Quantification of muscle morphology and stiffness in spastic hemiplegic cerebral palsy versus typically developing controls using magnetic resonance elastography

Abstract

Background:

Increased muscle stiffness is common in cerebral palsy, but accurate and reliable methods for quantitative assessment are not established for cerebral palsy. Magnetic resonance elastography provides a robust approach to mapping skeletal muscle properties. This study aimed to assess muscle morphology and stiffness in the gastrocnemius-soleus complex in children with hemiplegic cerebral palsy, comparing more affected and less affected sides, and to legs of typically developing children.

Methods:

Eleven children (age 11.5 ± 4.1 years) with spastic hemiplegic cerebral palsy (Gross Motor Function Classification System levels I/II; no botulinum toxin injections within 6 months) and fifteen typically developing children (age 13.1 ± 4.5 years) were prospectively enrolled. Magnetic resonance elastography measured both legs in participants with cerebral palsy and the dominant leg in controls. Additional imaging assessed muscle volume, cross-sectional area, and contractile to non-contractile tissue ratio. Normalized stiffness and anatomical features were compared between groups. Pearson correlations evaluated associations with age and between imaging measures.

Findings:

Normalized stiffness in soleus was significantly higher in the more affected leg in unilateral cerebral palsy than the less affected leg (471 vs 328 Pa/cm², p = 0.010) and the contractile tissue ratio also differed significantly (7.44 vs 10.25, p = 0.002). Gastrocnemius and soleus cross-sectional areas were reduced in the more affected side relative to controls, more markedly in soleus. Soleus stiffness showed significant positive correlations with age.

Interpretation:

Results suggest increased stiffness and reduced contractile tissue in the soleus of the more affected leg in cerebral palsy, supporting the utility of magnetic resonance elastography in characterizing muscle pathology.

1. Introduction

Cerebral palsy (CP) occurs secondary to an insult to the developing brain, resulting in musculoskeletal impairments that are progressive with growth. The spastic type of CP is the most common and is associated with a velocity-dependent increase in muscle stiffness that precedes the development of fixed muscle contracture – a permanent shortening of the muscle tendon unit (MTU) even when relaxed (Howard and Herzog, 2021). Increased muscle stiffness is common in CP, and while emerging imaging studies (Åhblom et al., 2024; Joshi et al., 2025) have begun to quantify whole-muscle biomechanical properties in vivo alongside clinical measures, reliable non-invasive methods that provide muscle-specific biomechanical information in the lower limb (e.g., gastrocnemius vs. soleus) are still needed to optimally inform surgical management.

Clinical assessment of fixed muscle contracture in CP has historically been achieved by goniometer measurement with identification of a reduction in joint range of motion (ROM). Recent work in ambulatory children with CP has quantified significantly reduced passive ankle dorsiflexion in the most-affected leg compared to the less affected leg, with median values of 2.5° vs. 12.5°, respectively (Åhblom et al., 2024). At the microscopic level, ex vivo mechanical testing of single muscle fibers has quantified underlying tissue changes; the Young’s modulus of fibers in children with CP (55 ± 6.6 kPa) is nearly double that of typically developing peers (28.3 ± 3.3 kPa) (Friden and Lieber, 2003). Further investigations into the extracellular matrix (ECM) have quantified an even higher Young’s modulus of approximately 205 kPa in CP muscle bundles, which can be reduced to 100 kPa following collagenase treatment, highlighting the role of non-contractile tissue in muscle stiffness in CP (Howard et al., 2023; Smith et al., 2011).

While these biopsy-based findings provide a precise quantification of the biological drivers of stiffness, translating these micro-scale measurements to a functional, macro-scale context remains a challenge. For example, in the gastrocnemius and soleus, stiffening of individual muscle fibers does not always correlate with fiber bundle stiffness measured ex-vivo (Howard and Herzog, 2021; Mathewson et al., 2014). To bridge this gap, in vivo investigations have sought to quantify the mechanical properties of the whole muscle-tendon unit. Using isokinetic dynamometry and ultrasound, Barber et al. (2011) quantified 51% higher ankle stiffness and 47% lower medial gastrocnemius fascicle strain in CP, indicating a functional inability of the tissue to elongate. More recently, ultrasound shear wave elastography (SWE) has been used in this context, as a non-invasive tool to quantify in vivo stiffness; however, reports of greater stiffness in the gastrocnemius and soleus exhibit high heterogeneity, ranging from 13% to 200% higher than controls (Brandenburg et al., 2016; Casellas-Vidal et al., 2025; Lallemant-Dudek et al., 2021; Lee et al., 2016). This variability, combined with the limited field-of-view and depth penetration of ultrasound, underscores the need for a comprehensive, 3D volumetric approach to accurately map stiffness across the entire calf muscle volume in hemiplegic CP.

