Transcranial Magnetic Stimulation Induces White Matter Microstructure Alterations in Patients with Mild Cognitive Impairment Open Access

The rapid aging of society causes a sharp increase in the prevalence of Alzheimer’s disease (AD), which is a progressive neurodegenerative disorder [1]. Repeated failures of pharmaceutical trials seem to suggest that intervention after the onset of symptoms may be unable to resist decades of pathological accumulation [2, 3]. On the pathological continuum of AD development, mild cognitive impairment (MCI) is defined as a transition phase when objective cognitive tests can earliest detect cognitive decline but do not meet the diagnostic criteria of dementia [4]. Patients with MCI are at high risk of developing AD, with an annual conversion rate of 5–22% widely reported [5]. But MCI has the opportunity to reverse the course of the disease and return to normal cognitive older adults, accordingly, MCI may provide an appropriate window for delaying the progression of AD and reducing the conversion rate.

In recent years, noninvasive, safe, and effective rehabilitation techniques, including repetitive transcranial magnetic stimulation (rTMS), have provided new access for neurological and psychiatric disorders. rTMS transmits pulses to subcortical and distant brain regions by stimulating a special target of the scalp, generating neurophysiological and behavioral effects [6]. TMS is considered safe when applied according to the recommended safety and application guidelines endorsed by the International Federation of Clinical Neurophysiology [7]. An increased number of trials have found cognitive benefits of rTMS in patients with MCI [8, 9]. Currently, there is no consensus on the optimal site for rTMS in the treatment of cognitive impairment. But most of them chose the left dorsolateral prefrontal cortex (DLPFC) as the stimulation target, for it is a key brain region involved in the cognitive control network and frontoparietal network [8, 10]. However, the mechanism of rTMS-induced cognitive improvement is poorly known, which is at the center of the current debate. A meta-analysis [11] on rTMS stimulation parameters showed that 10 Hz rTMS has a greater effect size than 20 Hz in improving long-term memory. Additionally, research [12] on different frequencies (5 Hz, 10 Hz, 15 Hz) of rTMS applied to MCI patients showed that, compared to the other two frequencies (5 Hz, 15 Hz), 10 Hz rTMS has a more significant therapeutic effect on MCI patients. Therefore, we chose to apply 10 Hz rTMS to the DLPFC of MCI patients to observe its effects on their cognitive function and the microstructure of white matter (WM) in the brain.

The combination of neuroimaging technology and rTMS provides a powerful tool for observing the changes in brain structure and function. The development of AD is accompanied by continuous degeneration, damage and loss of neural structures, and changes in the function of the brain [13]. At present, several functional magnetic resonance imaging (fMRI) studies have reported the effects of rTMS on spontaneous neural activity and functional connectivity in MCI patients [14]. Although AD is currently primarily considered a disorder of cognitive decline caused by abnormal gray matter functional connectivity and abnormal gray matter morphology, conventional structural MRI mainly focuses on the gray matter neuronal loss in AD. Besides, previous study [15] suggests that progressive WM degeneration and demyelination are significant pathological features of AD. Another study [16] found that patients exhibit impaired WM structural integrity even in the preclinical stage of AD – the subjective cognitive decline phase. Many neuroimaging studies have provided evidence for extensive WM lesions in the MCI stage, which may occur independently and earlier than gray matter [17‒19], and decreased WM integrity was associated with a higher conversion rate of MCI to AD [20, 21]. These findings suggest that the degeneration of WM fiber tracts may be one of the causes of cognitive decline. At present, few studies have investigated whether rTMS can cause neural structural changes, especially of the WM. Therefore, this study aimed to explore the effects of rTMS on the WM structure in patients with MCI.

