MRI-based Differentiation of Parkinson's Disease by Cerebellar Gray Matter Volume

doi:10.21203/rs.3.rs-4307066/v1

Download PDF

Research Article

MRI-based Differentiation of Parkinson's Disease by Cerebellar Gray Matter Volume

https://doi.org/10.21203/rs.3.rs-4307066/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

The underlying mechanism of Parkinson's disease is associated with the neurodegeneration of the basal ganglia, the cerebellum plays a significant role together with the basal ganglia in non-motor and motor functions. Morphological changes in the cerebellum can have a great impact on patients' clinical symptoms, especially on their motor control symptoms, and may also help distinguish patients from healthy subjects. This study aimed to explore the potential of cerebellar gray matter volume, related to motor control function, as a neuroimaging biomarker to classify patients with Parkinson's disease (PD) and healthy controls (HC) by using voxel-based morphometric (VBM) measurements and support vector machine (SVM) methods based on independent component analysis (ICA).

Methods

Cerebellar gray matter volume was measured by using VBM in patients with PD (n = 27) and HC (n = 16) from the Neurocon dataset. ICA analysis was performed on the gray matter volume of all subregions, resulting in 7 independent components. These independent components were then utilized for correlation analysis with clinical scales and trained as input features for the SVM model. PD patients (n = 20) and HC (n = 20) from the TaoWu dataset were used as test data to validate our SVM model.

Results

Among patients with PD, 3 out of the 7 independent components showed significant correlation with clinical scales. The SVM model achieved an accuracy of 86% in classifying PD patients and HC, with a sensitivity of 72.2%, specificity of 88%, and F1 Score of 76.5%. The accuracy of the SVM model verification analysis used the TaoWu dataset is 70%, the sensitivity 62.5%, the specificity 100%, and the F1 Score 76.9%.

Conclusions

The results suggest that abnormal cerebellar gray matter volume, which is highly correlated with motor control function in Parkinson's patients, may serve as a valuable neuroimaging biomarker which is capable of distinguishing Parkinson's patients from healthy individuals. We observed that the combination of the ICA method and SVM method produced an improved classification model. This model may function as an early warning tool which enables the clinicians to conduct preliminary identification and intervention for patients with PD.

Parkinson's Disease

Cerebellum

Voxel-based Morphometry

Independent Component Analysis

Support Vector Machine

Parkinson's disease (PD) is the second most common multifactorial neurodegenerative disease after Alzheimer's disease (AD). The increasing aging of the global population has caused its prevalence to increase year by year, with the incidence higher than any other age-related neurological disease [1]. The number of individuals afflicted with PD was 6 million in 2015, and it is predicted to double by 2040 [2]. As a complicated progressive neurodegenerative disease, patients with PD have a wide range of premotor, non-motor and motor symptoms at different stages of this disease [3]. Progressive impairment involving voluntary motor control is the major feature of this disease, impairment of voluntary motor control manifests through multiple symptoms and signs including rigidity, hypokinesia, akinesia, postural instability, stooped posture, and tremor at rest. These motor abnormalities are often accompanied with poor coordination and balance, gait impairment, and bilateral vocal cord paralysis in severe cases [4, 5]. These are the main features for clinical diagnosis of Parkinson's Disease and for assessing disease progression and treatment response. In addition to the motor signs of PD, there are also a wide range of non-motor sign abnormalities, which may cause great distress and burden to PD patients, including but not limited to dream enactment, cognitive dysfunction, autonomic nervous system dysfunction, executive dysfunction, and emotional disorders such as anxiety and depression [6-9].

Although the etiology of PD remains partially unknown and clinical diagnosis of this condition during its early pre-motor stages is challenging, various non-motor and motor symptoms have been found to be associated with changes at different levels of PD patient's brains [10]. One of the primary pathological hallmarks of Parkinson's Disease is the progressive neurodegeneration of dopaminergic neurons in the substantia nigra pars compacta (SNc) [11]. The substantia nigra compacta is one of the important nuclei that comprises the basal ganglia. In anatomy, the substantia nigra compacta is divided into the dorsal pars compacta and the ventral pars reticulata. The former consists of dopaminergic neurons which are important for the transmission of dopamine to the basal ganglia, as well as to the striatum, which is highly involved in movement planning and execution [12, 13]. The depletion and degeneration of these neurons will result in dysfunction in neuronal circuitry, primarily affect the motor cortex areas and basal ganglia, and ultimately it will lead to abnormal voluntary movements at the individual's behavioral level [14]. The basal ganglia are considered as a highly organized network, different parts are activated for specific functions under different conditions. They are the primary regions responsible for processing signals from the cerebral cortex related to the execution of voluntary movements and cognitive functions [15, 16]. A study based on causal structural covariance networks (CaSCN) demonstrated that gray matter (GM) atrophy progresses from the basal ganglia to the hippocampus, temporal regions. Eventually, as PD advances, the GM atrophy spreads to other brain regions through the cortico-cortical networks, and the early selective vulnerability of basal ganglia could make a great contribution in modulating late-onset non-motor and motor circuit disorder [15]. A meta-analysis based on neuroimaging studies revealed that activation in the basal ganglia (putamen and pallidum) in PD patients decreased, along with a trend of reduced activity in the mirror system. This reduction in activation might be associated with the disruption of cognitive resonance mechanisms, potentially causes impairments in perception of others' emotions and behaviors [17]. A voxel-based morphometry (VBM) longitudinal study in multiple system atrophy revealed that shorter disease duration in PD is connected with progressive atrophy in the striatum, while longer disease duration correlates with increased atrophy in the cerebellum, which suggests that early degeneration of the basal ganglia might lead to subsequent cortical atrophy in the cerebellum [18]. Although it is previously believed that they are isolated from each other with different functions, increasing evidence has suggested that the cerebellum is anatomically connected to the basal ganglia [19, 20]. Another research showed that the connection between the subthalamic nucleus and cerebellum is implicated in cerebellar functions and basal ganglia can integrate in both motor and non-motor domains. The basal ganglia and cerebellum have substantial communications between each other and are interconnected to form an integrated network [21]. The cerebellum receives dopaminergic projections from the basal ganglia, and dopamine receptors are also found to be presented in the cerebellum. Consequently, degeneration of these dopaminergic neurons will lead to alterations in cerebellar activation [22]. The discovery of this reciprocal connection between basal ganglia and cerebellum provides an anatomical foundation to elucidate the important role of the cerebellum in PD. Similar to the basal ganglia, the subcortical system of the cerebellum influences cortical activity through the thalamus [23]. It was traditionally considered to relate to pure motor control. However it has been confirmed crucial for the development of "forward models," which involve predicting the sensory consequences of motor actions [24]. In addition to its primary function in motor control, the cerebellum also plays a significant role in cognitive, sleep, emotional recognition and processes [25-27]. As an important contributor to the overall pathophysiology of PD, the cerebellum is sometimes neglected. The role of the basal ganglia in PD has received extensive research and attention. However, our knowledge on the role of the cerebellum in PD remains limited [19], more attention should be paid to the cerebellum.