Magnetic resonance elastography (MRE) provides an alternative, non-invasive, approach to characterizing soft tissue stiffness (Muthupillai et al., 1995). Over the past three decades, MRE has emerged as a robust method for mapping spatially resolved mechanical properties across multiple organs and has demonstrated sensitivity to pathology, particularly in the liver and brain (Hiscox et al., 2021; Mariappan et al., 2010; Murphy et al., 2019). In the liver, MRE is widely used for fibrosis assessment, with published comparisons indicating diagnostic performance that is often superior to ultrasound-based elastography (Yin and Venkatesh, 2018). A recent multicenter phantom study further demonstrated excellent same-session precision and test-retest repeatability of MRE stiffness measurements across sites and scanners when protocols were standardized (Ozkaya et al., 2024). More recently, MRE has been applied to study skeletal muscle under passive and active loading conditions (Ghatas et al., 2021; Green et al., 2012, 2013; Guo et al., 2016; Kennedy et al., 2017; Kennedy et al., 2020; Smith et al., 2023). MRE is a phase contrast MRI technique that utilizes externally applied cyclic perturbations, synchronized to an oscillating motion-encoding MR gradient, to quantify shear wave propagation (displacement fields), from which shear modulus and stiffness measures can be derived. MRE has been used in both children and adults to characterize passive mechanical properties of skeletal muscle (Debernard et al., 2011; Kennedy et al., 2020) and to detect alterations in neuromuscular disease, including spastic paraplegia (Basford et al., 2002) and Duchenne muscular dystrophy (Bensamoun et al., 2015). Hence, MRE shows promise as a sensitive technique to quantify muscle stiffness in CP, leveraging volumetric coverage of multiple muscles within a single field of view using standardized protocols with minimal operator dependence.

The purpose of this study was to use MRI and MRE to measure muscle morphology and stiffness in the gastrocnemius-soleus muscle complex from children with hemiplegic CP, comparing more affected (MA) to less affected (LA) sides, and to typically developing control participants (TDC). We hypothesized that calf muscle stiffness, as measured by MRE, will be increased in the MA side in hemiplegic CP as compared to the LA side and TDC. This change will be associated with shifts in tissue composition, represented by the ratio of contractile to non-contractile tissue.

2. Methods

2.1. Participants

Participants were prospectively recruited from the Cerebral Palsy Clinic (CP group) and the Fracture Clinic (TDC group) at a tertiary level children’s hospital (Nemours Children’s Hospital, Wilmington, DE). Inclusion criteria for the CP group were children with spastic hemiplegic CP, aged 5–18 years, Gross Motor Function Classification System (GMFCS) scores of I or II (Palisano et al., 1997), no use of walking aids other than below-knee bracing, and no requirements for sedation during imaging. Participants were excluded if they had received botulinum toxin type A (BoNT-A) injections into the more affected calf within the preceding 6 months, consistent with prior studies suggesting substantial recovery of muscle tissue over this interval (Deschrevel et al., 2024). Two participants in the CP group received BoNT-A injections during early childhood, approximately 10 years prior to imaging. No participants received injections within the 6 months preceding the study. This prior exposure is noted for transparency, and its potential long-term implications are considered in the Discussion. Inclusion criteria for the TDC group were children, aged 5–18 years, who were generally healthy. Some participants in the TDC group had previously incurred upper extremity fractures, but these injuries were unrelated to lower extremity function or mobility and did not affect the in vivo, intrinsic muscle properties being studied. This study was approved by the Nemours Delaware Valley Institutional Review Board, and informed written consent was obtained from all parents or guardians of participants prior to study commencement.

The age range of 5–18 years was selected based on two key considerations: developmental phases of muscle growth and the clinical availability of pediatric patients with CP. Muscle development undergoes significant changes during two rapid growth phases, with notable increases in muscle strength, size, and morphology occurring particularly during adolescence (Chow et al., 2024; Peeters et al., 2023). This period is critical for understanding the musculoskeletal differences between typically developing children and those with CP, where neuromuscular adaptations may be most pronounced (Kruse et al., 2018; Willerslev-Olsen et al., 2018). Clinically, the majority of pediatric patients with CP undergoing regular monitoring and intervention are in this age range, making it a practical and relevant cohort for studying disease impact and treatment outcomes.

Of the 15 children with cerebral palsy initially recruited for imaging, one participant was unable to complete the scan protocol due to discomfort and did not return for a rescan. Two additional participants exhibited excessive motion artifacts and image degradation caused by restlessness and spasticity, rendering their data unusable. One further participant underwent partial image acquisition, but the session was interrupted by a scanner malfunction, and the session could not be completed due to time constraints. As a result, 11 participants with CP successfully completed the full imaging protocol and were included in the final analysis.

