The Temporal Muscle of the Head Can Cause Artifacts in Optical Imaging Studies with Functional Near-Infrared Spectroscopy

Abstract

Background: Extracranial signals are the main source of noise in functional near-infrared spectroscopy (fNIRS) as light is penetrating the cortex but also skin and muscles of the head. Aim: Here we performed three experiments to investigate the contamination of fNIRS measurements by temporal muscle activity. Material and methods: For experiment 1, we provoked temporal muscle activity by instructing 31 healthy subjects to clench their teeth three times. We measured fNIRS signals over left temporal and frontal channels with an interoptode distance of 3 cm, in one short optode distance (SOD) channel (1 cm) and electromyography (EMG) over the edge of the temporal muscle. In experiment 2, we screened resting state fNIRS-fMRI (functional magnetic resonance imaging) data of one healthy subject for temporal muscle artifacts. In experiment 3, we screened a dataset of sound-evoked activity (n = 33) using bi-temporal probe-sets and systematically contrasted subjects presenting vs. not presenting artifacts and blocks/events contaminated or not contaminated with artifacts. Results: In experiment 1, we could demonstrate a hemodynamic-response-like increase in oxygenated (O₂Hb) and decrease in deoxygenated (HHb) hemoglobin with a large amplitude and large spatial extent highly exceeding normal cortical activity. Correlations between EMG, SOD, and fNIRS artifact activity showed only limited evidence for associations on a group level with rather clear associations in a sub-group of subjects. The fNIRS-fMRI experiment showed that during the temporal muscle artifact, fNIRS is completely saturated by muscle oxygenation. Experiment 3 showed hints for contamination of sound-evoked oxygenation by the temporal muscle artifact. This was of low relevance in analyzing the whole sample. Discussion: Temporal muscle activity e.g., by clenching the teeth induces a large hemodynamic-like artifact in fNIRS measurements which should be avoided by specific subject instructions. Data should be screened for this artifact might be corrected by exclusion of contaminated blocks/events. The usefulness of established artifact correction methods should be evaluated in future studies. Conclusion: Temporal muscle activity, e.g., by clenching the teeth is one major source of noise in fNIRS measurements.

Keywords: NIRS; artifact; clenching teeth; fNIRS; noise; optical topography; temporal muscle.

Figure 1

Measurement setup and findings of experiment 1. (A) Probe-set arrangement over the left frontotemporal area (red dots = light emitters, blue dots = detectors, numbers = channels), EMG electrodes in green, and short optode distance NIRS channel in pink. (B) Trajectories of O₂Hb (red lines) and HHb (blue lines) in temporal and frontal channels of the probe-set showing significant changes. The dark gray box indicates a phase of an initial parallel dip of both chromophores induced by probe-set movement. The light gray box indicates the trajectory of the temporal muscle artifact which was used for further analyses. (C) Topographies of significant channels for O₂Hb and HHb for all head sizes and for O₂Hb for small and large head sizes. Scatterplot and correlation coefficient for the correlation between the number of artifact channels (spatial extent of the artifact) with head size. Please note that three dots are overlapping.

Figure 2

Measurement setup and findings for experiment 2. (A) Temporal muscle extraction of one healthy subject and three fNIRS channels covering the edge of the temporal muscle. (B) Resting state measurement of the healthy subject. Mean trajectories of O₂Hb and HHb of the three fNIRS channels, of muscle and gray matter BOLD of extra- and intracranial voxels in the area of the temporal muscle. The gray-green color bar indicates the relationship of the explained variance of the correlation between O₂Hb and muscle and between O₂Hb and gray matter BOLD signal. Based on visual inspection and correlation analyses, two time windows were identified showing the temporal muscle artifact. (C) For the two time windows of the artifact, O₂Hb signal was correlated with all voxels showing correlations in extra-cranial layers mirroring the temporal muscle on both head sides (lower on the right side, indicated by blue color).

Figure 3

Temporal muscle artifacts for experiment 3. (A) Exemplary fNIRS data with no, one and four artifacts resulting in the exclusion of none, two and seven blocks, respectively, from data analysis. (B) Contrasts of the standard deviation of events/blocks with and the standard deviation of events/blocks without artifacts in a channel-wise manner in the group of subjects with artifacts showing the typical topography of the temporal muscle artifact.

Figure 4

Sound-evoked activity of experiment 3. Trajectories of O₂Hb (red lines) and HHb (blue lines) for the block and event-related design are shown on the top. Topographies of model-based analyses of block and event-related design are shown on the bottom. Please note, that for block design two model-based approaches were analyzed. Findings show short-lasting increases in areas over the auditory cortex and Broca's area for the block and event-related design and stable increases over the 20 s of auditory stimulation in inferior frontal areas for the block design.

Figure 5

Sound-evoked activity in anatomically defined regions of interests (ROIs) of experiment 3. Anatomically defined ROIs were the auditory cortex and Broca's area as shown on the right bottom. Sound-evoked activity (bars) in these areas was defined by t-tests against 0. Contrasts between groups with and without artifacts and between conditions with and without exclusion of artifacts are shown by effect sizes over the bars.