Resting-state functional magnetic resonance imaging (R-fMRI) has provided an effective means to investigate intrinsic spontaneous brain activity without the need for external task engagement, it has been widely applied in many neurodegenerative diseases, such as Parkinson's disease and Alzheimer's disease [28-30]. A meta-analysis of fMRI research on cognitively impaired PD patients indicates that cognitive dysfunction in PD is related to decreased connectivity within cognitive-related networks, with the most prominent effect observed in the default mode network (DMN) [31]. Another fMRI meta-analysis revealed a series of brain perfusion and regional functional abnormalities in PD, which are predominantly centered around brain regions related to the striato-thalamo-cortical circuits, are associated with the clinical manifestations of patients with PD [32]. MRI-based assessment of brain gray matter volume is considered as an effective method to diagnose PD and to evaluate its progression. Voxel-based morphometry analysis (VBM) provides an effective approach that allows for a fair and comprehensive assessment on the anatomical differences and changes across the entire brain. It has been well recognized and widely applied in PD [33]. A voxel-based meta-analysis on 2867 patients with PD and 1990 healthy controls (HC) detected significant gray matter volume (GMV) abnormalities in patients with PD. These aberrant brain regions including the basal ganglia, visual network, and auditory network, have provided morphological evidence for the pathophysiology of PD [34]. A recent study which combined VBM and surface-based morphometry (SBM) in PD patients with mild cognitive impairments (MCI) showed reduced gray matter volume in the frontal cortex, extending to the cerebellum, and cortical thinning in the temporal lobes, extending to the parietal cortex [35]. Further substantiating the association between PD and alterations in brain structure, morphometric analysis focused on subcortical and cortical regions may serve as important biomarkers for disease progression in PD patients.

Recently, an increasing number of studies have focused on exploring the underlying pathophysiological mechanisms of PD by using pattern recognition and machine learning methods [36]. As a supervised machine learning method used for multivariate pattern recognition, the Support Vector Machine (SVM) method utilizes statistical learning theory and the principle of structural risk minimization to find the optimal separating hyperplane in the feature space of data samples, maximizing the margin between the hyperplane and different types of samples [37]. In the field of psychiatry and neuroimaging, SVM method is able to address challenges such as high dimensionality and local minima, and stands out among many classification algorithms due to its excellent performance and usability [38, 39]. Multiple studies have demonstrated the significant value of SVM method in exploring early detection and subtyping classification of Parkinson's disease. Compared to the random forest method, SVM method can better integrate multiple indicators such as GMV and resting-state functional connectivity (RSFC) to predict PD. These findings demonstrate the potential to support radiological diagnosis and to achieve high classification accuracy in clinical diagnostic systems for patients with PD [40]. A rs- fMRI research revealed aberrant dynamic brain activity in the left precuneus of PD patients. By using an SVM classifier, PD patients can be accurately distinguished from HC based on the variability of dynamic amplitude of low-frequency fluctuations (dALFF) [41]. SVM classification model is also used in another study recently to distinguish PD patients from HC and accurately classify PD patients into different subtypes (the tremor-dominant subtype and the postural instability gait difficulty subtype), with an accuracy rate of 89.47% [42]. In addition, Independent Component Analysis (ICA), as a crucial blind signal separation and data-driven technique, provides an effective method for analyzing neuroimaging data. It can be combined with machine learning models to improve their performance and is widely utilized in various psychiatric disorders and neurodegenerative diseases including PD [43, 44].

Currently, there is no research combining SVM method with cerebellar GMV to classify PD patients. In this study, we utilized VBM analysis to explore cerebellar GMV in PD patients, using it as feature input to train the multivariate machine learning classification model. We hypothesized that the SVM model combined with ICA method would demonstrate good performance and generalization ability, allowing cerebellar GMV independent components to serve as potential neuroimaging biomarkers in distinguishing PD patients and HC.

