Ultrasound radiomics features for identifying the masseter muscle in patients with masticatory muscle tendon-aponeurosis hyperplasia: A pilot study.

Purpose: To investigate whether quantitative radiomics features derived from ultrasound images of the masseter muscle, combined with machine learning, can identify masticatory muscle tendon-aponeurosis hyperplasia (MMTAH). Materials and Methods: Seven patients clinically diagnosed with MMTAH (3 men, 4 women; mean age, 52.7 ± 14.3 years) and 7 age- and sex-matched healthy controls were included. A total of 84 ultrasound image slices (42 MMTAH and 42 controls) were analysed. The masseter muscle was manually segmented using 3D Slicer, and radiomics features were extracted with the SlicerRadiomics extension. Logistic regression and random forest models were constructed, and their performance was evaluated using internal validation. Results: A total of 107 radiomics features were extracted, from which informative features were selected for model development. The logistic regression model showed a sensitivity of 0.83, specificity of 0.90, accuracy of 0.87 and an area under the curve (AUC) of 0.95. The random forest model showed a sensitivity of 0.83, specificity of 0.93, accuracy of 0.88 and an AUC of 0.92. There was no significant difference in AUC between the two models ( Conclusion: Radiomics-based analysis of ultrasound images may provide a quantitative and reliable approach for differentiating MMTAH from normal masseter muscles.