2.2. Image acquisition and processing

Participants completed an MRI scanning protocol on a GE Signa 3 T PET/MR scanner (GE Healthcare; Waukesha, WI). The imaging protocol consisted of an MR elastography sequence to measure mechanical properties and additional imaging to determine the anatomical characteristics of muscles. All imaging data were acquired using a flexible RF receive coil covering the lower leg from the knee to the ankle along the vertical direction. During the MR elastography acquisition, a passive paddle driver is used to deliver small-amplitude vibrations to the calf via a pneumatic actuator (Resoundant; Rochester, MN), generating shear waves that displace the muscle tissue at an 80 Hz frequency (Fig. 1). The MRE scans were acquired with a phase contrast, single-shot 2D-echoplanar imaging (EPI) sequence that incorporates motion encoding gradients synchronized to the external vibrations to image full vector displacement fields throughout the gastrocnemius-soleus muscle complex at 2.25 × 2.25 mm2 in-plane resolution and 4 mm slice thickness, and a TR/TE = 5602/49.7 ms. During the MRE experiment, mechanical vibrations were applied using a paddle driver, positioned against the proximal portion of the lateral head of the gastrocnemius muscle belly, with the participant’s leg resting on top of the driver. This placement likely resulted in localized compression of the lateral gastrocnemius during acquisition. Because such compression may introduce a prestrain prior to steady-state shear wave generation and bias stiffness estimates toward higher apparent values, the lateral gastrocnemius was excluded from subsequent quantitative analysis. MRE was performed on more affected (MA) and less affected (LA) legs for the CP group, and on the dominant leg in the TDC group, with representative images shown in Fig. 2. The approximate scan durations for each sequence (single-leg acquisition) were: MRE = 4:48 min, diffusion weighted imaging (DWI) = 4:29 min, T1-weighted = 3:12 min, and T2-weighted Dixon = 5:14 min. DWI data was collected for all participants but was not included in the analysis for this study. For participants with CP who underwent bilateral scans, the total protocol time was approximately 36 min. For the typically developing control participants with only one leg scanned unilaterally, the protocol duration was approximately 18 min.

Fig. 1.

Experimental setup for the MR Elastography acquisition of the lower leg muscles. Participant is positioned in 3 T PET/MR scanner in feet-first supine position with a flex coil secured over the lower leg, covering from below the knee to the ankle. A paddle driver connected to a pneumatic actuator (Resoundant; Rochester, MN) delivers 80 Hz steady-state vibrations to the calf muscle belly during MRE acquisition.

Fig. 2.

Example of MRE stiffness maps of both more affected (MA) versus less affected (LA) legs from a patient with hemiplegic CP (17 y, M) compared to a typically developing control (TDC) (15 y, F). Shear wave motion in micrometers and shear stiffness measured in kilopascals derived from MRE image acquisition are depicted in the top and middle rows, respectively. Note the higher amount of red shading on the LA leg MRE shear stiffness image, consistent with a stiffer soleus primarily. Dixon-based fat fraction images from MRI are also shown (bottom row). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Prior to estimating muscle mechanical properties, the data was processed using FSL (FMRIB Software Library) version 6.0.5.1 (Smith et al., 2004) to perform phase unwrapping. After phase unwrapping, the data was treated with a subzone based nonlinear inversion algorithm (NLI) to estimate tissue mechanical properties (McGarry et al., 2012). NLI produces spatial maps of mechanical properties from imaged wavefields through an optimization-based inverse solution of the governing wave equation. This implementation of NLI assumes the material is viscoelastic as represented by a complex shear modulus, G* = G’ + iG”, and isotropic such that properties are assumed equivalent for any direction of displacement. The outcome of NLI is an estimate of G* at each image point across the muscle. From G* we calculate and report the value of shear stiffness, μ = 2*|G*|2/(G’ + |G*|), which is a composite measure describing the wave speed in a viscoelastic solid, measures the resistance of the tissue to shearing deformations, and is commonly reported in MRE literature (Manduca et al., 2021).

Anatomical T1-weighted images were acquired using a 2D spin-echo sequence with a field-of-view (FOV) of 180 mm², a matrix size of 256 × 256, and an in-plane resolution of 0.35 × 0.35 mm², 4 mm slice thickness, and TR/TE = 522/8.24 ms. These high-resolution images were used to aid the segmentation process and delineate muscle boundaries. T2-FLEX Dixon-based MRI scans were performed using a 2D fast spin echo (FSE) sequence with the following parameters: FOV of 180 mm², a matrix size of 256 × 256, an in-plane resolution of 0.70 × 0.70 mm, and a slice thickness of 5 mm, and TR/TE = 2893/89.04 ms. These parameters were optimized to produce separate “Fat,” “Water,” and “In-Phase” images. For each dataset, the Dixon-based fat and water images were reoriented and registered to the MRE magnitude image using FSL FLIRT before segmentation. After registration, Dixon-based images were manually traced to segment regions of interest (ROI) and anatomical boundaries for each muscle: the medial head of the gastrocnemius and the soleus. All subsequent references to the gastrocnemius in the Results and Discussion sections refer specifically to the medial head. The water (W) and fat (F) images are combined to estimate fat fraction (FF) using the equation FF = F/(F + W).

The mean cross-sectional area (CSA) of the muscle was determined from MR images using a multi-step process. First, the muscle regions of interest (ROIs) for the medial head of the gastrocnemius and the soleus were segmented manually from the Dixon-based images. Muscle segmentation was performed on 16 slices at the midpoint of the muscle’s field of view, capturing the central portion of the muscle belly. The cross-sectional area for each slice was calculated, and these values were then averaged across all segmented slices to determine the mean CSA for each muscle.

