Importance of neural network complexity for the automatic segmentation of individual thigh muscles in MRI images from patients with neuromuscular diseases

ObjectiveSegmentation of individual thigh muscles in MRI images is essential for monitoring neuromuscular diseases and quantifying relevant biomarkers such as fat fraction (FF). Deep learning approaches such as U-Net have demonstrated effectiveness in this field. However, the impact of reducing neural network complexity remains unexplored in the FF quantification in individual muscles.Material and MethodsU-Net architectures with different complexities have been compared for the quantification of the fat fraction in each muscle group selected in the central part of the thigh region. The corresponding performance has been assessed in terms of Dice score (DSC) and FF quantification error. The database contained 1450 thigh images from 59 patients and 14 healthy subjects (age: 47 +/- 17 years, sex: 36F, 37M). Ten individual muscles were segmented in each image. The performance of each model was compared to nnU-Net, a complex architecture with 4.35 x\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 107 parameters, 12.8 Gigabytes of peak memory usage and 167 h of training time.ResultsAs expected, nnU-Net achieved the highest DSC (94.77 +/- 0.13%). A simpler U-Net (5.81 x\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 105 parameters, 2.37 Gigabytes, 14 h of training time) achieved a lower DSC but still above 90%. Surprisingly, both models achieved a comparable FF estimation.DiscussionThe poor correlation between observed DSC and FF indicates that less complex architectures, reducing GPU memory utilization and training time, can still accurately quantify FF.