Thirty-seven consecutive PSA inpatients were screened from October 2020 to May 2021, in the rehabilitation department of the Jing’an Branch of Huashan Hospital of Fudan University, a tertiary care teaching hospital in Shanghai. Nineteen patients were enrolled according to the following inclusion criteria: (1) age more than 18 years old; (2) native Chinese speakers; (3) clinically diagnosed as ischemic or hemorrhagic stroke confirmed with brain images; (4) be able to follow instructions in order to cooperate with the assessments and study treatments; (5) diagnosed as aphasia according to clinical assessments and the Chinese version of the western aphasia battery (WAB) (aphasia quotient below 93.8). The exclusion criteria were: (1) a previous stroke history; (2) severe and uncontrolled psychiatric disorders; (3) uncontrolled epilepsy, (4) other neurological disorders that might interfere with the post-stroke functional impairments; (5) brain tumor(s) or brain injury history; (6) unstable vital signs, or severe organ failure; (7) any metal implant or device in the body, or any other condition that would restrict the application of iTBS or MRI scanning. The age and time from onset of stroke were comparable between the recruited participants and those patients who did not meet the above inclusion and exclusion criteria, while the mean aphasia quotient (AQ) of WAB and educational level were higher in the included patients compared with those that were not included. Three participants’ brain MRI data were excluded from the analyses because of excess head motion; thus, sixteen participants were included in the final analyses. All the recruited participants were right-handed native Chinese speakers.

The flowchart of the participants and the analyses is shown in Fig. 1. This study adopted a within group longitudinal design, without a control group. After the recruitment, each participant underwent baseline behavioral assessments, as well as the single TMS tests for identifying the resting motor threshold (RMT) and the hot spot of the hand area in the left M1. Then, the fMRI scanning was applied for each participant, before and immediately after the first session of iTBS, respectively. Each participant also underwent structural MRI scanning before the first iTBS session. Then the participants continued to undergo daily iTBS until being discharged from the hospital, or until twenty iTBS sessions were completed. Ten of the nineteen participants completed twenty sessions of iTBS treatments during their hospital stay and underwent another structural MRI and fMRI scanning and behavioral assessments after the twentieth session of iTBS. After excluding the brain images not eligible for analyses (due to artifacts or excessive head motion), the MRI data of sixteen participants were included for one-session iTBS effect analyses (step 1), and the data from six participants were included for multiple-session iTBS effect analyses (step 2). During the study, all the participants also underwent the conventional rehabilitation treatments in addition to the iTBS. Our previous study has reported other analyses based on the one-session data of the same group of participants [55].

Fig. 1.

Participants and analyses flowchart. Abbreviations: PSA, post-stroke aphasia; TMS, transcranial magnetic stimulation; (f) MRI, (functional) magnetic resonance imaging; iTBS, intermittent theta burst stimulation; FC, functional connectivity.

This study was approved by the ethical committee of Huashan Hospital, Fudan University. Written informed consents were acquired from all the participants or their legal representative(s). The trial was registered on the Chinese Clinical Trial Registry website (http://www.chictr.org.cn/index.aspx), and the registration number is ChiCTR2100041936.

Standardized scales were used to assess the language, cognitive and motor performances, as well as activities of daily living of each participant. The Chinese version of WAB was used for language performance assessment, and an aphasia quotient (AQ) was calculated to measure the severity of aphasia for each participant. The non-language-based cognitive assessment (NLCA) was used to measure the potential cognitive impairments of the participants, and the Fugl–Meyer motor assessment–upper extremity (FMA-UE) was used to assess the severity of the motor deficits of the participants. The Barthel Index (BI) was used to measure the participant’s activities of daily living. The details of these assessments are described in our previous study [55].

We used the MagPro X100 magnetic stimulator (Medtronic Co, Copenhagen, Denmark) with a figure-eight coil (MC-B70) to apply the iTBS intervention. Before the first session of iTBS, the RMT was measured for each participant as follows: first, surface electrodes were attached to the abductor pollicis brevis muscle of the participant’s right hand (i.e., the hemiplegic hand) [33, 55]; then the location of the hot spot of the hand area was identified by recording the magnitudes of motor evoked potentials (MEP) when stimulating the left M1 by single pulse TMS. Next, the RMT of the left M1 was identified by stimulating the hot spot with stepwise increasing intensities of single pulse TMS, until the minimum intensity that could induce at least five MEPs of no less than 50 $\mu$v out of ten consecutive stimulations was reached [55]. The hot spot in the left M1 was used as the stimulation site of the iTBS paradigm. When no MEP could be induced from the left M1, the same procedures were applied to the right M1 to identify an alternative RMT and hot spot; and the mirrored location of the hot spot on the left hemisphere was identified with a location cap provided by the Yiruide Co. (Wuhan, China), which was used as the simulation site. The stimulation site of each participant was marked on a cap, which was used in the daily iTBS session to locate the coil. The intensity of the iTBS was set as 80% of the RMT for each participant, and the iTBS parameters were: 3-pulse bursts delivered at 50 Hz and repeated at 5 Hz, given in consecutive short trains of 10 bursts (lasting for 2 s) every 10 s for 20 cycles. This resulted in a total of 600 pulses per session [7, 8], lasting for about three minutes.