2.1. Participants and MRI data Acquisition

The MRI data used in this study is derived from two publicly accessible datasets: the Neurocon dataset and the TaoWu dataset [45]. The Neurocon dataset comprises rs-fMRI and T1-weighted structural images from 27 PD patients and 16 age-matched HC. The scanner is Siemens AVanto 1.5 T, resting-state scanning was conducted by using the echo-planar imaging sequence (EPI), TE = 50 ms, TR = 3480 ms, Flip angle = 90°, Voxel size = 3.8 × 3.8 × 5 mm³ (without slice gaps), Scan time = 8.05 min, volume = 137. T1-weighted scanning was conducted using the magnetization prepared-rapid gradient echo (MPRAGE) sequence, TE = 3.08 ms, TR = 1940 ms, Voxel size = 0.97 × 0.97 ×1 mm³. According to the ethical standards set forth in the 1964 Helsinki Declaration and its subsequent amendments, the Neurocon dataset has obtained approval from the Ethics Committee of the Emergency Hospital of Bucharest University. All participants have provided written informed consent to participate in this study. The TaoWu dataset comprises rs-fMRI and T1-weighted structural images from 20 patients PD with and 20 age- and gender-matched HC. The scanner is Siemens Magnetom 3 T, resting-state scanning was conducted by using the EPI sequence, TE = 40 ms, TR = 2000 ms,, Flip angle = 90°, Voxel size = 4 × 4 × 5 mm³ (64 x 64 matrix, 28 slices, field of view = 256 mm × 256 mm), Scan time = 8 min, volume = 239. T1-weighted scanning was conducted by using the MPRAGE sequence, TR = 1100 ms, TE = 3.39 ms, Voxel size = 1 × 1× 1 mm³. For more detailed information regarding MRI acquisition, please refer to the official Parkinson's Disease Datasets website (Parkinson's Disease Datasets (nitrc.org)). The datasets also include demographic information of all participants, such as gender, age, and cognitive scores. All PD patients are in the early or moderate stages of PD symptoms, with the majority receiving treatment (most commonly treated under levodopa therapy) [45].

2.2. Voxel-based morphometric analysis

We utilized the CAT12 toolbox (http://www.neuro.uni-jena.de/cat) based on the SPM12 software (http://www.fil.ion.ucl.ac.uk/spm/software/spm12) to pre-process the T1-weighted cerebellum structural images [46]. CAT12 toolbox is a widely adopted program that contains a variety of templates to perform voxel-based morphometry (VBM) analysis. We employed SUIT template to automatically separate and segment gray matter (GM) and white matter (WM) of the cerebellum [47, 48]. Cerebellar GM images were inspected visually for any misclassification of brain tissue, and we corrected them manually if errors were present. Afterwards, the segmented WM images and GM images of all subjects were then used to create the study-specific template by using DARTEL toolbox (Diffeomorphic Anatomical Registration using Exponentiated Lie algebra) [49]. The SUIT template was utilized to achieve normalization of the standard space using modulation, and the voxel size of normalized images was 1.5×1.5×1.5 mm³. It improved the alignment of the fissures and reduced the spatial spread in comparison to the Montreal Neurological Institute (MNI) template [50]. Lastly, the cerebellar GM tissue segments underwent smoothing by using an 8-mm full width at half maximum (FWHM) isotropic Gaussian kernel. We acquired the modulated and smoothed cerebellar GM images of each subject to be used in further analysis. To exclude the influence of total intracranial volume (TIV) on statistics, we calculated the relative GMV obtained by dividing the absolute volume by the TIV including the cerebrospinal fluid volume (CFV), the relative GMV of 28 cerebellar subregions in SUIT template were obtained and then used in subsequent analysis.

2.3. Statistical analysis

Chi-square tests and two-tailed two-sample t-tests were used to compare the clinical and demographic data between HC and PD patients in the Neurocon dataset and TaoWu dataset. We used the Fast ICA method to analyse the relative volumes of cerebellum subregions in the Neurocon dataset, and identified 28 cerebellar subregions as 7 distinct independent cortical network components as input to train our machine learning classifier [51, 52]. Partial correlation analysis with controlled covariates were utilized to examine the relationship between the 7 independent components and Unified Parkinson’s Disease Rating Scale motor assessment scores (UPDRS motor on state and off state).

One of the widely-used supervised classification algorithms in the realm of rs-fMRI is the Support Vector Machine (SVM) [53]. This algorithm aims to discover a hyperplane within the feature space that segregates positive and negative samples effectively, maximizing the margin between them while minimizing misclassification errors. Because our classification task was to divide all participants into the PD patient group and the HC group, we leveraged the LIBSVM package (http://www.csie.ntu.edu.tw/~cjlin/libsvm/) and custom code developed on the MATLAB platform (The MathWorks, Inc., Natick, MA, USA). Before training the SVM model, we optimized the parameters of the model, which had been proved to have a significant impact on the accuracy, efficiency and performance of the classifier[54]. The main parameters in SVM are the cost (C) and gamma (g). The parameter C is called the penalty coefficient, which controls the trade-off between maximizing the margin and minimizing the classification error. A higher value of C will make the model more prone to overfitting. And g is a parameter for choosing the radial basis function (RBF) kernel, which influences the speed of both the training and prediction processes. There are many ways to optimize parameters. We chose the grid search method, changing values of the two parameters within a certain range during the model training process. We used k-fold cross-validation to assess how changes in these parameters affected the classification accuracy of the SVM model [55]. Finally, we compared the results to determine the optimal parameters for training the SVM model. Figure 2 shows the SVM parameter selection results.

After selecting the optimal parameters C and gamma, the 7 independent components derived from the grey matter volume of the cerebellum were used as feature inputs to train the SVM classifier. Multiple metrics, such as sensitivity, specificity, and the F1 score, were used to assess the performance of the SVM classifier. We utilized leave-one-out cross-validation (LOOCV) and data from an external dataset, the TaoWu dataset, as the validation dataset to test the generalization ability of the SVM model.