To estimate the proportions of contractile muscle relative to intramuscular fat, we calculated the contractile-to-noncontractile tissue ratio (CTR), using the equation: CTR = CSA × (1 - FF). This equation is a modified version of contractile CSA reported previously (Wokke et al., 2014). To account for inter-individual differences in muscle size and overall anthropometry, shear stiffness values were additionally normalized to muscle cross-sectional area (CSA). Prior MRE studies have demonstrated a positive association between skeletal muscle stiffness and muscle size, including CSA, suggesting that apparent stiffness may scale with muscle geometry and loading conditions during vibrations (Ghatas et al., 2021). Normalization by CSA therefore reduces variability related to growth and body size and is consistent with approaches used in musculoskeletal studies of children with CP (De Beukelaer et al., 2022, 2023). For both shear stiffness and fat fraction, quantitative values were computed by averaging voxel-wise measurements across the entire segmented muscle volume for each muscle.

2.3. Statistical analysis

Statistical analyses were conducted using JMP Pro v.17 (SAS Institute, Cary, NC). Mixed linear models were used to compare muscle properties between limbs (MA vs. LA) within participants with CP and between groups (CP vs. TDC). For comparisons within the CP group (MA vs. LA), a paired design was used, with participant ID included as a random effect. For between-group comparisons (MA vs. TDC, LA vs. TDC), age and group were modeled as fixed effects to account for variability in participant age. Due to the limited sample size, sex and GMFCS level were not included in the models but are reported in Supplementary Tables S1–S2 for reference. All statistical tests were two-tailed, with significant assessed at p < 0.05. When statistically significant effects were identified, the direction of the difference was determined from model estimates and group means and is reported descriptively in the Results. Pearson correlation analyses were performed to examine associations between imaging derived measures and age, as well as between mechanical properties and anatomical characteristics. Specifically, we evaluated correlations between (1) shear stiffness, (2) fat fraction (FF), and (3) cross-sectional area (CSA) with participant age to characterize developmental trends. In addition, we assessed the relationship between shear stiffness and contractile tissue ratio (CTR) to explore whether increased contractile content was associated with greater muscle stiffness.

To account for observed age-related trends in shear stiffness (μ) and muscle cross-sectional area (CSA), as an exploratory post hoc age-standardization of CSA-normalized shear stiffness (μ/CSA) was performed for visualization purposes only. Specifically, for each group (MA, LA, TDC) and each muscle (gastrocnemius, soleus), a linear regression was fitted between age and μ/CSA. The resulting slope was used to adjust individual values to a reference age of 13 years, using the equation: μadjusted = μobserved-slope × (age-13). This procedure reduces the influence of age-related variability when visually comparing groups. Age-corrected μ/CSA values were used only for descriptive visualization (Fig. 5) and were not used for hypothesis testing, correlation analyses, or statistical inference.

Fig. 5.

Comparisons of age-corrected CSA-normalized shear stiffness between patients with CP (MA/LA sides) and typically developing controls (TDC) shown for gastrocnemius (medial head) and soleus muscles. Normalized stiffness values were corrected to a reference age of 13 years to account for age-related variations. This correction standardizes the CSA-normalized shear stiffness, allowing for group comparisons independent of age effects. (a) For gastrocnemius, there were no significant differences identified between MA/LA and TDC sides (p < 0.05; tested before age-correction). (b) For soleus, there were significant differences between MA and LA sides only (p = 0.014*; tested before age-correction).

3. Results

MRE/MRI scans were performed for 26 patients [CP: n = 11, age 11.5 (4.1) years, TDC: n = 15, age = 13.1(4.5) years; p = 0.34]. CP participants had median LA side dorsiflexion with knee flexed(extended) at 13.7°(5.1°) and MA side dorsiflexion with knee flexed(extended) 4.1° (−4.0°), in contrast to TDCs which had average dorsiflexion for left/right side of 15.5°(9.4°)/16.5°(9.5°). Participant demographics for patients with CP are listed in Table S1 and demographics for typically developing children in Table S2.

3.1. Anatomical characteristics

A significantly smaller CSA of the gastrocnemius muscle was observed in MA side of patients with CP compared to TDCs (4.2 ± 1.5 vs 7.4 ± 3.8 cm2, p = 0.015) but not between MA and LA with the CP group (4.2 ±1.5 vs 5.8 ±2.1 cm2, p = 0.09) nor LA and TDC (p = 0.34) (Fig. 3 (a)). Significantly smaller soleus CSA was found for MA compared to TDC (9.2 ± 2.0 vs 13.0 ± 6.2 cm2, p = 0.046) and MA compared to LA (9.2 ± 2.0 vs 12.6 ± 3.8 cm2, p = 0.002), but not between LA and TDC (p = 0.998) (Fig. 3(b)).

Fig. 3.

