P.136 - Random forest approach to assess relationships of subjective muscle fatty infiltration with age at clinical onset and time of disease evolution in LMNA-related muscle disorders

Abstract

Muscle MRI is a popular approach to orientate genetic studies in clinical practice. Systematic description of muscle infiltration with semiquantitative scores is important to define specific MRI profiles in the different muscle disorders. Recently, two studies have described fatty infiltration patterns in patients with LMNA mutations. One has described whole body MRI in congenital muscular dystrophy (Gómez-Andrés et al. 2016); the other has assessed fatty infiltration of lower limbs by CT or MRI in Emery-Dreifuss and LGMD phenotypes (Díaz-Manera et al. 2016). We aim to combine data from both groups to evaluate which muscles are related with age at clinical onset and time of disease evolution. We collected data from both studies selecting those muscles that were represented in both. In those cases that one study represented data for muscle groups, a median score was calculated. Random forests were trained to predict both age at onset and disease evolution based on imaging data. We calculated the importance of each muscle in the prediction to point out which muscle fatty infiltration is related with onset or evolution. Fatty infiltration patterns discriminates age at onset (rho = 0.71). The relevant infiltrations were those of tibialis posterior, semitendinosus, iliopsoas, soleus and rectus femoris. Fatty infiltration patterns are also related to disease evolution but the prediction capacity is lower (rho = 0.53). Infiltration of semitendinosus, medial and lateral gastrocnemii, long head of biceps femoris and semimembranosus are the potential indicators of disease evolution. These results may be helpful in the design of longitudinal studies with quantitative assessment of fat fraction. Based on this result, rational designs would include muscles whose semiquantitative fatty infiltration is related with significant events in the disease course. Moreover, this study may represent a proof of concept that supports the use of systematic evaluation of fatty infiltration and descriptive techniques such as heatmaps that allow meta-analytical approaches.