ORIGINAL RESEARCH published: 08 July 2021 doi: 10.3389/fneur.2021.675781 Different Approaches to Analyze Muscle Fat Replacement With Dixon MRI in Pompe Disease Alicia Alonso-Jiménez 1,2 , Claudia Nuñez-Peralta 3 , Paula Montesinos 4 , Jorge Alonso-Pérez 1,5 , Carme García 6 , Elena Montiel 6 , Izaskun Belmonte 6 , Irene Pedrosa 6 , Sonia Segovia 1,5 , Jaume Llauger 3 and Jordi Díaz-Manera 1,5,7* 1 Neuromuscular Disorders Unit, Neurology Department, Departament de Medicina, Hospital de la Santa Creu I Sant Pau, Universitat Autònoma de Barcelona, Barcelona, Spain, 2 Neuromuscular Reference Center, Neurology Department, University Hospital of Antwerp, Edegem, Belgium, 3 Radiology Department, Hospital de la Santa Creu i Sant Pau, Universitat Autònoma de Barcelona, Barcelona, Spain, 4 Philips Healthcare Iberia, Madrid, Spain, 5 Biomedical Network Research Centre on Rare Edited by: Diseases (CIBERER), Barcelona, Spain, 6 Rehabilitation and Physiotherapy Department, Hospital de la Santa Creu i Sant Pau, Massimiliano Filosto, Universitat Autònoma de Barcelona, Barcelona, Spain, 7 John Walton Muscular Dystrophy Research Centre, Newcastle NeMO-Brescia Clinical Center for University, International Centre for Life, Newcastle upon Tyne, United Kingdom Neuromuscular Diseases, Italy Reviewed by: Antonio Toscano, Quantitative MRI is an increasingly used method to monitor disease progression in University of Messina, Italy muscular disorders due to its ability to measure changes in muscle fat content (reported Sabrina Ravaglia, Neurological Institute Foundation as fat fraction) over a short period. Being able to objectively measure such changes is Casimiro Mondino (IRCCS), Italy crucial for the development of new treatments in clinical trials. However, the analysis of the *Correspondence: images involved continues to be a daunting task because of the time needed. Whether Jordi Díaz-Manera a more specific analysis selecting individual muscles or a global one analyzing the whole jordi.diaz-manera@newcastle.ac.uk thigh or compartments could be a suitable alternative has only been marginally studied. In Specialty section: our study we compare three methods of analysis of 2-point-dixon images in a cohort of 34 This article was submitted to patients with late onset Pompe disease followed over a period of one year. We measured Neuromuscular Disorders and Peripheral Neuropathies, fat fraction on MRIs obtained at baseline and at year 1, and we calculated the increment a section of the journal of fat fraction. We correlated the results obtained with the results of muscle function tests Frontiers in Neurology to investigate whether the three methods of analysis were equivalent or not. We observed Received: 03 March 2021 significant differences between the three methods in the estimation of the fat fraction at Accepted: 08 June 2021 Published: 08 July 2021 both baseline and year 1, but no difference was found in the increment in fat fraction Citation: between baseline and year 1. When we correlated the fat fraction obtained with each Alonso-Jiménez A, Nuñez-Peralta C, method and the muscle function tests, we found a significant correlation with most tests Montesinos P, Alonso-Pérez J, García C, Montiel E, Belmonte I, in all three methods, although in most comparisons the highest correlation coefficient was Pedrosa I, Segovia S, Llauger J and found with the analysis of individual muscles. We conclude that the fastest strategy of Díaz-Manera J (2021) Different analysis assessing compartments or the whole thigh could be reliable for certain cohorts Approaches to Analyze Muscle Fat Replacement With Dixon MRI in of patients where the variable to study is the fat increment. In other sorts of studies, an Pompe Disease. individual muscle approach seems the most reliable technique. Front. Neurol. 