Abstract
Effective rehabilitation in Parkinson's disease (PD) is related to brain reorganization with restoration of cortico–subcortical networks and compensation of frontoparietal networks; however, further neural rehabilitation evidence from a multidimensional perspective is needed. To investigate how multidisciplinary intensive rehabilitation treatment affects neurovascular coupling, 31 PD patients (20 female) before and after treatment and 30 healthy controls (17 female) underwent blood oxygenation level-dependent functional magnetic resonance imaging and arterial spin labeling scans. Cerebral blood flow (CBF) was used to measure perfusion, and fractional amplitude of low-frequency fluctuation (fALFF) was used to measure neural activity. The global CBF–fALFF correlation and regional CBF/fALFF ratio were calculated as neurovascular coupling. Dynamic causal modeling (DCM) was used to evaluate treatment-related alterations in the strength and directionality of information flow. Treatment reduced CBF–fALFF correlations. The altered CBF/fALFF exhibited increases in the left angular gyrus and the right inferior parietal gyrus and decreases in the bilateral thalamus and the right superior frontal gyrus. The CBF/fALFF alteration in right superior frontal gyrus showed correlations with motor improvement. Further, DCM indicated increases in connectivity from the superior frontal gyrus and decreases from the thalamus to the inferior parietal gyrus. The benefits of rehabilitation were reflected in the dual mechanism, with restoration of executive control occurring in the initial phase of motor learning and compensation of information integration occurring in the latter phase. These findings may yield multimodal insights into the role of rehabilitation in disease modification and identify the dorsolateral superior frontal gyrus as a potential target for noninvasive neuromodulation in PD.
SIGNIFICANCE STATEMENT Although rehabilitation has been proposed as a promising supplemental treatment for PD as it results in brain reorganization, restoring cortico–subcortical networks and eliciting compensatory activation of frontoparietal networks, further multimodal evidence of the neural mechanisms underlying rehabilitation is needed. We measured the ratio of perfusion and neural activity derived from arterial spin labeling and blood oxygenation level-dependent fMRI data and found that benefits of rehabilitation seem to be related to the dual mechanism, restoring executive control in the initial phase of motor learning and compensating for information integration in the latter phase. We also identified the dorsolateral superior frontal gyrus as a potential target for noninvasive neuromodulation in PD patients.
- cerebral blood flow
- dynamic causal modeling
- functional magnetic resonance imaging
- neurovascular coupling
- Parkinson's disease
- rehabilitation
Introduction
Parkinson's disease (PD) is a progressive neurodegenerative disorder characterized by motor symptoms and nonmotor symptoms (Bloem et al., 2021) because of the loss of dopamine in striatal pathways. Rehabilitation is a promising supplemental treatment for PD as it reduces functional disability and improves quality of life in both motor and nonmotor symptoms among patients with PD (Mak et al., 2017; Marinelli et al., 2017; Nackaerts et al., 2019). In particular, multidisciplinary intensive rehabilitation treatment (MIRT), which includes aerobic, motor-cognitive, and goal-based components, and relies on cognitive engagement and motor learning, showed potential in improving patients' quality of life (Frazzitta et al., 2013, 2015; Ferrazzoli et al., 2016; Franciotta et al., 2019; Radder et al., 2020; K. K. Chen et al., 2021).
In rehabilitation, motor learning is defined as a practice-related change or improvement in motor performance (Marinelli et al., 2017). The initial phase of motor learning refers to connections between the basal ganglia and the prefrontal cortex, which play an important role in goal-directed learning (Petzinger et al., 2013; Silveira et al., 2018). Then, in the latter phase, the shift from goal-directed learning to habit-based learning leads to decreased activation of circuits in the prefrontal and basal ganglia and increased activation of circuits in the sensorimotor regions of the basal ganglia with the sensorimotor cortex (Petzinger et al., 2013; Baglio et al., 2021). Preliminary evidence indicates an association between the effective rehabilitation and brain reorganization with restoration of the cortico–subcortical pathway and activation of compensatory networks, that is, the frontoparietal network (Marinelli et al., 2017; Baglio et al., 2021). However, further evidence of neural mechanisms underlying rehabilitation from a multidimensional neural perspective is needed.
To elucidate the neural mechanisms that underlie rehabilitation effects, neurovascular coupling (NVC) is measured to evaluate the coupling between supply (blood flow) and metabolic demand (neural activity; Girouard and Iadecola, 2006; Phillips et al., 2016; Ahmad et al., 2020) by combining arterial spin labeling (ASL) and resting-state blood oxygen level-dependent (BOLD) functional magnetic resonance imaging (fMRI), respectively. Several studies have assessed NVC using two different indices, namely, the correlation between ASL and BOLD-fMRI at the global level and the ratio of perfusion and neural activity at the regional level (Liang et al., 2013; Zhu et al., 2017; Kim et al., 2020; Baller et al., 2022). These measures indicated tight coupling between perfusion and neural activity in normal (Liang et al., 2013; Phillips et al., 2016; Baller et al., 2022) and pathologic conditions, such as schizophrenia (Zhu et al., 2017) and Alzheimer's disease (Kim et al., 2020). Consequently, investigating the rehabilitation-induced alterations of NVC can help to elucidate treatment mechanisms, thus allowing the identification of potential neural pathways that orchestrate brain function and metabolism.