3.1. Clinical and demographic characteristics

Demographic and clinical characteristics of the Neurocon dataset and TaoWu dataset were shown in Table 1 and Table 2. In the Neurocon dataset, there were no significant differences in age (P = 0.759, two-tailed two-sample t-test) and TIV (P = 0.06, two-tailed two-sample t-test) between HC and PD patients, while significant differences were found in sex (P = 0.027, chi-square test). In the TaoWu dataset, no significant differences were found in sex (P = 0.99, chi-square test), age (P = 0.78, two-tailed two-sample t-test) and TIV (P = 0.47, two-tailed two-sample t-test) between patients with PD and HC.

3.2 Correlation analysis

Figure 1 shows the correlation analysis results between the independent components and UPDRS motor scores (on state and off state). Significant partial correlations were detected between the second component and UPDRS motor off state score (r = 0.43, p = 0.05), the fourth component and UPDRS motor on state score (r = 0.41, p = 0.05), the seventh component and UPDRS motor on state score (r = -0.4, p = 0.05) as well as UPDRS motor off state score (r =-0.38, p = 0.05).

-------------------------------- Insert Table 1 here -------------------------------

Table 1. Clinical and demographic data of Neurocon dataset

Characteristics

Neurocon PD

Neurocon HC

P-value

Sex (Male/Female)

Age (years)

TIV (mm³)

UPDRS motor on

UPDRS motor off

27(17/10)

68.7 ± 10.6

1537.8 ± 164.2

9.22 ± 5.26

28.33 ± 9.27

16(4/12)

67.6 ± 11.9

1450.4 ± 105.2

.027^a

.759^b

.06^b

The data were presented as the mean ± standard deviation.

^a The p-value was obtained by a chi-square test;

^b The p-value was obtained by a two-tailed two-sample t-test.

Abbreviations: PD, Parkinson's disease; HC, healthy controls; TIV, total intracranial volume; UPDRS motor on, Unified Parkinson’s Disease Rating Scale motor assessment during on state; UPDRS motor off, Unified Parkinson’s Disease Rating Scale assessment during off state.

-------------------------------- Insert Table 2 here -------------------------------

Table 2. Clinical and demographic data of TaoWu dataset

Characteristics

TaoWu PD

TaoWu HC

P-value

Sex (Male/Female)

Age (years)

TIV (mm³)

20(11/9)

65.2 ± 4.4

1445.2 ± 131.2

20(12/8)

64.8 ± 5.6

1474.8 ±122.0

.99^a

.78^b

.47^b

^a The p-value was obtained by a chi-square test;

^b The p-value was obtained by a two-tailed two-sample t-test.

Abbreviations: PD, Parkinson's disease; HC, healthy controls.

3.2 SVM classifier

The seven independent components based on grey matter volume of cerebellar subregions were used as features to train the SVM model. The best parameter c (2)

To the best of our knowledge, this study is the first to utilize the SVM classification model combined with ICA algorithm, using cerebellar gray matter volume to differentiate between PD patients and HC. The results found that the independent components of GMV in patients with PD were significantly correlated to the patients' motor control scale scores and using the independent components as input features for the SVM classifier resulted in strong model performance. Furthermore, the validation analysis proved the generalization ability of the model.

Multiple MRI studies have confirmed abnormalities in the function and the structure of cerebellum in PD and its subtypes. William et al. used the cerebellar vermis as the seed point and revealed increased functional connectivity between cerebellar nodes of the somatomotor network (SMN) and corresponding cortical nodes compared to HC. Additionally, cerebellar connectivity between DMN and frontoparietal was found to correlate with the severity of motor function, which highlights the alterations of cerebellar cortical FC in patients with PD and emphasizing the important role of cerebellar functional connectivity of cognitive and motor functions in PD [56]. One research conducted in PD patients on medication revealed an increase of ALFF in both the left and right cerebellum of all PD patients, as well as in the left cerebellum of newly diagnosed PD patients, but no significant changes of ALFF were found in medicated PD patients. All PD patients, including newly diagnosed ones, had significantly lower FC between the left medial frontal gyrus and the left cerebellum. However, medicated PD patients showed significantly higher FC between the right cerebellum and left precentral gyrus. These findings suggested that anti-PD medications could alter changes in local neural activity and FC in the cerebellum caused by Parkinson's disease, which provided new insights into the role of the cerebellum in early medication-treated PD [57]. A study examining various subtypes of PD such as tremor-predominant PD (PDT), akinetic/rigidity-predominant PD (PDAR), and four other affective subtypes revealed different degrees of GMV decrease in multiple subregions of the cerebellum. It suggested there was a close association between gray matter changes in the cerebellum and different symptoms of PD, and it supported the crucial role of the cerebellum in cognitive and emotional processes [58]. PD patients with MCI have been found to have gray matter reduction in multiple brain regions, particularly in the cerebellum. Even in early-stage PD patients with a short disease duration, reduced gray matter density in the cerebellum has been observed. This heterogeneous pattern of gray matter atrophy suggests that cerebellar changes differ from HC even in the early stages of Parkinson's Disease [59]. One recent study also found that PD patients with MCI displayed gray matter reduction compared to HC and PD patients without MCI. This reduction extended from the frontal lobe, which was closely associated with cognitive function, to the cerebellum. Furthermore, the subcortical cerebellum area could serve as a neural biomarker to distinguish between HC, PD patients with MCI and PD patients without MCI [35]. Previous studies have demonstrated the discriminative power of the cerebellum as a feature in SVM to classify PD patients and different subtypes. Zeng utilized cerebellar gray matter density in the SVM classification to distinguish patients with PD from HC at the individual level, achieving an accuracy of 95% through various cross-validation strategies [60]. Cherubini et al. successfully differentiated tremor-dominant Parkinson's disease (tPD) patients from PD patients with essential tremor and resting tremor with 100% accuracy by combining cerebellar gray and white matter volumes [61]. These studies, along with the findings of our current study, collectively proved the discriminative capability of cerebellar structural alterations in PD patients.

