MR Elastography-based Slip Interface Imaging for Assessment of Myofascial Interface Mobility in Chronic Low Back Pain: A Pilot Study

Abstract

Purpose:

Abnormal restrictions at myofascial interfaces are a key factor in chronic low back pain (CLBP) with myofascial pain syndrome (MPS), but no established imaging method exists for quantitative assessment. This pilot study evaluated the clinical feasibility of MR elastography (MRE)-based slip interface imaging (SII) for visualizing myofascial mobility in CLBP-MPS patients versus age-/sex-/BMI-matched healthy participants. A new SII biomarker, maximal normalized displacement discontinuity ($D_{norm}$), was introduced, and the previously developed normalized octahedral shear strain (NOSS) map was used for comparison. Preliminary assessments of vibration-frequency effects, repeatability, and group differences in inter-muscular mobility were conducted.

Methods:

MRE was conducted at 30, 60, and 90 Hz to compare $D_{norm}$ with NOSS for visualizing inter-muscular slip interfaces at the quadratus lumborum-erector spinae (QL-ES) and erector spinae-multifidus (ES-M) muscle boundaries at L3-L4. Repeatability of $D_{norm}$ and NOSS was assessed at 30 Hz (identified as the best-performing frequency) using within-subject coefficients of variation and a linear mixed-effects model from within- and between-day scans. Clinical feasibility was then assessed via Wilcoxon rank-sum tests in six CLBP-MPS patients and matched controls.

Results:

$D_{norm}$ at 30 Hz exhibited superior delineation of slip interfaces compared to higher frequencies and NOSS. $D_{norm}$ demonstrated significantly lower variation and better repeatability than NOSS. Healthy participants showed continuous, well-defined slip interfaces, whereas CLBP-MPS patients displayed disrupted interfaces with significant mobility reduction at L4 ES-M.

Conclusion:

SII with lower vibration frequency and $D_{norm}$ map enhances visualization of lower back myofascial mobility and shows potential for assessing differences between healthy individuals and CLBP-MPS patients.

1. Introduction

Chronic lower back pain (CLBP) is a widespread public health issue, affecting approximately 619 million people globally in 2020 (1). Among those with CLBP, Myofascial Pain Syndrome (MPS), which arises from the skeletal muscles and surrounding fascia (2), is particularly common, with prevalence rates ranging from 63.5% to 90% (3). Despite its high prevalence, diagnosing CLBP within the MPS spectrum largely depends on physical examinations and self-reported symptoms (4), highlighting the need for objective, quantitative methods to standardize diagnosis and guide evidence-based treatment strategies.

The underlying mechanisms of CLBP associated with MPS are still under investigation, with growing attention to how impairments in the thoracolumbar fascia (TLF) and the surrounding lumbar muscles contribute to pain (5). The TLF plays a key role in stabilizing the trunk, supporting the lower back musculature, transmitting force, and facilitating spinal motion (6). One theory linking myofascial layer abnormalities to CLBP is that repeated inflammation from muscle misuse or overuse leads to myofascial layer adhesion, which restricts the normal sliding motion between layers, alters mechanical force distribution, and potentially triggers abnormal pain and sensory signals (7). Previous ultrasound studies have shown reduced TLF gliding capabilities in CLBP patients compared to healthy individuals (8). In addition to the TLF, the surrounding muscles including the multifidus (M) and erector spinae (ES) also undergo changes in CLBP such as alterations in morphology (9,10) and stiffness (10–14). Moreover, an ultrasound study has shown that paraspinal muscle characteristics, including those of the ES, affect TLF mobility, with muscle contraction reducing shear strain in CLBP patients (7). Furthermore, the quadratus lumborum (QL) muscle is considered another common source of CLBP within MPS, although its functions and biomechanics are not fully understood (15). These studies suggest that beyond the TLF, myofascial interfaces, referring to the apposing fascial and aponeurotic planes between lumbar musculature, may also be affected in CLBP, potentially becoming more adherent in patients compared to healthy individuals with intact myofascial layers. Currently, there is a significant knowledge gap in understanding how myofascial mobility across the lower back myofascial system relates to pain. Developing imaging techniques to assess myofascial mobility could provide objective biomarkers, enhancing our understanding of its relationship with pain, and ultimately advancing patient care for those with CLBP within the MPS spectrum.

MR Elastography (MRE)-based slip interface imaging (SII) shows great potential as an imaging biomarker for noninvasively assessing myofascial interface mobility. SII was originally developed to evaluate the adherence between brain tumors and adjacent tissue layers (16). SII-assessed normalized octahedral shear strain (NOSS) was developed as a biomarker to quantify the degree of adhesion at the brain-tumor interface (17,18). In the musculoskeletal system, a previous study demonstrated the feasibility of using the SII-assessed NOSS map to evaluate the mobility of inter-muscular myofascial interfaces in the quadriceps (19), where a healthy myofascial interface exhibits hyperintense NOSS contrast, indicating freely gliding/sliding motion between adjacent muscles. However, our exploration of direct NOSS implementation in the lower back suggests that NOSS may be less effective for assessing inter-muscular interfaces in this region. This may be due to the complexity of multilayered and multidirectional myofascial structures in the lower back, as well as the relatively decreased range of motion in trunk muscles compared to muscles in the extremities. Thus, a new SII metric is required to better characterize inter-muscular interface mobility in the lower back.