Therefore, we collected both ASL and BOLD-fMRI data from 31 PD patients before and after rehabilitation and 30 healthy controls (HCs) matched for sex, age, and education. Cerebral blood flow (CBF) derived from ASL data was used to measure perfusion, and fALFF derived from fMRI data was used to measure neural activity because of its sensitivity and specificity for spontaneous activity (Zou et al., 2008; Zuo et al., 2010) and its correlation with CBF (Bray, 2017; Li et al., 2021). We analyzed NVC alterations characterized by the across-voxel CBF–fALFF correlations and the CBF/fALFF ratio and explored relationships between regions with significant NVC alterations and clinical indicators. We further performed dynamic causal modeling (DCM) to estimate the causal influence and determine the strength and directionality of information flow among treatment-related regions (Friston et al., 2014). We hypothesized that NVC alterations are mainly located in regions important for motor learning and might exhibit different trends, supporting the hypothesis that the effective rehabilitation is reflected through a dual mechanism of restoration and compensation (Frazzitta et al., 2013; Baglio et al., 2021).
Materials and Methods
Demographic and clinical evaluation
Data from this study were drawn from a prospective real-world cohort study (Multidisciplinary Rehabilitation Registration Study on Parkinson's disease, registration number ChiCTR2000033768). Thirty-three patients (20 female) and 30 HCs (17 female) matched for sex, age, and education were included in this study. The PD patients (before and after a 2 week rehabilitation period) and HCs underwent both ASL and BOLD fMRI scans. Two of 33 patients were excluded after preprocessing, resulting in 31 patients being included. All patients were diagnosed by the Movement Disorder Society diagnostic criteria in the early to middle stage of the disease (Table 1). Inclusion criteria were the following: The Movement Disorder Society diagnostic criteria for primary PD (Postuma et al., 2015) were met; patients were <75 years old; the presence of comorbidities did not require special in-hospital treatment; there was no deep brain stimulation therapy and no implantable medical device; the Mini-Mental State Examination (MMSE) score was >24 (education level higher than or equal to secondary school) or >20 (less than or equal to elementary school); and patients had the ability to stand unsupported for >20 s and walk independently. Exclusion criteria were the presence of other diseases that hinder one's walking ability, severe cognitive impairment affecting comprehension, antipsychotic drug use, considerable visual or auditory deficit, or combined serious complications contraindicated to rehabilitation (K. K. Chen et al., 2021). This research was approved by the ethics committee of the Beijing Rehabilitation Hospital (2020bkky010). All participants signed informed consent following the Declaration of Helsinki.
Demographic and clinical characteristics of the participants
The third part of the Unified Parkinson's Disease Rating Scale (UPDRS-III) was used to evaluate the alterations in motor performance (Movement Disorder Society Task Force on Rating Scales for Parkinson's Disease, 2003). Additionally, a series of neuropsychological tests was administered to patients before treatment, including the MMSE, Montreal Cognitive Assessment (MoCA), Hamilton Depression Scale (HAMD), and Hamilton Anxiety Scale (HAMA).
MRI acquisition
Magnetic resonance data were acquired using a 3.0 Tesla GE SIGNA Pioneer Scanner. All participants underwent the resting-state ASL and the resting-state BOLD scan, remaining awake with their eyes closed during the scans.
High-resolution T1-weighted images were acquired using a three-dimensional brain volume (3D BRAVO) sequence with the following parameters: echo time (TE) = 3.06 ms, repetition time (TR) = 8.06 ms, inversion time (TI) = 450 ms, flip angle (FA) = 15°, field of view (FOV) = 300 × 300 mm2, matrix = 512 × 512, slice thickness = 1.0 mm, and slice number = 160.
Resting-state perfusion imaging was acquired using a pseudocontinuous ASL sequence with 3D fast spin-echo acquisition and background suppression. Interleaved control and label images were acquired with the following parameters: TE = 10.85 ms, TR = 4904 ms, FA = 111°, FOV = 260 × 260 mm2, matrix = 128 × 128, slice thickness = 4.0 mm, slice number = 40, and postlabel delay = 2000 ms.
Functional BOLD images were collected using a gradient-recalled echoplanar imaging sequence with the following parameters: TE = 35 ms, TR = 2000 ms, FA = 90°, interslice gaps = 0 mm, FOV = 280 × 280 mm2, matrix = 128 × 128, slice thickness = 4.0 mm, and slice number = 40.
MRI preprocessing
The ASL difference images were created by pairwise subtraction of the label and control images. Then, the CBF maps were calculated in combination with ASL difference images and proton density weighted reference images (Ma et al., 2010). CBF maps were normalized to Montreal Neurologic Institute (MNI) space using Statistical Parametric Mapping (SPM8; https://www.fil.ion.ucl.ac.uk/spm) software with the following steps: (1) The native CBF images were coregistered to structural T1-weighted images of each participant; (2) the T1-weighted images were spatially normalized to MNI space using the deformation fields generated during segmentation and normalization; (3) for each participant, the CBF image was transformed into MNI space using the deformation parameter derived from the registration of T1-weighted images and was resampled to 2 × 2 × 2 mm3; (4) since normalized CBF was more sensitive to small changes in regional perfusion, we corrected for individual variations in global perfusion by scaling each voxel in the CBF map by the mean whole-brain CBF; and (5) standardized maps were spatially smoothed with a Gaussian kernel of 6 × 6 × 6 mm3 full-width at half maximum (FWHM). We masked the CBF maps using the whole-brain gray matter (GM) tissue priors to minimize the inclusion of white matter and CSF.
