Measuring the cortical tracking of speech with optically-pumped magnetometers

• Published in: NeuroImage (Orlando, Fla.) Fla.), 2021-06, Vol.233, p.117969-117969, Article 117969
• Main Authors: de Lange, Paul, Boto, Elena, Holmes, Niall, Hill, Ryan M., Bowtell, Richard, Wens, Vincent, De Tiège, Xavier, Brookes, Matthew J., Bourguignon, Mathieu
• Format: Article
• Language: English
• Subjects: Brain; Cortical tracking of speech; Developmental disabilities; Electroencephalography; Handedness; Life span; MEG; Neural networks; Neurodevelopmental disorders; Optically pumped magnetometer; Principal components analysis; Scalp; Sensors; Spatial discrimination; Speech
• ISSN: 1053-8119; 1095-9572
• DOI: 10.1016/j.neuroimage.2021.117969

•The cortical tracking of speech (CTS) was assessed with wearable MEG.•Optically-pumped magnetometers (OPMs) were used to record brain activity.•Significant CTS was uncovered at phrasal, word and syllabic rates.•Levels of CTS were similar to those previously reported using classical MEG sensors.•This suggests that OPM-MEG is suitable to measure brain activity below 4 Hz. During continuous speech listening, brain activity tracks speech rhythmicity at frequencies matching with the repetition rate of phrases (0.2–1.5 Hz), words (2–4 Hz) and syllables (4–8 Hz). Here, we evaluated the applicability of wearable MEG based on optically-pumped magnetometers (OPMs) to measure such cortical tracking of speech (CTS). Measuring CTS with OPMs is a priori challenging given the complications associated with OPM measurements at frequencies below 4 Hz, due to increased intrinsic interference and head movement artifacts. Still, this represents an important development as OPM-MEG provides lifespan compliance and substantially improved spatial resolution compared with classical MEG. In this study, four healthy right-handed adults listened to continuous speech for 9 min. The radial component of the magnetic field was recorded simultaneously with 45–46 OPMs evenly covering the scalp surface and fixed to an additively manufactured helmet which fitted all 4 participants. We estimated CTS with reconstruction accuracy and coherence, and determined the number of dominant principal components (PCs) to remove from the data (as a preprocessing step) for optimal estimation. We also identified the dominant source of CTS using a minimum norm estimate. CTS estimated with reconstruction accuracy and coherence was significant in all 4 participants at phrasal and word rates, and in 3 participants (reconstruction accuracy) or 2 (coherence) at syllabic rate. Overall, close-to-optimal CTS estimation was obtained when the 3 (reconstruction accuracy) or 10 (coherence) first PCs were removed from the data. Importantly, values of reconstruction accuracy (~0.4 for 0.2–1.5-Hz CTS and ~0.1 for 2–8-Hz CTS) were remarkably close to those previously reported in classical MEG studies. Finally, source reconstruction localized the main sources of CTS to bilateral auditory cortices. In conclusion, t his study demonstrates that OPMs can be used for the purpose of CTS assessment. This finding opens new research avenues to unravel the neural network involved in CTS across the lifespan and potential alterations in