Compensatory Brainstem Functional and Structural Connectivity in Patients with Degenerative Cervical Myelopathy by Probabilistic Tractography and Functional MRI

Abstract

Degenerative cervical myelopathy (DCM) is the most common cause of spinal cord impairment in adults. Previous supraspinal investigations have primarily focused on cortical changes in this patient population. As the nexus between the brain and the spinal cord, the brainstem has been understudied in patients with DCM. The current study examined the structural and functional connectivity between the brainstem and cortex in DCM patients using probabilistic tractography and resting-state functional MRI. A total of 26 study patients and 32 neurologically intact, healthy volunteers (HCs) participated in this prospective analysis. The study cohort included DCM patients (n=18), as well as neurologically asymptomatic patients with evidence of cervical spine degenerative changes and spinal cord compression (n=8). Results of the study demonstrated significant differences in fiber density (FD), fiber cross-section (FDC), and the functional connectivity (FC) between the study cohort and HCs. Through seeding the brainstem, the study cohort showed reductions in FD and FDC along the corticospinal tract, including regions extending through the corona radiata and internal capsule. By correlating FD and FDC with the Neck Disability Index (NDI), and the modified Japanese Orthopaedic Association (mJOA), we identified increasing total volume of projections to the thalamus, basal ganglia, and internal capsule, and increased functional connectivity to visual network and the posterior parietal cortices. These results support our hypothesis that DCM patients tend to have long-term FC reorganization not only localized to sensorimotor regions, but also to regulatory and visual processing regions, designed to ultimately preserve neurological function.

1. Introduction

Degenerative cervical myelopathy (DCM) is the most common cause of spinal cord impairment in adults, and results from spinal cord injury secondary to the progressive degeneration of cervical osseous and soft tissue structural components (Kalsi-Ryan et al., 2013; Tracy and Bartleson, 2010). The classic myelopathic symptoms include gait dysfunction, hand incoordination, upper extremity dysesthesias, and bladder disturbance, and these findings are often used to evaluate the severity of DCM. Early surgical intervention is often advocated as prolonged duration of symptoms may be associated with poorer outcomes (Holly et al., 2009). Previous studies have also evaluated the association between T2 and T1 weighted signal changes of the spinal cord, and the relationship between these findings with baseline severity and neurological outcome (Nouri et al., 2017). However, limitations still exist in our ability to monitor the progression of this disorder and predict the potential for recovery by conventional imaging modalities.

Magnetic resonance imaging (MRI) has been critical in the diagnosis and management of patients with cervical degenerative changes, including those with DCM and neurologically asymptomatic patients with evidence of spinal cord compression, and recent enhancements have allowed for increased sensitivity to detect spinal cord injury induced neural tissue changes. Quantifiable metrics obtained from different MR imaging modalities have been used to reveal both spinal and cerebral changes in DCM patients. Cervical spinal cord compression induces upstream alteration in both white and gray brain matter, manifested in part by cortical representational changes and regions of atrophy (Bernabeu-Sanz et al., 2019; Holly et al., 2019; Sun et al., 2017; Woodworth et al., 2018; Yoon et al., 2013; Zdunczyk et al., 2018). Resting-state functional MRI (rs-fMRI) has revealed that cerebral connectivity is associated with altered sensorimotor function and neck disability in DCM patients (Holly et al., 2019; Woodworth et al., 2018). Further, the application of diffusion weighted imaging (DWI) has provided reliable biomarkers for detecting both cerebral and spinal white matter changes associated with DCM (Banaszek et al., 2014; Dong et al., 2008; Holly et al., 2007; Hrabalek et al., 2015; Hrabalek et al., 2018; Mamata et al., 2005). More recently, probabilistic tractography was proposed to facilitate fiber tract-specific comparisons (Raffelt et al., 2017). By characterizing the specific fiber population within a voxel, referring to fixel-based analysis (FBA) (Raffelt et al., 2015), total intra-axonal volume of white matter axons can be measured regardless of direction. This technique not only allows for estimation of differences in fiber density (FD), but also for cross-sectional fiber bundles (FDC), thus providing a combinational method to detect degeneration of white matter tracts.

