Quantitative evaluation of cervical muscles in patients with degenerative cervical myelopathy using diffusion tensor imaging metrics

Abstract

Aims

The objective of this research was to assess the diffusion tensor imaging (DTI) parameters of deep cervical muscles in patients diagnosed with degenerative cervical myelopathy (DCM) before and after surgery at different follow-up stages, and to analyze their relationship with functional scores.

Methods

This study involved 30 patients with DCM who underwent cervical laminoplasty. DTI imaging and clinical assessments were performed preoperatively as well as at three-month and six-month follow-up. DTI parameters (fractional anisotropy (FA); mean diffusivity (MD); radial diffusivity (RD); and axial diffusivity (AD)) of cervical paraspinal muscles (multifidus and semispinalis cervicis) were measured, and mean values were calculated. Preoperative and postoperative changes in DTI parameters were analyzed using the Friedman test. Spearman correlation and linear regression analyses was conducted to assess the relationship between preoperative cervical paraspinal muscle DTI measurements and changes in modified Japanese Orthopedic Association (mJOA) scores and visual analogue scale (VAS) at preoperative and three-month and six-month follow-up. Multivariate logistic regression analysis was performed to assess the relationship between recovery rates of mJOA and VAS and changes of DTI metrics.

Results

Preoperative paraspinal muscle FA is associated with preoperative mJOA scores (r = 0.372, p = 0.043). Preoperative paraspinal muscle FA significantly correlated with VAS change at three-month (r = −0.461, p = 0.010) and six-month follow-up (r = −0.477, p = 0.008), and also correlated with mJOA recovery rate at six-month follow-up (r = 0.368, p = 0.045). Higher preoperative paraspinal muscle FA was a significant predictor of decreased VAS scores at three-month and six-month follow-up (β = −0.411, p = 0.024; β = −0.385, p = 0.035) and increased mJOA scores at six-month follow-up (β = 0.401, p = 0.028). Changes in FA (p = 0.033, OR 1.05 (95% CI 1.01 to 1.11)) and RD (p = 0.028, OR 1.27 (95% CI (1.03 to 1.58)) at six months’ follow-up may be variables influencing the VAS change.

Conclusion

In DCM patients, changes in DTI parameters from preoperative to postoperative follow-up stages can indicate paraspinal muscle recovery after cervical surgery. DTI examination before surgery can predict patient recovery outcomes.

Cite this article: Bone Joint Res 2026;15(2):179–189.

Article focus

• Investigate the changes in diffusion tensor imaging (DTI) metrics (fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), radial diffusivity (RD)) of paraspinal cervical muscles in degenerative cervical myelopathy (DCM) patients during preoperative and postoperative periods.

• Assess the relationship between DTI metrics and functional scores (modified Japanese Orthopedic Association (mJOA) and visual analogue scale (VAS) scores).

Key messages

• Higher preoperative FA values were associated with improved functional recovery in mJOA and reduced pain measured by VAS score.

• Changes in FA and RD values at six-month follow-up were variables influencing neck pain.

• DTI imaging offers insights into the microstructural dynamics of cervical musculature, potentially optimizing surgical candidates’ selection and care planning.

Strengths and limitations

• The main strength of this study was used DTI metrics (FA, MD, RD, AD) to provide novel insights into the microstructure and recovery of cervical muscles after surgery, with functional correlations to mJOA and VAS scores.

• The main limitation was a small sample size (n = 30), and a short follow-up period of six months may restrict the generalizability and assessment of long-term outcomes.

