Ten healthy volunteers were recruited from the Ohio State University campus. Participants were required to be 18–35 years of age, native English-speakers, report normal or corrected-to-normal vision (including visual acuity, binocular depth perception, and color vision), report normal audition, no history of neurological or psychiatric diseases, and have a primary care practitioner (required by the Center for Cognitive and Behavioral Brain Imaging center). Further, participants were required to have no contraindications for MRI or TMS, no pregnancy, and not be taking psychoactive substances or medications that induce light-sensitivity. Participants were informed of any risks and provided informed consent prior to participation. Monetary compensation was provided for voluntary participation in the study. The protocol was approved by the Institutional Review Board (protocol number: IRB- 2022H0044). Two participants failed to complete all study sessions, resulting in a final sample of eight healthy participants (two female; M = 23.23 years, SD = 2.96 years).

Each participant underwent four neuromodulation visits, wherein they received one of the following treatments: tPBM, sham tPBM, TMS, or sham TMS, with order counterbalanced across subject (Table 1). We included sham stimulation to control for the effects of experience and task-learning on collected brain measures. Days between adjacent lab visits varied from two to 64 days (M = 13 days; SD = 15 days). For both tPBM and sham tPBM, participants were comfortably seated facing away from the tPBM system and were instructed to keep still during stimulation. Both the participants and experimenters wore dark safety goggles during both sham and active sessions. The tPBM system (shown in Figure 1) consisted of a 1,064 nm (Zhao et al., 2022) continuous wave laser (EVO FX Tri-wavelength Class-4 Therapy Laser). The circular laser beam was positioned over the vertex of the cranium according to the 10–20 international system. Importantly, the vertex was used as the point of stimulation in order to balance the headset, as participants found it uncomfortable to wear if the laser were positioned directly over the somatomotor region; we felt comfortable doing this because previous work has demonstrated that tPBM stimulation has broad focality, making precision of application less essential for tPBM than for TMS (Wang et al., 2019). Furthermore, prior work from our group simulating the optical scattering of the tPBM laser using Monte Carlo eXtreme from MATLAB (Fang and Yan, 2019) confirmed that the bihemispheric somatomotor regions, including the right hand area of M1, fell within the range of stimulation when the laser was applied at the vertex. The laser was held in place by a modified OpenBCI 3D printed headset with a strap under the participant’s chin. Active tPBM sessions consisted of laser output of 0.6 W for 11 min for a total dosage of 396 J. Sham tPBM sessions consisted of 0.1 W for 11 min, and black electrical tape placed over the beam to assist in blocking the laser.

Transcranial magnetic stimulation was delivered using a MagPro X100 stimulator with a Cool B6 Coil. Participants first underwent the motor cortex mapping and motor threshold determination. Single pulses of biphasic waves, starting at 30% of maximum stimulator output, were provided to a point corresponding to C3 on the participant’s head, 45° to the midsagittal line. Stimulation was titrated until 50% of trials elicited a visible contraction of the first dorsal interosseus muscles of the hand, and the pulse amplitude and location of the coil were recorded for each participant (Badran et al., 2019). The coil contained an active stimulation face, where the electromagnetic field could pass freely to the subject’s cortex, and an opposing, shielded face, with an insert that blocked the electromagnetic field during sham sessions. The faces were not marked for the subject, so the subject remained blinded to the condition. During TMS sessions, biphasic stimulation was provided to this spot using the excitatory intermittent theta burst protocol (Oberman et al., 2011) at 80% of the participant’s resting motor threshold for 600 pulses. Sham TMS consisted of the same protocol, but with the shielded face of the coil applied to the participant’s head.

