Using MRI to study the back muscles of DMD patients, fat infiltration was found larger on the convex side of the deviations of pre- and scoliotic patients while no appreciable difference was observed in non-scoliotic patients. A different result was obtained by Stern and Clark  when they studied the back muscles of 16 scoliotic DMD patients with CT scans. They reported that fat infiltration was more important on the concave side of the spinal deviation but when their density readings (Table 2 of their paper) are closely scrutinized, one can be lead to consider that fat infiltration was larger on the concave side for only half of their patients. Also, their density readings were taken only at T9 and L3 levels in the medial (i.e. SP) and lateral (i.e. IL) portions of the ES and fat infiltration in the LO was probably not taken into account. In contrast, our data were continuously collected at every 3 mm from T8 to L4 and all the constituents of the ES muscle were analyzed. Through the reliability offered by the use of our 3D segmentation procedure, the accuracy of volume determination can be considered similar to the one of the cross section area (i.e. ≤ 1%). For the ES muscle of the two controls, the smallest differences in muscle volume was 1.4 ± 1.2% and for fat infiltration it was 2.6 ± 0.8% while for the DMD patients, these differences were larger. All of our results are then outside the segmentation accuracy. Segmentation was easier to achieve for ES due to its size and for SP because fat infiltration was minimal. For more severely infiltrated LO and IL muscles, an additional segmentation cycle (axial, sagittal, then coronal plane) was done compared to the number of iterations used to delineate ES and SP.
Some of the differences between our results and those of Stern and Clark may be associated to the effects of the DFZ: all of our patients had been and some were still treated (n = 6) with DFZ while none of their patients received such a treatment. DFZ has an overall positive impact on DMD patients' quality of life: it improves cardiac function, prolongs walking, and seems to eliminate the need for spinal surgery, although vertebral fractures and stunted growth occur . Weight gain is a side effect of the medication but no correlation was found here between BMI and the duration of DFZ medication.
The presence of an uneven fat infiltration can induce a force unbalance on the spine which could lead to its buckling toward the weaker side. As the spine buckles, the skin becomes compressed on the concave side and stretched on the convex one. Our results indicate that the thickness difference was smaller for the PS than for the S patients (and absent for the NS patients). Due to the filtering effects of the overlying skin thickness, smaller EMG signals would be expected on the concave side. However, in another set of data collection with the same patients , a larger surface EMG signals were recorded on the concave side of the PS patients and on the convex side of a S patient. While smaller signals on the convex side of the PS may have result from the greater fat infiltration on that side, it cannot explain the larger signal on the convex of the S patient. It can thus be hypothesized that a more important neural input on one side of the spine can contribute to some extent to the presence of larger EMG signals on that side . In our pre-scoliotic DMD patients, a deficit in muscle fibers on one side of the spine and an imbalance in the neural input could thus be two factors leading to the deformation of the spine.
As the Cobb angle gets larger than 15°, it seems that compensatory mechanisms to prevent further deviation of the spine are set in action. For instance, for scoliotic DMD P7, a larger EMG activity was detected on his convex side  as it is frequently reported in AIS [22, 23]. The stretching of the skin on the convex side of a deviation causes a reduction in the overlying skin thickness, favoring the presence of a larger EMG signal but a greater fat infiltration on the convex side would counteract to some level such an increase. The presence of a larger EMG signal on the convex side may thus be attributed to a greater feedback from the more stretched muscle spindles and/or a larger neural input from the central nervous system. Imbalanced neural input was also considered to have a pathogenic importance in the etiology of AIS .
For Shimada , the trunk muscle imbalance would be one of the most important factors in the onset and progression of AIS since EMG amplitude is higher on the convex side when the deviation is progressive while no differences are observed in non progressive situation. According to the medical records of our PS and S patients, their spinal deformation did not progress over an 18 months period preceding our images acquisition. Either DFZ halted the progression of the deformation or their scoliosis was non progressive. Since higher EMG signals were still detected on the convex side, it seems that the scoliotic mechanism present in a DMD is somewhat different than in AIS.
To predict the progression of scoliosis in DMD patients, approaches such as Griffiths developmental scales, the Reynell language scales and the British picture vocabulary scales have been proposed [24, 25]. but their reliability seems to be limited as compared to the vital capacity at age 10 when ambulation usually ceases . As for Gibson and Wilkins , they proposed a clinical mobility index based on muscle weakness which is essentially caused by the progressive replacement of muscle tissue by fat. In our study, that mobility index was found correlated with fat infiltration in the ES (R2 = 0.70, p < 0.01) but not with BMI nor with the age of the patients. This lack of correlation with age may be specific to the ES and its constituents since for thigh muscles (hip, mid-thigh, and knee), older boys demonstrated more prominent fatty infiltration than younger ones . However, it has to be considered that despite a severe infiltration, NS P4, which had a very limited mobility, did not develop scoliosis since infiltration was similar on both sides of his spine. Therefore, an asymmetry in fat infiltration (Fig. 4) could help predict the development of a spinal deformation.
Only the ES muscle and its constituents were studied here since they can greatly influence the development of a scoliosis and their activity can be easily monitored from the skin surface. Investigation of less accessible trunk muscles, such as the abdominal and external oblique, psoas and quadratus lumborum, could be worth of investigation as they can also influence the stability of the spine.
DMD patients are frequently solicited to participate in a research protocol and we experienced difficulties with patients' recruitment. The small number of patients participating in our protocol constitutes the main limitation of the study. With fat infiltration appearing initially in the L2 region, images acquisition could be restricted in that zone thus reducing the time required for the execution of the MRI protocol. This could contribute to facilitate the recruitment of patients.