A key parameter of muscle MRE is the frequency at which acoustic mechanical waves are generated in the targeted muscle by applying a muscle-specific passive driver. MRE frequency selection requires a balance between wavelength and wave attenuation: higher frequencies generate shorter wavelengths for finer wave resolution but attenuate more quickly, while lower frequencies penetrate deeper with longer wavelengths. Compared to the typical 60 Hz used in clinical MRE protocols for the liver (20) and the brain (21), previous muscle MRE studies have generally applied higher frequencies. For instance, prior MRE studies have measured the mechanical properties of the erector spinae and multifidus muscles using vibration frequencies of 90 Hz (22), 100 Hz (23,24), 120 Hz (25), and used 90 Hz for thigh stiffness measurements (26–30). Multifrequency MRE approaches have also been explored to characterize the frequency-dependent viscoelastic properties of skeletal muscle (31). To further investigate the complexities of wave propagation across the inter-muscular interfaces, Chakouch et al. (2015) developed a muscle-mimicking phantom that simulates thigh muscle architecture, offering a controlled setting to study shear wave behavior and optimize MRE acquisition strategies (32).

In contrast to higher frequencies typically used in muscle MRE for stiffness estimation, in our previous SII thigh study, a lower frequency such as 40 Hz (19) was used, as lower frequencies were observed to be more suitable for SII imaging. This may be because lower frequencies provide greater wave penetration, enabling the induction of relative motion between muscle boundaries, which is essential for capturing displacement at inter-muscular interfaces. However, the effects of frequency on SII evaluations have not been fully explored in previous studies.

The objective of the study was to develop a new SII biomarker optimized for assessing lower back inter-muscular interface mobility and to evaluate its clinical feasibility in CLBP patients within the MPS spectrum. First, the effect of MRE vibration frequency on SII imaging of lower back muscles was investigated by comparing three vibration frequencies: 30 Hz, 60 Hz and 90 Hz. Second, a new SII biomarker, the maximal normalized displacement discontinuity map ($D_{norm}$), was introduced and analyzed to assess its repeatability and determine whether it demonstrated better repeatability in measuring inter-muscular interface mobility compared to the existing NOSS map. Lastly, $D_{norm}$ was evaluated in both patients with CLBP within the MPS spectrum and age-, sex-, and BMI-matched healthy participants, focusing on the inter-muscular interfaces between the QL and ES muscles (QL-ES), which anatomically aligns with the middle layer of the TLF, as well as ES-M interfaces at the third and fourth lumbar (L3-L4) vertebral levels. The hypothesis was that patients would exhibit varying degrees of myofascial restriction at these muscle boundaries, resulting in a loss of the slip interface on the $D_{norm}$ map compared to their matched healthy participants.

2. Methods

2.1. Participants

Approval for the study was obtained from the Institutional Review Board (IRB). A total of 13 participants were enrolled for this pilot prospective study, including 7 healthy individuals without a history of musculoskeletal disorders or pain prior to the study (5 females, 2 male; age range = 24–54 years; mean age = 42.0 ± 11.18 years; mean BMI = 26.29 ± 2.75 kg/m²), and 6 patients with CLBP within the MPS spectrum (5 females, 1 male; age range = 27–52 years; mean age = 42.83 ± 10.17 years; mean BMI = 26.60 ± 1.93 kg/m²; mean PEG score = 5.50 ± 2.70 and mean pain duration = 87 ± 75.68 months). The two groups were matched for age, sex, and BMI. For this pilot study, the BMI cutoff was set at 30 kg/m² to avoid potential issues with limited wave penetration in individuals with higher BMI. All participants provided written informed consent.

To assess both within-day and between-day repeatability of the SII biomarkers, five healthy participants underwent three repeated MRE scans: two same-day scans on Day 1 (Exams 1-2, passive driver reset between scans), and a third scan approximately two weeks later on Day 2 (Exam 3).

A questionnaire was administered to CLBP patients within the MPS spectrum to assess both pain duration and the Pain, Enjoyment, and General Activity (PEG) score (33). The PEG score evaluates pain intensity, its impact on daily life, and interference with activities, ranging from 0 to 10, with 10 representing the greatest severity. The final PEG score is calculated as the average of the three assessments (33).