The brainstem has been understudied in DCM patients, as neural plasticity research has primarily been focused on cerebral connectivity and the spinal cord microstructural reorganization in this patient population. Functioning as the nexus of the brain and the spinal cord, the brainstem plays a critical role in conveying sensorimotor signals. One of the major goals of this investigation was to elucidate how brainstem connectivity responds to changing cerebral and spinal networks in DCM patients, as this critical information is unknown, and has yet to be investigated.

In this present study, group connectometry analysis using probabilistic tractography was first performed on a cohort of 26 patients with spinal cord compression and 32 neurologically intact, healthy controls (HCs) to detect the cerebral white matter changes associated with spinal cord injury. Statistical differences in voxel-wise FD measurement were explored, and correlated with the degree of neurological impairment as measured by the modified Japanese Orthopedic Association (mJOA) scale, and neck disability as measured by the neck disability index (NDI). By seeding the brainstem, we also tested for the difference in FD and FDC in areas of the brain associated with integration of sensory information and pain modulation. Functional connectivity (FC) of the brainstem in the study cohort was further correlated with the mJOA and NDI scales. We hypothesized that the brainstem of the study cohort would demonstrate altered neurological connectivity depending on the severity of neurological impairment and neck disability. Such changes were not only localized to the primary sensorimotor network, but also in regions responsible for sensorimotor regulation and visual processing to compensate for insufficient input sensory information. We aimed to define a distinct set of imaging biomarkers based on brainstem neuronal connections that clinicians may use to successfully diagnose and treat DCM patients.

2. Results

2.1. Correlation of white matter with symptom scores

The mean mJOA score for the study cohort was 16.1 (range from 12 to 18). Of the 26 study patients with spinal cord compression, 8 were neurologically asymptomatic (mJOA=18), and 18 were symptomatic (mean mJOA = 15.3). High quality whole brain FD images were obtained from all 58 subjects that participated in the current study. To test the hypothesis that the extent of specific microstructural changes within the brain may correlate with degree of neurological impairment and neck disability, we explored the relationship between FD values and the mJOA and NDI scales respectively, in both the study cohort and HCs. The resulting correlation matrix identified several anatomical regions which may potentially serve as biomarkers in the stratification of symptom severity. As shown in Fig. 2A, FD values tended to show a positive correlation with mJOA score within regions related to the sensory and motor system. The cerebral and cerebellar peduncles, the posterior thalamic radiation, the corticospinal tract, along with pathways extending through the corona radiata and internal capsule, demonstrated higher fiber track density with higher mJOA score. When examining the association between whole brain FD values and the degree of neck disability, FD also tended to show a positive correlation with NDI score within pathways connecting to motor regions, specifically the bilateral anterior, posterior, and superior corona radiata within the study cohort and HCs, as illustrated in Fig. 2B. In addition, the corpus callosum, the internal capsule, and external capsule also exhibited higher FD values with large cluster volumes associated with higher NDI scores. Scatter plots of the representative correlation between symptoms and FD values from specific cerebral white matter tracts are illustrated in the right column of Fig. 2, where both the right corticospinal tract and the middle cerebellar peduncle show a positive linear correlation with mJOA with R² values of 0.2600 (p<0.0001) and 0.2249 (p=0.0002), respectively, and both the right anterior corona radiata and the right superior corona radiata show a positive linear correlation with mJOA with R² values of 0.2608 (p<0.0001) and 0.2729 (p<0.0001), respectively. All white matter fiber tracts, together with identified cluster volumes, are summarized in Table 2.

Fig. 2.

Anatomical localization of regions correlated with A) mJOA and B) NDI in FD measurement. Analysis was performed within the study cohort and HC’s by controlling subject ages. Representative correlations between clinical measures (mJOA and NDI) and cerebral fiber density. Significant clusters were determined by thresholding based on level of statistical significance (p < 0.05) and cluster-based corrections using permutation test.

Table 2.

Cluster volume [uL] of regions showing significant association of fiber density (FD) with modified Japanese Orthopaedic Association (mJOA) score and neck disability index (NDI) score within cervical spondylosis patients and healthy controls.

