Radiomics analysis of intraoral ultrasound images for prediction of late cervical lymph node metastasis in patients with tongue cancer.

Abstract

We investigated the predictability of late cervical lymph node metastasis using radiomics analysis of ultrasonographic images of tongue cancer. We selected 120 patients with tongue cancer who underwent intraoral ultrasonography, 30 of which had late cervical lymph node metastasis. Radiomics analysis was used to extract and quantify the image features. Bootstrap forest (BF), support vector machine (SVM), and neural tanh boost (NTB) were used as the machine learning models, and receiver operating characteristic curve analysis was conducted to determine diagnostic performance. The sensitivity, specificity, accuracy, and AUC in the validation group were, respectively, 0.600, 0.967, 0.875, and 0.923 for the BF model; 0.700, 0.967, 0.900, and 0.950 for the SVM model; and 0.900, 0.967, 0.950, and 0.967 for NTB model. Radiomics analysis and machine learning models using ultrasonographic images of pretreated tongue cancer could predict late cervical lymph node metastasis with high accuracy.