RESUMEN
OBJECTIVE: In this study we address the automatic segmentation of selected muscles of the thigh and leg through a supervised deep learning approach. MATERIAL AND METHODS: The application of quantitative imaging in neuromuscular diseases requires the availability of regions of interest (ROI) drawn on muscles to extract quantitative parameters. Up to now, manual drawing of ROIs has been considered the gold standard in clinical studies, with no clear and universally accepted standardized procedure for segmentation. Several automatic methods, based mainly on machine learning and deep learning algorithms, have recently been proposed to discriminate between skeletal muscle, bone, subcutaneous and intermuscular adipose tissue. We develop a supervised deep learning approach based on a unified framework for ROI segmentation. RESULTS: The proposed network generates segmentation maps with high accuracy, consisting in Dice Scores ranging from 0.89 to 0.95, with respect to "ground truth" manually segmented labelled images, also showing high average performance in both mild and severe cases of disease involvement (i.e. entity of fatty replacement). DISCUSSION: The presented results are promising and potentially translatable to different skeletal muscle groups and other MRI sequences with different contrast and resolution.
Asunto(s)
Aprendizaje Profundo , Procesamiento de Imagen Asistido por Computador , Procesamiento de Imagen Asistido por Computador/métodos , Pierna/diagnóstico por imagen , Imagen por Resonancia Magnética/métodos , Músculo Esquelético/diagnóstico por imagen , Muslo/diagnóstico por imagenRESUMEN
Purpose: Quantitative Muscle MRI (qMRI) is a valuable and non-invasive tool to assess disease involvement and progression in neuromuscular disorders being able to detect even subtle changes in muscle pathology. The aim of this study is to evaluate the feasibility of using a conventional short-tau inversion recovery (STIR) sequence to predict fat fraction (FF) and water T2 (wT2) in skeletal muscle introducing a radiomic workflow with standardized feature extraction combined with machine learning algorithms. Methods: Twenty-five patients with facioscapulohumeral muscular dystrophy (FSHD) were scanned at calf level using conventional STIR sequence and qMRI techniques. We applied and compared three different radiomics workflows (WF1, WF2, WF3), combined with seven Machine Learning regression algorithms (linear, ridge and lasso regression, tree, random forest, k-nearest neighbor and support vector machine), on conventional STIR images to predict FF and wT2 for six calf muscles. Results: The combination of WF3 and K-nearest neighbor resulted to be the best predictor model of qMRI parameters with a mean absolute error about ± 5 pp for FF and ± 1.8 ms for wT2. Conclusion: This pilot study demonstrated the possibility to predict qMRI parameters in a cohort of FSHD subjects starting from conventional STIR sequence.
RESUMEN
PURPOSE: Quantitative MRI (qMRI) plays a crucial role for assessing disease progression and treatment response in neuromuscular disorders, but the required MRI sequences are not routinely available in every center. The aim of this study was to predict qMRI values of water T2 (wT2) and fat fraction (FF) from conventional MRI, using texture analysis and machine learning. METHOD: Fourteen patients affected by Facioscapulohumeral muscular dystrophy were imaged at both thighs using conventional and quantitative MR sequences. Muscle FF and wT2 were calculated for each muscle of the thighs. Forty-seven texture features were extracted for each muscle on the images obtained with conventional MRI. Multiple machine learning regressors were trained to predict qMRI values from the texture analysis dataset. RESULTS: Eight machine learning methods (linear, ridge and lasso regression, tree, random forest (RF), generalized additive model (GAM), k-nearest-neighbor (kNN) and support vector machine (SVM) provided mean absolute errors ranging from 0.110 to 0.133 for FF and 0.068 to 0.115 for wT2. The most accurate methods were RF, SVM and kNN to predict FF, and tree, RF and kNN to predict wT2. CONCLUSION: This study demonstrates that it is possible to estimate with good accuracy qMRI parameters starting from texture analysis of conventional MRI.