A total of 4 male participants with complete SCI, aged 30 (SD 8) years, were recruited from the Istituto di Ricovero e Cura a Carattere Scientifico Eugenio Medea. The inclusion criteria required participants to have a complete SCI (more than 6 month and less than 5 years after the lesion; American Spinal Injury Association Impairment Scale of A or B; lesion level ≤ T₃), an age between 18 and 65 years old, and the ability to engage in the FES-cycling training program.

At baseline (T₀) demographic and clinical data were collected from each participant, including age, height, BMI, time since injury, lesion level (American Spinal Injury Association Impairment Scale), previous experience with FES and trike or cycling after the injury, current pharmacological therapy. Furthermore, spasticity was assessed using the Modified Ashworth Scale (MAS) and no severe levels were found at T₀ for all participants (MAS score for each participant <2).

The participants underwent a 6-month FES-cycling training program performed using a recumbent trike (ICE VTX, 2017), adapted for participants with reduced mobility, combined with two 4-channels current-controlled stimulators (RehaMove3 and Hasomed GmbH) [29]. The training program included 2 weekly sessions, each lasting 30 minutes of stimulation, over 6 months. Each session consisted of 6 sets of 3‐6 minutes of duration at a cycling rate between 30 and 50 revolutions per minute, with brief rest periods in between. The cadence for each pilot was chosen to allow independent pedaling throughout the set duration [28]. Throughout the training sessions, continuous monitoring of heart rate (HR) was conducted with the only purpose of ensuring participant safety, allowing for the immediate cessation of exercise if HR exceeded safe thresholds. This was achieved using the Polar H10 chest strap, which captures HR data at a sampling rate of 1 Hz. HR during training sessions ranged from a minimum average HR of 75 bpm to a maximum average HR of 113 bpm.

The stimulators delivered biphasic square pulses with a maximum current amplitude of 130 mA, a frequency of 40 Hz and a pulse width that ranged between 400 and 500 µs. The stimulation targeted 4 muscle groups per leg such as quadriceps, hamstrings, gluteal muscles, and calf muscles.

MRI scans were performed at 4 time points, at the beginning of the study (T₀, n=4), after 3 months of training (T₁, n=4), at the end of the training (T₂, n=4), and 1-month posttraining deconditioning (T₃, n=3). One of the 4 participants was scanned only up to T₂, since he did not interrupt the training program. The longitudinal MRI assessment was designed to track the SM alterations that occur throughout the training period. In addition, this timeline allows us to assess the impact of short-term deconditioning, which can frequently arise in real-life situations, such as during vacation intervals.

A 3T Achieva dStream MRI scanner (Philips) was used for imaging. The participant was positioned on the examination table, feet facing the scanner (“feet-first” orientation), with the pelvis slightly shifted to align the thigh being scanned closer to the midline. Positioning cushions were used to improve comfort, stabilize the limb and keep the legs separated. Finally, the multichannel Philips dStream Torso coil is placed over the targeted thigh and secured with velcro straps.

The MRI protocol included a T₁w turbo spin echo (TSE) sequence for volume quantification, a 6-point Fast Field Echo m-Dixon Quant sequence for fat fraction quantification, 15-echo multi-echo turbo spin echo (multi-TSE) sequence for T₂ relaxation time quantification, and a single shell diffusion tensor imaging (DTI; 16 directions at b=400 s/mm²; 5 b=0 s/mm² volumes acquired also in opposite phase encoding direction) for the diffusion parameters assessment. Regarding the anatomical region covered in the magnetic resonance (MR) images, the scans encompassed thigh volume in a 30 cm range along the head-to-feet axis, starting from the midpoint of the femoral head. Further details regarding the acquisition protocols are displayed in Figure 1.

As shown in Figure 2, to derive muscle volume and CSA, regions of interest (ROIs) were delineated on T₁w images for specific thigh muscles, including the vastus lareralis, vastus medialis, vastus intermedius, rectus femoris, sartorius, gracilis, adductor magnus, semimembranosus, semitendinosus, biceps femoris caput longum, biceps femoris caput breve, and adductor longus. The semiautomated segmentation was performed using the Deep Anatomical Federated Network [30], which combines automated and manual refinement processes. Volumes for individual muscles and the overall average were computed using 3DSlicer software. The largest CSA (CSA-max) was obtained selecting the slice exhibiting the maximum muscle area.

For the quantitative analysis of MRI derived parameters, the same ROIs used for the evaluation of volume and CSA, ranging from the beginning of the semimembranosus to the last available slice of the rectus femoris were used. Consistency was maintained in the ROIs across all longitudinal scans for each participant as shown in Figure 3.