Cross-sectional areas for (a) gastrocnemius (medial head) and (b) soleus of the typically developing controls (TDC), versus the less affected (LA) and more affected (MA) sides of patients with CP. A significant difference in CSA between the MA side and TDC was observed in the gastrocnemius (p = 0.015*), whereas the MA side was significantly different than the LA side (p = 0.002*) and the TDC (p = 0.046*) in the soleus.

A significantly higher FF in the gastrocnemius was found in the LA side vs TDC (0.19 ± 0.04 vs 0.17 ± 0.02, p = 0.024) and for MA vs TDC (0.19 ± 0.04 vs 0.17 ± 0.02, p = 0.017), but no difference was found between MA and LA side (p = 0.96). For soleus, the only significant difference in FF was between MA side vs TDC (0.21 ± 0.04 vs 0.19 ± 0.03, p = 0.03).

The CTR was significantly lower for MA vs TDC sides for gastrocnemius (3.4 ± 1.3 vs 6.2 ± 3.1, p = 0.01) and soleus (7.4 ± 1.8 vs 10.8 ± 5.1, p = 0.024) (Fig. 4(a–b)). For MA vs LA sides, a significant lower CTR was found for soleus in the MA side (7.4 ± 1.8 vs 10.3 ± 3.3, p = 0.002) but not for gastrocnemius (p = 0.09).

Fig. 4.

Contractile-Non-contractile Tissue Ratio (CTR) comparisons between patients with CP (MA/LA sides) and typically developing controls (TDC) by muscles. (a) For gastrocnemius (medial head), a significant difference was identified between MA vs TDC sides only (p = 0.01*). (b) For soleus, there were significant differences between MA vs TDC (p = 0.024*) and MA vs LA (p = 0.002*).

3.2. Mechanical properties

There was no significant difference in mean CSA-normalized stiffness between MA and LA sides for gastrocnemius (804 ± 338 vs 586 ± 256 Pa/cm2, p = 0.1), but a significantly higher stiffness was found for soleus in the MA side (471 ± 180 vs 328 ± 63 Pa/cm2, p = 0.014). There were no significant differences in CSA-normalized shear stiffness between the MA or LA sides of the gastrocnemius or soleus muscles when compared to the TDC group (p > 0.05) (Fig. 5(a–b)).

Imaging parameters of the gastrocnemius and soleus across patient groups are summarized in Table S3.

3.3. Correlations with age and between tissue properties

There were significant correlations between age and soleus shear stiffness for LA (r = 0.63, p = 0.037), MA (r = 0.66, p = 0.020), and TDC (r = 0.82, p < 0.001), such that stiffness increases with age. But no associations were found between age and gastrocnemius stiffness for LA (r = 0.30, p = 0.376), MA (r = 0.22, p = 0.488), or TDC (r = 0.41, p = 0.141). No significant relationships were found for FF relative to age (all p > 0.05). Results from correlations between imaging parameters and age are summarized in Table 1.

Table 1.

Pearson correlation coefficients (r) and associated p-values for imaging parameters measured in the gastrocnemius and soleus muscles for all measurement groups with age: less affected (LA) and more affected (MA) sides of patients with CP and typically developing controls (TDC).

Imaging Parameter	Less Affected	More Affected	Typically Developing Controls
Gastrocnemius
Shear Stiffness	r = 0.30 p = 0.376	r = 0.22 p = 0.523	r = 0.41 p = 0.133
Cross Sectional Area	r = 0.65 p = 0.030*	r = −0.07 p = 0.830	r = 0.88 p < 0.001***
Fat Fraction	r = −0.06 p = 0.867	r = −0.27 p = 0.422	r = −0.17 p = 0.541
Soleus
Shear Stiffness	r = 0.63 p = 0.037*	r = 0.72 p = 0.012*	r = 0.77 p = 0.001**
Cross Sectional Area	r = 0.77 p = 0.006**	r = 0.30 p = 0.363	r = 0.86 p < 0.001***
Fat Fraction	r = 0.32 p = 0.338	r = 0.12 p = 0.726	r = 0.31 p = 0.255

Statistically significant correlations are indicated in bold, with significance levels denoted by asterisks

^*for p < 0.05,

^**for p < 0.01,

^***for p < 0.001.

Analyzing the relationship between stiffness and anatomical properties, we found significant correlations between shear stiffness vs CTR for the soleus only; LA side (r = 0.69; p = 0.02) and TDC (r = 0.88; p < 0.001), but not for the MA side (r = 0.06; p = 0.852). No significant correlations for shear stiffness vs CTR for gastrocnemius were identified (p > 0.05).