2.2. MR scanning

All MRI scans were conducted using a 3.0 T MRI system (Signa Premier, GE Healthcare, Wisconsin, USA). The custom-made passive belt driver used in the study consists of three parts: the flexible belt, the connecting tube, and a hook-and-loop fastener. The belt measures 71.12 cm in length, 7.62 cm in width, and 1.67 cm in thickness (Figure 1a). It is constructed from a nylon mesh filling surrounded by a neoprene rubber form frame, with a polycarbonate pad on the outer side, where three MRI-safe fiducial markers are placed at the center of the belt to facilitate imaging localization (Figure 1b). The inner side of the belt, which is in contact with the participant’s body, is sealed with a neoprene rubber strip. All components are enclosed to prevent air leakage. The silicone tube measures 59.05 cm in length, with an inner diameter of 1.27 cm. Two identical-sized tubes are symmetrically attached to the belt and connect to a Y-shaped connector at the other end, which is linked to a commercially available pneumatic active driver (Resoundant Inc., Minnesota, USA) (Figure 1a). The connecting sections between the belt and the Y connector are tightly secured to ensure no air leakage. The hook-and-loop fastener is adjustable to accommodate different body sizes and is used to secure the belt when wrapped around the participants’ fourth and fifth lumbar (L4-L5) vertebrae regions in the supine position. Either 30 Hz with 95-100% amplitude or 90 Hz with 60-80% amplitude of the mechanical vibrations, considering the participant’s comfort vibration level, was induced to assess the effect of vibration frequency for the SII imaging. During scanning, a knee bolster was placed under the participants’ knees to prevent passive stretching of the lumbar region throughout the process (Figure 1c). Care was taken to position the belt driver symmetrically at the L4-L5 level to ensure consistent generation of the pneumatic wave. This alignment was checked using the scout localizer MRI images, with reference to MRI-safe fiducial markers placed at the center of the belt to assist in localizing the body’s midline (Figure 1d–f).

Figure 1:

SII acquisition setup for the lower back. Design of the custom-made passive belt driver for SII acquisition of the lower back (a). The details of the materials used in the belt, magnified in the green-outlined region in (a): (1) nylon mesh filling, (2) neoprene rubber foam, (3) polycarbonate pad, (4) MRI-safe fiducial markers, and (5) neoprene rubber strip (b). A participant was in the supine position, with a belt driver wrapped around the participant’s L4-L5 vertebrae region (c). Fiducial markers were used to check the placement prior to MRE imaging. Anatomical locations of the right (R) and left (L) Quadratus Lumborum (QL), Erector Spinae (ES), and Multifidus (M), along with the position of the belt driver, are shown in the axial (d) and sagittal (f) T2-weighted images. Red arrows in (d) and (f) indicate the center of the belt driver, positioned at the intervertebral disc between L4 and L5, while yellow arrows denote the top and bottom fiducial markers. The anterior signal dropout observed in (d) is due to the placement of an anterior saturation band, which was used to suppress respiratory motion artifacts. The white-outlined rectangle in (f) indicates the slice coverage.

3D MRE scanning was performed to acquire both magnitude and phase difference images using motion encoding gradients (MEGs) synchronized with the given vibrations using a spin-echo (SE) echo-planar-imaging (EPI) MRE pulse sequence that incorporated the dual-saturation dual-sensitivity motion encoding (DSDM) scheme (34), which acquires both water and fat MRE signals in the lower back to enhances the visualization, particularly in regions where fascia or fat infiltration is more prominently represented in the fat contrast. (Figure 2a and Supplementary Figure S1). 3D MRE images were acquired in the 36 oblique axial slices, centering the slice coverage at the L3-L4 region and parallel to the disk (see the white outlined rectangle in Figure 1f) with acquisition matrix = 80 × 80 reconstructed to 256 × 256, FOV = 38.4 cm, slice thickness = 4 mm, TE = 46 ms, TR = approximately 4800 ms, 6 motion encoding directions with ±x/y/z, 4 phase offsets sampled equally in one course of the motion, and a total scan time of less than 4 minutes. T2-weighted (T2W) images were acquired in the same orientation as MRE with the following imaging parameters: acquisition matrix = 512 × 512, FOV = 38.4 cm, slice thickness = 4 mm, and TE = 50 ms, TR = approximately 5300 ms. Other anatomical MRI sequences, including Axial T1-weigthed (T1W), Axial T2W Fat Suppression (T2FS), Sagittal T1W, and Sagittal T2W Fat Suppression (T2FS) were also included to assess potential musculoskeletal abnormalities in the spinal structures and surrounding tissues of the lower back.

Figure 2:

SII pipeline for lower back imaging: 1. MRE acquisition: magnitude and paired phase images in the 3D MRE wavefield are acquired (a). 2. SII Post-processing: Phase unwrapping is performed first (b), followed by first harmonic extraction using Fourier Transform. Shear wave amplitude (c), OSS (d), and 9 displacement gradient components (e) are calculated from the 3D displacement. Each of the x, y, and z displacement datasets undergo 3D spatial-temporal directional filtering (DF) with 10 bi-directional filters (f). 3. SII biomarkers: NOSS (g), is calculated from (c) and (d). $D_{norm}$ maps (h) are generated using DF (f) and the 9 displacement gradient components (e). All images from (a) to (h) are overlaid on the magnitude image shown in (a), with images (c) to (f) being zoomed in to highlight specific regions, featuring a brighter background for visualization purposes. Color scales in figures (a) to (g) indicate values in the specified units, including arbitrary units (a.u.), as labeled on each scale.