Diffusion tensor imaging (DTI) technology indirectly reflects the integrity of WM fiber bundles through the dispersion of revealed water molecules in nervous tissues, which can provide important information for subtle neurological structural damage [13]. This approach has various indicators, among which fractional anisotropy (FA) is one of the most sensitive, describing the degree of anisotropy of water molecular dispersion. A higher FA value indicates a more intact structure of the WM fiber tracts. A trial observed increased FA value of the bilateral frontal lobe, the callosum, and the bilateral hippocampus brain regions in schizophrenia patients compared with the control group after 4 weeks of 10 Hz rTMS stimulation treatment, which showed the active influence of rTMS on WM [22]. Another trial has shown the network measures of the high executive ability group demonstrated greater WM integrity [23]. Few studies have explored the influence of rTMS on the WM structure of MCI patients through FA and its relationship with cognitive improvement. To our knowledge, only one study has reported the effects of TMS treatment on FA values in patients with MCI and AD. This study observed that 4 weeks of TMS treatment could increase the FA values of right anterior thalamic radiation among amnestic mild cognitive impairment (aMCI) patients [24]. However, this article does not report the relationship between FA changes and improvements in cognitive function in MCI and AD. Additionally, the study did not conduct a comprehensive cognitive assessment in patients with MCI and AD, involving memory, language, and executive function. We plan to address the above issues through a more comprehensive study.

Therefore, we propose the following hypothesis: cognitive dysfunction in MCI may be related to WM structural damage, and rTMS might improve cognitive function in MCI patients by enhancing the integrity of WM microstructure. To investigate this question, this study employs a double-blind, randomized, controlled experiment to compare changes in WM fiber structure between MCI patients treated with rTMS (real group) and those not treated with rTMS (sham group), providing imaging evidence to support the potential neuropathological mechanisms underlying cognitive impairment in MCI patients.

This study was approved and supervised by the Medical Ethics Committee of the Second Clinical Medical College of North Sichuan Medical College, Nanchong, China. A total of 22 MCI participants aged from 55 to 80 years were recruited from communities and signed informed consent. Inclusion criteria were based on the core diagnostic criteria of “MCI due to AD” and Bondi’s quantitative indicators [25, 26]: (1) complaints of cognitive decline were expressed by participants or their families or recalled by clinicians; (2) all concerns must be confirmed by standardized test using Bondi’s criteria; (3) patients still maintain independence in daily life abilities; (4) Clinical Dementia Rating Scale of 0.5 points; (5) not dementia yet. Participants were excluded if they existed any following condition: (1) left-handed; (2) any significant disease suspected to cause cognitive decline (Parkinson’s disease, vascular dementia, stroke, depression); (3) magnetic field safety concerns such as mental implants in the body, claustrophobia; (4) severe illness, inability to complete scale assessment and rTMS treatment.

In this randomized, double-blind study, researchers who did not take part in the treatment randomly divided patients into one of two groups: the real group or the sham group. Each of them underwent a systematic cognitive test and MRI scan before and after 10 consecutive days of rTMS intervention. The scale evaluation was conducted by two professional clinicians in a quiet environment, and neither the evaluator nor the subjects were aware of the grouping.

The therapy was delivered by using a MagPro R30 magnetic stimulator (Denmark Tonica Company, MCF-B65 coil). Before treatment, we measured the resting motor threshold (RMT) of each patient to set the appropriate stimulus intensity of individual tolerance. Then, the patient was asked to lie flat on the treatment bed with the rTMS handle fixed tangentially on the left DLPFC region. The treatment parameters were preset as follows: 10 Hz frequency, 90% of RMT intensity, 30 stimulation trains per day, 50 pulses per stimulation train, a total of 1,500 pulses, and 25 s of inter-train interval. The sham group could not produce effective stimulation with the sham stimulation coil, but other parameters were completely consistent with the real group. All patients received a total of 10 consecutive days of treatment.

To assess the effect of rTMS on cognitive function, we performed a series of comprehensive cognitive assessments at each time point (baseline and post-intervention) to assess which specific domain was influenced. Specifically, these scales focused on the overall cognitive function and three main cognitive domains, including memory, language, and executive function. Each cognitive domain was evaluated by two kinds of independent tests or items. Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) were chosen as overall cognitive function tests for they are the most widely used cognitive screening scales. Long-term recall (N5) and recognition (N7) of the Auditory Verbal Learning Test (AVLT) were used to test memory function. Similarly, Boston Naming Test (BNT) and Animal Verbal Fluency Test (AFT) were used for language function and Trail Making Test (TMT-A/B) for executive function.