4. Discussion

The measurement of muscle stiffness using MRE in spastic hemiplegic CP provides novel insights into muscle pathology, particularly for the gastrocnemius and soleus muscles. Prior studies using ultrasound elastography and shear wave elastography have generally reported elevated gastrocnemius stiffness in CP relative to typically developing children; however, these approaches can be limited in their ability to reliably assess deeper muscles such as the soleus and to disentangle mechanical changes from shifts in tissue composition (e.g., contractile versus non-contractile contributions) (Mathewson et al., 2014). Given the critical role of the soleus in gait mechanics, including push-off and stability during walking (Wren et al., 2004), these limitations motivate methods that can standardize passive mechanical measurements across muscles within a single acquisition. Accordingly, we applied MRE within a multimodal MRI protocol to quantify passive shear stiffness in both the gastrocnemius and soleus during one scan session, alongside complementary measures of muscle structure and composition including cross-sectional area (CSA), contractile tissue ratio (CTR), and Dixon-derived fat fraction. Imaging both muscles in the same field-of-view enabled direct, internally consistent comparisons without the need for separate acquisitions per muscle. Using this framework, we found that the more affected (MA) limb exhibited higher stiffness and lower CTR compared with the less affected (LA) limb and typically developing controls (TDC), consistent with the sensitivity of MRE to disease-related muscle remodeling reported in other neuromuscular conditions (Bensamoun et al., 2015; Ringleb et al., 2007). Notably, the soleus demonstrated more pronounced differences than the gastrocnemius, including higher stiffness, reduced CSA, and greater non-contractile tissue burden in the MA limb relative to LA and TDC. Collectively, these results underscore the value of muscle-resolved mechanical assessment and integrated structure-composition metrics when interpreting plantarflexor involvement in hemiplegic CP.

Leveraging the cross-sectional coverage inherent to MRI-based MRE, we were able to quantify muscle mechanical properties of the gastrocnemius and soleus simultaneously. Unlike point-based or surface-constrained measurements, our approach ensures the delivery and imaging of the shear wave distribution throughout the entire field-of-view in a single acquisition. By capturing wave propagation across the full calf volume including the tibialis anterior, lateral and medial gastrocnemius, and the soleus, we can evaluate multiple muscle compartments simultaneously. This capability is particularly important for the lower leg, where the soleus lies deep to the gastrocnemius and is less readily evaluated with ultrasound-based elastography. We show that there are strong biomechanical effects in the soleus, and not the gastrocnemius in our sample of children with CP, highlighting the importance of MRE in being able to resolve this important muscle.

Prior work has more commonly quantified lower-limb muscle stiffness in CP using ultrasound elastography, including shear wave elastography (SWE), with most studies focusing on the gastrocnemius (Bilgici et al., 2018; Brandenburg et al., 2016;Lallemant-Dudek et al., 2021; Lee et al., 2016), and fewer studies reporting soleus measurements (Casellas-Vidal et al., 2025; Vola et al., 2018). Because elastography methods differ in both excitation and readout, it is important to distinguish what is being reported: US-SWE typically measures shear-wave speed (m/s) and may also report an elastic modulus (often in kPa) derived from wave speed under simplifying assumptions (e.g., material isotropy and a mapping between shear modulus and Young’s modulus). By contrast, MR elastography (MRE) measures motion fields induced by harmonic shear waves and uses an inversion to estimate a mechanical properties, conventionally reported as a shear modulus or shear stiffness (kPa), with the specific parameters depending on the material model and inversion algorithm. As a result, differences in joint angle (rest vs stretch), passive vs active state, transducer/driver placement, inversion assumptions, and reporting convention contribute to substantial variability in the magnitude of stiffness contrasts across studies even when the qualitative pattern is consistent. Across CP cohorts, these studies consistently report elevated gastrocnemius stiffness relative to typically developing controls or increased stiffness on the more affected side. Reported medial gastrocnemius stiffness differences between MA and LA limbs in CP span roughly ~13–76% across study protocols and reported metrics (Lallemant-Dudek et al., 2021; Lee et al., 2016), and MA limb versus control differences span ~118–203% (Bilgici et al., 2018; Lallemant-Dudek et al., 2021). Available soleus measurements likewise suggest elevated stiffness in CP relative to controls, with differences spanning ~37–69% (Casellas-Vidal et al., 2025; Vola et al., 2018). In our study, CSA-normalized shear stiffness differed by ~37% (MA vs LA) in the gastrocnemius (not significant) and ~ 43.6% (MA vs LA) in the soleus (p = 0.014), illustrating that muscle-specific contrasts observed with MRE can be placed in the broader context of elastography findings. Ultrasound-based elastography has also shown utility in CP for detecting stiffness changes tied to treatment effects, with the added advantages of portability and repeat bedside assessment for longitudinal monitoring (Cesaro et al., 2024). In contrast, MRE enables standardized, volumetric assessment across the full calf compartment and can be acquired alongside Dixon-based fat fraction and anatomical measures within a single protocol.