All the participants underwent a multimodal MR scanning prior to the first iTBS session, including the T1-weighted, T2-weighted, rs-fMRI and diffusion tensor image scans. The rs-fMRI scanning was again applied immediately after the first iTBS session for each participant. The participants who completed twenty sessions of iTBS underwent another multimodal MRI scanning after the twentieth iTBS session. A 3.0T GE MR750 scanner was used for MRI scanning at the Jingan Branch of Huashan Hospital, Fudan University. These analyses were based on the rs-fMRI data, and the T1-weighted images were used for lesion identification and rs-fMRI normalization.

The T1-weighted images were acquired using 3D FSPGR sequence with the following parameters: matrix size = 260 $\times$ 224, FOV = 200 $\times$ 200 mm${}^2$, layer thickness = 1 mm, voxel size = 1 $\times$ 1 $\times$ 1 mm${}^3$, repetition time (TR) = 7800 ms, echo time (TE) = 5 ms, tilt angle = 12°, number of layers = 248.

The rs-fMRI images were acquired in the horizontal plane by EPI echo-planar imaging with the following parameters: TR = 2000 ms, TE = 30 ms, inclination = 90°, matrix size = 96 $\times$ 96, and voxel size = 3 $\times$ 3 $\times$ 3 mm${}^3$. A total of 240 volumes were acquired during each rs-fMRI scanning procedure, lasting for 480 s. Each participant was instructed to keep still with eyes closed, relaxed, and was required to do no systematic thinking during the scanning.

Each participant’s lesion was manually segmented using ITK-SNAP tools [62] on the individual T1-weighted structural images, and then registered to the Montreal Neurological Institution (MNI) space using spm12 (https://www.fil.ion.ucl.ac.uk/spm/). The group-level voxel-wise map of the lesions were created by summing up all the individual lesions, which are shown in Fig. 2.

Fig. 2.

Lesions’ location of the participants. The overlapped lesion map of the sixteen participants (A) with post-stroke aphasia (PSA) who underwent at least one session of intermittent theta burst stimulation (iTBS), and of the six participants (B) who underwent twenty sessions of iTBS treatment. The color scales represented the number of individual lesions overlapped.

The preprocessing of the rs-fMRI data followed conventional procedures: (1) discarding the first 10 volumes of the 240 volumes; (2) slice timing for systematic time shift; (3) head motion correction using Friston 24 parameters [63, 64]; (4) image reorientation for better normalization; (5) normalizing to the MNI space using unified segmentation on T1-weighted images, and being resliced into 3 mm $\times$ 3 mm $\times$ 3 mm resolution; (6) removing the linear signal trend; (7) regression of the nuisance variables including Friston 24 parameters, the white matter, and cerebrospinal fluid signals; (8) band-pass filtering of 0.01–0.10 Hz signal. Since spatial normalization might be influenced by the lesions, an enantiomorphic approach [65] was used to replace the lesioned brain tissues with the contralateral mirrored images for each participant. These preprocessing procedures were applied using the data processing and analysis for (resting-state) brain imaging (DPABI) procedures [66].

We performed two-step FC analyses (Fig. 1). In the first step, we attempted to reveal the functional connections (i.e., edges) that exhibited significant FC strength alteration following one session of iTBS, in the brain networks of the semantic and mirror-neuron systems, respectively. The nodes connected with the most altered connections were thus identified in each of the networks, which were then defined as seeds in the second step analyses. The second step analyses were applied in the sub-group of six participants who underwent twenty sessions of iTBS. The voxel-wise seed-based FC map in the whole brain gray matter was then calculated by correlational manner, before (time 1), immediately after (time 2) the first session of iTBS, and after the twenty iTBS sessions (time 3), respectively. Then the trends in longitudinal FC patterns’ were observed for these seeds. The details of the analyses are described below.