The BOLD-fMRI preprocessing was conducted using SPM8 and the Data Processing & Analysis for Brain Imaging (www.rfmri.org/dpabi). For each run, the first 10 time points were discarded to account for signal equilibrium and participants' adaptation to the circumstances. The remaining functional images were first corrected for timing and then realigned to the first volume to correct for head motion, which did not exceed 3.0 mm of displacement or 3.0° of rotation in any direction in any participant. Subsequently, functional images were spatially normalized to the standard MNI template and resampled to a resolution of 3 × 3 × 3 mm3. To reduce the effects of motion and nonneuronal BOLD fluctuations, changes because of head motion, the CSF signal, and white matter signals were further removed as nuisance covariates. Finally, fMRI images were smoothed with a Gaussian filter of 4 × 4 × 4 mm3 FWHM.
In this study, we only used the fALFF maps to determine spontaneous neural activity because fALFF minimizes artifacts because of body motion, respiration, and cardiac noise and improves sensitivity when detecting spontaneous neural activity (Zuo et al., 2010; Yang et al., 2018). We computed fALFF values based on the preprocessed data without bandpass filtering. As described by Zou et al. (2008), Zuo et al. (2010), the time series of each voxel was transformed into the frequency domain using the fast Fourier transform, and the square root was calculated for each power spectrum frequency to derive an amplitude spectrum. Then, the sum of amplitudes within a specific frequency range (0.01–0.10 Hz) was divided by the sum of amplitudes across the entire frequency range (0–0.25 Hz) to obtain the fALFF maps. To prevent the influence of individual differences on brain activity levels, each fALFF value per voxel was divided by the whole-brain mean fALFF value to yield normalized fALFF maps. The whole-brain GM mask was applied to fALFF maps to minimize the inclusion of white matter and CSF (Sarmiento et al., 2020).
Global CBF–fALFF coupling analysis
To quantitatively evaluate the coupling between CBF and fALFF within the whole GM, correlation analysis was performed across voxels for each participant (Liang et al., 2013). The CBF–fALFF correlation represents global NVC and reflects the coordination between the requirement of oxygen and the blood supply (Phillips et al., 2016; Baller et al., 2022).
Regional CBF/fALFF ratio analysis
Voxel-wise comparisons were performed to identify alterations in CBF and fALFF. To quantify amount of blood flow or metabolic energy per unit of neural activity, we computed the regional CBF/fALFF ratio in a voxel-wise manner. The ratio represents the regional NVC across the brain and reflects a balance between CBF and neural activity (Liang et al., 2013). Regions with higher values tend to have more metabolic demands as they communicate with the rest of the brain (Liang et al., 2013). The ratio was then divided by the global mean value to improve normality. For each participant, a 6-mm-radius kernel size sphere around the peak voxel within clusters showing significant treatment-related differences in the CBF/fALFF ratio was used to extract ratio values for the subsequent correlation analysis and DCM.
Correlation analysis between regional NVC and clinical variables
Pearson correlation analysis was conducted to examine the correlations between the CBF/fALFF and UPDRS-III scores. As mentioned above, we used a 6-mm-radius sphere within clusters displaying significant CBF/fALFF ratios to extract ratio values. To determine whether treatment-related changes in NVC were associated with motor improvement, the CBF/fALFF values of treatment-related regions and UPDRS-III scores after treatment were subtracted from the values at baseline. Next, the differential values of CBF/fALFF and scores were divided by the corresponding value at baseline to obtain the percentage changes. Finally, Pearson's correlation analysis was conducted to examine the associations of percentage alterations between changes in the CBF/fALFF ratio and UPDRS-III scores. We further analyzed the correlations between UPDRS-III scores and CBF or fALFF values in regions showing significant correlations the CBF/fALFF and UPDRS-III scores.
Effective connectivity based on dynamic causal modeling
DCM is the predominant analysis framework for inferring effective connectivity; this method estimates the causal architecture from the observed BOLD activity on fMRI, namely, the direct causal influences of one brain area on another. The effectivity connectivity between regions of interest (ROIs) represents the change in the activity of the destination area caused by changes in the source area, with positive values indicating excitation and negative values indicating inhibition. Conversely, self-connections are inhibitory, with positive values indicating greater inhibition and negative values indicating reduced inhibition (Jamieson et al., 2022).
To further evaluate treatment effects and identify potential therapeutic targets, DCM was used to estimate the causal influence of effective connectivity among five ROIs based on prior clusters showing significant differences in the CBF/fALFF ratio and important roles in the neuropathology and motor learning circuitry of PD (Nackaerts et al., 2019; Baglio et al., 2021; Bloem et al., 2021). These ROIs were as follows: the left angular gyrus (MNI coordinates, −44, −68, 44), the right inferior parietal gyrus including the supramarginal and angular gyri (MNI coordinates, 50, −60, 44), the right dorsolateral superior frontal gyrus (MNI coordinates, 20, −2, 52), and the bilateral thalamus (left MNI coordinates, −10, −28, 6; right MNI coordinates, 22, −26, 18).