All MRI scans were conducted by using a 1.5-Tesla MRI machine (General Electric Company, Signal HDxt, America) using an eight-channel head coil. Prior to the MRI scan, the subject’s head was fixed as much as possible to reduce head motion, and they were told to try their best to keep still. In addition to conventional structural images (T₁WI, T₂WI, T2-FLAIR, DWI), DTI sequences were obtained with the following protocol: TR = 8,500 ms; TE = 96 ms; flip angle = 90°; FOV = 24.0 cm × 24.0 cm; matrix = 256 × 256; slice thickness = 5.0 mm; voxel size = 0.94 × 0.94 × 5.00 mm³; a total of 32 slices with no gap. The diffusion-sensitive gradients were applied along 30 noncollinear directions with B = 1,000 s/mm² to obtain the weighted images and one unweighted B0 image with B = 0 s/mm². When the scan was completed, a professional radiologist would check conventional sequences to rule out those with organic cranial lesions.

The original DTI images were converted into 3D NIfTI hdr/img format by dcm2nii software. Image preprocessing was performed by Statistical Parametric Mapping 12 (SPM 12, http://www.fil.ion.ucl.ac.uk/spm) toolkit under MATLAB 2013b platform with the following steps: (1) head motion correction: to eliminate the head motion generated by various factors during the scanning process, images in same sequence were matched to the first reference image; (2) fiber tracking and the reconstruction of the diffusion parameter maps: Diffusion Toolkit software was used to perform the WM fiber tracking based on the interpolated streamline algorithm with the stopping condition of FA <0.20 or tracking angle < 40°; (3) normalization: all of the FA images were normalized to the Montreal Neurological Institute (MNI) spatial coordinate system and resampled to a voxel of 2.0 × 2.0 × 2.0 mm³ size, in order to unified analysis and comparison; (4) smoothing: a Gaussian kernel with a full width at half maximum of 4 mm × 4 mm × 4 mm was used for spatial smoothing to improve the signal-to-noise ratio. Then the comparison of FA images before and after treatment was conducted. All brain regions with significant differences in FA values were considered regions of interest (ROIs). Each ROI was generated in MNI space through peak coordinates. The Resting-State fMRI Data Analysis Toolkit (REST) was used to extract the FA values of ROIs.

Demographic data and scale scores were statistically analyzed by using SPSS 24.0. All results were considered statistically significant with p < 0.05. Population data and baseline scales between the two groups were compared to ensure the feasibility of the experiment. Except for the χ² test used in gender analyses, other measurement data were compared by independent-sample t test. Then to testify to the effects of rTMS on MCI patients, a paired-sample t test was applied to compare the cognitive scales within the two groups. As for the comparison between groups, we chose to conduct an independent-sample t test on the changes in scores before and after treatment rather than directly comparing the scores, which may better explain the therapeutic effect.

The comparison of the FA images was conducted by using the SPM 12 toolkit, and all results were corrected by the AlphaSim method with p < 0.05 as the significance threshold. Finally, Spearman correlation analysis between FA values in ROIs and cognitive scores was carried out through SPSS 24.0 software.

A total of 22 eligible MCI patients were screened in this study (real group = 11; sham group = 11). No significant difference was observed in age, gender, education level (summarized in Table 1). All cognitive scores were comparable between the real and the sham groups at baseline. After accomplishing 10 days of rTMS treatment, the real group showed a significant increase in the MMSE, MoCA, AVLT (N5, N7), and BNT scores (p < 0.05). Notably, the changes of the same scales above mentioned and the TMT-B score in the real group were higher than those in the sham group (p < 0.05). Only the MoCA scores were significantly improved in the sham group (summarized in Table 2).

As shown in Figure 1, compared with baseline, the FA values of the real group had increased significantly in the left middle frontal gyrus and bilateral parahippocampal gyrus after rTMS treatment. At the same time, the FA values in the right superior frontal gyrus had decreased observably (AlphaSim correction, p < 0.05, cluster size >395). In the sham group, we did not find any brain regions that showed significant changes in FA values. As shown in Figure 2, the FA values in the left middle frontal gyrus of the real group were higher than the sham group after treatment (AlphaSim correction, p < 0.05, cluster size >388). Table 3 shows the detailed information of the brain regions above.