A study emphasized that the gray matter loss in multiple brain regions of PD patients, such as the parahippocampal gyrus, lingual gyrus, cingulate gyrus, and frontal regions was correlated with overall cognitive scores. However, in most of these regions, gray matter loss was unrelated to the patients' motor impairments [62]. These are consistent with our correlation analysis results. The cerebellum is responsible for autonomous movement, gait, posture, and coordination of motor functions, plays a crucial role in motor control in PD [63]. Despite research has demonstrated a direct impact of the cerebellar dentate nucleus network of PD on cognitive performance, it is also associated with motor performance, and partially mediated by motor abilities [64]. Wen et al. found that using repetitive transcranial magnetic stimulation (rTMS) to increase the fractional amplitude of low-frequency fluctuations (fALFF) in the left Crus II of the cerebellar hemisphere could improve motor symptoms in PD patients, which provided new evidence for the involvement of cerebellar activity in the motor response to rTMS therapy [65]. The cerebellum also makes an important contribution to motor inhibition, which is a fundamental skill for adaptive behavior, requiring higher-order functions of motor control and motor cognition. The contribution of the cerebellum to motor inhibition is related to patients' performance and cerebellar activation during motor inhibition tasks [66]. A behavioral study conducted a conflict task on patients with focal cerebellar lesions, resulting in lower performance compared to HC, and the patients showed even poorer performance when the lesions affected the deep nuclei of the cerebellum, which demonstrated the involvement of the cerebellum in the control and monitoring of voluntary behavior [67]. Neuroimaging studies have provided further evidence for the involvement of cerebellar structures in the inhibition of motor responses. Patients with multiple sclerosis (MC) showed increased errors in the go/no-go task and demonstrated significant neural compensation during response inhibition [68]. In another fMRI study, researchers measured cerebellar-cortical interactions during response inhibition over two days of task performance and examined the changes in interactions. The results showed enhanced performance on the second day, and the decrease in cerebellar interaction with the right inferior frontal cortex and the increase in interaction with the primary motor area were found [69]. All these findings indicated the significant role of the cerebellum in executive control and motor inhibition. One research on PD patients with postural instability and gait disorders (PD-PIGD) revealed a decline in both non-motor and motor areas of the cerebellum. In motor tasks, increased activity in the cognitive region of the cerebellum (Crus I) was found to be related to better motor performance in PD-PIGD patients. This suggested that increased activity in non-motor regions of the cerebellum in patients might be a result of cerebellar gray matter atrophy or an attempt to compensate for the dysfunction of the motor regions of the cerebellum and basal ganglia [70]. Yu detected excessive activation in both the right and left cerebellum and contralateral motor cortex of PD patients through experiments of automatic and cognitively controlled thumb pressing movements [71]. Sen et al. also observed a progressive recruitment of the cerebello-thalamo-cortical loop over time [72]. A meta-analysis revealed that 60% of functional neuroimaging literature reported cerebellar overactivation in PD, encompassing motor paradigms with significant cognitive task demands. Additionally, negative correlation between motor scale scores and activation in cerebellar subregions associated with motor functions was found, as the severity of the disease increases, the extent of involvement of these regions decreases [73]. Excessive activation of the cerebellum and motor cortex in PD patients has been confirmed by multiple studies. Currently, it is widely believed that this abnormal activation pattern is compensatory, aiming to offset functional abnormalities or deficits in brain regions such as the basal ganglia by altering the structure or physiological response of compensatory brain areas [71, 74]. It has suggested that this compensatory mechanism may be mediated through the cerebello-thalamo-cortical loop. PD patients maximize recruitment of normal motor networks when the task is relatively easy. However, as the task becomes more challenging, compensatory activity in the cerebello-thalamo-cortical network increases steadily [75]. This compensatory effect may contribute to maintaining better non-motor and motor functions in patients with PD, deserves further exploration in future study.

This study still has some limitations that need to be considered. First, although we utilized an external dataset to validate and assess the generalization ability of our classification model, both the sample size of the training set and the validation set were relatively small, which may pose a risk of reduced statistical power and affect the stability of the results. Second, this study was cross-sectional in nature and lacked longitudinal data on PD patients. Future studies should consider incorporating longitudinal data over longer timeframes to better explore the complicated physiological and pathological mechanisms. Last but not least, special attention should be paid to the heterogeneity among participants, especially regarding the different treatments and medications received by PD patients, which may have varying impacts on their brain function or structure. Therefore, our results should be interpreted with caution, and future studies should aim to include consistent treatment regimens or first-episode drug-naïve PD patients.

In conclusion, this study demonstrates that the cerebellar GMV independent component in PD patients is highly correlated with their motor control function and may serve as a potential neuroimaging biomarker to distinguish PD patients from HC. This paper highlights the role of the cerebellum as a biomarker in PD, providing new evidence and insights in understanding the pathophysiological mechanisms of PD and aiding in clinical diagnosis and identification.

Acknowledgements

Not applicable.

Authors’ contributions

Dacong Zhao: conception and study design, data curation, fMRI data formal analysis. Chao Tian: conception and study design, validation, drafting the manuscript work. Jiang Guo: methodology. Guanghua Lu: visualization. Rui Jiang: review and edit the manuscript. Xu Liang: supervision and critical revision. All authors contributed to the article and approved the submitted version.

Funding

None

Availability of data and materials

Public access to the databases is open. The Neurocon dataset and TaoWu dataset can be found at: https://fcon_1000.projects.nitrc.org/indi/retro/parkinsons.html. The code that supports the findings of this study are available from the corresponding author upon reasonable request.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing of interest

No conflicts of interest, financial or otherwise, are declared by the authors.

