MRI can detect the most significant pathological changes of muscle-fat replacement and muscle edema in muscular dystrophies. MRI-derived texture analysis is superior to conventional MRI for identifying pathological changes in muscular dystrophies. Furthermore, the combination of selected radiomics features and clinical biomarkers can enhance the diagnostic accuracy for muscular dystrophies. Muscular dystrophies are difficult to discriminate from their mimickers, so further research is warranted to identify the optimal feature combination and validate the performance of the combined model in myopathies prone to misdiagnosis. This study evaluated the diagnostic accuracy of radiomics features and clinical biomarkers in differentiating muscular dystrophies from their mimickers in a cohort of 161 myopathy patients using machine learning techniques. Multiple machine learning algorithms were jointly applied to screen robust features. The results showed that the combined nomogram exhibited better performance than the individual model using clinical or radiomic features, achieving an AUC of 0.955 in the training set and 0.923 in the validation set. Decision curve analysis confirmed the clinical utility of the nomogram. This multiparametric approach, combining texture features from MRI T1-weighted sequences and STIR sequences with clinical biomarkers (age, gender, and creatine kinase), significantly enhanced the discriminative power. The better performance of the nomogram compared with expert evaluations demonstrated its potential application in distinguishing muscular dystrophies from their mimickers.
Niu et al. (Wed,) studied this question.