12:675781. doi: 10.3389/fneur.2021.675781 Keywords: quantitative MRI, Dixon, Pompe, fat replacement, outcome measures Frontiers in Neurology | www.frontiersin.org 1 July 2021 | Volume 12 | Article 675781 Alonso-Jiménez et al. Approaches to Analyze Muscle MRI INTRODUCTION We collected the following epidemiological and clinical data: date of birth, gender, age at diagnosis, time of evolution of the In the past two decades, muscle MRI has been increasingly disease, gene mutations, age at start of enzyme replacement used for diagnosis of neuromuscular disorders. More recently, therapy (ERT) treatment, current disease stage (ambulant or quantitative techniques such as Dixon or spectroscopy have been non-ambulant), and the need of (non)-invasive ventilation. progressively implemented, as they provide the exact amount Patients were evaluated at baseline and 1 year after (±2 of fat present in skeletal muscles and can be used to follow up months). At each visit, muscle function tests and quantitative patients (1). In muscle spectroscopy, a region of interest (ROI) muscle MRI were performed. All patients provided written is drawn in the muscle when the patients are inside the MRI, informed consent to participate in the study. The HSCSP ethics and the analysis is only done in that specific region. In the committee approved the study, and all participants signed an case of Dixon, images are acquired and analyzed later on using informed consent form. All study procedures were performed in specific software. Muscle fat content quantified using muscle accordance with Spanish regulations. MRI correlates with results of muscle function tests (2–4), and therefore, it is considered a good biomarker for neuromuscular MRI Acquisition and Analysis diseases. Furthermore, quantitative muscle MRI is able to detect Patients were examined in a 1.5T Ingenia MR system (Philips subtle changes in the amount of fat in muscles even before Healthcare, Best, the Netherlands) at HSCSP. Axial 3D fast field it impacts muscle function (3, 5). Quantitative muscle MRI is echo (FFE) Dixon sequence was acquired with the following reproducible among different centers, and it is also harmless, as parameters: repetition time/echo time (TR/TE) = 5.78/1.84 ms, it does not use radiation. For its advantages, it has been proposed flip angle = 15◦ , voxel size = 1 × 1 × 3 mm and field of view as a reliable outcome measure for natural history studies and (FOV) 520 × 340 × 300 mm. We used the same position protocol clinical trials (6, 7). However, there is no standardized method for all patients: a supine position with the legs stretched out. to assess the fat content in skeletal muscle MRI acquired using The percentage of fat in muscle or fat fraction (FF) Dixon sequence. Although several software solutions have been was calculated with the PRIDE tool (Philips Research Image proposed to perform automatic or semiautomatic segmentation Development Environment) developed for this purpose. ROIs (8, 9), they are not widely accepted and used, and therefore, the were manually drawn by one investigator (AA-J) on one analysis is continued manually, drawing ROIs in selected slices slice in the right leg for each patient with three different of the muscle. Analysis of Dixon images consumes a lot of time approaches: firstly, by individual muscles (rectus femoris, vastus and requires a high degree of expertise and a profound knowledge intermedius, vastus lateralis, vastus medialis, adductor magnus, of the anatomy of the muscles. Additionally, whether evaluating sartorius, gracilis, biceps femoris long head, semitendinosus, individual muscles is better than assessing compartments or even and semimembranosus); secondly, by compartments (anterior the whole limb has only been addressed in one study (10). Our compartment including the rectus femoris, vastus intermedius, aims for this study were as follows: (1) to check if different approaches to the quantification of fat replacement in thigh muscles of a cohort of patients with late-onset Pompe disease (LOPD) studied using Dixon quantitative muscle MRI showed differences at baseline and after 1-year follow-up, (2) to study whether there were significant differences in the increase of fat content after 1-year follow-up depending on the method used to analyze fat replacement, and (3) to identify which method of analysis correlated better with the results of the muscle function tests performed. MATERIALS AND METHODS Cohort and Study Design A total of 34 LOPD patients were included in this study. They were part of a larger prospective observational study following up LOPD patients registered in the webpage ClinicalTrials.gov with the identifier NCT01914536 (11). Inclusion criteria for the study were as follows: (1) diagnosis of LOPD based on recommendations recently proposed FIGURE 1 | Schema of the three different analysis approaches at baseline visit by the European Pompe Consortium (12) and reduced and visit 1. VL, vastus lateralis; Vi, vastus intermedius; RF, rectus