Anatomic Regions	Cluster Volume [uL] of Regions Showing Association of Fiber Density with
	mJOA	NDI
Lt. / Rt. Anterior Limb of Internal Capsule	- / 347	356 / 301
Lt. / Rt. Anterior Corona Radiata	-	778 / 862
Body of Corpus Callosum	-	32
Lt. / Rt. Cerebral Peduncle	337 / 1185	75 / 13
Lt. / Rt. Cingulum (Cingulate Gyrus)	-	79 / -
Lt. / Rt. Corticospinal Tract	496 / 714	-
Lt. / Rt. External Capsule	3 / 135	571 / 1596
Lt. / Rt. Fornix /Stria Terminalis	3 / 713	48 / 138
Genu of Corpus Callosum	-	1285
Lt. / Rt. Inferior Cerebellar Peduncle	194 / 361	-
Lt. / Rt. Medial Lemniscus	336 / 640	-
Middle Cerebellar Peduncle	8660	-
Pontine Crossing Tract	1001	-
Lt. / Rt. Posterior Limb of Internal Capsule	107 / 677	452 / 273
Lt. / Rt. Posterior Corona Radiata	55 / 193	481 / 728
Lt. / Rt. Posterior Thalamic Radiation	360 / -	252 / 23
Lt. / Rt. Retrolenticular Part of Internal Capsule	1150 / 4	1405 / 317
Lt. / Rt. Sagittal Stratum	14 / -	-
Splenium of Corpus Callosum	-	197
Lt. / Rt. Superior Cerebellar Peduncle	451 / 859	-
Lt. / Rt. Superior Corona Radiata	10 / -	247 / 477
Lt. / Rt. Superior Longitudinal Fasciculus	7 / 6	- / 199
Lt. / Rt. Tapetum	-	112 / 21
Rt. Uncinate Fasciculus	6	99 / 126

2.2. Probabilistic tractography analysis by seeding brainstem

By selecting the brainstem as the seed ROI, and the sensorimotor cortex as the target ROI, streamline segments associated with fixels that had a significant change (FEW-corrected P-values < 0.05) were identified for both FD (Fig. 3) and FDC (Fig. 4) in the study cohort compared to HCs. Connections of the study cohort showed a significant percentage decrease (~20%) in microstructural fiber density versus HCs. These connections start from the medulla, pass through the pons and midbrain along the corticospinal tract, to the primary motor and somatosensory cortex. With respect to the bundle cross-section, the FDC metric of the study cohort presented a larger widespread percentage of decrease compared to HCs in both microstructure and macrostructure. The internal capsule, basal ganglia, and white matter pathways extending through the corona radiata, exhibited a decrease in the study cohort in total intra-axonal volume occupied by fiber bundles. In particular, supratentorial streamline segments close to the pre- and postcentral gyrus exhibited a higher percentage decrease (> 40%).

Fig. 3.

Fixel-based analysis of fiber density (FD) showing regions and pathways of statistical significant difference in probabilistic tractography between the study cohort and HCs, with covariates including subject ages and mJOA. A) Tract-specific percentage changes in the study group compared to HCs. B) Sagittal view of brainstem with streamlines projected onto population template map. C) Axial view of brainstem with streamlines projected onto population template map. Streamline segments were cropped from the template tractogram to include only streamline points that correspond to significant fixels (FWE-corrected P-value <0.05). Streamlines were colored by absolute effect size, indicating percentage changes in the study group compared to the HCs for FD.

Fig. 4.

Fixel-based analysis of fiber density and fiber cross-section (FDC) showing regions and pathways of statistical difference in probabilistic tractography between the study cohort and , with covariates including subject ages and mJOA. A) Tract-specific percentage changes in the study cohort compared to HCs. B) Sagittal view of brainstem with streamlines projected onto population template map. C) Axial view of brainstem with streamlines projected onto population template map. Streamline segments were cropped from the template tractogram to include only streamline points that correspond to significant fixels (FWE-corrected P-value <0.05). Streamlines were colored by absolute effect size, indicating percentage changes in the study cohort compared to the HCs for FDC

