Safety and data quality of EEG recorded simultaneously with multi-band fMRI

Purpose Simultaneously recorded electroencephalography and functional magnetic resonance imaging (EEG-fMRI) is highly informative yet technically challenging. Until recently, there has been little information about EEG data quality and safety when used with newer multi-band (MB) fMRI sequences. Here, we measure the relative heating of a MB protocol compared with a standard single-band (SB) protocol considered to be safe. We also evaluated EEG quality recorded concurrently with the MB protocol on humans. Materials and methods We compared radiofrequency (RF)-related heating at multiple electrodes and magnetic field magnitude, B1+RMS, of a MB fMRI sequence with whole-brain coverage (TR = 440 ms, MB factor = 4) against a previously recommended, safe SB sequence using a phantom outfitted with a 64-channel EEG cap. Next, 9 human subjects underwent eyes-closed resting state EEG-fMRI using the MB sequence. Additionally, in three of the subjects resting state EEG was recorded also during the SB sequence and in an fMRI-free condition to directly compare EEG data quality across scanning conditions. EEG data quality was assessed by the ability to remove gradient and cardioballistic artifacts along with a clean spectrogram. Results The heating induced by the MB sequence was lower than that of the SB sequence by a factor of 0.73 ± 0.38. This is consistent with an expected heating ratio of 0.64, calculated from the square of the ratio of B1+RMS values of the sequences. In the resting state EEG data, gradient and cardioballistic artifacts were successfully removed using traditional template subtraction. All subjects showed an individual alpha peak in the spectrogram with a posterior topography characteristic of eyes-closed EEG. The success of artifact rejection for the MB sequence was comparable to that in traditional SB sequences. Conclusions Our study shows that B1+RMS is a useful indication of the relative heating of fMRI protocols. This observation indicates that simultaneous EEG-fMRI recordings using this MB sequence can be safe in terms of RF-related heating, and that EEG data recorded using this sequence is of acceptable quality after traditional artifact removal techniques.

Electrode heating has previously been characterized as a function of the Specific Absorption Rate (SAR) [19,32]. However, the whole-body SAR estimates can differ amongst scanners 1 0 4 [34,35], rely on assumptions about the body being scanned [36], and typically provide biased 1 0 5 measurements [34]. Given these problems it has been suggested to characterize safety limits RMS refers to a root-mean-square, or quadratic mean of the B 1+ field calculated over time. This 1 1 0 quantity is proportional to the oscillating electric fields responsible for RF heating. The heating associated with a given experiment can be characterized in absolute or relative 1 1 6 terms. Absolute heating experiments attempt to measure the temperature change due to scanning 1 1 7 for a given experimental condition. The accuracy of these measurements is complicated by the simultaneous EEG-fMRI with the same MB sequence to assess EEG data quality. University of Illinois at Urbana Champaign. Anatomical information was obtained using a high-resolution 3D structural MPRAGE scan (0.9 1 6 9 mm isotropic, TR = 1900 ms, TI = 900 ms, TE = 2.32 ms, GRAPPA factor = 2).  12,43,44]. Therefore, the MB factor and slice number were chosen to maintain maximal brain 1 7 9 coverage at a short TR while moving the RF excitation repetition artifact outside of the report of a task-based functional localizer using the MB sequence in a single human subject (S1 properties of the object being scanned. Because RF energy deposition scales as the square of ‫ܤ‬ ଵ 1 9 7 [46], the MB sequence is expected to exhibit lower RF energy deposition than the SB protocol 1 9 8 by a factor of (8/10) 2 = 0.64. This estimate is in line with time-averaged RF power deposition 1 9 9 values reported by the scanner during scanning of a standard 2 liter aqueous Siemens phantom 2 0 0 designed to mimic typical body coil loading, where the RF power was 3.2 for the MB protocol 2 0 1 and 5.0 for the SB, with the ratio being 3.2/5.0 = 0.64. Given that the heating scales with the RF 2 0 2 power deposition, it is expected that heating during the MB protocol will be lower than that of 2 0 3 the SB protocol by a factor of 0.64. It is important to note that the square of the ratio of is only predictive of relative heating for comparisons of the same sample. Samples that load the 2 0 5 coil differently due to differences in volume or conductivity could require more RF power and experimental conditions, such as hardware and body types.  On' markers to verify synchronization. The RTBox was directly connected to the scanner and 2 2 0 placed scanner pulse markers in the EEG file at the time of delivery of every RF pulse. The scalp 2 2 1 electrodes had 10 kOhm built-in resistors (5 at amplifier + 5 at tip) and were recorded with a 0.5 2 2 2 µV resolution. The drop-down electrodes had 20 kOhm built-in resistors (5 at amplifier + 15 at 2 2 3 tip) and were recorded with a 10 µV resolution. All electrodes had a low cutoff filter of 10s, fMRI to ensure that the peaks of the artifact can be detected for GA/BCG artifact rejection.

8
Additionally, during all human subject runs the scanner's helium pump was turned off to 2 2 9 eliminate vibration artifacts at 42 Hz from the pump.