femoris; Vm, enzymatic activity in leukocytes, fibroblasts, or skeletal muscle vastus medialis; Sa, sartorius; AL, adductor longus; Gr, gracilis; AM, adductor and/or the presence of two mutations in the GAA gene; major; SM, semimembranosus; ST, semitendinosus; BF, biceps femoris; AC, (2) no contraindications to MRI; and (3) willingness to anterior compartment; PC, posteromedial compartment; WT, whole thigh; FFB , fat fraction at baseline; 1FF, fat fraction increment between baseline and complete all muscle function tests, respiratory assessment, and visit 1; FF1 , fat fraction at visit 1. patient-reported outcome measures. Frontiers in Neurology | www.frontiersin.org 2 July 2021 | Volume 12 | Article 675781 Alonso-Jiménez et al. Approaches to Analyze Muscle MRI vastus lateralis, vastus medialis, and sartorius; and posteromedial disorders at HSCSP in Barcelona. Each patient was examined compartment including the adductor magnus, gracilis, biceps by the same professional at baseline and after 1 year. The femoris long head, semitendinosus, and semimembranosus); and physiotherapists evaluated muscle function using the following thirdly, including the whole thigh (Figure 1). To obtain the tests: the 6-min walking test (6MWT), time to walk 10 m global thigh FF with the first two approaches, weighted averages (10MWT), timed up-and-go test (TuGo), time to climb up were calculated by normalizing the FF by the area of the (Tup4) and down four steps (Tdo4), the motor function measure- muscle/compartment (bigger muscles had higher weight). All −20-item scale (MFM20), and the Muscle Research Council ROIs were drawn in the same slice just below the biceps femoris (MRC) scale in lower limbs. Patients also completed the self- short head. Neurovascular bundles and the femur were avoided. reported questionnaire ActivLim (portmanteau of “ACTIVity The analysis was performed in MRIs at baseline and after 1- LIMitations”). Global muscle function tests were selected instead year follow-up. of specific muscle tests to correlate with the muscle MRI because our analysis approach measured the global FF of the thigh. We Muscle Function Tests obtained forced vital capacity, both seated and lying down, using All patients were studied by four physiotherapists (IB, CG, the Carefusion Microlab ML 3500 MK8 spirometer (Care Fusion, EM, and IP) with considerable experience in neuromuscular Yorba Linda, CA, USA). TABLE 1 | Clinical characteristics of the patients. Patient Sex Age at study Wheelchair Phenotype GAA mutation 1 GAA mutation 2 Age at ERT Respiratory (years) dependent support 1 F 50 No Muscular IVS1-13T>G c.1076-1G>C 47 No 2 F 48 Yes Muscular + respiratory IVS1-13T>G c.2173C>T 37 Yes 3 F 26 No HyperCKemia IVS1-13T>G c.1889-1G>A - No 4 F 63 No Muscular IVS1-13T>G c.2600_2604delTGCTGinsA 59 No 5 F 47 No Muscular IVS1-13T>G c.15323c>A 42 No 6 F 51 No Muscular IVS1-13T>G c.236_246del 47 No 7 M 66 No Muscular + respiratory IVS1-13T>G c.1933G>T - No 8 F 59 No Muscular IVS1-13T>G c.1637A>G 52 No 9 F 55 No Muscular IVS1-13T>G c.2173C>T 48 No 10 M 42 No Muscular + respiratory IVS1-13T>G c.573C>A 39 Yes 11 F 31 Yes Muscular + respiratory IVS1-13T>G c.1637A>G 24 Yes 12 F 47 No Muscular IVS1-13T>G c.1192dupC 39 No 13 M 47 No Muscular + respiratory c.2173C>T c.2173C>T 45 Yes 14 M 51 No Muscular + respiratory IVS1-13T>G c.1657C>T 45 Yes 15 F 51 No Muscular + respiratory IVS1-13T>G c.1657C>T 46 Yes 16 M 24 No HyperCKemia IVS1-13T>G c.1802C>T - No 17 M 51 No HyperCKemia c.271G>A c.2510G>A - No 18 M 14 No HyperCKemia IVS1-13T>G c.281_282delCT - No 19 F 67 No Muscular c.1781G>A c.1194mas5G>A 64 No 20 F 35 No Muscular IVS1-13T>G c.1A>T 29 No 21 F 40 No Muscular + respiratory IVS1-13T>G c.1889-1G>A - Yes 22 F 53 No Muscular c.1781G>A c.1194+5G>A 45 No 23 M 66 No Muscular + respiratory IVS1-13T>G c.2481+102_2646+31del 57 Yes 24 M 8 No HyperCKemia IVS1-13T>G c.1889-1G>A - No 25 F 57 No Muscular + respiratory IVS1-13T>G c.1447G>T 55 Yes 26 M 46 No Muscular + respiratory IVS1-13T>G c.15323c>A 43 Yes 27 M 51 No Muscular + respiratory IVS1-13T>G c.1933G>T 51 Yes 28 M 51 No Muscular IVS1-13T>G c.1933G>T - No 29 M 43 No Muscular IVS1-13T>G c.1408a1410delC>T 43 No 30 F 54 No Muscular c.-113+2T>A c.