It was found that the FA value of the left middle frontal gyrus was positively correlated with the BNT score after treatment (r = 0.72, p = 0.01), and the changes in FA value in the right superior frontal gyrus were positively correlated with the shift in TMT-B score (r = 0.84, p = 0.001) (Fig. 3).

Transcranial magnetic stimulation (TMS) can induce an induced current in the brain, which can elicit inhibitory or excitatory potentials in neural tissue, and these changes in focal neural activity may spread to anatomically or functionally interconnected cerebral regions. The therapeutic effects of rTMS are strongly dependent on the preset parameters. Although there is still no consensus on parameters, most studies have applied high-frequency stimulation over DLPFC. The study by Haosu Zhang’s et al. [27] has shown high-frequency (>5 Hz) rTMS can produce excitatory effects and may generate more neural activity than low frequency, which may lead to better influence. Moreover, DLPFC is a popular target for the current rehabilitation of neuropsychiatric diseases because it is a central area for working memory, cognitive control, and decision processing [28]. Our study explored a kind of rTMS treatment protocol (10 Hz, 90% of RMT, 10 days) over the left DLPFC in MCI patients. From Table 2, we can discover that scores of MMSE, MoCA, AVLT (N5, N7), and BNT in the real group were significantly improved. And when compared to the sham group, the variation in these and TMT-B scores in the real group was greater than that in the sham. These results demonstrated the positive effects of rTMS on overall cognition, memory, language, and executive function in MCI patients. Our behavioral results lead support to prior studies that high-frequency rTMS over DLPFC enhanced cognitive functions in MCI patients [14].

A study [29] investigated the neurobiological effects of rTMS applied to the precuneus in 16 patients with mild to moderate AD. Their results showed that after 24 weeks of rTMS treatment, the degree of gray matter volume atrophy in the stimulated areas was significantly lower in the real stimulation group compared to the sham stimulation group. These results suggest that rTMS can effectively slow down hippocampal structural atrophy in patients with MCI and improve cognitive function, indicating its efficacy in the recovery of brain structure of MCI. Therefore, these structural changes may also exist in the WM structure. The effects of rTMS on improving the integrity of WM structure could potentially be effectively detected and evaluated using DTI imaging technology. Many previous studies have reported the impairment of the functional connectivity within the default mode network (DMN) in AD and MCI. For instance, a recent study observed a breakdown of signal propagation within the DMN in AD by using TMS-evoked electroencephalogram technology [30]. Therefore, the functional connectivity of DMN is often used as an indicator for assessing the treatment efficacy. Cui et al. found that rTMS induced deactivation of functional connectivity within the DMN; this rTMS-induced hypoconnectivity was associated with clinical cognitive improvement in MCI [31]. Another dual-targeted rTMS study also reported the significant correlation between improved cognitive performance and improved functional connectivity of the precuneus and caudate nucleus [32]. On the one hand, these studies suggest that changes in functional connectivity within the DMN after rTMS treatment could represent a valuable indicator of treatment response [33]. On the other hand, these researches also indicate that TMS treatment can restore the disruptions in functional connectivity caused by MCI, which also confirms the findings obtained in our study. Our study, utilizing DTI imaging technology to reflect changes in WM structure in MCI patients following rTMS treatment, demonstrated that after rTMS treatment, the integrity of WM structure in certain brain regions of the real group of MCI patients may have changed, and the behavioral performance of the subjects also improved. For example, the FA value in the left middle frontal gyrus of the real stimulation group increased, which was significantly higher than that in the sham group and positively correlated with the BNT score. The left frontal lobe is one part of the complex web of language procession and supports different aspects of pragmatic processing [34]. Task-state language studies using DTI or fMRI found that the left frontal lobe is activated and the language performance was significantly correlated with the degree of myelin bloom in the cerebral fiber tracts [35], another retrospective and cross-sectional study has found the network measures of the high executive ability group demonstrated greater WM integrity [23]. Therefore, WM fiber Bundles play an important role in the delivery of information between different brain regions, and the integrality of the structure is extremely crucial for rapid and complete transmission of information. The frontal lobe would undergo structural damage, which may contribute to corresponding cognitive deficits due to aging or disease [36]. Cross-sectional studies have described that MCI emerged with a broad decline in WM FA value, localized to the cingulum, corpus callosum, superior lateral fasciculus, and uncinate fasciculus and throughout temporal, occipital, and frontal lobes, and associated with reduced episodic memory and executive function [11, 37]. A recent meta-analysis revealed that decreased FA values coexist in AD and MCI in the left frontal lobe, corpus callosum, hippocampus, and cingulate gyrus. Moreover, these changes in the two phases showed a strong correlation, suggesting that the WM nerve damage in these regions is involved in the progression of MCI to AD [38].