2.3. SII Post-imaging data processing

The acquired images were processed using MATLAB (The MathWorks, Massachusetts, USA). Phase difference images were first unwrapped using the DSDM unwrapping algorithm (34) (Figure 2(b)), followed by a Fourier transform to extract the first harmonic signals from the unwrapped phase data. The 3D phase signal was then converted to the 3D displacement, denoted as $U=(u_x,u_y,u_z)$ in micrometers, where $u_x,u_y,u_z$ represent displacements in the x-, y-, and z-directions, respectively. The SII algorithm is developed under the assumption that, in healthy tissue, myofascial layers glide relative to one another under shear loading, resulting in greater displacement discontinuity or “slip” across interfaces. When mobility is reduced (e.g., due to myofascial adhesions or restrictions), this relative motion diminishes, producing smoother and more continuous displacement fields (16,19). These changes are captured by spatial gradients of the displacement field, allowing detection of myofascial slip as a biomarker of tissue mobility. In this study, two SII biomarkers were generated during the processing: NOSS and a newly developed biomarker, $D_{norm}$, which was developed to improve the detection of displacement discontinuities (e.g., slip) across myofascial interfaces in lower back muscles that are highly organized and anisotropic in nature.

NOSS Calculation:

The NOSS map quantifies voxel deformation by first calculating the octahedral shear strain (OSS), which represents the norm of the deviatoric strain tensor, following the method of McGarry et al. (35) (Figure 2(d)). The shear wave amplitude is calculated as the combined displacement amplitude (square root of the sum of squares) of the first harmonic of the complex shear waves along the x-, y-, and z-axes (17) (Figure 2(c)). The NOSS map normalizes OSS by the shear wave amplitude (Figure 2(g)) (17) .

$D_{norm}$ Calculation:

Unlike NOSS, which normalizes OSS by the 3D wave amplitude, the $D_{norm}$ map measures the displacement gradients to directly capture displacement discontinuities in the x-, y-, and z-directions. These displacement gradients are then normalized by the wave amplitudes from directional filtered wave components, addressing potential wave interferences caused by the multi-directional and multi-layered structure of lower back muscles. More specifically: (1) The 3D displacement gradients are calculated as $\nabla U=(\frac{\partial U}{\partial x},\frac{\partial U}{\partial y},\frac{\partial U}{\partial z})$ (Figure 2(e)). (2) A 3D spatial-temporal direction filter (DF) (36) (order 4, frequency cutoff = 1-40 cycles/FOV), originally comprising 20 evenly spaced directions (36), was modified to merge opposing vectors into 10 bidirectional filters, decomposing displacement into 10 directionally filtered 3D wave components ($u_{xi},u_{yi},u_{zi})$, $i=1\text{to}10$. The wave amplitudes in x-, y-, and z-directions are now calculated as the combined magnitude of all 10 directionally filtered components along each respective axis: ${DF}_k=\sqrt{{\sum }_{i=1}^{10}{u}_{ki}^2}$ for $k\in (x,y,z)$ (Figure 2(f)). (3) The magnitude of the normalized displacement gradient in x-, y-, and z-direction is then calculated as $D_{norm\_k}=\sqrt{{\left(\frac{\partial u_x}{\partial k}/{DF}_x\right)}^2+{\left(\frac{\partial u_y}{\partial k}/{DF}_y\right)}^2+{\left(\frac{\partial u_z}{\partial k}/{DF}_z\right)}^2}$ for $k\in (x,y,z)$. (4) Finally, $D_{norm}$ is defined as the maximal intensity projection (MIP) of these magnitudes: $MIP\left\{D_{norm\_x},D_{norm\_y},D_{norm\_z}\right\}$ (Figure 2(h)).

A comparison between the 10 merged directions and the original 20 directions is included and discussed in Supplementary Figure S2 to evaluate their effects on $D_{norm}$ performance. Empirically, the 10-direction approach yields improved contrast and clearer detection of myofascial slip interfaces compared to the 20-direction method.

Mobility Index Calculation:

Beyond qualitative visualization, a mobility index was derived from each SII biomarker ($D_{norm}$ and NOSS) to quantify myofascial mobility for this preliminary study (37). Briefly, following manual segmentation of the QL-ES and ES-M interfaces at the L3 and L4 level, normal line profiles orthogonal to each interface were generated. Peaks within each profile were defined as local maxima exceeding the mean voxel intensity along each profile. The area within 3 pixels of each peak center was calculated and normalized by the total area under the profile, providing a localized mobility measure robust to variations in peak position and width of the profile. Profiles lacking identifiable peaks, indicative of minimal localized mobility, were assigned to a value of zero. The mobility index was then defined as the mean of these per-profile peak values across all slices, serving as a summary metric for each interface (Supplementary Figure S3).

2.4. SII image and statistical analysis

To evaluate the effects of vibration frequency on SII imaging of the lower back, NOSS maps calculated at 30 Hz, 60 Hz, and 90 Hz in healthy participants were qualitatively compared to determine which frequency provided greater contrast (i.e., a more distinct slip interface) at the muscle boundary. The ability to delineate a clear slip interface was further assessed by comparing $D_{norm}$ and NOSS maps at those frequencies in healthy participants.

Repeatability was assessed using repeated scans acquired at 30 Hz (identified as the best-performing frequency in the frequency comparison). Mobility indices derived from both $D_{norm}$ and NOSS maps were computed across 12 regions of interest (ROIs), corresponding to the QL-ES and ES-M interfaces at the L3, L4, and consecutive L3-L4 levels bilaterally. The within-subject coefficient of variation (wCV, %) was calculated for each subject and averaged per ROI (38). Within-day repeatability was assessed using data from Day 1 Exams 1-2, while between-day repeatability was assessed using data from Exam 1 (Day 1) and Exam 3 (Day 2). Overall repeatability was compared using a linear mixed-effects model with biomarker type ($D_{norm}$ vs. NOSS) and measurement timing (within-day vs. between-day) as fixed effects, and ROI as a random effect to account for variability across anatomical regions. Post-hoc pairwise comparisons were performed to assess differences between biomarkers within each repeatability condition, where applicable. A threshold of wCV < 10% was considered indicative of excellent repeatability (38).