Introduction

Degenerative cervical myelopathy (DCM) represents the leading cause of chronic spinal cord impairment globally, resulting in spinal canal narrowing and chronic compression of the spinal cord, which can cause pain and neurological dysfunction.^1-3 The incidence of DCM is rising year by year, therefore it is necessary to diagnose and treat DCM patients promptly.⁴ It is widely recognized that moderate to severe DCM requires prompt surgical intervention. Posterior cervical laminectomy or laminoplasty is widely used for the treatment of multilevel cervical spondylotic myelopathy caused by ossification of the posterior longitudinal ligament.^5,6 The benefits of posterior surgery include achieving decompression indirectly without compromising the intervertebral spaces, clear visualization during operation on multiple segments, effective improvement of spinal cord blood supply and neurological function, and creation of favourable conditions for nerve recovery.⁷

Muscle health represents a potential biological mechanism affecting the onset, persistence, and recovery of DCM patients. Cervical muscles are often considered a source of dysfunction in various conditions. Previous studies have typically described changes in paraspinal cervical muscles using size (cross-sectional area or imaging-measured volume),⁸ morphology,⁹ and muscle composition (e.g. muscle fat infiltration (MFI)).¹⁰ These studies suggest that muscle morphology and composition, particularly MFI, may be associated with changes in cervical function and pain, or could potentially predict pain recovery. However, functional MRI studies on the effects of surgery on cervical muscles in DCM patients have been largely neglected.

Diffusion tensor weighted imaging (DTI) is a functional MRI technique that enables the study of tissue microstructure at the micrometre scale. This approach serves as an effective non-invasive means to characterize the composition of various anisotropic tissues.^11,12 The principle of DTI relies on the measurement of anisotropic diffusion of water molecules within tissues, indirectly revealing microstructural changes and providing detailed information about the fine structure and anatomy of local tissues. Among the primary DTI parameters used are fractional anisotropy (FA, 0 to 1), mean diffusivity (MD, × 10⁻³ mm²/s), radial diffusivity (RD, × 10⁻³ mm²/s), and axial diffusivity (AD, × 10⁻³ mm²/s),¹³ which are highly sensitive to microstructural changes, making them capable of identifying pathological and stress-related alterations in the composition of tissues.

Skeletal muscle fibres exhibit high anisotropy, allowing its microstructure to be quantitatively assessed using DTI.^14-17 Recent studies have shown that DTI can non-invasively characterize the microstructural architecture of muscle tissue.^18-21 Furthermore, microscopic changes in muscle tissue caused by intense activity can be detected using DTI, whereas these changes cannot be identified with conventional T2-weighted fat-suppressed MRI sequences.^22,23 DTI of spinal muscles has been shown to be consistent with DTI parameters of other muscles, has a lower correlation with BMI, and possesses the ability to predict muscle strength.²⁴ However, to our knowledge, most studies related to muscle DTI have focused on the legs or lumbar region, with some recent studies examining the shoulder and upper limb areas,^25,26 while longitudinal studies on DTI of paraspinal muscles in the cervical spine remain lacking. Currently, no research has explored the impact of surgery on DTI metrics of cervical paraspinal muscles or the relationship between DTI metrics and postoperative functional recovery. This study aims to investigate whether the DTI parameters of cervical paraspinal muscles in DCM patients change during the preoperative and postoperative follow-up stages, and whether preoperative and postoperative DTI parameters are associated with cervical functional recovery and neck pain. This may help in the selection of surgical candidates.

Methods

Participants

Between January 2017 and December 2017, with written consent, 54 DCM patients were enrolled. The mean age of subjects was 59.30 years (44 to 71; SD 6.88), with 21 males (70%). Table I lists the patients’ characteristics and functional scores.

Table I.

Characteristics of the patient population (n = 30).