In addition, we also found increased FA values in bilateral parahippocampal gyrus in the real group. As an essential part of the limbic system, the parahippocampal gyrus, and other limbic lobe structures are interconnected with each other through Papez circuit, participating in the expression and control of emotion, cognition, and memory function [39]. Researchers have provided evidence on the role of parahippocampal gyrus in cognitive disorders [40]. A study compared the whole brain DTI indicators and cognitive scores of the healthy controls, cognitive complaints but without psychometric impairment, and the MCI group, found that the MCI group showed lower FA values in bilateral parahippocampal WM, the FA values of the cognitive complaint group was intermediate between MCI and healthy controls, and the overall level of the parahippocampal WM diffusion measures was associated with the memory function [41]. Thus, MCI progression showed selective bilateral parahippocampal gyrus impairment. Some authors who drove machine learning methods to classify AD and healthy controls found that the FA values of the parahippocampal lobe were the best discriminates [42]. Our findings raised the possibility that the impaired parahippocampal gyrus WM microstructure of MCI is actively regulated by rTMS. Future studies with large samples are needed to explain it for we did not find any correlation with cognitive scores.

Notably, the right superior frontal gyrus showed decreased FA values in the real group, which was contrary to the left. Correlation analysis showed that the more the FA values decreased in the right superior frontal gyrus, the more the TMT-B score decreased which represented a better executive function. Language and executive functions are both highly lateralized and need hemispheric specialization and left hemisphere linguistic predominance has been recognized as a unique aspect of human brains [43]. DTI imaging found asymmetric FA values in the frontal lobe in healthy adults. Higher FA values in the left frontal lobe than the right were associated with better language comprehension and memory [44]. FA asymmetry in the frontal region is also vital for the normal operation of executive function [45]. This is mainly maintained by the contact fibers that connect the two hemispheres, such as the corpus callosum which can keep functional asymmetry of homologous regions between the two hemispheres [46]. As the previous meta-analysis stated, the FA value in the left frontal lobe was significantly reduced in MCI patients, but no changes were found on the right side [38]. More severe destruction of dominant hemispheres leads to the weakening of asymmetrical relationships between hemispheres, which may be detrimental to the maintenance of language and executive function. The rTMS pulses can spread to regions with anatomic or functional connections to the stimulation site, and induce bidirectional effects through the corpus callosum fiber tracts [47]. Some researchers elaborated that the rTMS pulse passing through the corpus callosum mainly produces inhibitory effects [48]. Therefore, we hypothesized that rTMS through the corpus callosum pathway may reduce the FA value of the contralateral frontal lobe in MCI patients, but reshaped the trend of the left frontal lobe dominance, which was beneficial to the maintenance of language and executive function.