-113+2T>A 48 No 31 M 51 No HyperCKemia IVS1-13T>G c.1637A>G - Yes 32 M 43 No Muscular + respiratory IVS1-13T>G c.655G>A - No 33 F 20 No HyperCKemia IVS1-13T>G c.1551+1G>A - No 34 F 41 No Muscular IVS1-13T>G c.1655T>C 40 No ERT, enzyme replacement therapy. Frontiers in Neurology | www.frontiersin.org 3 July 2021 | Volume 12 | Article 675781 Alonso-Jiménez et al. Approaches to Analyze Muscle MRI Statistical Analysis strength and were considered asymptomatic. Among the 27 We used the Shapiro–Wilk-test to check if our variables followed symptomatic patients, 14 had exclusively limb muscle weakness, a normal distribution. As they did not, we used non-parametric whereas 13 had respiratory and limb muscle involvement. A total statistic tests for the analysis. We used the Friedman test to of 32 patients were ambulant (two of whom used a walker), and compare the FF and the increment of FF obtained with the only two needed a manual wheelchair. A total of 23 patients were three methods of analysis as well as the time necessary to treated with ERT at the time of the baseline visit. perform the analysis. In case of a global significant effect, pairwise comparisons are performed with Wilcoxon signed-rank tests. Differences in Fat Fraction Among the We used the paired Wilcoxon signed-rank test to investigate Three Methods of Analysis at Baseline and if there were significant changes in the FF obtained between baseline and visit 1. To investigate the correlations between the Visit 1 At baseline, the mean FF of the thigh calculated by analyzing muscle function tests and FF, we used Spearman’s correlation. individual muscles and performing then a weighted average We considered that correlations were good if p was lower than (analysis by individual muscles) was 38.98% (SD: 23.58). The FF 0.05 and the correlation coefficient (rho) was 0.65 or higher. obtained by analyzing anterior and posteromedial compartment To investigate whether the differences between the correlations and then doing a weighted average (analysis by compartments) with the three different approaches were significant or not, we was 39.08 (SD: 22.45). The FF obtained by the whole-thigh used the web utility provided by quantpsy.org (13). Graphs and approach was 42.48 (SD: 23.55). Friedman’s-test showed that statistical analysis were performed using IBM R SPSS R Statistics the differences in the average FF among the three methods Version 21. at baseline were significant (p ≤ 0.001). Pairwise comparisons showed a significant difference between the analysis by individual RESULTS muscles and the analysis of the whole thigh, and between the analysis by compartments and the analysis of the whole thigh, but Clinical Features of Patients not between analysis by individual muscles and compartments We included a total of 34 LOPD patients in this study of whom 19 (respectively, p < 0.001, p < 0.001, and p = 0.825). were women (55.9%) and 15 were men (44.1%). Clinical data of At visit 1, the mean FF in the individual muscles analysis the patients are summarized in Table 1. The mean age at first visit was 40.81 (SD: 24.05), by compartments was 40.79 (SD: 23.27), was 45.6 years (SD: 14.22 years), and the mean time from onset of and by the whole thigh was 43.76 (SD: 24.34). These differences symptoms was 14.03 years (SD = 9.47 years). Two patients were were also significant between the analysis by individual muscles homozygous, one for the mutation c.2173C>T and the other and the analysis of the whole thigh, and between the analysis by one for the mutation c.−113+2T>A. The rest were compound compartments and the analysis of the whole thigh (p < 0.001) heterozygous. Most patients carried the classical inversion IVS1- and not significant between analyses by individual muscles and 13T>G. Seven patients had only hyperCKemia with normal compartments (respectively, p < 0.001, p < 0.001, and p = 0.996). FIGURE 2 | Fat fraction obtained with the three methods of analysis for each patient. (A) At baseline. (B) At visit 1. The variability is higher in patients with middling fat replacement and lower in both patients with very high and very low fat replacement. Frontiers in Neurology | www.frontiersin.org 4 July 2021 | Volume 12 | Article 675781 Alonso-Jiménez et al. Approaches to Analyze Muscle MRI Figure 2 shows the FF obtained with the three methods for each showed strong correlations with the FF both at baseline and at patient of the study. visit 1 with the three methods of analysis (Figure 4). 