Muscle-water T2 (T2m) obtained by MRI relaxometry may be highly sensitive to pathophysiological changes in neuromuscular diseases: T2 increases due to oedema and inflammation as well as fat infiltration, and may decrease due to eventual muscle fibrosis [1]. In many neuromuscular diseases the final pathological pathway is fatty infiltration, and to determine T2m under these circumstances it is necessary to separate the water and fat signal components. One method proposed to achieve this is multi-exponential analysis of CPMG multi-echo imaging data (e.g.[1]–[6]). An appropriate model for the transverse-magnetization decay is required to fit to the experimental data; due to clinical scanning acquisition constraints, to obtain robust T2 estimates it is necessary in practice to restrict the number of free parameters in the fitting process. Recently a constrained tri-exponential model including bi-exponential fat components has been demonstrated to provide successful T2m estimates from 17-echo CPMG muscle MRI data [5], [6]. We developed a simpler signal decay model appropriate when fewer CPMG echo-time images are available, and demonstrated its use to estimate T2m from 12-echo CPMG measurements obtained in the forearms of a cohort of non-ambulant DMD boys and age-matched controlsThe forearms of 9 healthy boys (mean age: 14.6y, range: 13-17y; mean BMI 21.5, range: 16.5-25.4) and 14 non-ambulant DMD patients (mean age: 13.3y, range: 10.8-17.3y; mean duration of non-ambulation 20.2 months, range: 4.7-41.6 and mean BMI 26.5, range: 20.8-41.7) were examined at 3T (Siemens Skyra) using a flexible matrix-coil with a non-fat suppressed multi-echo spin-echo sequence (TR=3000ms, 12 echo-times (TEs) from 10-120ms with 10ms interval, 9 x 6 mm slices, matrix 320x320, or 320x190, in-plane resolution 0.5625x0.5625 mm). The patients were examined at study baseline, 6 and 12 months. Ten forearm muscles were manually segmented on the central slice of a reference image set using ITK-snap software. The first TE image was excluded from the analysis to ensure a constant stimulated-echo-coherence background before a bi-exponential function [S(TE) = a.(1-ff).exp(-TE/T2m)+ a.ff.(0.75.exp(-TE/76)+0.25)], where a represents overall amplitude, and ff the fat-fraction (i.e. fat component amplitude as a proportion of the total signal amplitude) was fitted pixel-wise to the data using nonlinear least-squares minimization in a custom-written Matlab tool. The parameters of the second, mono-exponential component approximating the fat signal were determined separately as the mean values from 4 subcutaneous fat ROIs in 8 representative subjects. For quality control, values were excluded from the maps for pixel data which failed to meet empirically determined fit-quality criteria [R2 (goodness of fit) > 0.8, amplitude a lower confidence interval (CI) > a/2 , amplitude a < 10 x the 1st TE image amplitude, T2m-CI-width/T2m < 100%, fat fraction ff<50% and T2m<100ms]. T2m maps were generated and histogram metrics (mean, full width at quarter maximum, 75th percentile and skewness) for all the individual muscle ROIs in the cross-section combined were extracted; for comparison mean cross-sectional muscle ff was also determined independently by 3-point Dixon measurements. One-way ANOVA tested against the null hypothesis of equality of histogram metrics between patients longitudinally, and healthy volunteers.We obtained analysable T2 maps from all 23 subjects, with a reasonable number of pixel T2m values surviving fit quality-control criteria in most ROIs. Figure 1 shows histogram metrics (including the mean) for patients longitudinally, and for healthy controls. Patient cross-sectional mean T2m remained in the normative range while in general exhibiting wider distributions. ANOVA suggested significant difference in the full width at quarter maximum (p=0.0032) and skewness (p=0.0016) between patients and healthy controls. Figure 2 shows that cross-sectional mean T2m was largely independent of 3-point Dixon-determined ff for both groups. Figure 3 shows a representative control-subject T2m map, and Figure 4 a representative patient T2m map and corresponding independently obtained ff map. In initial exploratory analyses the tri-exponential model of Azzabou et al. [5] did not provide reliable parameter estimates with our data, presumably due to the different acquisition conditions in our case. The signal model used here was the most simple of those we investigated providing stable T2m estimates from our data. More sophisticated approaches are likely to be appropriate for analysing CPMG data with more TEs and higher SNR, particularly where determining a physically meaningful model of lipid compartmentalisation is a primary objective. Nevertheless, using our method we obtained forearm T2m values in both controls and DMD patients which are consistent with those previously reported [6], and appear largely independent of muscle fat content.