Predicting persistent/recurrent cervical lymph node metastasis in papillary thyroid carcinoma with PET/CT-based...

Area of Science:

• Nuclear Medicine
• Radiology
• Oncology

Background:

• Assessing persistent/recurrent cervical lymph node metastasis (CLNM) in papillary thyroid cancer (PTC) using radiomics, especially with PET/CT, is not well-established.
• Limited evidence exists for radiomics in predicting CLNM in PTC post-treatment.

Purpose of the Study:

• To develop and validate a prediction model for persistent/recurrent CLNM in postoperative PTC patients.
• To utilize fluoro-18-deoxyglucose (18F-FDG) PET/CT multimodal radiomics for this prediction.

Main Methods:

• Retrospective analysis of 158 postoperative PTC patients with 425 lymph nodes (CLNs) using 18F-FDG PET/CT.
• Development of five radiomics models (clinical, CT, PET, PET/CT, integrated) and a nomogram combining radiomics and clinical features.
• Feature selection using Pearson correlation and LASSO regression.

Main Results:

• The integrated PET/CT radiomics and clinical features model achieved the highest diagnostic accuracy (AUC 0.944 training, 0.922 test).
• PET/CT, PET, CT, and clinical features alone models showed AUCs of 0.887, 0.834, 0.775, and 0.783, respectively.
• The nomogram model demonstrated strong predictive performance (C-index 0.937 training, 0.895 test).

Conclusions:

• Integrating PET/CT radiomics with clinical features offers a promising noninvasive tool for predicting persistent/recurrent CLNM in PTC.
• This approach has the potential to significantly support clinical decision-making for postoperative PTC management.