Time series were extracted from fMRI signals with 6-mm-radius spheres around the aforementioned ROIs for DCM. At the first level, we used spectral DCM to estimate fully connected effective connectivity among these ROIs and constructed a directed and weighted effective connectivity network for each subject (Friston et al., 2014). At the second level, parametric empirical Bayes (PEB) was used to examine whether the treatment altered certain connections (i.e., a continuous difference in connectivity; Friston et al., 2016); this method estimates the effects of group mean and group differences on each effective connectivity link. To evaluate how regions interact, Bayesian model comparison and greedy search were used to automatically remove redundant connections that did not contribute to the model evidence (Friston and Penny, 2011). Finally, the Bayesian model average was calculated by averaging the final parameters of the selected models and weighting their model evidence by the posterior probabilities to visualize the connectivity corresponding to the treatment-related differences.
Experimental design and statistical analysis
Experimental design
Aiming to combine motor relearning with external and internal cue strategies, multidisciplinary intensive rehabilitation treatment has been considered an effective rehabilitation with multidisciplinary, aerobic, motor-cognitive, intensive, and goal-based strategies in PD (Frazzitta et al., 2013, 2015; K. K. Chen et al., 2021). In this study, patients were admitted to a 2-week MIRT program in the hospital with 5 d per week in the ON state (1–2 h after medication).
The treatment consists of four daily rehabilitation sessions, and each session lasted for 30–60 min, including (1) one-on-one treatment with a physical therapy for half an hour, (2) goal-directed balance and gait training by augmented reality treadmill training twice daily (in the morning and afternoon) for 30 min each time, (3) aerobic training for 30 min to improve autonomy in everyday activities using an upper and lower limb trainer (T5XR; NuStep), and (4) speech therapy for 30–60 min (K. K. Chen et al., 2021). Each session was conducted by the same well-trained physical therapist (sessions 1, 2, and 3) or one speech therapist (session 4).
During all the activities, the heart rate of patients was kept at 70–80% of the maximum rate to avoid risk. Patients had at least a 30-60 min break between sessions. Also, patients reported no feelings of discomfort or fatigue in the next day after treatment.
Statistical analysis
Statistical analyses were performed using MATLAB 2016a (MathWorks) and IBM SPSS Statistics 25.0. The results are reported as the mean ± SD.
The demographic and clinical variables between HCs and PD patients at baseline were examined with a two-sample t test for parametric variables and a chi-square test for dichotomous variables. The paired t test was used to examine treatment effects on the UPDRS-III scores and the global CBF–fALFF correlation coefficients.
To determine treatment-induced differences in the CBF, fALFF, and the CBF/fALFF ratio at the local level, voxel-wise comparisons were performed with paired t test and corrected by a Gaussian random field (GRF) with a combination of voxel-level p < 0.001 and cluster-level p < 0.05 for multiple comparisons. Then, to explore whether these CBF/fALFF ratio changes reflected a restoration of the normal pattern or a compensatory mechanism of brain function, an ANCOVA was used to estimate CBF/fALFF ratio differences between HCs and PD patients (HCs vs before treatment, HCs vs after) on these regions with age, age squared, sex, and education as covariates of no interest. A false discovery rate (FDR)–corrected threshold of p < 0.05 was used for multiple comparisons.
Pearson's correlation analysis was conducted to examine the associations between CBF/fALFF ratio and UPDRS-III scores with age, age squared, sex, and education as covariates and statistical significance of p < 0.05. We performed 10,000 permutation tests for correlation analysis. Multiple comparisons were not corrected because the aforementioned analyses were exploratory in nature. Hence, a significant relationship was considered at an uncorrected p < 0.05.
Finally, connections that survived a threshold of 95% posterior probability (i.e., strong evidence of the parameters being present vs absent) were included in the DCM. The positive estimated parameters indicate stronger connectivity after treatment than before, and negative parameters indicate stronger connectivity before treatment. Thus, we divided treatment-induced alterations in effective connectivity into four categories—increased or reduced excitatory excitation and increased or reduced inhibitory connectivity. Increased excitatory connectivity was defined as a group mean and a group difference both greater than zero, whereas reduced excitatory connectivity was defined as a group mean value greater than zero and a group difference smaller than zero. Increased inhibitory connectivity was defined as a group mean smaller than zero and a group difference smaller than zero, whereas reduced inhibitory connectivity was defined as a group mean smaller than zero and a group difference greater than zero.
Data availability
The data that support the findings of this study are available from the corresponding authors on reasonable request (yantianyi{at}bit.edu.cn or fangboyanv{at}ccmu.edu.cn). The data are not publicly available because of privacy restrictions.
Results
Demographic and clinical results
In this study, we found that PD patients obtained benefits from MIRT as demonstrated by a reduction in UPDRS-III scores (t(30) = 4.413; 95% CI, 2.84–7.74; before, 30.5 ± 13.2; after, 25.2 ± 12.7; p < 0.001; paired t test; Table 1; Fig. 1A). There were no significant differences between HCs and PD patients in terms of age, sex, or education (Table 1).
MIRT-induced motor improvement, global and regional neurovascular coupling distribution in PD patients before and after treatment and HCs. The threshold value for establishing significance was set at p < 0.05. A, Violin plots of the UPDRS-III scores before and after treatment. B, Top three rows, The spatial distributions of CBF and fALFF values and the CBF/fALFF ratios in the HCs and PD patients before and after treatment. The maps were averaged across subjects. Despite subtle differences, HCs and PD patients before and after treatment exhibited highly similar spatial distributions in the three measures. The results were mapped on cortical surfaces using BrainNet (https://www.nitrc.org/projects/bnv). Bottom, Scatter plot of the spatial correlations across voxels between CBF and fALFF values in an HC and a PD patient before and after treatment. C, Box plot distributions of global CBF-fALFF coupling in the HCs, PD patients before and after treatment. The solid lines represent the median, boxes represent lower and upper quartiles, and whiskers the minimum and maximum. Patients with PD who underwent treatment exhibited reduced CBF–fALFF coupling (p = 0.020). before, before treatment; after, after treatment; *p < 0.05; **p < 0.01; ***p < 0.001.