One key advantage of MRE is the ability to capture more detailed anatomical data and measures of tissue composition via other MRI sequences in the same scan session. In this current study, we measured fat fraction (FF) using Dixon-based imaging as a proxy for adipose infiltration within the muscle. From this, we derived the contractile tissue ratio (CTR), calculated as CTR = CSA × (1 - FF), to estimate the proportion of contractile muscle tissue. Prior biomechanical studies have indicated that a passive muscle stiffness reflects not only myofiber properties but also the surrounding extracellular matrix (ECM), which can bear a substantial fraction of passive load (Alnaqeeb et al., 1984; Friden and Lieber, 2003; Gillies and Lieber, 2011; Smith et al., 2011). In this study, use MRI-derived tissue composition measures to estimate and contractile tissue content (via FF and CTR) assumed to be indirect measures of ECM content and fibrosis. Consistent with greater non-contractile content in the more affected limb, the CTR for the MA side was approximately 45% lower in the gastrocnemius and 32% lower in the soleus compared to typically developing children; relative to the LA side, CTR was reduced by 31% in the gastrocnemius and 29% in the soleus, suggesting more severe pathological changes in the MA limb. Indeed, the soleus displayed significantly more pathologic involvement than gastrocnemius, particularly with respect to decreased muscle CSA, increased shear stiffness, and increased non-contractile tissue identified. It is important to note that FF measurements do not directly measure signal from fibrotic tissue components such as collagen, and thus, CTR does not directly quantify fibrotic tissue. Future work could also integrate recent advancements in diffusion MRI to quantify muscle architecture (e.g., pennation/fascicle geometry and physiological CSA) and provide complementary microstructural markers that may be sensitive to extracellular matrix remodeling in CP (Åhblom et al., 2024; Joshi et al., 2025).

In addition to an upregulation of extracellular matrix with its purported effect on overall muscle stiffness, muscle in CP is known to be smaller in both cross-sectional area and volume than TDC, consistent with reduced muscle growth (Howard and Herzog, 2021). Despite the observed reductions in soleus CSA on the MA side compared to the LA and TDC limbs, a positive age-related growth trend was still evident in our data for all groups. Previous studies have suggested that the development of clinical contracture may be due to a mismatch in muscle and bones growth rates, with muscle growing more slowly—potentially as a result of impaired satellite cell function and reduced myogenic capacity (Dayanidhi et al., 2015; Smith et al., 2013). While our cross-sectional study does not capture longitudinal growth trajectories, the presence of age-associated increases in CSA, including in participants with CP, is consistent with ongoing, albeit reduced, muscle development. Although the LA side is typically considered “normal” in spastic hemiplegia, the lack of significant differences in cross-sectional area—and fat fraction—between the MA and LA sides for gastrocnemius in our study suggests that this may not be entirely the case.

Contrary to our hypothesis, no significant differences in shear stiffness were found between the MA or LA limbs and the TDC group for the gastrocnemius muscle. This unexpected result may be due to the wide heterogeneity in stiffness within the TDC group, potentially influenced by variations in age, physical fitness, and tissue density rather than purely connective tissue differences. Normalizing stiffness values to CSA likely mitigates some of these variances but does not entirely account for the inherent physiological diversity.

In the current study, despite reduced ankle dorsiflexion on the MA side versus LA and TDC limbs, shear stiffness was found to be positively correlated with CTR for soleus on the LA side and in TDC, but not on the MA side. Additionally, soleus shear stiffness increased with age across both CP (MA and LA sides) and TDC groups, while no significant correlation with age was found in the gastrocnemius. These findings are counterintuitive given our original hypothesis that increased non-contractile tissue proportion (i.e., decreased CTR) would be associated with increased shear stiffness on the MA side in hemiplegia, akin to the clinical presentation where an overall shortening of the muscle-tendon unit would lead to decreased ankle range of motion. One possible explanation lies in the nature of MRE-derived shear stiffness, which may reflect intrinsic muscle mechanical properties—such as fiber and matrix composition—rather than directly capturing functional muscle excursion or clinical joint range of motion.

In typically developing children, the observed increase in soleus stiffness with age likely reflects healthy musculoskeletal maturation. Prior work has shown that muscle fibers undergo hypertrophy during growth, particularly in slow-twitch (Type I) fibers, which predominate in the soleus (Aherne et al., 1971; Andries et al., 2024). As fiber cross-sectional area increases, the muscle becomes more densely packed, contributing to shear modulus even in the absence of pathology. This mechanical response may also be influenced by increased muscle loading and activity with age, both of which contribute to passive stiffness. Additionally, developmental changes in the extracellular matrix such as collagen content, remodeling of collagen architecture, and increased cross-linking, have been shown to elevate passive stiffness in healthy muscle (Pavan et al., 2020; Wohlgemuth et al., 2024; Wood et al., 2014). Maturation of cytoskeletal proteins such as titin may also alter the elastic properties of muscle fibers (Mathewson et al., 2014). Together, these biological factors help explain why soleus stiffness may increase with age in typically developing children and underscore the importance of age-matched control groups in pediatric musculoskeletal imaging studies.