2.8.1 Immediate FC Changes in the Semantic and Mirror Neuron Networks Following one Session of Left M1 iTBS (Step 1)

To investigate the immediate FC modulation effects of the left M1 iTBS in the semantic and mirror neuron system, we constructed three functional networks in the PSA brain, based on the previously reported locations of these two systems [56, 67]. A brain network comprises two types of elements, which are the key regions (i.e., node) in the system, and the functional relationship between each pair of nodes (i.e., edges) [68]. The locations of the nodes in each network (shown in Figs. 3,4) were defined as follows:

Fig. 3.

Edges showing functional connectivity (FC) strength changes following one-session left M1 iTBS in the SN+ (A) and SmN+ (B). The matrices in the left two columns illustrated the mean FC strengths of the edges in the SN+ (A) and SmN+ (B), before (time 1) and immediately after (time 2) the first session of left M1 iTBS, in the sixteen PSA participants. The two figures in the right column illustrated the locations of the nodes (spheres) in the SN+ (A) and SmN+ (B), and the edges (sticks) exhibiting significant FC strength changes after the iTBS in these two networks, respectively. The color bars indicated the mean FC strength of each edge. The color bars indicated the mean Fisher’s z scores of each edge in the PSA participants. Abbreviations: SN+, semantic network (plus the two M1 nodes); SmN+, semantic and mirrored network (plus he two M1 nodes); L, left; R, right.

Fig. 4.

Edges showing functional connectivity (FC) strength changes following one-session left M1 iTBS in the AoN+. The left two matrices illustrated the mean FC strength of the edges in the AoN+ before (time 1) and immediately after (time 2) the first session of left M1 iTBS, in the sixteen PSA participants. The figure on the right illustrated the locations of the nodes (spheres) in the AoN+, together with the edges (sticks) exhibiting marginally significant FC strength changes after the iTBS treatment. The color bar indicated the mean FC strength of each edge. The color bar indicated the mean Fisher’s z scores of each edge in the PSA participants. Abbreviations: AoN+, action observation network (plus the two M1 nodes); L, left; R, right.

• Semantic network (SN+)

Sixty nodes comprised in this network were defined according to the template reported by Xu and colleagues [69], which was based on the peaks of the activation likelihood estimate (ALE) map reported in a meta-analysis by Binder and colleagues [56]. These nodes represented the brain regions activated in the general semantic tasks [56]. Each node was built as a 6mm-radius sphere around the peak voxels according to the MNI coordinates reported by Xu and colleagues [69]. The 61st and 62nd nodes were built to represent the hand areas of the left and right M1, respectively, representing the iTBS stimulation site and its right homologous region. The MNI coordinates of these two nodes were identified according to the study by Alkadhi and colleagues [30].

• Semantic and mirrored network (SmN+)

While the semantic network identified in the healthy adults exhibited prominent left lateralization, previous studies have revealed a right hemispheric reorganization of the language network in the patients with PSA [70, 71, 72, 73]. Considering this potential right hemispheric reorganization of the language system, we included the contralateral mirrored nodes of the above semantic network into the original SN+, and after excluding eighteen nodes in the SN+ next to the midline, which involved overlapping voxels between the original and mirror nodes, a semantic and mirrored network (SmN+) comprising 86 nodes were constructed. The two nodes representing the left and right M1 areas were included in this SmN+.

• Action observation network (AoN+)

The AoN+ was constructed to represent the location of the mirror neuron system in the PSA brains. The spheric nodes were built in the same way as described above, around the MNI coordinates reported in the ALE meta-analyses conducted by Caspers and colleagues [67], as the peaks activation of the “action observation” tasks.

• Edges

After locating the above nodes, the edges were defined by a correlation manner in each of the networks, respectively. Specifically, the Pearson’s correlation coefficient was calculated between the rs-fMRI signal courses of each pair of nodes, which lasted for 460s after removing the first ten volumes in the preprocessing procedures. Then each correlation coefficient was Fisher-z transformed to get approximately normally distributed data [74], which was defined as the strength of the edge.

After constructing each individual network, we then compared the FC strength of each edge before and after the first iTBS session by paired sample t test at group level, in each of the networks, respectively. The statistical significance level was defined as individual edge p 0.005 for height, with extent-based multiple comparisons correction following network-based statistic (NBS) methods (component p 0.05) [75]. A marginal statistical significance level was defined as individual edge p 0.005, with no multiple comparison correction. The network construction and statistical tools of GRETNA [76] were used for the edge calculation and network statistical analyses.

In the subgroup of participants who underwent twenty sessions of iTBS, we assessed the potential relationship between their aphasia severities and the FC strength changes of each of the three networks, respectively. One FC metric was calculated for each network at time 1 and time 3, respectively, by averaging the FC strength of all the edges which exhibited significant FC changes following one session iTBS (as revealed in step 1), respectively. Then the spearman’s correlation analyses were used to assess the relationships between the WAB-AQ change (from time 1 to time 3) and each of the network FC metric changes (from time 1 to time 3), respectively.