Characteristic	Value
Mean age, yrs (SD)	59.30 (6.88）
Sex, n (%)
Male	21 (70）
Female	9 (30）
T2 Hypersignal grade, n (%)
Grade 0	10 (33.3)
Grade 1	11 (36.7)
Grade 2	9 (30)
Functional scores
Mean mJOA (SD)
Preoperative	13.0 (2.0)
Three-month follow-up	15.3 (1.5)
Six-month follow-up	15.8 (1.2)
Median VAS (IQR)
Preoperative	0 (0 to 4.5)
Three-month follow-up	1 (0 to 3)
Six-month follow-up	0 (0 to 2)
Mean AD (× 10-³ mm²/s) (SD)
Preoperative	1.165 (0.218)
Three-month follow-up	1.301 (0.258)
Six-month follow-up	1.210 (0.303)
Mean FA (0 to 1) (SD)
Preoperative	0.141 (0.026)
Three-month follow-up	0.114 (0.024)
Six-month follow-up	0.168 (0.036)
Mean MD (× 10-³ mm²/s) (SD)
Preoperative	1.036 (0.189)
Three-month follow-up	1.187 (0.236)
Six-month follow-up	1.054 (0.229)
Mean RD (× 10-³ mm²/s) (SD)
Preoperative	0.967 (0.186)
Three-month follow-up	1.126 (0.232)
Six-month follow-up	0.973 (0.213)

AD, axial diffusivity; FA, fractional anisotropy; MD, mean diffusivity; mJOA, modified Japanese Orthopedic Association; RD, radial diffusivity; VAS, visual analogue scale.

This study was approved by the Ethics Review Committee of our Hospital. Inclusion criteria were as follows: spinal cord compression involving at least 0.3 vertebral levels due to herniated disc, ligament ossification, and spinal stenosis; age between 18 and 80 years; failure of conventional conservative treatment and scheduled for posterior cervical laminoplasty; and no clear surgical contraindications. The exclusion criteria were inability to undergo MRI and/or history of spinal surgery or spinal cord/nerve root injury caused by surgery, with incomplete recovery at the final follow-up. During the postoperative follow-up, we excluded patients lost to follow-up (14 patients: seven were unable to return for scheduled follow-up visits due to distance issues, four refused to continue participating in the study for personal reasons, and three could not be contacted due to changes in contact information) and those with incomplete examinations during follow-up (ten patients: six failed to complete all required tests, three had scan ranges that did not meet requirements, and one could not complete all examinations on time due to equipment failure), resulting in 30 patients being included in this study after screening, as shown in Figure 1. We compared the baseline data of 24 dropped out patients with 30 enrolled patients, as shown in Supplementary Table i.

Fig. 1.

Flowchart of inclusion and exclusion criteria. DCM, degenerative cervical myelopathy.

Posterior cervical laminoplasty was performed on all participants, during which non-metallic coral implants were used. Consequently, no artifacts were present on the MRIs. At the time of enrolment, clinical symptoms of cervical spondylosis and cervical spine functional test scores were collected for each patient. All patients underwent MRI examinations, with additional MRI and clinical function evaluations performed at three-month and six-month follow-up. Prognosis and postoperative functional recovery at various intervals were assessed using the following indicators: the modified Japanese Orthopedic Association (mJOA) score²⁷ and the visual analogue scale (VAS). The mJOA is a 17-point scale used to assess movement and sensory capabilities of the arms and legs. The VAS is a widely used unidimensional tool for measuring pain intensity. Clinical evaluations were scored by two spine surgeons (JHL, QPS), with the same surgeon scoring each patient at different follow-up stages. Additionally, the mJOA recovery rate, representing follow-up outcomes, was determined for each patient: Postoperative mJOA change / (17-preoperative mJOA) × 100%. VAS recovery was defined as postoperative VAS − preoperative VAS. AD, FA, MD, RD recovery were defined as (postoperative-preoperative) / preoperative × 100%

MRI data acquisition

All patients underwent MRI using an Ingenia 3 T scanner (Philips Healthcare, UK) equipped with a 16-channel head and neck coil. DTI data were acquired in the axial view using a single-shot EPI sequence, with a total of 17 slices covering the vertebral levels from C2 to C7. The imaging plane was aligned parallel to the intervertebral discs. The parameters were as follows: FOV = 220 × 140 × 80 cm³; number of signal averages = 2; number of diffusion directions = 32; sensitivity encoding factor = 2; partial Fourier factor = 0.75; TE/TR = 77/4,500 ms; in-plane resolution = 1.5 × 1.5 mm²; acquisition matrix size = 40 × 106; slice thickness = 4 mm; inter-slice gap = 2 mm; total scan time = five minutes. In addition to DTI, sagittal and axial T1-weighted, sagittal T2-weighted turbo spin echo (T2W-TSE), and axial T2*-weighted multi-echo fast field echo (T2*W-MFFE) sequences were used for structural imaging.