6MWT We measured the time needed to draw the ROIs by each and TuGo showed good correlation at visit 1 but not at the approach in a subset of 15 patients. The mean time per patient to baseline. The correlation coefficient was slightly higher in the analyze the MRI by individual muscles in both visits was 330.67 s approach by individual muscles for all the correlations except (SD: 22.71), by compartments 179.09 s (SD 12.87), and by the MRC in lower limbs at visit 1, where the coefficient was whole thigh 114.7 s (SD 12.97). These differences were significant slightly higher by whole-thigh analysis (Table 2). We investigated for the three approaches (Friedman-test, p < 0.001). whether these differences among the correlation coefficients were significant or not and found that in eight out of the Fat Fraction Changes After 1 Year of 13 comparisons with strong correlation coefficients, they were Follow-Up and Differences Among the indeed significantly different. Three Methods of Analysis The mean increment in FF, calculated by subtracting the FF at baseline visit from the FF at visit 1, was 1.83% (SD: 2.48) DISCUSSION for the individual muscles analysis, 1.71% (SD: 2.68) for the compartments analysis, and 1.28% (SD: 2.44) for the whole- In the present paper, we compared three methods of analysis thigh approach. A Wilcoxon signed-rank test showed that this of FF in the thigh in a cohort of LOPD patients. There is no difference in FF between baseline and visit 1 was significant for standardized method to perform this analysis to date. In most the three approaches (by individual muscles Z = −3.462, p = published studies, analysis is performed by manually drawing 0.001; by compartments Z = −3.171, p = 0.002; and by whole ROIs in individual muscles in several slices of the thighs thigh Z = −2.727, p = 0.006). In fact, the mean FF increment was (between three and five in most of the studies), which is highly not statistical significantly different among the three methods of time-consuming (3, 5, 14, 15). Drawing ROIs over the whole analysis (Figure 3). thigh or over the anterior and posteromedial compartments could reduce the time of analysis. Whether these different Correlation With Muscle Function Tests approaches are clinically or statistically significant has only After our observation that there were statistically significant recently been approached by one group (10). We have shown differences between the three analysis strategies, we wondered here that there is a significant difference in the time used whether these differences were clinically significant or not. between the three strategies: compared with the analysis by Correlations between FF and muscle function tests have been individual muscles, it was 45.84% faster when performed by previously described in LOPD patients (2). We confirmed compartments and 65.31% by the whole-thigh approach. In that there was a correlation between FF and some muscle studies involving a large number of MRIs, this difference can function tests and patient-reported outcome measures, which mean a lot of hours. we considered could be affected by fat replacement in the In their study, Reyngoudt et al. (10) performed a comparative thigh. 10MWT, MRC in lower limbs, MFM20, Tup4, and Tdo4 analysis of different segmentation approaches in 102 patients with different muscle disorders. They conclude that the increment in FF calculated by global analysis of the whole thigh is reliable in most muscle disorders. Our study supports these findings, as we did not find a significant difference in the increment of FF between the three approaches. The added value of our study is that we compared the three methods of analysis at baseline and visit 1, and we correlated them with muscle function tests. The comparisons between the average FF using the three methods showed statistical differences at both baseline and visit 1. The pairwise comparison showed that the differences were found between the analysis by individual muscles and by the whole thigh, and by compartments and the whole thigh, but not between individual muscles and compartments. Interestingly, the mean FF was higher for the whole-thigh analysis in both visits. We hypothesized that this higher content in fat could be due to intermuscular fat, which can be affected by other variables besides the muscle disease such as the index body mass (16). This could explain the statistically significant difference with the other two approaches. However, when we looked into each patient in detail (Figure 2), we observed that the