2.3. ROI-to-ROI FC analysis by seeding the brainstem

Accounting for the differences in age, the occipital cortex of visual network, the left lateral, and the right lateral visual network exhibited increased connectivity to the subthalamic nucleus of brainstem with increased neurological impairment, with R² values of 0.4162 (p=0.0012), 0.4422 (p=0.0007), and 0.4023 (p=0.0015), respectively (Fig. 5). In addition, the correlation matrix associating FC with NDI scores revealed stratification of symptoms with other functional regions (Fig. 6). Increased connectivity between the bilateral posterior parietal cortices and the brainstem, including pons, the red nucleus, and medulla is associated with a higher NDI score, and increased connectivity between the medial frontal cortex and the subthalamic nucleus of brainstem is also associated with a higher NDI score. Moreover, the right caudate and the right cerebellum were identified as having increased functional connectivity with higher NDI score. Scatter plots of the representative association between FC and NDI scores are illustrated in the right column of Fig. 6, where the entire brainstem and the bilateral posterior parietal cortices show a positive linear correlation with NDI with R² values of 0.3131 (p=0.0068) and 0.3324 (p=0.0050), respectively, and the specific brainstem region, such as the pons, showed a positive linear correlation with NDI with R-square values of 0.1490 (p=0.0376) and 0.2020 (p=0.0174), respectively. Additional analysis within the study cohort by seeding the brainstem did not identify any microstructural effect of T2 weighted spinal cord signal change on fiber density, but compared to study patients with T2 weighted signal changes in the cord, study patients without T2 weighted spinal cord signal change showed stronger FC between the right putamen and the red nucleus of brainstem.

Fig. 5.

ROI-to-ROI functional connectivity (FC) association with modified Japanese Orthopedic Association (mJOA) score for the study cohort by seeding the brainstem. Colors denote the T-statistics, with yellow-red representing a positive association (increasing FC with improved neurological status), light aqua-blue representing a negative association (decreasing FC with improved neurological status). Positions of ROIs are displayed on mid-axial slices: r = Right Hemisphere; l = Left Hemisphere.

Fig. 6.

ROI-to-ROI functional connectivity (FC) association with Neck Disability Index (NDI) score for the study cohort by seeding the brainstem. Colors denote the T-statistics, with yellow-red representing a positive association (increasing FC with worsened neck disability), light aqua-blue representing a negative association (decreasing FC with worsened neck disability). Positions of ROIs are displayed on mid-axial slices: MedFC = Frontal Medial Cortex; Cereb = Cerebellum; PPC = Posterior Parietal Cortex; r = Right Hemisphere; l = Left Hemisphere.

3. Discussion

DCM is the leading cause of progressive spinal cord impairment and neurological deficit in elderly patients (Kalsi-Ryan et al., 2013; Tracy and Bartleson, 2010). While the vast majority of DCM research has focused on the spinal cord (Avadhani et al., 2010; Ellingson et al., 2014; Ellingson et al., 2015; Nouri et al., 2016; Nouri et al., 2017; Tetreault et al., 2013), recent interest has moved to the upstream microstructural and functional changes within the sensorimotor network in DCM patients (Aleksanderek et al., 2017; Dong et al., 2008; Holly et al., 2007; Holly et al., 2019; Peng et al., 2020; Takenaka et al., 2020; Tan et al., 2015; Woodworth et al., 2018). Although incompletely understood, the cerebral cortex reorganizes in response to spinal cord compression for the preservation of neurological function (Holly et al., 2007; Zdunczyk et al., 2018). Thus, the connectivity of the brainstem and its alterations with DCM is worthy of investigation.

The brainstem forms the connection between the brain and the spinal cord, playing a critical role in neural communication, especially for motor and sensory signals traveling up- and downstream between the brain and the rest of body. Studies have shown that the brainstem can function to both facilitate and inhibit the degree of nociceptive processing, thus affecting the level of pain a patient experiences (Mills et al., 2018; Moulton et al., 2010). In this current study, we applied probabilistic tractography and rs-fMRI to investigate neurological connectivity of the brainstem in DCM patients and neurologically asymptomatic spinal cord compression patients. We observed significant widespread differences in FD and FC in the study patients when compared to HCs, and by seeding the brainstem, and such alterations were not only limited to the primary sensory and motor regions, but were also found in the visual network and the posterior parietal cortices when associated with symptom severity. We propose potential biomarkers that may provide a non-invasive method to monitor the progression of neurological deterioration, ultimately impacting the treatment paradigm for DCM patients.