The parameters derived from the MR images were calculated by averaging their values across the 12 distinct ROIs defined before. The corresponding coefficients of variation were also calculated to characterize the intrinsic variability of each parameter for each time point and each participant (Multimedia Appendix 1).

Fat fraction (FF) was estimated from the mDixon sequence [31] using the Quant model implemented in the scanner which improves the classical bicompartmental exponential model by including a 7-peak fat modeling and T₂* correction to produce FF maps.

T₂ relaxation times were estimated from the multi-TSE images using the extended phase graph approach [32], which models spin behavior and predicts MR signals at various time points, accounting for T₁ and T₂ relaxation processes. Specifically, an open-source toolkit for water T₂ mapping was used [33]. The algorithm was applied to the multi-TSE acquisition after generating a dictionary containing 200 linearly spaced values for water T₂ (range 5‐80 ms), 50 values for the B1 factor (range 50%‐120%), and 101 values for the FF (range 0%‐100%). The fat T₂ was assumed constant at 151 ms. Maps derived from the extended phase graph method, constrained by the external proton-density-weighted fat fraction, were produced.

DTI parameters, including fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AD), were calculated using the MRtrix3 package [34]. Images were denoised with a method based on random matrix theory [35] and the Gibbs ringing artefacts were removed using the method of local subvoxel-shifts [36]. To mitigate susceptibility artifacts, b₀ images were collected with the reversed phase-encode directions, resulting in pairs of images with distortions going in opposite directions. From these pairs the susceptibility-induced off-resonance field was estimated using a method similar to that described in [37] as implemented in FMRIB Software Library [38]. Afterwards, images were corrected for eddy current-induced distortions and participant movements registering each volume in the data set to the reference b₀ volume. Finally, the diffusion tensor was fitted to the log-signal using an iterative weighted least-squares with weights based on the empirical signal intensities (2 iterations were be performed) [39].

Considering the small sample size, a descriptive statistical analysis was performed using R Statistical Software (v4.1.2, R Foundation for Statistical Computing) [40,40]. In particular, changes in muscle volume, CSA, FF, T₂ relaxation times, and DTI parameters across the 4 time points were described reporting medians and median absolute deviations (MAD).

All participants provided written informed consent before enrollment. The research protocol was approved by the ethics committee of Istituto di Ricovero e Cura a Carattere Scientifico Medea (N. 14/22 CE, approved on February 17, 2022) and the protocol was registered on ClinicalTrials.gov (NCT06321172).

Table 1 provides clinical and demographic characteristics of participants in the pilot study. All participants completed the training program with a compliance greater than 78%, performing a minimum of 40 sessions over the 52 foreseen.

As shown in Figure 4, analysis of the MR images revealed relevant changes in muscle volume and CSA-max over the 24-week FES-bike training program. At T₁ (after 3 months of training), participants showed an average increase in muscle volume of 23% compared with baseline (T₀). This growth continued through T₂ (after 6 months of training), with an overall increase of 37% in muscle volume. One month after the training (T₃), a slight reduction was observed, yet the muscle volume remained 23% higher than at baseline. Similarly, as displayed in Figure 5, CSA-max increased by 27% at T₁ and 37% at T₂, with a decrease to 16% above baseline at T₃. It is noteworthy that the observed trend was consistent in all the 4 participants involved in this study as shown in Figures 3 and 4.

As shown in Figure 6, the FF in the muscle tissue showed a notable decrease from T₀ to T₂. The baseline measurements indicated a median FF of 10.9% (MAD=4%). By T₁, the FF decreased to 9.4% (MAD=3%), representing a reduction of 13%. By T₂, this reduction continued to 8.9% (MAD=4%), representing a total reduction of approximately 19%. The FF reduction was consistent in all the 4 participants.

At T₃, a rebound effect was observed, with FF measurements showing an increase to 11.3% (MAD=4%), a 3% higher than the baseline. Even in this case the trend was consistent in all the participants.

As shown in Figure 7, qT₂ times demonstrated a decreasing trend over the course of the training. Starting at a median of 24.3 ms (MAD=2.2 ms) at T₀, the qT₂ reduced to 23.7 ms (MAD=1.1 ms) by T₁, reflecting a decrease of 2%. By T₂, the qT₂ relaxation time further reduced to 22.7 ms (MAD=0.9 ms), reflecting a total decrease of approximately 6%. At T₃, the qT₂ relaxation time slightly reduced to 22.0 ms (MAD=1.8 ms) and it was still lower than the initial value of about 9%. As to qT_2, there was a high variability, and the trend was not consistent among participants. Just at T₂, 4 out of 4 participants showed a reduction in qT₂ as shown in Figure 7.

As shown in Figure 8, DTI parameters revealed subtle but meaningful changes in muscle tissue microstructure.