Image processing

Motion between DWIs was corrected using the method presented by Xu et al,²⁸ which was implemented in the Spinal Cord Toolbox (SCT).²⁹ Subsequently, all images were identically preprocessed with the Python DIPY Library,³⁰ which included noise correction through the Marchenko-Pastur principal component analysis method,^31,32 Gibbs‐ringing artifact correction based on local subvoxel‐shifts,³³ and computation of the DTI metrics (FA, MD, AD, and RD). An experienced radiologist (BH), blinded to patient clinical information, manually segmented the multifidus and semispinalis cervicis on the axial T2-weighted images and registered them to processed DTI images using the free open-source software ITK-SNAP (v4.0.0). For each patient, the muscles on both the left and right sides were manually delineated at the C3/4, C4/5, and C5/6 intervertebral discs levels (along the contours of the deep cervical muscles – multifidus, semispinalis cervicis – on both sides of the vertebral bodies, avoiding fascia, connective tissue, and subcutaneous fat). Since the fascial boundary between the multifidus and semispinalis cervicis is not distinct, these muscles were combined to reduce measurement errors. An example of the ROI delineation and a representative result of AD, FA, MD, and RD maps is shown in Figure 2 and Figure 3. For each ROI, we extracted pertinent DTI metrics and the mean (SD) values were recorded. To assess the reliability of the method, the same physician (BH) repeated the measurements after a one-month interval and calculated the consistency using the intraclass correlation coefficient (ICC). Additionally, to assess the consistency of ROI delineation, another experienced radiologist (YY) performed ROI delineation of the multifidus and semispinalis cervicis muscles on the same slice for each patient. The consistency between the two delineations was calculated, with ROI area consistency assessed using ICC.

Fig. 2.

Region of interest including bilateral multifidus, semispinalis cervicis muscles at the C3/4, C4/5, and C5/6 intervertebral disc levels.

Fig. 3.

Representative results of a) axial diffusivity (AD), b) mean diffusivity (MD), c) radial diffusivity (RD), and d) fractional anisotropy (FA) maps from one subject.

Statistical analysis

All data following a normal distribution are presented as the mean (SD), whereas data that do not follow a normal distribution are reported as the median (IQR). The Friedman test was used to evaluate changes in DTI metrics at different follow-up stages before and after surgery. Spearman correlation analysis was used to assess the relationships between preoperative mJOA and VAS scores and paraspinal muscle DTI. To evaluate the association between preoperative cervical paraspinal muscle DTI metrics and recovery rates of mJOA and VAS at three-month and six-month follow-ups, Spearman correlation and linear regression analyses were performed. We divided participants into two groups based on VAS recovery of > 0 and ≤ 0, and mJOA changes of > 3 and ≤ 3. The VAS grouping is based on the patients’ direct responses to postoperative pain symptoms, while the mJOA grouping was based on the minimal clinically important differences (MCID) established in previous studies,^34,35 and clinical experience with moderate to severe DCM patients. Multivariate logistic regression analysis was performed to assess the relationship between recovery rates of mJOA and VAS and changes of DTI metrics. The reproducibility of DTI metrics and ROI areas between two measurements was assessed using ICC, with values greater than 0.75 signifying good reproducibility. All data analyses were conducted using SPSS v26.0 (IBM, USA), and p < 0.05 was considered statistically significant.