Spatial distributions of CBF/fALFF ratios
As shown in Figure 1B, despite subtle differences, PD patients at baseline and after treatment and healthy controls exhibited a highly similar spatial distribution of CBF, fALFF, and CBF/fALFF ratios. Specifically, high CBF was mainly distributed over the posterior cingulate gyrus/precuneus and the lateral prefrontal cortex. High fALFF values were mostly located in the lateral prefrontal cortex and motor cortex. Brain regions with a high CBF/fALFF ratio were observed in the medial prefrontal cortex and the lateral temporal cortex, whereas low ratios were observed in the motor and sensorimotor cortex. These patterns were largely consistent with the findings in healthy participants (Liang et al., 2013).
Global NVC changes after treatment
Representative global CBF–fALFF coupling maps from a patient before and after treatment are shown in Figure 1B. Although high spatial correlations across voxels between CBF and fALFF values were found in HCs (r = 0.9502 ± 0.011), PD patients at baseline (r = 0.9525 ± 0.012) and after treatment (r = 0.9455 ± 0.009), we found reduced CBF–fALFF coupling after treatment (t(30) = 2.246; 95% CI, 0.001–0.013; p = 0.020; paired t test; Fig. 1C). No significant changes were found between HCs and PD patients at baseline or after treatment (Fig. 1C).
Regional NVC changes after treatment
A paired t test based on whole-brain voxel analyses was performed to compare regional CBF, fALFF, and CBF/fALFF ratios in PD patients before and after treatment. Then ANCOVA was used to explore the CBF/fALFF ratio differences mentioned above between HCs and PD patients at baseline and after treatment.
After treatment, patients exhibited increased CBF in the left lingual gyrus (MNI coordinates, −18, −86, −4; t = 4.47) as well as increased perfusion in the bilateral superior frontal gyrus (dorsolateral; left, MNI coordinates, −12, 58, 20; t = −4.63; right, MNI coordinates, 18, 62, 10; t = −4.89) and the left superior frontal gyrus (medial; MNI coordinates, −6, 70, 16; t = −5.05; voxel p < 0.001; cluster p < 0.05; paired t test; GRF corrected; Fig. 2A; Table 2). Patients had significantly decreased fALFF in the left calcarine fissure (MNI coordinates, −4, −82, 10; t = 6.71), the right superior occipital gyrus (MNI coordinates, 26, −88, 32; t = 5.43), and the bilateral angular gyri (left, MNI coordinates, −44, −68, 44; t = 7.45; right, MNI coordinates, 52, −60, 44; t = 8.69). In contrast, the regions with significantly increased fALFF were mainly located in the right superior temporal gyrus (MNI coordinates, 62, −16, 20; t = −4.81), the median cingulate and paracingulate gyri (MNI coordinates, 16, −42, 34; t = −4.85), and the bilateral thalamus (left, MNI coordinates, −10, −28, 8; t = −7.48; right, MNI coordinates, 4, −40, −10; t = −6.84; voxel p < 0.001; cluster p < 0.05; paired t test; GRF corrected; Fig. 2B; Table 3).
Brain regions with significant differences in CBF after treatment
Brain regions with significant differences in fALFF after treatment
Significant alterations in the CBF, fALFF, and CBF/fALFF ratios before and after treatment (GRF corrected, voxel p value < 0.001, cluster p value < 0.05). A–C, Significant alterations in the CBF values, fALFF values, and CBF/fALFF ratios, respectively. Lingual_L, left lingual gyrus; Frontal_Sup_Medial_L, left superior frontal gyrus (medial); Frontal_Sup_L, left superior frontal gyrus (dorsolateral); Frontal_Sup_R, right superior frontal gyrus (dorsolateral); Calcarine_L, left calcarine fissure and surrounding cortex; Occipital_Sup_R, right superior occipital gyrus; Angular_L, left angular gyrus; Angular_R, right angular gyrus; Temporal_Sup_R, right superior temporal gyrus; Cingulum_Mid_R, right median cingulate and paracingulate gyri; Thalamus_L, left thalamus; Thalamus_R, right thalamus; Cingulum_Mid_L, left median cingulate and paracingulate gyri; Parietal_Inf_R, right inferior parietal with supramarginal and angular gyri; before, before treatment; after, after treatment.
The CBF/fALFF ratio was used to quantify the blood flow consumption per unit of neural activity and represented the regional NVC. After treatment, patients showed increased ratio values in the left angular gyrus (MNI coordinates, −44, −68, 44; t = −8.30) and the right inferior parietal with supramarginal and angular gyrus (MNI coordinates, 50, −60, 44; t = −6.71), as well as reduced values in the right superior frontal gyrus (dorsolateral; MNI coordinates, 20, −2, 52; t = 5.72), the median cingulate and paracingulate gyri (MNI coordinates, −8, −30, 26; t = 5.35), the right superior temporal gyrus (MNI coordinates, 62, −14, 22; t = 5.45), and several subcortical regions, such as the bilateral thalamus (left, MNI coordinates, −10, −28, 6; t = 5.72; right, MNI coordinates, 22, −26, 18, t = 6.62; voxel p < 0.001; cluster p < 0.05; paired t test; GRF corrected; Fig. 2C; Table 4).