variability between the three FIGURE 3 | Increment in fat fraction obtained with the three methods of analysis. There were no significant differences among them (Friedman p = 0.2). methods of analysis is higher for patients with intermediate amount of fat replacement (20–60%), whereas patients with Frontiers in Neurology | www.frontiersin.org 5 July 2021 | Volume 12 | Article 675781 Alonso-Jiménez et al. Approaches to Analyze Muscle MRI FIGURE 4 | Correlations between fat fraction and muscle function tests with the three approaches. (A) Correlation between MRC in lower limbs and FF at baseline. (B) Correlation between MRC in lower limbs and FF at visit 1. (C) Correlation between the time to climb four steps test and FF at baseline. (D) Correlation between the time to climb four steps test and FF at visit 1. (E) Correlation between the time to descend four steps test and the FF at baseline. (F) Correlation between the MFM20 and FF at visit 1. FF, fat fraction; BV, baseline visit; V1, visit 1; MRC, Muscle Research Council scale; MFM20, motor function measure−20-item scale. low (<20%) or very high (>60%) fat content tend to be more obtained by each of the methods and muscle function tests homogenous. This finding might indicate that in studies with and the ActivLim. We found strong correlation coefficients cohorts of patients with very low fat replacement (such as (rho ≥ 0.65) between FF and 10MWT, MRC in lower limbs, asymptomatic and pauci-symptomatic patients), a more general MFM20, Tup4, and Tdo4 in both visits and 6MWT in visit 1. approach with analysis by compartments of the whole thigh may Although the correlation was found for the three methods of be appropriated. analysis, the coefficient was higher for the individual muscles We wondered whether the difference in FF between the approach in all but one of the comparisons with a high three methods could also be clinically significant. In order coefficient. Furthermore, the differences between the correlation to clarify it, we investigated the correlations between the FF coefficients were significant for the majority of functional Frontiers in Neurology | www.frontiersin.org 6 July 2021 | Volume 12 | Article 675781 Alonso-Jiménez et al. Approaches to Analyze Muscle MRI TABLE 2 | Correlations between fat fraction and muscle function tests with the three methods of analysis. Baseline Significant Visit 1 Significant difference? difference? Type of By muscles By compartments By whole thigh By muscles By compartments By whole thigh analysis Test p Spearman’s p Spearman’s p Spearman’s p Spearman’s p Spearman’s p Spearman’s rho rho rho rho rho rho 10MWT 0.00 0.8 0.00 0.769 0.00 0.780 Yes 0.00 0.796 0.00 0.738 0.00 0.753 Yes MRC LL 0.00 −0.885 0.00 −0.866 0.00 −0.881 No 0.00 −0.891 0.00 −0.880 0.00 −0.892 No MFM20 0.00 −0.681 0.00 −0.656 0.00 −0.661 No 0.00 −0.725 0.00 −0.681 0.00 −0.682 Yes TuGo 0.039 0.373 0.056 0.054 No 0.00 0.791 0.00 0.729 0.00 0.744 Yes Tup4 0.00 0.846 0.00 0.822 0.00 0.835 Yes 0.00 0.840 0.00 0.777 0.00 0.805 Yes Tdo4 0.00 0.764 0.00 0.756 0.00 0.749 No 0.00 0.754 0.00 0.694 0.00 0.716 Yes 6MWT 0.00 −0.658 0.00 −0.634 0.00 −0.641 No 0.00 −0.773 0.00 −0.724 0.00 −0.734 Yes Activlim 0.00 −0.607 0.00 −0.619 0.00 −0.604 No 0.00 −0.494 0.00 −0.497 0.01 −0.469 No The highest correlation coefficients have been remarked in bold for each comparison. 10MWT, time to walk 10 m; MRC LL, Muscle Research Council Scale in lower limbs; MFM20, motor function measure−20-item scale; TuGo, timed up-and-go test; Tup4, time to climb up four steps; Tdo4, time to climb down four steps; 6MWT, 6-min walking test; Activlim, Activity Limitations questionnaire. tests, pointing out that the differences between the different chosen, taking into account the goal of the study as well as approaches of analysis are not only statistically significant but the characteristics of the patients. In transversal studies where also clinically relevant. the variable to study is the FF, or in subjects with moderate fat Our study suggests that, in general, Dixon MRI analysis replacement, an analysis by individual muscles seems to be more seems to correlate better with muscle function tests if the accurate. On the other hand, in longitudinal studies where the manual input is executed by drawing ROIs in individual variable of interest is the increment in FF, or in cohorts of patients muscles and obtaining later the global thigh FF