To explore the clinical feasibility of SII in assessing myofascial interface mobility, $D_{norm}$ maps and $D_{norm}$-derived mobility indices from CLBP-MPS patients (n = 6) were compared to those of age-, sex-, and BMI-matched healthy participants (n = 6, selected from a pool of 7). Based on the hypothesis, patients were expected to exhibit reduced mobility, reflecting varying degrees of myofascial restriction and diminished slip interface visibility on the $D_{norm}$ maps. Because the current patient cohort did not report consistent unilateral pain, mobility indices from the left and right sides were averaged to yield a single global mobility index per region of interface for each participant. Group comparisons focused on the QL-ES and ES-M interfaces at both L3 and L4 levels. Given the small sample size and non-normal data distribution (assessed via the Kolmogorov-Smirnov test), non-parametric Wilcoxon rank-sum tests were used to compare mobility indices between groups at each ROI.

The significance level was set at p < 0.05. All statistical analyses were performed using MATLAB R2023a.

3. Results

3.1. The effect of MRE vibration frequency on NOSS and $D_{norm}$ analysis in the lower back

Figure 3 presents examples of NOSS, corresponding $D_{norm}$, and displacement maps for 30 Hz, 60 Hz and 90 Hz scans. The NOSS map obtained at 30 Hz provides clearer visualization of slip interfaces (Figure 3(b)) than the NOSS map acquired at 60 and 90 Hz (Figure 3(c–d)), where the interfaces appear less distinct. The $D_{norm}$ map at 30 Hz further enhances visualization (Figure 3(e)), offering improved delineation of all inter-muscular myofascial interfaces compared to NOSS. In contrast, the $D_{norm}$ obtained at 60 and 90 Hz shows similar convoluted patterns as observed in the NOSS map at 60 and 90 Hz (Figure 3(f–g)), suggesting a decrease in visualization clarity at the higher frequency. The displacement map at 30 Hz shows higher displacement amplitudes with a more uniform displacement distribution across muscle compartments (Figure 3(h)). In contrast, the displacement map at 60 and 90 Hz exhibits lower displacement amplitudes and more low-amplitude wave nodes, likely due to wave interference (Figure 3(i–j)), which appears to contribute to the noise amplification in the corresponding NOSS calculation.

Figure 3:

Effect of frequency on SII biomarkers: The anatomical locations of the right (R) and left (L) Quadratus lumborum (QL), Erector spinae (ES), and Multifidus (M) are shown in the T2W anatomical image with colored outlines (a). NOSS maps obtained at 30 Hz (b), 60 Hz (c) and 90 Hz (d) for one of the healthy participants are displayed, as well as the corresponding $D_{norm}$ maps at 30 Hz (e), 60 Hz (f) and 90 Hz (g). Displacement maps, representing shear wave amplitude (the combined amplitude of the first harmonic of complex waves in x-, y-, and z-directions) are shown for 30 Hz (h), 60 Hz (i) and 90 Hz (j), with the different scales noted. NOSS, $D_{norm}$, and displacement maps in (b-j) are overlaid on the corresponding T2W image. Blue, pink, white, and yellow arrows highlight the right QL-ES, right ES-M, left ES-M, and left QL-ES muscle interfaces, respectively.

Using the same 30 Hz data, the $D_{norm}$ map highlights myofascial interfaces with greater sharpness and more distinct layer separation than NOSS for all inter-muscular myofascial interfaces in the lower back, whereas NOSS appears more diffuse and less defined. These findings were consistent across all volunteers, as shown in the comparison between 30 Hz, 60 Hz, and 90 Hz in Supplementary Figure S4. The results consistently demonstrate that the $D_{norm}$ map at 30 Hz provides superior slip boundary clarity than the NOSS map, supporting the observation that $D_{norm}$ at lower frequencies (e.g., 30 Hz) is more effective for SII visualization of lower back muscles.

3.2. Within-and Between-day repeatability of ${\mathit{D}}_{\mathit{n}\mathit{o}\mathit{r}\mathit{m}}$ and NOSS biomarkers

Across the 12 anatomical ROIs (Figure 4), both biomarkers showed high repeatability, with all mean wCV values below 10% except for the left side of L3 ES-M when using NOSS. Despite this overall performance, $D_{norm}$ exhibited systematically lower variability than NOSS across regions and for both time scales. The mixed-effects analysis confirmed a significant main effect of biomarker (p = 0.0001), indicating that $D_{norm}$ and NOSS differed in their repeatability performance. No significant effects were found for measurement timing or for the interaction between biomarker type and timing. Post-hoc pairwise comparisons showed that $D_{norm}$ exhibited significantly lower mean wCVs than NOSS in both within-day (p = 0.033) and between-day (p = 0.0001). These results suggest that $D_{norm}$ provides more repeatable estimates of intermuscular mobility than NOSS while maintaining excellent repeatability across all examined ROIs.