Results

Changes in DTI parameters of paraspinal muscles at different follow-up stages are shown in Figure 4. The repeatability of DTI parameters measured twice by the same doctor was good, with the following results: AD 0.793 (95% CI 0.724 to 0.862), FA 0.838 (95% CI 0.696 to 0.914), MD 0.876 (95% CI 0.767 to 0.934), and RD 0.823 (95% CI 0.780 to 0.866). The reproducibility of ROI area measurements between the two measurements was good, with an ICC of 0.823 (95% CI 0.732 to 0.914).

Fig. 4.

Diffusion tensor imaging (DTI) parameters of cervical paraspinal muscles at different follow-up stages. a) Changes in paraspinal muscle axial diffusivity (AD) at different follow-up stages, with no significant differences observed. b) Mean diffusivity (MD) at different follow-up stages; both preoperative and six-month follow-up MD values (p = 0.002) were lower than those at three-month follow-up (p = 0.007). c) Fractional anisotropy (FA) at different follow-up stages; preoperative FA values were higher than postoperative values at three-month follow-up (p < 0.001), and FA values at six-month follow-up were also higher than those at three-month follow-up (p = 0.014). d) Preoperative radial diffusivity (RD) values were lower than postoperative values at three-month follow-up (p = 0.007), and RD values at six-month follow-up were also lower than those at three-month follow-up (p = 0.005). **p < 0.01, ***p < 0.001.

Table II shows that the preoperative FA was correlated with preoperative mJOA scores (r = 0.452, p = 0.012), while the remaining preoperative DTI parameters (MD, AD, RD) did not demonstrate relationships with mJOA or VAS.

Table II.

Spearman correlation between preoperative paraspinal muscle diffusion tensor imaging and preoperative modified Japanese Orthopedic Association (mJOA) score and preoperative visual analogue scale (VAS).

Variable	mJOA		VAS
	r	p-value	r	p-value
AD	0.059	0.756	0.055	0.775
FA	0.452	0.012	0.268	0.152
MD	-0.056	0.773	0.019	0.922
RD	-0.084	0.659	-0.019	0.922

AD, axial diffusivity; FA, fractional anisotropy; MD, mean diffusivity; RD, radial diffusivity.

Table III shows the Spearman correlation analysis between preoperative DTI metrics of paraspinal muscles and mJOA recovery rates at three-month and six-month follow-up. The results indicate that the preoperative FA of paraspinal muscles correlates with the mJOA recovery rate at six-month follow-up (p = 0.045). Linear regression analysis showed that higher preoperative FA values are associated with increased mJOA scores at six-month follow-up (p = 0.028), as shown in Table IV.

Table III.

Spearman correlation between preoperative diffusion tensor imaging (DTI) metrics and changes in modified Japanese Orthopedic Association scores at three-month and six-month follow-up.

DTI metric	Three-month follow-up		Six-month follow-up
	r	p-value	r	p-value
AD	-0.164	0.386	-0.167	0.379
FA	0.135	0.477	0.368	0.045
MD	-0.132	0.486	-0.215	0.254
RD	-0.112	0.557	-0.245	0.192

AD, axial diffusivity; FA, fractional anisotropy; MD, mean diffusivity; RD, radial diffusivity.

Table IV.

Linear regression analysis of preoperative cervical paraspinal muscle diffusion tensor imaging (DTI) metrics and changes in modified Japanese Orthopedic Association scores at three-month and six-month follow-up. All regression coefficients, β, are standardized.

DTI metric	Three months		Six months
	β (95% CI)	p-value	β (95% CI)	p-value
AD	-0.188 (-0.568 to 0.192)	0.319	-0.199 (-0.578 to 0.181)	0.292
FA	0.109 (-0.276 to 0.494)	0.565	0.401 (0.005 to 0.756)	0.028
MD	-0.177 (0.150 to 0.204)	0.350	-0.253 (0.121 to 0.385)	0.177
RD	-0.173 (-0.554 to 0.208)	0.360	-0.301 (-0.670 to 0.068)	0.106

AD, axial diffusivity; FA, fractional anisotropy; MD, mean diffusivity; RD, radial diffusivity.