Specifically, FA showed a decrease from 0.271 (MAD=0.03) at T₀ to 0.249 (MAD=0.02) at T₁, a reduction of approximately 8%. By T₂, FA slightly increased to 0.258 (MAD=0.02), maintaining an overall reduction from baseline. At T₃, FA increased further to 0.261 (MAD=0.017), still showing a net decrease of 4% with respect to baseline. The FA reduction was consistent in all the participants at T₁ and T₂.

Regarding the MD, it showed minimal variation, starting at 0.00159 mm²/s (MAD=0.00008 mm²/s) at T₀, increasing to 0.00160 mm²/s (MAD =0.00004 mm²/s) at T₁ and decreasing to 0.00157 mm²/s (MAD=0.00003 mm²/s) by T₂, representing a total change of about 1%. By T₃, MD values were at 0.00160 mm²/s (MAD=0.00003 mm²/s), showing an increase of 1% from baseline. There was no consistency among partcipants regarding any relevant trend.

As to RD, it started at 0.00135 mm²/s (MAD=0.00008 mm²/s) at T₀, slightly increasing to 0.00138 mm²/s (MAD=0.00002 mm²/s) at T₁, an increase of 2%. By T₂, RD slightly decreased to 0.00136 mm²/s (MAD=0.00004 mm²/s), whereas at T₃, RD values still persisted to 0.00136 mm²/s (MA=0.00006 mm²/s). At T₁, 3 out of 4 participants exhibited the RD raise.

Finally, considering the AD, it began at 0.00207 mm²/s (MAD=0.00010 mm²/s) at T₀, decreasing to 0.00203 mm²/s (MAD=0.00006 mm²/s) at T₁. By T₂, AD further decreased to 0.00200 mm²/s (MAD=0.00001 mm²/s), representing a total decrease of 4%. At T₃, AD values were at 0.00209 mm²/s (MAD=0.00001 mm²/s), showing an increase of 1% from baseline. Even in this case, 3 out of 4 participants have shown a decrease in AD at T₁.

This pilot study aimed at evaluating the feasibility of the use of mpMRI to assess the effects of 6-month FES-cycling training on muscle health in individuals with complete SCI. The use of mpMRI to evaluate SM has emerged as a novel approach recently, providing comprehensive insights into both macroscopic and microscopic changes within muscle tissues [25-27].

This advanced approach is promising for evaluating the effects of FES-cycling training on muscle tissue, moving beyond traditional metrics based essentially on the assessment of muscle volume and strength to explore changes in muscle composition and microstructural environment.

The relevant growth in muscle volume, and similarly the CSA increase, documented in this study highlight the capability of FES-cycling training to counteract muscle atrophy. These findings are in line with previous research that demonstrates the benefits of FES in improving muscle mass. In particular, FES-cycling training was reported to be effective in increasing CSA up to 12% in individuals with complete SCI [13-15] and more strongly electrically-stimulated resistance training, focused mainly on muscle strength and hypertrophy, have reported a 20%‐72% increase of muscle size after 8‐16 week intervention at chronic timepoints [21,22,41].

A reduction in the FF, as observed in the study in all the participants from T₀ to T₂, generally indicates a decrease in intramuscular fat content. This is often associated with improved muscle quality and a shift toward more lean muscle mass. In the context of FES-cycling training, this reduction likely reflects the beneficial effects of increased physical activity and muscle contractions on reducing fat infiltration in the muscles [42-44], which is commonly increased in conditions of muscle disuse or atrophy [45] such as SCI. However, previous findings on the effects of FES training on muscle fat in individuals with SCI are contrasted. Some earlier studies have failed to find any relevant change in intramuscular fat [22,46], whereas others reported up to 53% decrease [15]. It should be noted that all previous studies evaluated fat infiltration using T₁w images in a “semiquantitative” approach, thus they did not fully exploit the capability of MRI to quantify fat infiltration using water-fat imaging techniques based on advanced Dixon techniques as proposed in our work.

Our results, although based on a limited number of participants and exhibiting significant variability, displayed a reduction in muscle water T₂ relaxation times more consistently after 6 months of training. This reduction is typically linked to decreased muscle inflammation and fluid content, usually indicating healthier muscle conditions. More broadly, the qT₂ of muscle water serves as an indicator of disease activity in SM [47]. Changes in qT₂ are nonspecific and can result from various mechanisms, including inflammation, necrosis, muscular dystrophy, acute denervation, or conditions causing intracellular or extracellular edema, or a combination of both. In rehabilitation contexts, particularly following interventions like FES-cycling training, a reduction in qT₂ may indicate positive responses to the physical activity imposed by the training regimen.