Figure 4:

Repeatability comparison of $D_{norm}$ and NOSS biomarkers across 12 anatomical ROIs (ES-M, QL-ES, Left (L) and Right (R) side, L3, L4 and L3-L4 level). Mean within-subject coefficient of variation (wCV, %) for each ROI is shown as colored jitter points, with box plots representing group-level distributions for each biomarker: pink for NOSS and blue for $D_{norm}$. The left panel displays within-day repeatability, and the right panel shows between-day repeatability. The p-values on each plot indicate that $D_{norm}$ exhibited significantly lower mean wCV values compared to NOSS, demonstrating superior measurement repeatability across regions.

3.3. Comparison of ${\mathit{D}}_{\mathit{n}\mathit{o}\mathit{r}\mathit{m}}$ for the CLBP patients within the MPS spectrum to healthy volunteers.

Representative $D_{norm}$ maps at 30 Hz for healthy participants and CLBP patients within the MPS spectrum at the L3-L4 vertebral levels are shown side by side in Figure 5 and Supplementary Figure S5. In healthy participants, $D_{norm}$ maps show clearer slip interfaces that can be easily traced along myofascial interfaces, suggesting free mobility with minimal restriction. In contrast, CLBP patients exhibit more disrupted patterns with partial loss of the slip interface along the muscle boundary, particularly toward the ES-M interface at the L4 level. This is better visualized in the coronal view comparison, indicating potential myofascial restrictions with restricted mobility.

Figure 5:

Example $D_{norm}$ map comparison between a CLBP patient (female (F), age 27, BMI 27.9, PEG 4, pain duration 48 months) in the first row and an age-, sex-, and BMI-matched healthy participant (F, age 28, BMI 27.9) in the second row. The first three columns show axial (Ax) L3-L4 $D_{norm}$ maps, sagittal (Sag) L3-L4 vertebrae images, and the corresponding L3-L4 level of coronal (Cor) $D_{norm}$ sections, respectively. The patient shows low intensity at L4 ES-M interface on $D_{norm}$. Blue, pink, white, and yellow arrows highlight the right QL-ES, right ES-M, left ES-M, and left QL-ES muscle interfaces, respectively. All axial and coronal images are overlaid on corresponding T2W images and MRE magnitude images, respectively.

Moreover, a significant group difference was observed at ES-M at L4 (p = 0.0411), with CLBP-MPS patients exhibiting lower mobility indices compared to controls (Figure 6, Left Panel). No significant differences were found at other interfaces (p > 0.05). In addition, Figure 6 illustrates that $D_{norm}$ provides better discrimination between patients and healthy volunteers, further supporting its superior performance demonstrated in the repeatability analysis (Section 3.2). Notably, all CLBP patients included in this study exhibited normal findings on conventional MRI, with no structural abnormalities detected.

Figure 6:

Box plots of $D_{norm}$ (left) and NOSS mobility indices (right) at the L4 ES-M interface in CLBP-MPS patients (PT, red) and healthy volunteers (HV, blue). The $D_{norm}$ mobility index showed clearer separation between groups than the NOSS mobility index.

4. Discussion

In this study, we advanced the SII technique for assessing myofascial mobility in the lower back, providing the first visualization of the inter-muscular myofascial slip interfaces between the QL-ES and ES-M boundaries at the L3-L4 vertebral levels. Our results suggest that a lower vibration frequency (e.g., 30 Hz) is more effective for SII imaging of the lower back, producing SII output with better contrast compared to higher frequencies (60, 90 Hz). Moreover, the $D_{norm}$ map, a newly proposed SII biomarker, addressed the limitations of the previously developed NOSS map in visualizing the complex inter-muscular boundaries of the lower back and demonstrated superior repeatability. Finally, in this pilot clinical study, preliminary findings suggest that the $D_{norm}$ map may help distinguish between healthy participants and MPS patients with CLBP, as patients exhibited significantly reduced mobility and more disrupted inter-muscular slip interfaces, particularly at the ES-M boundary L4 level. These findings further demonstrate the potential of the $D_{norm}$ map as an effective tool for assessing myofascial mobility in patient management.

The NOSS and $D_{norm}$ maps acquired at 30 Hz exhibited more distinct slip interfaces and greater contrast at the inter-muscular boundaries than those obtained at higher frequencies. This supports a previous SII study in the thigh (19), where lower frequencies (e.g., 40 Hz) were used for SII imaging, in contrast to the higher MRE frequencies (such as 60 Hz) typically used for the liver (20) and brain (21), or for muscle stiffness measurements above 90 Hz (27–29,39). Although the vibration amplitude was lower at 60 Hz (35-50%) and 90 Hz (60-80%) than at 30 Hz (95-100%), with the amplitude levels chosen to ensure patient comfort (since higher power at 60 or 90Hz (95-100%) can be intolerable), the difference in SII performance is not primarily due to amplitude. Instead, it reflects the fundamental wave propagation characteristics at different frequencies. While higher frequencies in MRE offer better resolution for stiffness estimation, they are limited by reduced wave penetration and increased attenuation, as well as greater wave interference due to shorter wavelengths and the formation of low-amplitude nodes, resulting in noise and less uniform displacement patterns. In contrast, the 30 Hz frequency generated greater and more uniform displacement in the muscle region, allowing for better assessment of potential discontinuities across the inter-muscular interfaces.