3.1. Altered structural connectivity of the brainstem: insight from probabilistic tractography

While DTI provides valuable information of white matter microstructure, its specificity decreases in pathological conditions and with increasing anatomical complexity. In order to better investigate fiber organization, we performed probabilistic tractography to obtain the structural connectivity index along white matter pathways, which is more sensitive to the white matter structure of pathological conditions. Generally, a decrease in FD indicates a decrease of the total intra-axonal volume of white matter axons in any particular direction, thus providing information of degeneration within specific white matter tracts (Raffelt et al., 2015; Raffelt et al., 2017; Tournier et al., 2007). After adjusting for age differences between the study cohort and HCs, the whole brain probabilistic tractography identified an overall positive correlation between FD values and mJOA score, suggesting increased structural connectivity for better neurological status. Such changes were observed with large cluster volumes in regions including the basal ganglia, internal capsule, cerebral and cerebellar peduncles, and fibers along the corticospinal tracts. By seeding the brainstem, FBA further confirmed our observation by identifying FD and cross-sectional reductions within specific white matter pathways in study patients compared to HCs, where FD and FDC metrics showed significant decreases of streamlines along the corticospinal tract, with regions close to the pre- and postcentral gyrus exhibiting a decrease of greater than 40%.

The brainstem, composed of the midbrain, pons, and medulla, links the brain with the spine, bridging connections from the cortex with the peripheral nervous system. By selecting the entire brainstem as the seed ROI, both the medulla and the midbrain were identified with a significant number of streamlines differences, predominately by FD and FDC. However, the pons was inconsistently identified by FDC, possibly due to limitations of streamline tractography with pontocerebellar fibers, nuclei, and tracts within the pons resulting in a large number of intravoxel crossing white matter tracts (Jeong et al., 2013).

Altered connectivity to cerebral motor regions is believed to be associated with a wide variety of chronic pain conditions, including fibromyalgia, neuropathic pain, and chronic back pain (Apkarian et al., 2004; Apkarian et al., 2011; Ellingson et al., 2013; Wand et al., 2011; Woodworth et al., 2015). We have also reported that worsening neck disability was associated with increased functional connectivity (Holly et al., 2019), particularly to the pre- and postcentral gyri, the superior frontal gyrus, and the supplementary motor area. Consistent with these results, we showed that the fiber density (FD) was positively correlated with NDI score within pathways connecting to motor regions. This may serve as an important biomarker in patients with chronic neck pain, implicating motor cortex involvement in both neurological symptoms and chronic pain.

Our results support the hypothesis that DCM causes significant structural changes, including brain atrophy, and functional reorganization that will be gradually adapted by the brain through cortical plasticity. By seeding the brainstem, we showed the progressive loss of axonal conduction along the corticospinal tract, suggesting a high degree of neurological connectivity changes in DCM patients in sensory and motor regions connected to the brainstem, emphasizing, that demyelination of the corticospinal tract fibers is the predominant pathological mechanism of DCM (Bernabeu-Sanz et al., 2019; Jang, 2014). However, chronic pain caused by cervical spondylosis can interrupt the normal transmission of neuronal signaling between the brain and the spinal cord, and such compensatory mechanisms can regulate motor movement and sensory processing. In particular, the reorganization occurs within regions of basal ganglia, thalamus, and primary sensorimotor regions, demonstrating an altered cortico-basal ganglia-thalamus-cortical loop in pain modulation (Borsook et al., 2010; Draganski et al., 2008; Herrero et al., 2002; Woodworth et al., 2015). Such observation is not only seen in other chronic pain syndromes, but also support our previous characterization of cortical plasticity in DCM patients using diffusion spectral imaging.

3.2. Associations between functional brainstem measurements and symptoms

Using rs-fMRI, the current study analyzed the association of brainstem functional connectivity with severity of DCM patients’ neurological impairment and neck disability. We identified increased FC between the brainstem and visual centers in the posterior parietal cortices with worsening neurological status and worsening neck disability. While this pattern may appear counterintuitive, given that DCM is typically known as dysfunction of the sensorimotor, it is conceivable that more input from the visual system is utilized in DCM patients to optimize body mechanics and balance, as the number of streamlines to primary somatosensory and motor regions has been demonstrated to decrease with more severe symptoms. Previous studies have reported similar findings highlighting connectivity differences to the visual cortex between DCM patients and HCs (Chen et al., 2018), suggesting compensation for insufficient sensory input to the proprioceptive system. This pattern was further confirmed by the identification of the superior longitudinal fasciculus and cingulum to be associated with higher NDI score by whole brain probabilistic tractography. Our results support the hypothesis that DCM patients tend to have long-term FC reorganizations not only localized to sensorimotor regions, but also to regions assisting in the regulation of movement and sensory perception, to preserve neurological function.