Dorsey ER, Bloem BR: The Parkinson Pandemic-A Call to Action. JAMA neurology 2018, 75(1):9-10.
Dorsey ER, Sherer T, Okun MS, Bloem BR: The Emerging Evidence of the Parkinson Pandemic. Journal of Parkinson's disease 2018, 8(s1):S3-s8.
Lu H, Li J, Zhang L, Meng L, Ning Y, Jiang T: Pinpointing the precise stimulation targets for brain rehabilitation in early-stage Parkinson's disease. BMC neuroscience 2023, 24(1):24.
Lotankar S, Prabhavalkar KS, Bhatt LK: Biomarkers for Parkinson's Disease: Recent Advancement. Neuroscience bulletin 2017, 33(5):585-597.
Mazzoni P, Shabbott B, Cortés JC: Motor control abnormalities in Parkinson's disease. Cold Spring Harbor perspectives in medicine 2012, 2(6):a009282.
Zhang W, Xia S, Tang X, Zhang X, Liang D, Wang Y: Topological analysis of functional connectivity in Parkinson's disease. Frontiers in neuroscience 2023, 17:1236128.
Marsh L: Depression and Parkinson's disease: current knowledge. Current neurology and neuroscience reports 2013, 13(12):409.
Roheger M, Kalbe E, Liepelt-Scarfone I: Progression of Cognitive Decline in Parkinson's Disease. Journal of Parkinson's disease 2018, 8(2):183-193.
Reich SG, Savitt JM: Parkinson's Disease. The Medical clinics of North America 2019, 103(2):337-350.
Saeed U, Compagnone J, Aviv RI, Strafella AP, Black SE, Lang AE, Masellis M: Imaging biomarkers in Parkinson's disease and Parkinsonian syndromes: current and emerging concepts. Translational neurodegeneration 2017, 6:8.
Chinta SJ, Andersen JK: Dopaminergic neurons. The international journal of biochemistry & cell biology 2005, 37(5):942-946.
Lehéricy S, Bardinet E, Poupon C, Vidailhet M, François C: 7 Tesla magnetic resonance imaging: a closer look at substantia nigra anatomy in Parkinson's disease. Movement disorders : official journal of the Movement Disorder Society 2014, 29(13):1574-1581.
Kumar S, Goyal L, Singh S: Tremor and Rigidity in Patients with Parkinson's Disease: Emphasis on Epidemiology, Pathophysiology and Contributing Factors. CNS & neurological disorders drug targets 2022, 21(7):596-609.
Lew M: Overview of Parkinson's disease. Pharmacotherapy 2007, 27(12 Pt 2):155s-160s.
Li R, Zou T, Wang X, Wang H, Hu X, Xie F, Meng L, Chen H: Basal ganglia atrophy-associated causal structural network degeneration in Parkinson's disease. Human brain mapping 2022, 43(3):1145-1156.
DeLong MR, Wichmann T: Basal Ganglia Circuits as Targets for Neuromodulation in Parkinson Disease. JAMA neurology 2015, 72(11):1354-1360.
Arioli M, Cattaneo Z, Rusconi ML, Blandini F, Tettamanti M: Action and emotion perception in Parkinson's disease: A neuroimaging meta-analysis. NeuroImage Clinical 2022, 35:103031.
Brenneis C, Egger K, Scherfler C, Seppi K, Schocke M, Poewe W, Wenning GK: Progression of brain atrophy in multiple system atrophy. A longitudinal VBM study. Journal of neurology 2007, 254(2):191-196.
Wu T, Hallett M: The cerebellum in Parkinson's disease. Brain : a journal of neurology 2013, 136(Pt 3):696-709.
Hoshi E, Tremblay L, Féger J, Carras PL, Strick PL: The cerebellum communicates with the basal ganglia. Nature neuroscience 2005, 8(11):1491-1493.
Bostan AC, Dum RP, Strick PL: The basal ganglia communicate with the cerebellum. Proceedings of the National Academy of Sciences of the United States of America 2010, 107(18):8452-8456.
Prakash KG, Bannur BM, Chavan MD, Saniya K, Sailesh KS, Rajagopalan A: Neuroanatomical changes in Parkinson's disease in relation to cognition: An update. Journal of advanced pharmaceutical technology & research 2016, 7(4):123-126.
Lewis MM, Galley S, Johnson S, Stevenson J, Huang X, McKeown MJ: The role of the cerebellum in the pathophysiology of Parkinson's disease. The Canadian journal of neurological sciences Le journal canadien des sciences neurologiques 2013, 40(3):299-306.
Jueptner M, Weiller C: A review of differences between basal ganglia and cerebellar control of movements as revealed by functional imaging studies. Brain : a journal of neurology 1998, 121 ( Pt 8):1437-1449.
Xu W, De Carvalho F, Clarke AK, Jackson A: Communication from the cerebellum to the neocortex during sleep spindles. Progress in neurobiology 2021, 199:101940.
Buckner RL: The cerebellum and cognitive function: 25 years of insight from anatomy and neuroimaging. Neuron 2013, 80(3):807-815.
D'Agata F, Orsi L: Cerebellum and Emotion Recognition. Advances in experimental medicine and biology 2022, 1378:41-51.
Biswal B, Yetkin FZ, Haughton VM, Hyde JS: Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magnetic resonance in medicine 1995, 34(4):537-541.
Xue SW, Kuai C, Xiao Y, Zhao L, Lan Z: Abnormal Dynamic Functional Connectivity of the Left Rostral Hippocampus in Predicting Antidepressant Efficacy in Major Depressive Disorder. Psychiatry investigation 2022, 19(7):562-569.
Meijer FJ, Goraj B: Brain MRI in Parkinson's disease. Frontiers in bioscience (Elite edition) 2014, 6(2):360-369.