The demographic and clinical information of the sixteen participants (mean $\pm$ SD: 55.56 $\pm$ 11.00 years of age, 12 males) are reported in Table 1. A subgroup of six participants (mean $\pm$ SD: 53.00 $\pm$ 12.26 years of age, 6 males) completed twenty sessions of iTBS. All the participants suffered from right hand hemiparesis as evaluated by the FMA-UE. The brain lesions were all in the left hemisphere (Fig. 2).

• SN+

Twelve edges in the SN+ were found to exhibit significantly decreased FC strength immediately following one session of left M1 iTBS, tested by paired sample t test (p 0.05, NBS corrected). The locations of the edges were illustrated in Fig. 3A and listed in Table 2. In summary, the twelve edges were among the left IFG, bilateral orbital frontal cortex, bilateral inferior, middle and superior temporal gyrus, left supramarginal gyrus and the left calcarine cortex, which involved thirteen nodes (Table 3) out of the 62 nodes in the SN+.

• SmN+

In the SmN+, the FC strengths of 24 edges were found to decrease significantly after one session of left M1 iTBS, tested by paired sample t test (p 0.05, NBS corrected). The locations of these edges were shown in Fig. 3B and Table 4. These edges involved 24 nodes out of the 86 nodes in the SmN+ (Table 5), covering the bilateral frontal, temporal, and parietal regions. More edges with altered FC strength were in the right than in the left hemisphere.

• AoN+

In the AoN+, no edge showed significant FC change after one session of left M1 iTBS. Two edges were revealed to exhibit marginally significant decreased FC strength after the iTBS session, tested by paired two sample t test (individual edge p 0.005) (Fig. 4, Table 6). Both two edges involved one node in the right inferior parietal lobule, and the other node in either of the connections were located close to each other, in the left and right SMA, respectively.

Based on the above results, we identified the nodes connected with the most edges showing significant FC strength changes following one session of iTBS, in each of the brain networks. Specifically, the nodes connected with at least four altered edges in the SN+ and SmN+ (seeds 1–6), or the nodes connected with either of the two edges reaching marginal significant change level in the AoN+ (seeds 7–9), were defined as seeds for the whole brain FC patten analyses. The anatomical locations of the seeds were as follows: seed 1, left medial orbital frontal gyrus (MNI coordinates: –6, –48, –6); seed 2, left inferior orbital frontal gyrus (MNI coordinates: –39, 24, –18); seed 3, right angular gyrus (MNI coordinates: 45, –69, 27); seed 4, right middle occipital gyrus (MNI coordinates: 45, –69, 27); seed 5, right fusiform gyrus (MNI coordinates: 30, –36, –15); seed 6, right inferior orbital frontal gyrus (MNI coordinates: 39, 42, –15); seed 7, right inferior parietal lobule (MNI coordinates: 30, –54, 48); seed 8, left supplementary motor area (MNI coordinates: –2, 18, 50); seed 9, right supplementary motor area (MNI coordinates: 4, 12, 58). The whole brain voxel-wise FC patterns of each seed identified by one sample t test (p 0.05, FWE corrected) at each of the three time points were shown in Fig. 5 (seed 1–6) and Fig. 6 (seed 7–9). By inspection of the FC maps, it was observed that although all the altered edges showing decreased FC strength from time 1 to time 2 in the step 1 analyses, the range of its FC in the whole brain underwent more diverse change trends.

Fig. 5.

Seed-based functional connectivity (FC) analyses in the multi-session left M1 iTBS sub-group (seeds from the SN+ and SmN+). The left column illustrated the location of each seed (left column). The right three columns showed the whole-brain voxel-wise FC maps of each seed at before (time 1), and immediately after (time 2) the first session of iTBS, and after twenty sessions of iTBS treatment (time 3), from left to right. The color bar represented the t value of the one-sample t test used for identifying the FC maps. R, right; L, left.

Fig. 6.

Seed-based functional connectivity (FC) analyses in the multi-session left M1 iTBS sub-group (seeds from the AoN+). The left column illustrated the location of each seed (left column). The right three columns showed the whole-brain voxel-wise FC maps of each seed at before (time 1), and immediately after (time 2) the first session of iTBS, and after twenty sessions of iTBS treatment (time 3), from left to right. The color bar represented the t value of the one-sample t test used for identifying the FC maps. R, right; L, left.

No significant correlation was revealed between the WAB-AQ changes (from time 1 to time 3) and the FC metric changes of the three networks, respectively, as tested by Spearman’s correlation analyses (p 0.05).