3.3. Clinical implications and limitations

Cortical alterations demonstrate a strong association with the degree of neurological impairment. By seeding the brainstem, FBA analysis identified significant reductions in structural connectivity along the corticospinal tract, and the loss of corticospinal tract fiber integrity may serve as a radiographic biomarker in the timing of surgical intervention. Although spinal cord injury leads to distal damage, ultimately causing cortical atrophy of primary motor and sensory cortices (Freund et al., 2011), our results provide evidence for cortical reorganization of the sensorimotor system to compensate for potential functional loss. We believe that neurological function is being preserved by increasing the total intra-axonal volume of projections to the thalamus, basal ganglia, and internal capsule. Information about cerebral functional connectivity with the brainstem provides new insights into cortical and spinal responses beyond the primary sensorimotor network. Future study will investigate whether these changes may be detected before symptoms become observable, potentially offering the promise of earlier diagnosis and intervention in patients with advanced cervical spondylosis. Although the statistical analysis of microstructural differences between the study and HC groups accounted for the variation in age of each patient, inclusion of additional older HC’s and additional younger study patients would be useful in making a further comparison in the future.

4. Experimental procedures

4.1. Patient population

The study cohort consisted of 26 subjects, including 18 DCM patients and 8 patients with cervical spine degenerative changes and spinal cord compression that were neurologically asymptomatic. These patients were prospectively enrolled in a cross-sectional study involving observational MRI and evaluation of neurological and neck disability from 2016–2018. The patients were recruited from an outpatient neurosurgery clinic, and each had spinal cord compression with evidence of spinal cord deformation, mass effect, and no visible cerebrospinal fluid signal around the spinal cord at the site of maximal compression on MRI. The neurologically asymptomatic patients initially presented with neck pain, were without radiculopathy, and had been referred for neurosurgical consultation for the diagnosis of cervical stenosis with spinal cord compression. All patients signed consent forms approved by the Institutional Review Board, and all analyses were performed in compliance with the Health Insurance Portability and Accountability Act (HIPAA). The study cohort had a mean age of 58.7 years (range from 35 to 82 years). The mJOA score was used as a measure of neurological function (Benzel et al., 1991), where lower values of mJOA represent worse neurological status. A cohort of 32 neurologically intact healthy control (HC) volunteers were included as a control group. The control group had a mean age of 32.0 years, with a range from 18 to 60 years. The patients and HCs demographic data is summarized in Table 1.

Table 1.

Cohort Demographics

Subject	Age (mean ± SD) [min, max]	Sex	Smoking	Subgroup	NDI (mean ± SD) [min, max]	mJOA (mean ± SD) [min, max]
Patients (N=26)	58.7 ± 12.9 [35, 82]	11 F	4 Smoking	DCM (N=18)	10.1 ± 8.2 [0, 28]	15.3 ± 1.5 [12, 17]
		15 M	22 Non-Smoking	Asymptomatic with CSD Changes (N=8)	7.5 ± 4.0 [2, 13]	18
HCs (N=32)	32.0 ± 12.2 [18, 60]	12 F	N/A	N/A	0	18
		20 M

DCM= Degenerative Cervical Myelopathy; HC = Healthy control; CSD=Cervical Spine Degenerative;

N= Number; SD = Standard deviation; F=Female; M=Male; N/A=Not Available;

mJOA = modified Japanese Orthopedic Association; NDI = Neck Disability Index

4.1.1. Past Medical History

Two of the patients were current tobacco smokers, and additional 4 had a past history of smoking. There was no history of illicit drug use. Two of the patients had Diabetes Mellitus (Table 1).

4.1.2. Presenting Symptoms

The most common presenting symptom was paresthesias or pain in the upper extremities, and was found in 22 patients. 16 patients noted a history of neck pain. 12 patients presented with deterioration of hand coordination. Gait dysfunction was encountered in 9 patients. 1 patient presented with recent changes in bladder function.