Table V presents the Spearman correlation analysis between preoperative DTI parameters of cervical paraspinal muscles and VAS changes at three-month and six-month follow-up. The results indicate that preoperative FA of the cervical paraspinal muscles is correlated with VAS changes at three-month and six-month follow-up (p = 0.010 and p = 0.008, respectively). Linear regression analysis shows that higher preoperative FA values are associated with lower VAS scores at the three-month and six-month follow-ups (p = 0.024 and p= 0.035, respectively), as shown in Table VI.

Table V.

Spearman correlation between preoperative diffusion tensor imaging (DTI) metrics and visual analogue scale changes at three-month and six-month follow-up.

DTI metric	Three months		Six months
	r	p-value	r	p-value
AD	-0.010	0.957	-0.067	0.725
FA	-0.461	0.010	-0.477	0.008
MD	0.045	0.815	0.022	0.910
RD	-0.081	0.670	-0.020	0.918

AD, axial diffusivity; FA, fractional anisotropy; MD, mean diffusivity; RD, radial diffusivity.

Table VI.

Linear regression analysis of preoperative cervical paraspinal muscle diffusion tensor imaging (DTI) parameters and visual analogue scale score changes at three-month and six-month follow-up. All regression coefficients, β, are standardized.

DTI metric	Three months		Six months
	β (95% CI)	p-value	β (95% CI)	p-value
AD	0.044 (-0.342 to 0.431)	0.816	0.118 (-0.266 to 0.502)	0.534
FA	-0.411 (-0.058 to –0.764)	0.024	-0.385 (-0.003 to –0.743)	0.035
MD	0.007 (-0.380 to 0.394)	0.971	0.072 (-0.314 to 0.458)	0.704
RD	-0.053 (-0.440 to 0.333)	0.780	0.009 (-0.378 to 0.396)	0.963

AD, axial diffusivity; FA, fractional anisotropy; MD, mean diffusivity; RD, radial diffusivity.

Table VII and Table VIII present the multivariate logistic regression analysis between changes of DTI parameters and clinical function scores at three-month and six-month follow-up. The results show that changes in FA and RD at six-month follow-up may be significant variables influencing neck pain (p = 0.033, OR 1.05 (95% CI 1.01 to 1.11); and p = 0.028, OR 1.27 (95% CI 1.03 to 1.58), respectively).

Table VII.

Multivariate logistic regression analysis of modified Japanese Orthopedic Association score change rates at three and six months grouped with axial diffusivity (AD), fractional anisotropy (FA), mean diffusivity (MD), and radial diffusivity (RD) change rates.

Variable	Three months		Six months
	p-value	OR (95% CI)	p-value	OR (95% CI)
AD rec	0.428	1.11 (0.86 to 1.42)	0.587	1.06 (0.87 to 1.28)
FA rec	0.477	0.98 (0.93 to 1.04)	0.453	1.01 (0.98 to 1.05)
MD rec	0.852	0.97 (0.68 to 1.37)	0.929	1.01 (0.78 to 1.31)
RD rec	0.490	0.95 (0.81 to 1.11)	0.501	0.96 (0.86 to 1.08)

rec, recovery rate.

Table VIII.

Multivariate logistic regression analysis of visual analogue scale change rates at three and six months grouped with axial diffusivity (AD), fractional anisotropy (FA), mean diffusivity (MD), and radial diffusivity (RD) change rates.

Variable	Three months		Six months
	p-value	OR (95% CI)	p-value	OR (95% CI)
AD rec	0.273	0.79 (0.52 to 1.20)	0.295	1.12 (0.91 to 1.37)
FA rec	0.218	1.05 (0.97 to 1.14)	0.033	1.05 (1.01 to 1.11)
MD rec	0.632	1.19 (0.58 to 2.42)	0.112	0.71 (0.46 to 1.08)
RD rec	0.621	1.10 (0.76 to 1.59)	0.028	1.27 (1.03 to 1.58)

rec, recovery rate.

