Development and validation of an ultrasound-based radiomics nomogram for predicting metastatic central cervical lymph...

The rate of cervical central lymph node metastasis (CLNM) in papillary thyroid microcarcinoma (PTMC) patients is relatively high. It is controversial that patients with clinically lymph node (LN)-negative entail prophylactic central neck dissection (pCND). The objective of this study was to develop and validate a predictive nomogram incorporating both radiomics signatures and ultrasonic features to aid in the individualized, preoperative assessment of central cervical LN metastasis in patients with clinically node-negative PTMC.

A total of 266 patients, each with a single malignant thyroid nodule, were enrolled and randomly split into a training set and a validation set at a ratio of 7:3. Radiomics signatures were constructed by applying the least absolute shrinkage and selection operator (LASSO) regression to features extracted from two-dimensional (2D) ultrasound (US) images. We then built three logistic regression models using clinical and US features (model 1), the radiomics signatures (model 2), and a combination of both (model 3). Finally, we presented the combined model as a nomogram and assessed its clinical utility with decision curve analysis (DCA).

The highest diagnostic performance was achieved by model 3, which showed robust discrimination in the training [area under the curve (AUC) =0.892], internal validation (AUC =0.772), and external validation (AUC =0.842) sets, outperforming model 1 and model 2.

A nomogram combining radiomics analysis based on US images of PTMC with clinical and imaging features showed a better diagnostic performance than conventional US imaging features alone in predicting metastatic central cervical LNs preoperatively in clinically LN-negative patients with PTMC and could assist in making advice associated with treatment plan.