Brain regions with significant differences in CBF/fALFF ratios after treatment
To determine whether these ratio changes reflected a restoration of the normal pattern or a compensatory effect, the ratio was extracted from the 6-mm-radius kernel size sphere within clusters showing significant treatment-related CBF/fALFF ratio differences among HCs and PD patients at baseline and after treatment (Fig. 3; FDR corrected). Compared with HCs, regions mainly located in the right superior frontal gyrus (dorsolateral; HC, 0.806 ± 0.119; before, 0.925 ± 0.139; after, 0.818 ± 0.151; HC vs before, F(1,55) = 14.24; p = 0.001; ANCOVA; η2 = 0.206; HC vs after, F(1,55) = 0.351; p = 0.649; ANCOVA; η2 = 0.006), the right superior temporal gyrus (HC, 0.969 ± 0.121; before, 1.137 ± 0.124; after, 1.017 ± 0.135; HC vs before, F(1,55) = 26.34; p < 0.001; ANCOVA; η2 = 0.324; HC vs after, F(1,55) = 2.234; p = 0.141; ANCOVA; η2 = 0.328), and the right thalamus (HC, 0.362 ± 0.061; before, 0.418 ± 0.092; after, 0.343 ± 0.077; HC vs before, F(1,55) = 6.79; p = 0.028; ANCOVA; η2 = 0.110; HC vs after. F(1,55) = 0.826; p = 0.514; ANCOVA; η2 = 0.015) showed significant increased ratio in PD patients at baseline but no differences in patients after treatment, indicating a restoration of the normal pattern in these regions (Fig. 3A). Furthermore, the left angular exhibited significant ratio changes between HCs and PD patients at baseline (HC, 0.973 ± 0.164; before, 0.881 ± 0.099; HC vs before, F(1,55) = 6.252; p = 0.027; ANCOVA; η2 = 0.102), as well as between HCs and PD patients after treatment (HC, 0.973 ± 0.164; after, 1.077 ± 0.181; HC vs after, F(1,55) = 7.59; p = 0.05; ANCOVA; η2 = 0.121), indicating the compensatory effect of treatment in this region (Fig. 3B). No differences were found in the left thalamus, cingulate gyrus, or the right inferior parietal with supramarginal and angular gyri between HCs and PD patients, indicating that treatment-related effects cannot be attributed to PD alone (Fig. 3C).
MIRT-induced alterations in the CBF/fALFF ratios across PD patients before and after treatment and HCs. A, B, The restoration of the normal pattern (A) in the right superior frontal gyrus (dorsolateral), the right superior temporal gyrus, and the right thalamus, as well as the compensatory effect (B) in the left angular gyrus. C, Treatment-related effects cannot be attributed to PD alone in the left thalamus, cingulate gyrus, or the right inferior parietal with supramarginal and angular gyri. The solid lines represent the median, boxes represent lower and upper quartiles, and whiskers the minimum and maximum. Frontal_Sup_R, right superior frontal gyrus (dorsolateral); Cingulum_Mid_L, left median cingulate and paracingulate gyri; Temporal_Sup_R, right superior temporal gyrus; Thalamus_L, left thalamus; Thalamus_R, right thalamus; Angular_L, left angular gyrus; Parietal_Inf_R, right inferior parietal with supramarginal and angular gyri; before, before treatment; after, after treatment; *p < 0.05; **p < 0.01; ***p < 0.001.
Correlation analysis between regional NVC and clinical variables
To determine whether treatment-associated alterations in the CBF/fALFF ratio were related to motor relevance, we conducted a correlation analysis on alterations in CBF/fALFF ratios and changes in UPDRS-III scores with age, age squared, sex, and education as covariates. After treatment, the percentage of change in the CBF/fALFF ratio in the right superior frontal gyrus (dorsolateral) was correlated with the change in UPDRS-III scores (Fig. 4A; r(25) = 0.39; 95% CI, −0.154–0.742; permutation test; p = 0.02; paired t test; uncorrected). No such correlation was found in this region between the CBF and UPDRS-III (Fig. 4B) or between the fALFF and UPDRS-III (Fig. 4C).
Correlation analysis between brain function and clinical variables. A, The percentage of change in the CBF/fALFF ratio in the right superior frontal gyrus (dorsolateral) was positively correlated with that of the UPDRS-III (r(25) = 0.39; p = 0.02; permutation test, uncorrected). B, C, No such correlation was found in the right superior frontal gyrus (dorsolateral) between the CBF and UPDRS-III (B) or between the fALFF and UPDRS-III (C). The translucent bands around the regression line represent the 95% confidence interval for the regression estimate. The contour lines show the kernel density estimations between brain function (CBF/fALFF ratio, CBF, fALFF) and clinical variables (UPDRS-III). Frontal_Sup_R, right superior frontal gyrus (dorsolateral); before, before treatment; after, after treatment.