In this pilot study, we observed subtle yet meaningful changes in DTI parameters that allow us to speculate on potential modifications occurring in muscle tissue microstructure. Specifically, FA decreased from T₀ to T₁, and although it slightly increased at T₂ and T₃, it still showed a net decrease from baseline. RD increased at T₁ and T₂ before returning to baseline levels at T₃. In contrast, AD decrease at T₁ and T₂, whereas MD showed minimal variations remaining relatively stable over the study period. These changes in DTI parameters may be related to alterations in the microstructural properties of muscle tissue following the FES-cycling training. Diffusion MRI has been widely used to assess the diffusivity of water molecules in tissue and the use of DTI was proposed to indirectly infer microstructural changes of muscle tissue [48,49]. Galbán et al [48] associated the first, second, and third eigenvalues to the diffusive transport along the long axis of a muscle fiber, within the endomysium perpendicular to the long axes of the muscle fibers, and within the cross-section of a muscle fiber, respectively. Furthermore, Hata et al [49] have associated changes in FA, AD, and RD to inflammation, regeneration and remodeling phase in a preclinical model of muscle injury. Interestingly, DTI parameters such as FA and RD have been found to be sensitive tools for monitoring muscle fiber size and can be useful in assessing muscle atrophy with some limitations in measuring muscle hypertrophy [50,51]. Furthermore, FA and RD parameters were associated with muscle fiber composition, with higher FA values (and lower RD values) indicating a higher proportion of type I fiber in muscle tissue [52]. Specifically, the decrease in FA observed in our study, primarily due to increased RD and reduced AD, may be associated with a change in muscle fiber diameters and a reconversion of fibers type, which are common responses to FES-training [53]. While it is important to acknowledge that DTI is one of the simplest methods for modeling diffusion MRI signals and that more complex modeling techniques have recently been developed for assessing SM using diffusion MRI [54-56], there remains room to speculate on interesting aspects of SM microstructure.

The findings from this study highlight the potential of mpMRI as a monitoring tool in rehabilitation settings. By providing a detailed assessment of muscle health, mpMRI can help clinicians develop more targeted and effective rehabilitation strategies. It also offers a method to supervise and adjust treatments based on individual responses, potentially leading to better outcomes and more personalized care strategies.

Furthermore, the application of mpMRI enables the exploration of the underlying mechanisms through which physical rehabilitation interventions, such as FES-cycling training, exert their effects. Understanding these mechanisms can guide the development of new interventions that target specific aspects of muscle health and function.

The limitations of this pilot study, including its small sample size and the absence of a control group, suggest caution in generalizing the findings. Future research should aim to confirm these results through larger-scale studies with diverse populations and control conditions.

The study exclusively involved male volunteers, as female participants were unavailable during the project’s s timeframe. This gender-specific enrollment, also considering the small sample size, aligns with the statistically higher occurrence of spinal cord injuries in men, which is threefold that of women, as detailed in the research by Lu et al [57]. While acknowledging this as a potential limitation, we maintain confidence in the validity of our findings. It is reasonable to expect that the observed trends in muscle parameter variations would be consistent across genders. However, it is important to note that the magnitude of changes, particularly in muscle mass, can differ in women, reflecting the distinct physiological characteristics between the sexes.

A further limitation of this study is also represented by the lack of longitudinal clinical data that, considering also the small sample size, cannot allow a reliable evaluation of relationships between MRI parameters and participant-specific characteristics (clinical, demographic etc). To strengthen the impact of the proposed approach on SM health in a rehabilitative context, larger cohort studies, including the collection of longitudinal parameters to describe patients’ characteristics, are needed.

In this study, the final timepoint, occurring one month posttraining, allowed us to evaluate the impact of deconditioning over a brief interval. This interval is representative of a potential pause in the training program, such as 1 that might occur during everyday life, for instance, a brief vacation break. Indeed, our aim was to understand the short-term reversibility of training effects and their implications for maintaining physical conditioning in real-world scenarios. It would be interesting to include an additional late MRI scan to monitor the progression of muscle atrophy and know the time required to return to baseline values. This additional data point could provide valuable insights into the recovery dynamics and inform future therapeutic strategies.

Finally, forthcoming studies should incorporate molecular and histological analyses to investigate changes at the cellular level, including muscle fiber type transitions, capillary density, and protein expression related to muscle hypertrophy and atrophy.

In conclusion, this study underscores the use of mpMRI in advancing our understanding of the physiological impacts of rehabilitation interventions on muscle health in individuals with SCI. The detailed insights provided by mpMRI suggest its integration into clinical practice to assess the efficacy of interventions like FES-cycling training. By advancing our understanding of the physiological impacts of FES-cycling training, this research paves the way for more effective and personalized rehabilitation protocols, ultimately improving the quality of life for individuals with SCI.