4.1.3. Physical Examination

6 patients were found to have weakness in the upper extremities on examination, and 1 had weakness in the lower extremities. 4 patients had decreased sensation in the upper extremities, and 1 had sensory changes in the lower extremities. Hyperreflexia was the most common upper motor neuron sign, and was elicited in 15 patients. Hoffman’s sign was the second most common long tract sign, and was observed in 12 patients. 5 patients had clonus in the lower extremities, and 2 had a positive Babinski reflex.

4.1.4. Radiographical Imaging

Standard MRI was obtained in all study patients, and revealed spinal cord compression in each case. The spinal canal narrowing was related to advanced cervical spondylosis manifested by a combination of facet arthropathy, ligamentum flavum hypertrophy, and varying degrees of ventral disc-osteophyte compression. 11 of the patients had a lordotic cervical spine, 11 had straight, and 4 had a kyphotic spinal alignment. 16 of the patients had T2-weighted signal abnormalities located within the spinal cord parenchyma and 10 were without signal changes. Ossification of the posterior longitudinal ligament was encountered in 4 patients.

4.2. Diffusion weighted imaging and resting-state fMRI acquisition

All diffusion-weighted images were collected using echo planar imaging on a Siemens Prisma 3T MR scanner (Siemens Healthcare, Erlangen, Germany) with a repetition time (TR) = 9500 – 11400 ms; echo time (TE) = 82 – 88 ms; flip angle 90°; field-of-view (FOV) of 256 x 256 mm with an acquisition matrix of 128 x 128 for a voxel size of 2 mm x 2 mm x 2 mm. Diffusion weighting was distributed along 65 directions using b value sets of 0 s/mm² and 1000 s/mm². All subject space data was resampled to 1 mm isotropic resolution for analyses in standard space.

All functional MR images were collected on a Siemens Prisma 3T MR scanner (Siemens Healthcare, Erlangen, Germany) with a repetition time (TR) = 2000 ms; echo time (TE) = 28 ms; slice thickness of 4 mm with no interslice gap; field-of-view (FOV) of 220 mm with an acquisition matrix of 64x64 for an in-plane resolution of 3.4 mm, interleaved acquisition; flip angle of 90°; parallel imaging via CAIPIRINHA with a factor of 4; and multi-band acceleration with a factor of 3.

Additionally, a 1 mm 3D isotropic MPRAGE sequence was acquired for alignment with functional MRI data using standard acquisition parameters (TR = 2300 ms, TE = 2000 ms, inversion time (TI) = 900 ms, flip angle 9 – 15°, FOV = 240 x 320 mm and matrix size of 240 x 320, slice thickness = 1 mm).

4.3. Image Processing

All diffusion weighted images were firstly denoised using the MRtrix3 software package (Brain Research Institute, Melbourne, Australia, http://www.brain.org.au/software/mrtrix) (Calamante et al., 2012). Then FMRIB’s Diffusion Toolbox (FDT) (www.fmrib.ox.ac.uk/fsl) was used to correct for eddy currents (which cause stretches and shears in images) and head motion, and images were affinely registered to the first no-diffusion weighted (b=0 s/mm²) image of each participant. Following skull extraction with BET, fixel-based analysis was performed via probabilistic tractography using the MRtrix3 software package. After seeding 1 million voxels randomly throughout brain, the number of fiber tracts passing through each image voxel was counted. Resulting fiber density images were associated with mJOA and NDI scales for correlation analysis. Additionally, we selected the brainstem as the seed region of interest (ROI), while target ROIs were selected from the Harvard-Oxford atlas based on studies that demonstrated cortical morphological and functional changes as a result of DCM (Holly et al., 2019; Woodworth et al., 2018; Wu et al., 2013). A total of 2 million individual pathways were generated throughout the entire brain, using a step length of 0.5 mm and a maximum of 2000 steps. The FD passing from any point in the seed ROI through any region of the target ROI was retained and compared between the study cohort and HC cohorts using statistical tests. To account for changes to both within-voxel fiber density and macroscopic atrophy, fiber density and fiber cross-section were combined to enable a more complete picture of group differences in white matter.