It should be noted that although 30 Hz was chosen in this study, the selection of a low frequency is not strictly limited to this value. Lower frequencies are generally preferable for improved wave penetration and interface visualization; however, they result in longer TE times, extended scan durations, and reduced patient comfort. Thus, 30 Hz was selected as a compromise to optimize image quality while maintaining a reasonable TE and ensuring patient comfort. Future work could explore the directional characteristics of slip interfaces to identify the most influential displacement components, which may allow for faster, direction-specific 2D acquisitions without compromising diagnostic performance. Additionally, the choice of excitation frequency is influenced not only by tissue characteristics but also by the mechanical performance of the MRE driver system (40). At lower frequencies, such as 30 Hz, the membrane of the pneumatic driver may operate near its eigenfrequency (40), potentially affecting the uniformity and amplitude of the transmitted vibrations. These variations can in turn influence wave propagation and displacement measurements. While 30 Hz was found to enhance wave penetration and facilitate interfacial motion in this study, future investigation is warranted to understand how driver mechanics across different frequencies may affect the accuracy and reproducibility of SII-based measurements.

Using the same 30 Hz data, the $D_{norm}$ map provided clearer and sharper contrast for the inter-muscular myofascial interfaces than the NOSS map. While NOSS has proven effective in other anatomical regions, its performance in the lower back appears limited, likely due to the complex wave patterns in the multilayered and multidirectional lumbar muscles. In contrast, $D_{norm}$ enhances interface visibility by emphasizing displacement gradients and reducing wave interference through directional filtering (36). Specifically, we found that merging opposing vectors into 10 bi-directional filters improved image clarity compared to the original 20-direction filter, likely by better preserving wave interactions across myofascial boundaries (Supplementary Figure S2). The final maximal intensity projection further highlights the most prominent displacement discontinuities. Together, these steps contribute to the improved visualization of inter-muscular structural integrity and mobility gradients in the lower back.

In addition to enhanced image clarity, the $D_{norm}$ map demonstrated greater measurement repeatability than the NOSS map, with significantly lower wCVs across 12 inter-muscular ROIs in both within-day and between-day scans. All wCVs indicated excellent repeatability (below 10%) for $D_{norm}$ biomarkers, and $D_{norm}$ consistently yielded more stable mobility index measurements. Despite the small sample size (n = 5), these preliminary results support its potential as a reproducible and reliable biomarker for longitudinal monitoring and clinical application.

The $D_{norm}$ map showed significantly reduced mobility in CLBP patients within the MPS spectrum at the ES-M interface at L4. The brighter intensity observed in the $D_{norm}$ maps of healthy participants likely reflect greater mobility and displacement discontinuities at the myofascial interfaces, whereas the absence or heterogeneous intensity observed in patients, particularly at the ES-M interface at the L4 level, may indicate the presence of myofascial restrictions or other dysfunctions contributing to persistent pain. This supports our hypothesis that myofascial abnormalities in CLBP patients within the MPS spectrum leads to decreased fascia mobility. Previous ultrasound shear wave elastography studies have shown significantly greater stiffness of the multifidus muscle at rest, measured at the L4 (13) and L3-L4 levels (14) in individuals with lower back pain. Murillo et al. reported a potential association between the increased stiffness and the accumulation of fibrotic proliferation in the adjacent connective tissue (14). This may be linked to the partial absence of slip interfaces observed at the ES-M L4 level in patients in this study, suggesting a potential correlation between increased stiffness and reduced mobility. In addition, other previous studies have reported that the axial section at the L3-L4 level shows an indistinct ES-M muscle boundary with interdigitating fibers that lead to complications in distinguishing where one muscle ends and the other begins (41,42), resulting in challenges for assessing potential dysfunction at the boundary using conventional MRI or ultrasound (42,43). Therefore, the different SII intensity patterns observed at the inter-muscular interfaces in CLBP patients may indicate potential functional abnormalities rather than mere anatomical muscle separation. SII revealed distinct impairments in myofascial mobility, indicating dysfunction that standard imaging fails to capture. These findings suggest that SII could be a sensitive tool for identifying subtle myofascial impairments that may contribute to chronic pain. Preliminary results also suggest that the $D_{norm}$ map could serve as a mobility index for clinical evaluation and management of CLBP patients within the MPS spectrum after further validation.

In addition to the ES-M interface, we also examined the QL-ES interface; however, no statistically significant differences were observed between groups. This may be due to the small sample size, the heterogeneity of the patient population, and both anatomical and functional factors. The QL contributes less to lumbar stabilization than the ES and M muscles, likely acting more as a passive force junction due to its position and fiber orientation (44). Further studies with larger patient populations are needed to evaluate the feasibility and clinical utility of assessing mobility across all myofascial interfaces in the lower back, beyond the ES-M interface.