Discussion

Our research shows that preoperative cervical paraspinal muscle DTI parameters in DCM patients are associated with cervical functional recovery and pain conditions. Performing DTI imaging is beneficial for understanding postoperative recovery, and provides guidance for treatment selection.

DTI assess the anisotropic properties by measuring the diffusion rates of water molecules along the principal axes,^36,37 represented by principal eigenvalues λ1, λ2, and λ3. Various other useful diffusion parameters can be derived from the principal eigenvalues.¹⁴ FA is a ratio-based metric calculated from the λ1, λ2, and λ3, ranging from 0 for completely isotropic diffusion to 1 for entirely anisotropic diffusion.³⁸ MD reflects the average diffusion coefficient of water across all directions within the tissue and is the mean of the three eigenvalues. RD reflects the diffusion coefficient along the anisotropic pathways, and is the mean of the two eigenvalues perpendicular to the axial direction (λ2, λ3). Finally, AD measures the rate at which water diffuses along the pathway’s axial direction, and is associated with the axial eigenvalue (λ1).³⁹

Zaraiskaya et al²¹ measured the DTI eigenvalues of injured calf muscles, which were significantly elevated compared with those in the healthy control group, indicating increased diffusivity in the damaged muscles. This finding is closely related to the theoretical hypothesis that muscle injury leads to increased diffusion of water within muscle tissues. Muscle injury disrupts the muscle cell membrane, resulting in the loss of the orderly fibre structure. In this study, the MD and RD values at three-month follow-up were all higher than those before surgery, and the FA value was lower than preoperative. MD is a sensitive marker of inflammation and tissue oedema, helping to evaluate early inflammatory responses after injury and the process of tissue fluid rebalancing after repair.^40,41 The increased MD may reflect expanded muscle tissue space due to oedema and inflammation, as well as loosened muscle matrix or increased local vascular permeability after injury. RD is mainly used to evaluate the state of transverse membrane damage and the recovery of fibrous transverse structure, serving as an important parameter reflecting muscle membrane integrity after injury.^42,43 The increased RD may reflect muscle transverse structural disorder caused by surgical injury. FA is the primary indicator for assessing muscle directionality and integrity. It is typically associated with the state of fibre alignment, the extent of damage, and the degree of repair.⁴² A decreased FA indicates muscle fibre damage, rupture, or disorganized alignment. At six-month follow-up, MD and RD values decreased while FA increased, indicating that muscle repair begins to progress over time, with muscle inflammation and oedema potentially subsiding, and compensatory muscle remodelling occurring.²² However, posterior approach surgery damages muscle ligament attachment points, inevitably leading to long-term muscle atrophy. Studies have shown that during long-term follow-up (18 months), muscles tend to show overall atrophy, with increased fatty infiltration and decreased area of the paraspinal muscles.⁴⁴

In our study, the preoperative paraspinal muscle FA values were positively correlated with preoperative spinal cord function and spinal cord function recovery at six-month follow-up, indicating that preoperative FA values can represent the preoperative disease severity, and serve as an important predictor of spinal cord recovery during follow-up. High preoperative FA values reflect better muscle health and structural integrity, which may be more conducive to postoperative spinal cord function recovery. Low FA values may indicate muscle fibre rupture and misalignment, and these changes could indirectly reduce mJOA scores by affecting cervical spine mechanical stability or spinal cord blood supply. The microscopic state of preoperative muscle tissue may indirectly support spinal cord function recovery.

Axial pain after surgery is an important complication of posterior cervical laminoplasty or laminectomy. Although posterior surgery is not the primary cause of axial pain, it can exacerbate the condition. Our research results show that preoperative FA values are negatively correlated with VAS changes at three- and six-month follow-up; higher preoperative FA values are associated with lower pain scores during follow-up. Identifying the importance of preoperative FA values in predicting patient prognosis is valuable – higher preoperative FA values may be more beneficial in reducing postoperative axial pain complications, while patients with lower FA values should receive additional attention during surgery and postoperative care to prevent complications.