The processing of functional MR images was carried out using the CONN Toolbox (https://www.nitrc.org/projects/conn) (Whitfield-Gabrieli and Nieto-Castanon, 2012), which implements functions from the Statistic Parametric Mapping (SPM, http://www.fil.ion.ucl.ac.uk/spm/) toolbox. The default pipeline embedded in the CONN Toolbox was used to pre-process individual structural and functional files in the ways: functional realignment (motion correction, 12 degrees of freedom) and unwarping; slice-timing correction; registration of functional data to the structural volume; and registration of the structural volume to the standardized space defined by the Montreal Neurological Institute (MNI) in averaged T1 brain. Segmentation of structural volumes, which included skull stripping and processing of tissue types (grey matter, white matter, and cerebrospinal fluid), were then performed. Artifacts Detection Tool (ART), a SPM package implemented in the CONN pipeline, was used to remove signal intensity spikes and fMRI volumes with excessive motion from the scan, with thresholds for signal intensity outliers set at 9 standard-deviations above or below the mean. A motion limit of 2 mm translation and 2° rotation in any direction was also enforced. Spatial smoothing of the functional data was performed using an 8 mm, full-width at half maximum (FWHM) Gaussian kernel. For denoising, signals from the white matter, cerebrospinal fluid, and motion parameters were regressed from the functional data. Additional signal filtering to reduce noise due to scanner drift and physiological effects, such as respiration and pulsation, was performed using a band-pass filter of 0.008 – 0.10 Hz.

4.4. Image Statistical Analysis

To evaluate potential associations between microstructural changes and severity of DCM, FD images for each participant were firstly registered to the Johns Hopkins University DTI atlas (ICBM-DTI-81 1mm FA atlas). After linear registration, elastic (nonlinear) registration was performed between individual FD maps and the same atlas space using the FNIRT command in FSL. Statistical parameter maps were created by performing a voxel-wise T-test to test for significant regional differences between study patients and HCs for FD measurements using AFNI. Additionally, age was included as a covariate in the statistical analysis of the microstructural changes to account for the difference in age between the study and HC groups and the effect of age on estimating structural connectivity. A cluster threshold was applied from permutation tests by estimating data’s smoothness through command 3dFWHMx in AFNI, and then estimating the cluster extent thresholds at a level of significance, p < 0.05, through command 3dClustSim in AFNI.

To evaluate the functional connectivity of the brainstem with cerebral networks, both ROI-to-ROI (also termed seed-to-seed) connectivity and ROI-to-voxel (also termed seed-to-voxel) functional connectivity analyses were performed within both the supratentorial and the infratentorial brain by seeding the brainstem. Target ROIs and networks were selected from the Harvard-Oxford atlas based on previous studies that have observed cortical morphological changes as a result of neck disability and neurological impairment. Functional connectivity between the brainstem and target ROIs was assessed for all 26 participants. Significance was set at p <0.05 (two-sided) for the individual connections with a false discovery rate (FDR) <0.05. The overall image processing and statistical analysis pipeline is shown graphically in Fig. 1.

Fig. 1.

Image processing and statistical parameter mapping pipeline for probabilistic tractography and resting state fMRI analysis. FA maps from both the study cohort and HC’s were calculated, then registered to the ICBM-DTI-81 1mm FA atlas template using both linear and nonlinear registration. The transformation matrices were subsequently applied to other measurement registrations. t-test with covariates was performed using AFNI (3dttest++) in order to compare the study cohort with HCs while considering subject ages, along with NDI and mJOA scores. Significance was set at p < 0.05 with a false discovery rate (FDR) < 0.05 to identify regions of statistical difference, and microstructures associated with neurological dysfunction measures. Connectivity matrix of the study cohort and HCs were generated by seeding the brainstem. Probabilistic fiber tracking was performed to identify regions of significant difference and microstructures associated with clinical measures.

Highlights.

• Altered neurological connectivity to the brainstem is associated with DCM.

• DCM patients had significant reduction in fiber density along corticospinal tract.

• Changes of functional connectivity to brainstem is associated with severity of DCM.

• Both sensorimotor and visual networks showed functional reorganization.

• Pathological changes in the spinal cord can induce cerebral functional alterations.

Abbreviations:

FD

Fiber Density

FDC

fiber cross section

FC

Functional Connectivity

NDI

Neck Disability Index

mJOA

modified Japanese Orthopaedic Association

DCM

Degenerative Cervical Myelopathy

rsfMRI

Resting State Functional MRI

ROI

Region of Interest