with weighted with little fat replacement, a global approach by analyzing averages. This approach has also the advantage of providing compartments or the whole muscle could be more efficient. information about individual muscles. Myopathies and muscular dystrophies are characterized by a heterogeneous muscle fat replacement, affecting different muscles in different diseases. DATA AVAILABILITY STATEMENT Furthermore, the rate of muscle degeneration is not the The raw data supporting the conclusions of this article will same for all the muscles: for example, in Pompe disease, the be made available by the authors under request, without adductor magnus is early affected, whereas the sartorius or undue reservation. gracilis is spared until late stages of the disease. Therefore, the former can be a good muscle to monitor progression at the beginning of the disease, while the latter could be used in ETHICS STATEMENT late stages. However, in specific situations, less time-consuming The studies involving human participants were reviewed and approaches such as drawing ROIs by compartments or in the approved by Hospital de la Santa Creu i Sant Pau ethics whole thigh might be used with good reliability, for instance, committee. Written informed consent to participate in this study in cohorts of patients with very low muscle fat replacement. was provided by the participants’ legal guardian/next of kin. These approaches could also be appropriated if we are interested in quantifying the increment in fat replacement between AUTHOR CONTRIBUTIONS several visits. Our study has two main limitations. Firstly, we have studied AA-J designed the protocol, visited the patients, analyzed only patients with Pompe disease; therefore, studies in other muscle MRIs, and wrote the paper. CN-P collaborated with type of muscle disorders and muscular dystrophies could show the design of the protocol, offered technical support for different results. Secondly, we have performed the study only in neuroimaging, and contributed to the final version of the one slice in the thighs. Analysis of more slices would probably paper. JA-P visited the patients and managed the database. provide a better approximation to the real amount of fat in JL offered technical support for neuroimaging and reviewed muscle. Despite its limitations, we think that our study is useful the paper. SS coordinated the study, managed the database, because it approaches both the statistical and clinically significant and reviewed the paper. IB, IP, CG, and EM visited the differences obtained with three different analysis strategies. patients, acquired motor function data, and contributed to In conclusion, the method of analysis used to quantify the fat the final version of the manuscript with their comments. replacement in muscle using Dixon sequences should be carefully PM offered technical assistance with the MRIs, verified the Frontiers in Neurology | www.frontiersin.org 7 July 2021 | Volume 12 | Article 675781 Alonso-Jiménez et al. Approaches to Analyze Muscle MRI quality of the acquired images and contributed to the final is generated within the European Reference Network for version of the manuscript. JD-M designed the protocol, visited Neuromuscular Diseases. the patients, wrote the paper, and obtained funding for the study. All authors contributed to the article and approved the ACKNOWLEDGMENTS submitted version. We would like to thank the Spanish Association of Patients FUNDING with Glycogenosis (www.glucogenosis.org) and all the patients who participated in the study for their collaboration and This work was supported by a grant from the Spanish Ministry understanding. We thank the MRI technician team for their of Health, Fondos FEDER-ISCIII PI18/1525 to JD-M. The help. We also thank Kristien Wouters for her support with the funder did not participate in the study design. This work statistical analysis and Mr. John Wilkos for his language advice. REFERENCES MRI-based fat fraction changes in neuromuscular diseases. Eur Radiol. (2021) 31:4264–76. doi: 10.1007/s00330-020-07487-0 1. Nuñez-Peralta C, Alonso-Pérez J, Díáz-Manera J. The increasing 11. Figueroa-Bonaparte S, Llauger J, Segovia S, Belmonte I, Pedrosa I, Montiel role of muscle MRI to monitor changes over time in untreated E, et al. Quantitative muscle MRI to follow up late onset Pompe patients: a and treated muscle diseases. Curr Opin Neurol. (2020) 