Our study quantified muscle shear stiffness with conventional MRE processing that assumes mechanical isotropy (i.e., direction-independent shear response). While this framework has enabled robust in vivo muscle stiffness mapping, emerging work increasingly supports modeling skeletal muscle as mechanically anisotropic, given its organized fiber architecture and extracellular matrix structure. Transversely isotropic inversion methods can estimate shear modulus components parallel and perpendicular to the local fiber direction (Guo et al., 2016; McGarry et al., 2021, 2022; Wang et al., 2025), improving physiological interpretability of stiffness differences by better separating fiber-associated mechanics from matrix-associated contributions. In lower leg muscles, anisotropic MRE has shown that these directional shear components can exhibit distinct responses under passive versus active loading conditions (Smith et al., 2023). In addition to inversion model selection, acquisition strategy is also likely to be important for future CP studies: delivering adequate, spatially distributed wave information in the presence of complex muscle geometry may benefit from multiple excitations and driver placements to improve directional coverage while avoiding compression of the muscle being evaluated. Prior work has demonstrated the value of multi-excitation approaches for providing richer wave fields needed for anisotropic reconstructions (Smith et al., 2020, 2022), and complementary strategies such as multi-frequency sampling may further stabilize parameter estimation (Guo et al., 2016). Collectively, incorporating anisotropic inversion alongside multi-directional excitation and acquisition strategies may improve sensitivity to muscle-resolved mechanical changes in CP.

4.1. Limitations and future directions

The primary limitations of this study include its small sample size, broad age range, and the lack of biopsy confirmation of MRE findings. No formal power analysis was performed given the exploratory nature of this pilot study. The sample size (n = 26) reflects feasibility constraints and is consistent with prior work in pediatric imaging studies. Effect sizes reported here may guide power calculations for future studies aiming to confirm these preliminary findings. The chosen age range of 5–18 years encompasses distinct phases of muscle development with significant change in muscle morphology, strength, and composition occurring naturally during growth and maturation (Parker et al., 1990). Nevertheless, the inclusion of a typically developing control group helps to address this limitation by providing a baseline for age-related muscle changes, allowing for better distinction between CP effects and normal growth patterns. Future studies may benefit from further stratifying analyses by narrower age groups or employing statistical techniques to adjust for age-related variability, enhancing the precision and applicability of the findings.

An additional limitation is that fat fraction measurements derived from Dixon imaging do not differentiate between adipose and fibrotic tissue. Consequently, the CTR measure reflects only an estimate of contractile muscle relative to fat, not the overall non-contractile tissue content. This distinction is particularly relevant in CP, where cellular and tissue-level studies of contractured muscle support a pro-fibrotic phenotype and altered regenerative capacity, consistent with increased extracellular matrix involvement beyond adipose infiltration (Loomis et al., 2025). T1 and T2 mapping techniques may offer complementary insights but are also limited in isolation—T2 values are elevated by both edema and fatty infiltration, while T1 values are influenced by the complex interplay of water and fat content, making it difficult to isolate fibrotic changes (Chianca et al., 2023). Future studies could incorporate T1ρ mapping—an emerging MRI technique sensitive to proteoglycan and collagen content—which has shown promise in selectively detecting fibrotic tissue and may enhance characterization of non-contractile muscle infiltration beyond what conventional imaging can provide.

Although no participants had received BoNT-A injections within 6 months of imaging, two children in the CP group were treated with BoNT-A during early childhood, approximately 10 years prior to participation in the current study. While direct pharmacologic effects of BoNT-A are unlikely to persist over such an interval, early exposure may influence long-term muscle development, including patterns of fat infiltration, stiffness, and extracellular matrix remodeling (De Beukelaer et al., 2022; Uzun Akkaya et al., 2025), which may impact the imaging measures described in this study. Future studies incorporating more recent BoNT-A treatment cohorts may help evaluate how muscle stiffness and composition evolve in response to injection-based interventions. Quantifying such changes non-invasively using MRE may offer valuable insight into treatment response and guide individualized therapeutic planning.

Finally, true passive conditions cannot be fully ensured during MRE acquisition in this population; residual activation (e.g., spastic tone or guarding) or passive tension related to positioning may increase measured stiffness and contribute to between-subject variability despite a standardized setup across participants (Smith et al., 2023). Future work incorporating MRE scans at multiple joint angles and controlled loading conditions would help disentangle intrinsic tissue mechanics from state-dependent effects and more fully characterize the biomechanical response in children with spastic hemiplegic CP.

5. Conclusions

Magnetic resonance elastography (MRE) has proven to be a valuable tool in this study, providing detailed and non-invasive quantification of muscle stiffness and morphology in the gastrocnemius-soleus complex of children with spastic hemiplegic CP and typically developing controls. In the soleus, stiffness increased with age across all groups, and significant correlations between stiffness and contractile tissue ratio were observed in the less affected limb and in controls, but not in the more affected limb; suggesting altered mechanical-compositional coupling in pathologic muscle. In contrast, while the gastrocnemius showed morphological differences between the more affected limb and controls, there were no significant differences from the less affected side, suggesting bilateral alterations to the muscle.

The ability of MRE to assess both superficial and deep muscles, and to integrate with complementary MRI-derived measures such as fat fraction and CSA, supports its potential to monitor muscle pathology in CP. Future applications may include longitudinal assessment of disease progression and response to interventions such as BoNT-A or surgical lengthening. Overall, MRE emerges as a promising modality for both clinical research and the management of pediatric neuromuscular disorders.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.clinbiomech.2026.106798.