Magnetic resonance imaging scans were conducted twice per visit, both pre- and post-stimulation or sham. Each MRI session lasted approximately 1 h, including all tasks and time getting the participant comfortably in and out of the scanner. Participants were scanned using Siemens 3T Magnetom Prisma with a 32-channel head coil. First, a high resolution (TE/TR = 2.9/2,300 ms, voxel size = 1.0 × 1.0 × 1.0 mm³) T1 weighted 3D magnetization-prepared rapid acquisition with gradient echo (MPRAGE) was collected. Next, 10 min of DWI data (to compute structural connectivity) were acquired using echo planar imaging (EPI; 78 slices, voxel size 2.0 × 2.0 × 20 mm³, 256 mm field of view, bandwidth of 1,698 Hz/Px, TR = 3,400 ms) with a multiband factor of two with b-values of 2,000 s/mm², and one b = 0 volume. Diffusion weighting was isotropically distributed along 96 directions. Then, 10 min of resting state fMRI data (to define functional connectivity) was collected using echo planar imaging on all participants while awake for approximately 10 min (TE/TR = 28/2,000 ms, voxel size = 2.0 × 2.0 × 2.0 mm³).

Task fMRI scans (EPI sequence, TR = 1,000 ms, TE = 28 ms, voxel size = 2.0 × 2.0 × 3.0 mm³) consisted of three 3.72 min runs of a block-design finger-tapping task per MRI session (Khoury et al., 2019). Each run of the motor task was comprised of four 18 s blocks of rest and four 18 s blocks of motor action, for a total of eight blocks per run. Rest was cued by presentation of a fixation cross, wherein participants were instructed to remain still and move as little as possible. Motor action was cued by a line drawing of a hand, whereafter participants were instructed to tap either the second or third digit of their right hand on a button box (following the finger indicated on the picture) at a frequency of 1 Hz. The output from the button box was monitored to ensure that the participant remained awake and complied with the task directions. A total of six finger-tapping task runs were collected per visit—three runs pre-stimulation, and three runs post-stimulation.

All structural (MPRAGE) images were preprocessed using the recon-all pipeline in FreeSurfer (FreeSurfer Wiki, 2024c) with default parameters. This pipeline includes intensity correction, skull stripping, surface co-registration, spatial smoothing, segmentation of white matter and subcortical areas, and parcellation of the cortex. All outputs were visually inspected to ensure successful reconstruction. The reconstructed images were used for multimodal registration within subject and across sessions, as well as to the group average template.

First, we examined whether tPBM would increase BOLD response during a finger-tapping task (Khoury et al., 2019) and how any observed change compared to that induced by TMS. Task data preprocessing was conducted in FreeSurfer using FS-Fast and included (a) motion correction, where timepoints with greater than 1 mm total vector moment between TRs were identified and regressed out as framewise nuisance regressors, (b) distortion correction, and (c) registration to anatomical space (using FreeSurfer’s bbregister) (Greve and Fischl, 2009). For each subject, the neural activation of motor activity (β estimates and contrast maps) were computed in FreeSurfer (using FS-Fast) using a first-level general linear model (GLM) analysis. Data were detrended and smoothed (FWHM = 3 mm kernel) with a regressor entered for the contrast of interest (finger tapping) as well as six motion parameters as nuisance regressors. A block design with a standard boxcar function (events on/off) convolved with a canonical hemodynamic response function was used. The resulting contrast maps for finger tapping over baseline were used for the probabilistic maps of activation. Individual contrast maps were registered from the subject’s native space (FreeSurfer Wiki, 2024a), and then projected onto the surface of each hemisphere and then to FreeSurfer’s FSAverage space. Contrast maps were thresholded and binarized for each subject to show any voxels where the effect of stimulation was 1.3x the effect of sham stimulation, and then concatenated across subjects to create a probabilistic map of activation, with the value of voxel corresponding to the number of subjects with significant motor activation.

We further explored the effect of stimulation within each individual’s somatomotor cortex. Subject-specific functional regions of interest (fROIs) were defined in MATLAB using the Group-Constrained Subject-Specific method (Fedorenko et al., 2010). Specifically, we extracted the top 10% of activated voxels (across pre- and post-stimulation runs) within the somatomotor cortex (precentral and postcentral gyri) for the automated Desikan/Killiany segmentation (Desikan et al., 2006) in individual subject space. The percent signal change (PSC) in identified fROIs was then computed in MATLAB for the three pre-stimulation motor task runs and three post-stimulation runs using the absolute value of first-level β estimates for the motor blocks divided by the baseline (intercept) for post- and pre- runs for each session (tPBM stimulation, tPBM sham, TMS, TMS sham). The effect of stimulation on motor activation was defined as the absolute change in PSC from pre- to post-stimulation minus the change in PSC from pre- to post-sham stimulation:

One-sample t-tests were conducted in MATLAB to assess whether normalized PSC was significantly different from zero for both tPBM and TMS. Paired t-tests were then conducted to assess whether PSC due to stimulation differed within subjects between tPBM and TMS.