Wolters AF, van de Weijer SCF, Leentjens AFG, Duits AA, Jacobs HIL, Kuijf ML: Resting-state fMRI in Parkinson's disease patients with cognitive impairment: A meta-analysis. Parkinsonism & related disorders 2019, 62:16-27.
Xie H, Yang Y, Sun Q, Li ZY, Ni MH, Chen ZH, Li SN, Dai P, Cui YY, Cao XY et al: Abnormalities of cerebral blood flow and the regional brain function in Parkinson's disease: a systematic review and multimodal neuroimaging meta-analysis. Frontiers in neurology 2023, 14:1289934.
Nemoto K: [Understanding Voxel-Based Morphometry]. Brain and nerve = Shinkei kenkyu no shinpo 2017, 69(5):505-511.
Xu X, Han Q, Lin J, Wang L, Wu F, Shang H: Grey matter abnormalities in Parkinson's disease: a voxel-wise meta-analysis. European journal of neurology 2020, 27(4):653-659.
Li L, Ji B, Zhao T, Cui X, Chen J, Wang Z: The structural changes of gray matter in Parkinson disease patients with mild cognitive impairments. PloS one 2022, 17(7):e0269787.
Khosla M, Jamison K, Ngo GH, Kuceyeski A, Sabuncu MR: Machine learning in resting-state fMRI analysis. Magnetic resonance imaging 2019, 64:101-121.
Zhou S: Sparse SVM for Sufficient Data Reduction. IEEE transactions on pattern analysis and machine intelligence 2022, 44(9):5560-5571.
Hosseini MP, Hosseini A, Ahi K: A Review on Machine Learning for EEG Signal Processing in Bioengineering. IEEE reviews in biomedical engineering 2021, 14:204-218.
Huang S, Cai N, Pacheco PP, Narrandes S, Wang Y, Xu W: Applications of Support Vector Machine (SVM) Learning in Cancer Genomics. Cancer genomics & proteomics 2018, 15(1):41-51.
Cao X, Wang X, Xue C, Zhang S, Huang Q, Liu W: A Radiomics Approach to Predicting Parkinson's Disease by Incorporating Whole-Brain Functional Activity and Gray Matter Structure. Frontiers in neuroscience 2020, 14:751.
Zhang C, Dou B, Wang J, Xu K, Zhang H, Sami MU, Hu C, Rong Y, Xiao Q, Chen N et al: Dynamic Alterations of Spontaneous Neural Activity in Parkinson's Disease: A Resting-State fMRI Study. Frontiers in neurology 2019, 10:1052.
Xiong J, Zhu H, Li X, Hao S, Zhang Y, Wang Z, Xi Q: Auto-Classification of Parkinson's Disease with Different Motor Subtypes Using Arterial Spin Labelling MRI Based on Machine Learning. Brain sciences 2023, 13(11).
Guo W, Su Q, Yao D, Jiang J, Zhang J, Zhang Z, Yu L, Zhai J, Xiao C: Decreased regional activity of default-mode network in unaffected siblings of schizophrenia patients at rest. European neuropsychopharmacology : the journal of the European College of Neuropsychopharmacology 2014, 24(4):545-552.
Yang W, Lui RL, Gao JH, Chan TF, Yau ST, Sperling RA, Huang X: Independent component analysis-based classification of Alzheimer's disease MRI data. Journal of Alzheimer's disease : JAD 2011, 24(4):775-783.
Badea L, Onu M, Wu T, Roceanu A, Bajenaru O: Exploring the reproducibility of functional connectivity alterations in Parkinson's disease. PloS one 2017, 12(11):e0188196.
Gaser C, Dahnke R, Thompson PM, Kurth F, Luders E, Initiative AsDN: CAT – A Computational Anatomy Toolbox for the Analysis of Structural MRI Data. 2022:2022.2006.2011.495736.
Diedrichsen J, Balsters JH, Flavell J, Cussans E, Ramnani N: A probabilistic MR atlas of the human cerebellum. NeuroImage 2009, 46(1):39-46.
Diedrichsen J, Maderwald S, Küper M, Thürling M, Rabe K, Gizewski ER, Ladd ME, Timmann D: Imaging the deep cerebellar nuclei: a probabilistic atlas and normalization procedure. NeuroImage 2011, 54(3):1786-1794.
Klein A, Andersson J, Ardekani BA, Ashburner J, Avants B, Chiang MC, Christensen GE, Collins DL, Gee J, Hellier P et al: Evaluation of 14 nonlinear deformation algorithms applied to human brain MRI registration. NeuroImage 2009, 46(3):786-802.
Diedrichsen J: A spatially unbiased atlas template of the human cerebellum. NeuroImage 2006, 33(1):127-138.
Yeo BT, Krienen FM, Sepulcre J, Sabuncu MR, Lashkari D, Hollinshead M, Roffman JL, Smoller JW, Zöllei L, Polimeni JR et al: The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of neurophysiology 2011, 106(3):1125-1165.
Soldati N, Calhoun VD, Bruzzone L, Jovicich J: ICA analysis of fMRI with real-time constraints: an evaluation of fast detection performance as function of algorithms, parameters and a priori conditions. Frontiers in human neuroscience 2013, 7:19.
Santana CP, de Carvalho EA, Rodrigues ID, Bastos GS, de Souza AD, de Brito LL: rs-fMRI and machine learning for ASD diagnosis: a systematic review and meta-analysis. Scientific reports 2022, 12(1):6030.
Matheny ME, Resnic FS, Arora N, Ohno-Machado L: Effects of SVM parameter optimization on discrimination and calibration for post-procedural PCI mortality. Journal of biomedical informatics 2007, 40(6):688-697.
Poldrack RA, Huckins G, Varoquaux G: Establishment of Best Practices for Evidence for Prediction: A Review. JAMA psychiatry 2020, 77(5):534-540.