Effective connectivity changes after treatment
We further used DCM analysis to explore the effective connectivity among the treatment-related brain regions found above in the regional CBF/fALFF ratio analysis (Figure 5A). We only displayed the suprathreshold effective connectivity related to treatment (posterior probability > 0.95). The treatment-related changes in effective connectivity are depicted in Figures 5B. The group mean effective connectivity in the HCs and PD patients before and after treatment are depicted in Figure 5, C–E. Specifically, the effective connectivity became more prominent from the right-side superior frontal but reduced connectivity from the bilateral thalamus to the left angular and right inferior parietal gyrus. Furthermore, we found increased connectivity from the right inferior parietal gyrus to the right thalamus. The inhibition effect of self-connections in the right thalamus was decreased after treatment. Changes in effective connectivity survived at the 95% threshold set for the posterior probability of the estimated parameters (Bencivenga et al., 2021; Jamieson et al., 2022).
Effective connectivity of DCM analysis. We displayed the suprathreshold effective connectivity (posterior probability > 0.95) related to treatment. A, Regions shown to be involved in DCM analysis and the corresponding MNI coordinates in millimeters used for the DCM analysis. B, Connections showing a significant association with treatment (before vs after). Purple, before < after; green, before > after. C–E, Group mean effective connectivity in the HCs (C), PD patients before (D), and after treatment (E). Arrows have been weighted to indicate the relative size of the effects. Frontal_Sup_R, right superior frontal gyrus (dorsolateral); Angular_R, right angular gyrus; Parietal_Inf_R, right inferior parietal gyrus with supramarginal and angular gyri; Thalamus_L, left thalamus; Thalamus_R, right thalamus; before, before treatment; after, after treatment.
Discussion
Rehabilitation is considered a complementary, effective treatment for PD. To the best of our knowledge, this is the first study to investigate MIRT-induced neurovascular coupling alterations by assessing CBF–fALFF correlations and CBF/fALFF ratios derived from ASL and BOLD-fMRI data. Patients showed reduced global CBF–fALFF correlations after treatment. The dual mechanism of altered regional CBF/fALFF ratios exhibited a decrease in regions responsible for executive control and an increase in regions related to information integration. Treatment-related decreases in the CBF/fALFF ratios in the right dorsolateral superior frontal gyrus were correlated with motor improvement (UPDRS-III scores). DCM further confirmed the effective connectivity among the aforementioned regions and identified the dorsolateral superior frontal gyrus as a critical treatment target in PD.
Reduced global NVC after treatment
The global CBF–fALFF correlation reflects comprehensive changes of NVC at the global level. In a healthy brain, NVC ensures a proportionate CBF response to neural activity (Liang et al., 2013; Baller et al., 2022), whereas abnormalities may lead to cerebral pathologies and neurologic disorders (Zhu et al., 2017; Kim et al., 2020). High across-voxel correlations between CBF and fALFF values were observed in patients before and after treatment as well as in healthy controls, but these values were lower after treatment. PD is characterized by inflammation and neurodegeneration that chronically affect brain tissue, which undergoes constant restructuring to compensate for damage (Baglio et al., 2021; Bloem et al., 2021). Increased activity in the primary motor cortex (Wu et al., 2011) and frontoparietal networks (Maidan et al., 2016) has been described as a compensatory mechanism to overcome motor difficulties in PD. Thus, MIRT-induced rehabilitation may promote brain reorganization by partially restoring connectivity within the corticostriatal network and normalizing hyperactivation (Filippi et al., 2018; Baglio et al., 2021), resulting in a reduction in coupling strength after treatment. The lack of differences observed between PD and healthy controls may because of the interaction between disruption caused by pathology and compensatory network activation to preserve global motor function (Maidan et al., 2016; Baglio et al., 2021).
Dual mechanisms of rehabilitation treatment
The CBF/fALFF ratio was used to reflect regional NVC and quantified the amount of blood flow or metabolic energy per unit of functional activity. The effective rehabilitation was demonstrated through a dual mechanism; in the initial phase of motor learning, regions related to executive control exhibited a decreased CBF/fALFF ratio to conserve energy, whereas in the latter phase, regions important for information integration exhibited higher values (and thus higher energy requirements) to maintain brain activity. This mechanism was further supported by the effective connectivity analysis as there was more connectivity from the dorsolateral superior frontal gyrus to the inferior parietal gyrus and less connectivity from the thalamus to the inferior parietal gyrus. These results provide theoretical support from a multimodal perspective for the dual-mechanism theory of motor rehabilitation, in which motor networks are restored and compensatory networks are activated (Baglio et al., 2021).
Partial restoration in executive control after treatment
The dorsolateral superior frontal gyrus and the thalamus, components of the corticobasal ganglia–thalamocortical circuits, exhibited partial restoration after treatment and were much more similar to HCs than those at baseline. The prefrontal cortex is highly interconnected to virtually all other subordinate cortical and subcortical structures (L. Chen et al., 2021; Liu et al., 2021; Luo et al., 2021) and controls executive functions (such as focused attention, cognitive flexibility, and rule-based regulation of behavior; Ferrazzoli et al., 2018; Ceccarini et al., 2019). The thalamus is involved in projecting information to specific motor, premotor, and prefrontal cortical areas (Ji et al., 2018). The prefrontal cortex works in concert with the thalamus to execute planned, motivated behaviors. Direct correlations have been observed between prefrontal metabolism and thalamic innervation (Orso et al., 2021), and disruption within parallel loops connecting these regions (Chen et al., 2015) is thought to be central to many executive impairments in PD (Ferrazzoli et al., 2018; Hill et al., 2020). Previous studies have confirmed that the frontal-subthalamic circuit is an important contributor to effective treatments for PD (Borchert et al., 2016; Rae et al., 2016; Habets et al., 2018; Orso et al., 2021), which is in line with our findings of treatment-related restorations in the prefrontal cortex and thalamus. The rehabilitation program refers to motor learning and requires high-level cognitive control (L. Chen et al., 2021), which is critically involved in the early phases of motor learning and executive control (Petzinger et al., 2013; Silveira et al., 2018). The benefits of rehabilitation may be mediated by enhancing both perfusion and functional activity in the prefrontal cortex (Silveira et al., 2018; Voss et al., 2013), resulting in energy perserve with the reduction energetic consumption required per unit to complete the early phases of motor learning. Our results indicate that the frontal-subcortical circuit related to motor and executive control was partially restored.