Structural connectivity describes how physically connected two brain regions are by neuronal white matter tracts (Warbrick et al., 2017). Similarly to functional connectivity, alterations in white-matter have been observed in both MDD and PD (Droby et al., 2022; Liberg and Rahm, 2015); thus, we examined the absolute and relative effects of tPBM and TMS on structural connectivity of the somatomotor cortex to connected nodes. Fractional anisotropy (FA) is the measurement of the directionality of water diffusion along axons and provides an approximation of structural integrity of white matter tracts connecting two regions (Elton et al., 2023). FA image preprocessing was conducted using FDT (FMRIB’s Diffusion Toolbox), which is a part of FSL (FMRIB’s Software Library)¹ (Behrens et al., 2003; Behrens et al., 2007). Preprocessing included an eddy current correction (eddycorrect from FDT) and brain extraction (BET from FDT). The images were then eroded and the end slices coded to zero in order to remove potential outliers. The images were then non-linearly registered to a standard template (Jenkinson et al., 2012) and thresholded at 0.2. Participant images were then concatenated into a single file for group analysis. A repeated measures ANOVA was used to calculate group statistics using threshold free cluster enhancement (TFCE) with a standard space white matter skeleton used to mask FA images.

To perform individual-subject region-to-region connectivity analyses (i.e., quantitative tractography), automated cortical parcellations were registered to each participant’s diffusion images. Image preprocessing was again conducted using FDT and included an eddy current correction (eddycorrect from FDT) and brain extraction (BET from FDT). Motor regions were used as seed regions for the analysis, while the thalamus and striatal regions were chosen as target regions. Principal diffusion directions were calculated for each voxel and probabilistic tractography was carried out using FDT with 10,000 streamline samples in each seed voxel to create a connectivity distribution to each of the target regions, while avoiding the ventricles as defined by a mask. Thus, every voxel within the motor seed region was described by a vector of connection probabilities to the target regions; we averaged these connection probabilities across voxels of the motor cortex per subject and then calculated the effect of stimulation on structural connectivity as above for functional connectivity. Inset of Figure 3 shows an example participant’s FDT_paths file, thresholded at 10% of maximum streamlines, as visualization of the tracts connecting motor to subcortex. While we had an a priori αlevel of 0.05, we found no differences in FA from pre- to post-tPBM stimulation at this level. Therefore, Figure 3 shows a cluster-corrected group map instead thresholded at p < 0.07.

Functional connectivity describes the extent to which two brain regions coactivate and is thought to reflect both direct and indirect connections between related brain regions (Bullmore and Sporns, 2009; Logothetis, 2003). We examined how tPBM compares to TMS in modulating the brain’s functional connectivity at rest (i.e., resting state functional connectivity; rsFC), as disturbances in typical rsFC networks have been seen in both MDD and PD (de Kwaasteniet et al., 2013; Petersen et al., 2018; Wang et al., 2017). RsFC data preprocessing was conducted using FreeSurfer’s FS-Fast (FreeSurfer Wiki, 2024b) and included motion correction, smoothing gray matter with a 3 mm FWHM kernel, linear interpolation for spikes greater than 0.5 mm, and bandpass filtering (0.0009–0.08 Hz). The functional connectivity changes were then explored on an individual subject-level within the following regions of interest (ROIs), defined from the Desikan-Killiany automated segmentation (Desikan et al., 2006) in subject space: motor cortex, thalamus, and striatum (caudate and putamen). Correlation analyses were run for signal extracted from the motor cortex to those of the thalamus and striatum. Time courses were averaged within each region of interest, then the Pearson product correlations between the time series were computed. The r values were then Fisher z-transformed and the effect of stimulation on connectivity was defined using the Fisher correlations as the difference from pre- to post-stimulation over difference from pre- to post-sham stimulation.