33:611–20. prospective study. Sci Rep. (2018) 8:10898. doi: 10.1038/s41598-018-29170-7 doi: 10.1097/WCO.0000000000000851 12. van der Ploeg AT, Kruijshaar ME, Toscano A, Laforêt P, Angelini C, Lachmann 2. Figueroa-Bonaparte S, Segovia S, Llauger J, Belmonte I, Pedrosa I, RH, et al. European consensus for starting and stopping enzyme replacement Alejaldre A, et al. Muscle MRI findings in childhood/adult onset pompe therapy in adult patients with Pompe disease: a 10-year experience. Eur J disease correlate with muscle function. PLoS ONE. (2016) 11:e0163493. Neurol. (2017) 24:768-e31. doi: 10.1111/ene.13285 doi: 10.1371/journal.pone.0163493 13. Lee IA, Preacher KJ. Calculation for the Test of the Difference Between Two 3. Willis TA, Hollingsworth KG, Coombs A, Sveen M-L, Andersen S, Stojkovic Dependent Correlations with One Variable in Common. Comput Softw. (2013) T, et al. Quantitative magnetic resonance imaging in limb-girdle muscular Available at: http://quantpsy.org/corrtest/corrtest2.htm (accessed January 14, dystrophy 2I: a multinational cross-sectional study. PLoS ONE. (2014) 2021). 9:e90377. doi: 10.1371/journal.pone.0090377 14. Gaeta M, Messina S, Mileto A, Vita GL, Ascenti G, Vinci S, et al. Muscle fat- 4. Rooney WD, Berlow YA, Triplett WT, Forbes SC, Willcocks RJ, Wang DJ, et al. fraction and mapping in Duchenne muscular dystrophy: evaluation of disease Modeling disease trajectory in Duchenne muscular dystrophy. Neurology. distribution and correlation with clinical assessments preliminary experience. (2020) 94:E1622–33. doi: 10.1212/WNL.0000000000009244 Skeletal Radiol. (2012) 41:955–61. doi: 10.1007/s00256-011-1301-5 5. Nuñez-Peralta C, Alonso-Pérez J, Llauger J, Segovia S, Montesinos P, 15. Morrow JM, Sinclair CDJ, Fischmann A, Machado PM, Reilly MM, Yousry Belmonte I, et al. Follow-up of late-onset Pompe disease patients with muscle TA, et al. MRI biomarker assessment of neuromuscular disease progression: magnetic resonance imaging reveals increase in fat replacement in skeletal a prospective observational cohort study. Lancet Neurol. (2016) 15:65–77. muscles. J Cachexia Sarcopenia Muscle. (2020) 11:1032–46. doi: 10.1002/jcsm. doi: 10.1016/S1474-4422(15)00242-2 12555 16. Addison O, Marcus RL, Lastayo PC, Ryan AS. Intermuscular fat: a review 6. Dahlqvist JR, Widholm P, Leinhard OD, Vissing J. MRI in neuromuscular of the consequences and causes. Int J Endocrinol. (2014) 2014:309570. diseases: an emerging diagnostic tool and biomarker for prognosis and doi: 10.1155/2014/309570 efficacy. Ann Neurol. (2020) 88:669–81. doi: 10.1002/ana.25804 7. Carlier PG, Azzabou N, de Sousa PL, Hicks A, Boisserie JM, Amadon Conflict of Interest: PM is employed by the company Philips Healthcare Iberia. A, et al. Skeletal muscle quantitative nuclear magnetic resonance imaging follow-up of adult Pompe patients. J Inherit Metab Dis. (2015) 38:565–72. The remaining authors declare that the research was conducted in the absence of doi: 10.1007/s10545-015-9825-9 any commercial or financial relationships that could be construed as a potential 8. Ogier A, Sdika M, Foure A, Le Troter A, Bendahan D. Individual conflict of interest. muscle segmentation in MR images: a 3D propagation through 2D non- linear registration approaches. In: Proceedings of the Annual International Copyright © 2021 Alonso-Jiménez, Nuñez-Peralta, Montesinos, Alonso-Pérez, Conference of the IEEE Engineering in Medicine and Biology Society, EMBS García, Montiel, Belmonte, Pedrosa, Segovia, Llauger and Díaz-Manera. This is an (Institute of Electrical and Electronics Engineers Inc.). Seogwipo (2017). p. 317– open-access article distributed under the terms of the Creative Commons Attribution 20. License (CC BY). The use, distribution or reproduction in other forums is permitted, 9. Li H, Luo H, Liu Y. Paraspinal muscle segmentation based on deep neural provided the original author(s) and the copyright owner(s) are credited and that the network. Sensors (Switzerland). (2019) 19:2650. doi: 10.3390/s19122650 original publication in this journal is cited, in accordance with accepted academic 10. Reyngoudt H, Marty B, Boisserie JM, Le Louër J, Koumako C, Baudin PY, practice. No use, distribution or reproduction is permitted which does not comply et al. Global versus individual muscle segmentation to assess quantitative with these terms. Frontiers in Neurology | www.frontiersin.org 8 July 2021 | Volume 12 | Article 675781