axis magnetometers produced by the company QuSpin, and an exemplary CAD model of one sensor holder. (C) Display of the subject’s head surface segmented from an MRI within the SQUID-MEG, and (D) OPM-MEG system. The SQUID-MEG system sensors cover the whole head; for this work we used only one hemisphere, the same as the OPM-MEG system. (E) Photo of one subject wearing the OPM-MEG sensor holder with OPMs placed on the right hemisphere. In the background, the SQUID-MEG system is visible.
Fig 2.
Schematic representation of two different transformation techniques.
(A) current source distribution located on the faces of an icosahedron calculated with the MNE method using the homogeneous spherical volume conductor model (MNE-SPH method), and (B) current source distribution on the subject’s reconstructed brain surface, calculated with the MNE method and individual geometry of the head (MNE-BEM method). In the figures, sensors from both MEG systems are drawn; however, we did not use them simultaneously.
Fig 3.
Display of measurements for three subjects using (A) SQUID and (B) OPM-MEG systems.
The first row displays the butterfly plots of averaged magnetic flux density due to acoustic stimulation for individual measurements. The second row represents the STD(t), where the characteristic M100 peak is marked with a green dot. In the third row, the MFMs at the time of the M100 peak can be found. For the OPM-MEG system, we show MFMs for each measured component (radial and tangential) separately. For the SQUID system, we show only 62 channels covering the right hemisphere.
Fig 4.
Transformation of the measured data between SQUID- and OPM-MEG for subject 2.
The upper part shows the originally measured OPM and SQUID MFMs at the time of the M100 peak for one subject. The bottom part shows MFMs transformed to the other MEG layout. For the transformation, we used the MNE-BEM method. Arrows indicate which measured MFM was used in the transformation. RE and CC were calculated between individual pairs of the original and the transformed data. For the SQUID data, we used only the right half of the channels.
Fig 5.
Transformation of the simulated data between SQUID- and OPM-MEG.
The upper part shows the originally simulated OPM and SQUID system data. The bottom part shows the transformed MFMs. For the simulations, the noise level was determined by calculating the average SNR of the measured data for each system (SNROPM = 11.1 dB, SNRSQUID = 13.8 db). Further details are identical to Fig 4.
Table 1.
Calculated values of SNR for each subject and each MEG system.
Fig 6.
Comparison of two transformation methods for the measurements.
The left image displays the relative error (RE), and the right the correlation coefficient (CC) between the measured original and the transformed magnetic field map from the other system. The orange column shows the results of the transformation of the SQUID data to the OPM-MEG system, and the blue column vice versa. The green and the red columns show the REs and CCs for comparison between the original and the reconstructed data. The black lines on top of the bars represent the standard deviation of the REs and CCs. Note that for both MNE methods the transformation cross-correlation CC is above 0.9, indicating that for the auditory M100 response both MEG devices record similar information.
Fig 7.
Transformation of simulated data using the method as indicated.
Each subfigure shows the averaged RE and CC vs. SNR for four cases: comparison of simulated and reconstructed SQUID data; comparison of simulated and reconstructed OPM data; comparison of simulated SQUID data and transformed simulated OPM data to the SQUID system; comparison of simulated OPM data and transformed simulated SQUID data to the OPM system. Each subfigure is the result of averaging 10 times for different random added noise. The dashed vertical lines show average values of SNR, which were calculated from the measured data for each MEG system.
Fig 8.
Impact of radial and tangential components during the transformation for the measured (A) and the simulated (B) MFM.
The transformation was done using the MNE-BEM approach. We show three different transformation scenarios: using only radial or tangential, and both sensor components of the OPM-MEG system. The blue bars show the averaged RE and CC when transforming the OPM measurements to the SQUID sensors, and the orange bars vice versa. The green and the red columns show REs and CCs for reconstructed data. The black lines on the top of the bars represent the standard deviation of REs and CCs. The subfigures for simulations were averaged 10 times for different random added noise. The noise level for each system was SNROPM = 11.1 dB, and SNRSQUID = 13.8 db. In both cases, (A) and (B), the best transformation results are obtained when using only the radial component.