Ultrasound scatteromics: A multimodal QUS-based solution for detecting ambulatory function deterioration in Duchenne muscular dystrophy

Abstract

Quantitative ultrasound (QUS) envelope statistics imaging has been investigated as a non-invasive method for evaluating Duchenne muscular dystrophy (DMD). This study introduces an ultrasound scatteromics approach to differentiate between early and late ambulatory stages of DMD. A total of 47 DMD subjects were divided into early (n = 23) and late (n = 24) ambulatory stages. Ultrasound images of the gastrocnemius muscle were acquired and processed to generate multimodal QUS envelope statistics images based on the Nakagami distribution parameter m, homodyned K-distribution parameters α and k, and information entropy H. A simplified feature set based on first-order statistical features of each QUS envelope statistics image was then used for classification with support vector machine (SVM), random forest (RF), and linear discriminant analysis (LDA). A total of 30 iterations of five-fold cross-validation, along with the area under the receiver operating characteristic curve (AUROC), were used for model evaluation. Individual QUS envelope statistics parameters produced average AUROC values below 0.7. Scatteromics achieved average AUROC values of 0.97, 0.98, and 0.83 for SVM, RF, and LDA models, respectively. The simplified scatteromics models, using m, α, k, and H as input to SVM, RF, and LDA, yielded average AUROC values of 0.98, 0.98, and 0.88, respectively. The scatteromics approach outperformed individual QUS envelope statistics imaging methods in monitoring ambulatory function deterioration in DMD in clinical settings (p < 0.05 for AUROC comparisons, DeLong test).