Application of decision tree algorithms to predict central lymph node metastasis in well-differentiated papillary thyroid carcinoma based on multimodal ultrasound parameters: a retrospective study

Abstract

Background: Prophylactic central neck dissection (pCND) in well-differentiated primary papillary thyroid carcinoma (PTC) patients have become controversial. Several attempts have been made to predict central compartment lymph node metastasis (CLNM) based on clinical and conventional ultrasonic parameters. This study seeks to develop a decision tree (DT) model for predicting the risk of CLNM in these patients based on clinical and preoperative multimodal ultrasound (US) characteristics.Methods: A total of 148 PTC nodules confirmed by surgical pathology at Beijing Tiantan Hospital were retrospectively analyzed. All nodules underwent multimodal US examinations from January 2020 to September 2021 before surgery. Correlation analysis of CLNM with clinical characteristics, multimodal US parameters of PTC lesions based on grayscale US, color Doppler flow imaging, superb microvascular imaging, contrast-enhanced US and shear wave elastography technology was carried out. Finally, the Chi-squared automatic interaction detector with a ten-fold cross test was used to establish DT models based on multimodal US parameters to predict CLNM. The area under the curve (AUC) was calculated for comparing prediction performance.Results: Univariate analysis indicated that CLNM was positively correlated with Tg level, maximum size, tall than wide, the number of microcalcifications greater than or equal to 5, contact capsule, US reported abnormal cervical lymph nodes, non-centripetal perfusion, delayed clearance, SWV mean and SWV ratio (p<0.05). The multimodal US algorithm based on tall than wide, contact the capsule, US reported abnormal cervical lymph nodes, centripetal enhancement as independent variables showed good discrimination: the sensitivity, specificity, accuracy and AUC were 80.0%, 76.7%, 78.4%, and 0.837 (95% CI: 0.771-0.902). There was a significant difference between multimodal US DT with conventional ultrasound DT (p<0.001). Conclusions: This study implies that the DT model based on preoperative multimodal US characteristics of PTCs has reasonable prediction ability for CLNM, which can be conveniently used for clinical decision-making of individualized treatment in well-differentiated PTC patients.