Proteomics combined with ultrasound radiomics predicts central lymph node metastasis in papillary thyroid carcinoma

Accurate preoperative prediction of central lymph node metastasis (CLNM) in papillary thyroid carcinoma (PTC) remains challenging. This study sought to develop and validate a combined model that combines ultrasound radiomics and proteomic features to predict CLNM in PTC patients as accurately as possible. A retrospective study was conducted on 155 PTC patients. Two experienced radiologists manually delineated the regions of interest (ROIs) on preoperative ultrasound images using ITK-SNAP, and 1322 radiomic features were extracted from each ROI using the PyRadiomics package, encompassing first-order, texture, and wavelet-based features. Proteomic profiling was then performed on formalin-fixed paraffin-embedded tissue samples via data-independent acquisition mass spectrometry. Feature selection was additionally performed using Pearson correlation and LASSO regression. Machine learning models were built using support vector machine, logistic regression, random forest, and XGBoost based on radiomics, proteomics, and combined features. The combined model significantly outperformed single-modality models, achieving an area under the curve of 0.84 in the validation set, compared to 0.75 (radiomics) and 0.75 (proteomics). Decision curve analysis confirmed superior clinical utility as well. Functional enrichment analysis revealed upregulation of metabolic pathways and downregulation of immune-related extracellular matrix organization in CLNM-positive patients. Integrating ultrasound radiomics and proteomics provides a powerful framework for predicting CLNM in PTC. Although its results are based on surgical specimens, this study lays a methodological foundation and provides a theoretical framework for future proteomic biomarker research using preoperative fine-needle aspiration biopsies, thereby facilitating personalized treatment strategies and offering insights into the underlying biology of metastasis.