Compensatory effect on information integration after treatment
Another main finding was a compensatory effect in the inferior parietal gyrus, including the angular gyrus (Deng et al., 2021), as reflected by a greater CBF/fALFF ratio and received effective connectivity from other regions. The inferior parietal gyrus integrates inputs from different sensorimotor modalities with motor signals to facilitate sensory guidance of movements and motor planning (Fogassi and Luppino, 2005; Wu et al., 2011; Wang et al., 2017) and is involved in the latter phase of motor learning (Baglio et al., 2021). Compared with healthy controls, several studies found that increased activation in the inferior parietal gyrus was related to a compensatory mechanism for impaired motor function (Sang et al., 2015; Borchert et al., 2016). Thus, the reduced effective connectivity from the thalamus to this region may be because of restoration of the parietal–thalamus pathway facilitated by rehabilitation, supported by previous results of reduced connectivity after treatment (Ballarini et al., 2018; Droby et al., 2020). Furthermore, consistent with the hypothesis of recruitment in compensatory networks to sustain motor activity (Agosta et al., 2017; Baglio et al., 2021), we found treatment-related increases in connectivity from the frontal to parietal cortices. Thus, we believe that parietal–thalamus connectivity restoration and frontal–parietal connectivity compensation may result in increased NVC in the inferior parietal cortex and may be associated with the compensatory mechanism needed to overcome reduced information integration in the depleted sensorimotor circuit in PD patients as well as the retention of rehabilitation-related benefits in the latter phases of motor learning.
The absent alteration in the striatum is most likely because treatment-related changes involved strengthening of attentional and sensorimotor compensatory circuits rather than increased efficiency of the posterior putamen (Nackaerts et al., 2019).
Relationship between alterations in regional NVC and UPDRS-III scores
Treatment-related CBF/fALFF decreases in the right dorsolateral superior frontal gyrus were correlated with motor improvement (UPDRS-III scores). We speculate that increased activity of the dorsolateral superior frontal cortex (Baglio et al., 2021) reflects the restoration of cognitive processes related to motor learning that underlie complex motor behaviors (Ceccarini et al., 2019; L. Chen et al., 2021). Thus, the reduced energetic demand per unit area found in this region may be associated with better motor control and executive performance, indicating a better response to rehabilitation; moreover, this decrease in demand may be related to the partial restoration of normal activity patterns in the early and middle stages of PD. In addition, this region is one of the stimulation sites for noninvasive brain stimulation as it has therapeutic potential for alleviating motor symptoms associated with UPDRS-III scores (Lee et al., 2014; Agarwal et al., 2018; Zhang et al., 2022). Overall, the treatment-related NVC pattern in the right dorsolateral superior frontal gyrus offers evidence as a treatment-related target for PD.
There were several limitations in the study. The present study did not include the analysis of the HC group before and after the training because of applicability or ethics. The HC groups performed MRI assessments only once without any MIRT in hospital to provide normative ranges of the CBF and fALFF; thus, our analysis was only concerned with regions with significant treatment-related differences of neurovascular coupling in PD and compared values among HCs and PDs within these regions to determine whether the changes resulted in restoration or compensation of brain function. The sample of participants was small, which limits the generalizability and statistical power of the results; thus, the findings need to be validated with a larger dataset. No assessments of nonmotor symptoms were collected before and after treatment, and more detailed neuropsychological assessments are needed to explore the rehabilitation-related NVC mechanisms underlying nonmotor symptoms in the future.
In conclusion, we combined ASL and BOLD-fMRI techniques to explore how MIRT affects neurovascular coupling. We found evidence of a dual mechanism of rehabilitation with restoration of executive control in the initial phase of motor learning and compensation of information integration in the latter phase, and identified the right dorsolateral superior frontal gyrus as a potential target for PD. Overall, by elucidating rehabilitation-induced alterations, we can obtain a better understanding of their role in improving motor performance.
Footnotes
This work was supported by STI 2030—Major Projects Grant 2022ZD0208500; National Natural Science Foundation of China Grants U20A20191, 61727807, 82071912, and 12104049; Beijing Municipal Science and Technology Commission Grant Z201100007720009; National Key Research and Development Program of China Grants 2020YFC2007305, and 2022YFC3602603; the Science and Technology Development Fund of Beijing Rehabilitation Hospital, Capital Medical University Grants 2020-069, and 2021-011. We thank the participants and clinical doctors at the Parkinson Medical Center of Beijing Rehabilitation Hospital, Capital Medical University.
The authors declare no competing financial interests.
- Correspondence should be addressed to Tianyi Yan at yantianyi{at}bit.edu.cn or Boyan Fang at fangboyanv{at}ccmu.edu.cn