Disruption of the original balance pattern between hemispheres has been documented in other diseases. For example, asymmetrical WM changes were found in the frontal tracts of schizophrenics. Higher mean FA values in the right frontal tract are associated with poorer verbal memory performance [49]. Likewise, patients with major depression disorder have reported that the left dorsolateral prefrontal lobe is inactivated while the right dorsolateral prefrontal lobe is overactivated. rTMS treatment can work by inhibiting the right and enhancing the left frontal lobe to rebuild bilateral balance [8]. Yet, we cannot draw firm conclusions about “rTMS regulates bilateral hemispheres to restore functional balance.” A massive number of studies are needed to further confirm the effects and mechanisms. Finally, DTI, as a noninvasive brain imaging technique, can reveal the three-dimensional morphology of nerve fibers and functional tracts in the WM, with higher FA values indicating more consistent diffusion direction of water molecules, representing greater structural integrity of WM fiber tracts. The structural integrity of WM is significantly correlated with cognitive decline. He et al. suggested that the FA value of the right cingulate gyrus is a sensitive indicator reflecting the progression of aMCI and may serve as a potential biomarker for predicting disease progression in patients [50]. Additionally, studies have found that the FA values in specific brain regions of patients (such as those with schizophrenia or cognitive impairment) change after rTMS treatment, demonstrating the impact of rTMS on the WM structure of the stimulated individuals. However, using DTI alone to evaluate the therapeutic effects of rTMS has certain limitations. On the one hand, DTI cannot describe water molecules with non-Gaussian distributions in the complex structure of the brain. In the future, combining the diffusional kurtosis imaging with DTI could analyze the state of water molecules that conform or do not conform to Gaussian distributions, thereby providing a more detailed depiction of gray and WM structures in the brain. On the other hand, FA values are local indicators and cannot comprehensively reflect changes in the entire WM fiber tracts. In the future, it will be necessary to thoroughly and deeply reflect the microstructural performance of WM and evaluate the efficacy of rTMS from the perspectives of diffusion fiber connectivity and structural networks.

Our study has several limitations. First, only a small sample size of 22 MCI patients was included in the study. Future studies with larger sample sizes are needed to further explore the effects of high-frequency rTMS on the WM microstructure in MCI patients. Second, MCI can be classified into different subtypes based on various criteria, and there are differences among these subtypes. For example, patients with aMCI have a higher probability of converting to AD compared to those with non-aMCI. Therefore, exploring different subtypes of MCI may lead to a more accurate and in-depth understanding of the neural mechanisms underlying rTMS treatment for MCI. And then, we only performed stimulation of left DLPFC, lack of other stimulation sites comparison to analyze the bidirectional effect of rTMS. At last, this study lacks the detection of positive biological markers in MCI patients. In the future, combining changes in biological markers with brain imaging analysis techniques could provide a more comprehensive reflection of the effects and impacts of rTMS treatment. Despite the shortcomings and limitations of this study, which may require a more cautious interpretation of the results, our findings suggest to some extent that rTMS treatment may have a positive impact on the structural integrity of WM in MCI patients, which could be one of the neural mechanisms by which rTMS improves cognitive impairment. Additionally, there are now various brain stimulation techniques for treating cognitive impairment, each proven to improve patients’ cognitive states. Therefore, future research could further compare the effects of different stimulation techniques on the WM fiber structure in MCI patients to identify the optimal treatment method.

In summary, this study provides new evidence for 10 Hz rTMS over DLPFC to regulate the abnormal WM microstructure in some special regions and causally ameliorate cognitive performance in MCI, which may be the underlying neural mechanism of intervention. Meanwhile, we found an asymmetric but coordinated regulation pattern of rTMS on the bilateral frontal lobe WM. Further studies with large sample sizes are required to verify our findings.

The studies involving human participants were reviewed and approved by the Medical Ethics Committee of Nanchong Central Hospital (2021089). Written informed consent was obtained from all vulnerable participants’ legal guardian for participation in this study.

There are no conflicts of interest in this study.

This work was supported by the State Administration of Foreign Experts Affairs, China (No. SYZ202105); Sichuan Medical Association (No. Q20043); and Bureau of Science and Technology Nanchong City (No. 21YFZJ0023).

All authors made substantial contributions to the conception and design of the study, acquisition of the data, analysis, interpretation of the data, involved in drafting and revising of the manuscript, and final approval of the submitted version. The conceptualization and experimental design of this study were proposed by Zhiwei Guo and Shengxue Song and jointly determined through discussions among all three authors (Shengxue Song, Zhiwei Guo, and Qiwen Mu). Shengxue Song was primarily responsible for data collection, data analysis, result interpretation, and manuscript drafting. Zhiwei Guo participated in experimental data collection and data analysis. The content verification, error correction, revision of the initial manuscript draft, and final version editing were mainly overseen by Qiwen Mu. All authors have read and approved this version of the article, and due care has been taken to ensure the integrity of the work.

The data that support the findings of this study are not publicly available due their containing information that could compromise the privacy of research participants but are available from the corresponding author (Qiwen Mu, E-mail: [email protected]).