There are several limitations in this study. First, although statistical analyses were performed for both biomarker repeatability and group comparisons, the small sample size limited the overall statistical power and the scope of inference. This reflects the pilot nature of the study, which focused primarily on the technical development and initial feasibility of the $D_{norm}$ biomarker. Additionally, heterogeneity within the small participant groups reduces the generalizability of the findings. Future studies should recruit a larger and more demographically balanced cohort matched for sex, age, and BMI for both healthy individuals and CLBP patients within the MPS spectrum to improve statistical power, support more robust quantitative analyses, and better characterize inter-muscular mobility across diverse populations, including analyses of associations between the $D_{norm}$ mobility index, PEG score (33), pain duration, sex, and age. Second, while a preliminary mobility index quantification method was introduced, future work should refine and validate this approach through repeatability assessments and pilot clinical investigations in a larger sample. Third, this study did not include MRI sequences to quantify fat infiltration or edema, which may influence muscle condition and myofascial mobility. Incorporating multi-echo Dixon for fat quantification (47) and T2 mapping for edema detection (48) in future studies could enhance the interpretation of SII-based biomarkers and help identify confounding physiological factors. Fourth, this study did not collect data on participants’ physical activity levels, which may influence both lower back pain (49) and myofascial interface mobility. The absence of a physical activity classification (e.g., sedentary vs. active) limits our ability to assess its potential confounding effects. While our primary focus of this study was technical feasibility and image-based biomarker development, future clinical studies aiming to explore the mechanistic underpinnings or treatment responsiveness of myofascial mobility should consider incorporating standardized physical activity assessments to better interpret individual variability in $D_{norm}$ metrics. Fifth, although this study focused on the QL-ES, a component of the TLF, and the ES-M interfaces, other regions of the lower back such as the posterior portion of the TLF that may be anatomically relevant and potentially contribute to impaired mobility should also be explored to gain a more comprehensive understanding of myofascial dysfunction in chronic low back pain. Further research is also needed to examine the clinical relevance of these imaging biomarkers in relation to treatment outcomes. Understanding how changes in myofascia mobility, as captured by SII imaging, correlate with patient-reported symptoms and functional improvements post-treatment could offer valuable insights into how these biomarkers can guide therapeutic interventions.

5. Conclusion:

The study demonstrated that MRE-based SII can effectively visualize mobility at inter-muscular myofascial slip interfaces in the lower back. By incorporating a lower vibration frequency and the newly introduced SII biomarker, the $D_{norm}$ map, SII provides clearer interface delineation and greater repeatability than the previously developed NOSS map and higher frequencies. Furthermore, SII shows significant potential in differentiating between healthy and adhesive myofascia in CLBP patients within the MPS spectrum. With further clinical validation, SII holds promise in both research and clinical practice, offering a unique biomarker for diagnosis, monitoring, and ultimately aiding improvement of treatment strategies for chronic low back pain.

Supplementary Material

Supporting_Information_FigureS1

Figure S1: Comparison of magnitude images highlighting the benefit of the dual-saturation dual-sensitivity motion encoding (DSDM) scheme in the lower back. The DSDM-derived magnitude image (center) is shown alongside the T2-weighted anatomical image (left) and the conventional MRE magnitude image (right). The thoracolumbar fascia (TLF), indicated by a yellow arrow and visualized adjacent to the posterior border of the erector spinae (ES) and the multifidus (M) muscles in the T2-weighted image, remains clearly visible in the DSDM magnitude image but becomes difficult to identify in the conventional MRE magnitude image.

Supporting_Information_FigureS2

Figure S2: Effect of directional filtering (DF) technique on $D_{norm}$ maps in seven healthy participants: The first and second columns show $D_{norm}$ maps acquired at 30 Hz with the use of 10-bidirectional DF (10 DF) and 20-direction DF (20 DF), respectively. All $D_{norm}$ maps are overlaid on the corresponding T2W image of each participant.

Supporting_Information_FigureS4

Figure S4: Effect of frequency on NOSS and $D_{norm}$ maps in six other healthy participants: The first three columns show NOSS maps at 30 Hz, 60Hz, and 90Hz, while the last three columns display $D_{norm}$ maps at 30 Hz, 60 Hz, and 90 Hz, respectively. A yellow box highlights the $D_{norm}$ map at 30 Hz, which shows superior results compared to the other cases. All NOSS and $D_{norm}$ maps are overlaid on the corresponding T2W image of each participant. Blue, pink, white, and yellow arrows highlight the right QL-ES, right ES-M, left ES-M, and left QL-ES muscle interfaces, respectively.

Supporting_Information_FigureS5

Figure S5: Comparison of $D_{norm}$ maps acquired at 30 Hz for the remaining five CLBP participants (the representative example is shown in Figure 4) within the MPS spectrum and age-, sex-, and BMI-matched healthy participants. The first and last three columns show healthy participants and patients, respectively, including axial (Ax) L3-L4 $D_{norm}$ maps, sagittal (Sag) L3-L4 vertebrae images, and the corresponding L3-L4 level of coronal (Cor) $D_{norm}$ sections. Each row represents a participant matched by sex (female (F) or male (M)), age, and BMI, with the first and second numbers indicating age and BMI, respectively. Additionally, the PEG score and pain duration in months (mth) are provided for each patient. Blue, pink, white, and yellow arrows highlight the right QL-ES, right ES-M, left ES-M, and left QL-ES muscle interfaces, respectively. All axial and coronal images are overlaid on corresponding T2W images and MRE magnitude images, respectively.