Setup of Heating Experiments in Phantom
During the electrode heating tests a watermelon 'phantom' was fit with the 64-channel cap. The electrode contacts [29,48]. We abraded the watermelon with sandpaper prior to placing the cap 2 3 5 on it and applying electrolytic gel to the electrodes, which allowed us to maintain impedances <5 small size of the watermelon, the ECG electrode was routed underneath the watermelon once and 2 3 8 placed in between the HEOG and IOG electrodes to ensure that no loop was created within the 2 3 9 magnet (Fig 1). The monitored electrodes were exposed to the ambient air, without cushioning  Temperature data were processed using MATLAB® 2020a. Temperature measurements were 2 6 8 smoothed using the "smoothdata" function using a Gaussian-weighted moving average filter temperature. This was apparent from the fact that during the first run, and the rest periods before smoothed temperature between the onset of the scan and the end of the scan by the scan duration. We obtained EEG data with simultaneous fMRI from nine human subjects during eyes-closed Following standard procedures, GA subtraction was performed first followed by BCG artifact trigger pulse markers obtained directly from the scanner (see methods) with a continuous artifact.

9 1
A baseline correction over the whole artifact was used with a sliding average calculation of 21 2 9 2 marker intervals. Bad intervals were corrected with the average of all channels. The data was not applied at 100 Hz. The data was then segmented to include only artifact-free resting-state data, 2 9 5 and BCG artifact rejection was performed using semi-automatic mode. After correcting all 2 9 6 marked heartbeats, artifact removal of the heartbeat template was performed using sequential 21 2 9 7 pulse templates as the template average. Hz using the Multitaper approach. We chose not to do any further processing of the data (i.e., To assure the safety of our MB sequence, we aimed at demonstrating that heating remained 3 0 9 below that of the recommended SB sequence previously established to be safe [38]. In all  shown in Table 1. The relative heating of the two sequences therefore is in approximate 3 1 8 agreement with the RF power deposition ratio of 0.64 derived from the scanner (see Methods).

1 9
This implies that the differences in heating between the two sequences are captured by the total 3 2 0 RF power deposition. paired t-test of MB sequences > SB sequences revealed significant differences, with p = 0.0054, that of the safe SB sequence, our MB sequence can be considered to be safe.    To check EEG signal quality, we assessed the spectral dominance and topography of the alpha demonstrate that each successive step substantially improves data quality, first showing the raw 3 5 5 data without GA or BCG artifact subtraction (Fig 4A), then with only the GA artifact cleaned related to the GA at the RF excitation repetition frequency (Fig 4D). The spectrogram is shown for the (A) raw EEG data without artifact removal, (B) the EEG data   was detectable in all individual subjects (Fig 5B), while other power peaks in lower frequencies 3 7 9

Fig 3. Superposition of the individual temperature measurements in the watermelon
(particularly vulnerable to BCG artifacts) or at 15.9 Hz (the RF excitation repetition frequency) were greatly attenuated. The large spikes shown in Subjects 1 and 5 near 47 Hz were due to the 3 8 1 scanner bore fan being on during the scan for subject comfort. GA causes stronger contamination in the high EEG frequencies (>~20Hz, cf. Fig 4), the brain 3 8 9 signal that can be recovered from this frequency range through GA removal is lower than in 3 9 0 slower EEG frequencies [54]. Note that the loss of high-frequency EEG signals observed for the 3 9 1 MB sequence were closely aligned with those observed in the SB sequence (Fig 6).

9 2
We conclude that EEG is of sufficient quality for cognitive neuroscience research using the MB 3 9 3 sequence after application of artifact rejection methods originally developed for SB sequences, Simultaneously recorded EEG-fMRI is a powerful tool that can provide information beyond an interleaved strategy designed to minimize the effect of temperature drift, we empirically the shape of the spatial distribution of heating will be the same for both sequences, and the 4 4 5 overall magnitude of the heating distribution will increase linearly with the RF power deposition.

6
This prediction is consistent with the expectation that the spatial distribution is largely In absolute terms, the heating rates we measured for all conditions were greater than 0.01 degrees C/min (see Table 1). These rates are higher than those observed in similar experiments 4 5 0 [19]. These differences may be due to slow drifts in the ambient temperature, like those observed 4 5 1 in some of our data during some of the rest periods (see Fig 2). The presence of drift is a serious factor of 4, a low TR of 440 ms, and 28 slices for whole-cerebrum coverage. This was achieved 4 5 8 by using a relatively low flip angle of 40°, and a moderately long pulse duration of 5300 µs.

5 9
Although temperature measurements were not performed on humans during scanning, our MB MB-fMRI (cf. Fig. 6 for EEG data) did not report any heating sensation in any of the conditions. These observations further support the safety demonstrated by the phantom heating experiment.

6 5
In addition to verifying heating safety, in our second experiment we conducted preliminary multiband EPI imaging for high-resolution, whole-brain, task-based fMRI studies at 3T: and EEG data quality of concurrent high-density EEG and high-speed fMRI at 3 Tesla.