Our results indicate that the FA values of paraspinal muscles decrease during the inflammation and increase during the repair. Furthermore, changes in FA during the follow-up stage are associated with neck pain. Several studies show that extracellular matrix proliferation (such as fibrosis) can lead to increased FA values, indicating that increased FA does not always indicate healthy fibres, but may also reflect enhanced collagen structural directionality in scar.^45,46 Additionally, FA can increase under compression or stretching conditions,^47,48 which could also explain our findings: passive muscle stretching caused by cervical decompression surgery. These changes may also be a primary cause of postoperative pain.

Additionally, various factors, including the selected measurement levels, may affect cervical paraspinal muscle measurements. In our study cohort, most compression levels were at C3-C6, so we selected paraspinal muscles at the C3/4, C4/5, and C5/6 intervertebral disc levels for DTI measurements. In Fortin et al’s⁴⁹ study, measurements were taken at C4-C6 levels since compression was highest at C4/C5 and C5/C6. Some studies have measured bilateral cervical muscles from C2-C7 levels.⁸ Our study outlined and measured the multifidus and semispinalis cervicis muscles, all of which are crucial for cervical spine stability. Research suggests that their pathology may be related in overlapping clinical presentations, quantified through mJOA scores.⁵⁰ Furthermore, while most current studies evaluate relationships between muscle characteristics and clinical outcomes, our research explores the association between muscle functional MRI indicators and clinical outcomes, explaining how cervical muscles relate to cervical function at a deeper level.

Our study has several limitations. First, the relatively high patient dropout rate may lead to selection bias and affect the generalizability of the research results. Although we compared baseline information between enrolled and dropout patients and found no significant differences, future studies should adopt effective retention strategies to improve completion rates and data representativeness. Second, we considered multifidus and semisplinalis cervicis as one muscle group, because the precise boundary between the two muscles cannot always be distinguished; we also did not separately analyze the deep neck muscles, which might influence the results, but prior research has indicated that the differences in DTI metrics across muscles are minimal,⁵¹ so we believe that these inter-muscular changes can be ignored. Third, the effects of pre-surgical conservative treatments on deep cervical muscles were not considered. Fourth, our study showed an unexpectedly low preoperative median VAS score (0 (IQR 0 to 4.5)) in our surgical DCM cohort. While this finding aligns with emerging evidence that a substantial subset of DCM patients may present without significant neck pain despite severe myelopathy,^52,53 we acknowledge potential methodological factors that might have influenced this result. This observation highlights the need for more comprehensive pain evaluation protocols in future DCM studies. Moreover, we only measured functional MRI parameters of cervical muscles without evaluating their morphology and composition. Future studies should consider additional measurements of cervical muscle morphology, composition, and function. Our study did not continue DTI measurements during follow-up beyond six months, so it is not possible to determine changes in DTI metrics in DCM patients over long-term follow-up. Other limitations include the limited sample size and single-centre design, which may impact the generalizability of the results. The lack of DTI comparison between cervical myelopathy patients and healthy individuals is also a significant limitation. Additionally, we did not assess the correlation between cervical alignment parameters and paraspinal muscle DTI indicators. Future studies should systematically collect parameters such as cervical Cobb angle and cervical lordosis to explore their potential correlation with paraspinal muscle DTI changes and impact on surgical outcomes.

Our research findings indicate that preoperative DTI parameters of cervical muscles, particularly FA, in DCM patients can serve as imaging predictors for improved surgical outcomes. This emphasizes the significance of including muscular parameters in the preoperative evaluation and management of DCM patients. Assessing cervical muscle function and structure before surgery could be crucial in patients’ management planning, to optimize postoperative care and improve surgical outcomes. This study provides potential support for cervical muscle strengthening in DCM patients prior to surgical intervention.

Data Availability

The datasets generated and analyzed in the current study are not publicly available due to data protection regulations. Access to data is limited to the researchers who have obtained permission for data processing. Further inquiries can be made to the corresponding author.