Palmer WC, Cholerton BA, Zabetian CP, Montine TJ, Grabowski TJ, Rane S: Resting-State Cerebello-Cortical Dysfunction in Parkinson's Disease. Frontiers in neurology 2020, 11:594213.
Xu S, He XW, Zhao R, Chen W, Qin Z, Zhang J, Ban S, Li GF, Shi YH, Hu Y et al: Cerebellar functional abnormalities in early stage drug-naïve and medicated Parkinson's disease. Journal of neurology 2019, 266(7):1578-1587.
Ma X, Su W, Li S, Li C, Wang R, Chen M, Chen H: Cerebellar atrophy in different subtypes of Parkinson's disease. Journal of the neurological sciences 2018, 392:105-112.
Donzuso G, Monastero R, Cicero CE, Luca A, Mostile G, Giuliano L, Baschi R, Caccamo M, Gagliardo C, Palmucci S et al: Neuroanatomical changes in early Parkinson's disease with mild cognitive impairment: a VBM study; the Parkinson's Disease Cognitive Impairment Study (PaCoS). Neurological sciences : official journal of the Italian Neurological Society and of the Italian Society of Clinical Neurophysiology 2021, 42(9):3723-3731.
Zeng LL, Xie L, Shen H, Luo Z, Fang P, Hou Y, Tang B, Wu T, Hu D: Differentiating Patients with Parkinson's Disease from Normal Controls Using Gray Matter in the Cerebellum. Cerebellum (London, England) 2017, 16(1):151-157.
Cherubini A, Nisticó R, Novellino F, Salsone M, Nigro S, Donzuso G, Quattrone A: Magnetic resonance support vector machine discriminates essential tremor with rest tremor from tremor-dominant Parkinson disease. Movement disorders : official journal of the Movement Disorder Society 2014, 29(9):1216-1219.
Melzer TR, Watts R, MacAskill MR, Pitcher TL, Livingston L, Keenan RJ, Dalrymple-Alford JC, Anderson TJ: Grey matter atrophy in cognitively impaired Parkinson's disease. Journal of neurology, neurosurgery, and psychiatry 2012, 83(2):188-194.
Li T, Le W, Jankovic J: Linking the cerebellum to Parkinson disease: an update. Nature reviews Neurology 2023, 19(11):645-654.
Kawabata K, Watanabe H, Bagarinao E, Ohdake R, Hara K, Ogura A, Masuda M, Kato T, Tsuboi T, Maesawa S et al: Cerebello-basal ganglia connectivity fingerprints related to motor/cognitive performance in Parkinson's disease. Parkinsonism & related disorders 2020, 80:21-27.
Wen X, Chi S, Yu Y, Wang G, Zhang X, Wang Z, Gesang M, Luo B: The Cerebellum is Involved in Motor Improvements After Repetitive Transcranial Magnetic Stimulation in Parkinson's Disease Patients. Neuroscience 2022, 499:1-11.
Picazio S, Koch G: Is motor inhibition mediated byd cerebello-cortical interactions? Cerebellum (London, England) 2015, 14(1):47-49.
Brunamonti E, Chiricozzi FR, Clausi S, Olivito G, Giusti MA, Molinari M, Ferraina S, Leggio M: Cerebellar damage impairs executive control and monitoring of movement generation. PloS one 2014, 9(1):e85997.
Smith AM, Walker LA, Freedman MS, DeMeulemeester C, Hogan MJ, Cameron I: fMRI investigation of disinhibition in cognitively impaired patients with multiple sclerosis. Journal of the neurological sciences 2009, 281(1-2):58-63.
Hirose S, Jimura K, Kunimatsu A, Abe O, Ohtomo K, Miyashita Y, Konishi S: Changes in cerebro-cerebellar interaction during response inhibition after performance improvement. NeuroImage 2014, 99:142-148.
Gardoni A, Agosta F, Sarasso E, Basaia S, Canu E, Leocadi M, Castelnovo V, Tettamanti A, Volontè MA, Filippi M: Cerebellar alterations in Parkinson's disease with postural instability and gait disorders. Journal of neurology 2023, 270(3):1735-1744.
Yu H, Sternad D, Corcos DM, Vaillancourt DE: Role of hyperactive cerebellum and motor cortex in Parkinson's disease. NeuroImage 2007, 35(1):222-233.
Sen S, Kawaguchi A, Truong Y, Lewis MM, Huang X: Dynamic changes in cerebello-thalamo-cortical motor circuitry during progression of Parkinson's disease. Neuroscience 2010, 166(2):712-719.
Solstrand Dahlberg L, Lungu O, Doyon J: Cerebellar Contribution to Motor and Non-motor Functions in Parkinson's Disease: A Meta-Analysis of fMRI Findings. Frontiers in neurology 2020, 11:127.
Wu T, Hallett M: A functional MRI study of automatic movements in patients with Parkinson's disease. Brain : a journal of neurology 2005, 128(Pt 10):2250-2259.
Palmer SJ, Ng B, Abugharbieh R, Eigenraam L, McKeown MJ: Motor reserve and novel area recruitment: amplitude and spatial characteristics of compensation in Parkinson's disease. The European journal of neuroscience 2009, 29(11):2187-2196.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

MRI-based Differentiation of Parkinson's Disease by Cerebellar Gray Matter Volume

Status:

Version 1

Abstract

Figures

1 Introduction

2 Materials and Methods

2.1. Participants and MRI data Acquisition

2.2. Voxel-based morphometric analysis

2.3. Statistical analysis

3 Results

3.1. Clinical and demographic characteristics

3.2 Correlation analysis

3.2 SVM classifier

4 Discussion

5 Conclusion

Declarations

References

Additional Declarations

Status:

Version 1