分子影像学杂志2024，Vol.47Issue(12)：1290-1297,8.DOI:10.12122/j.issn.1674-4500.2024.12.03
甲状腺结节良恶性鉴别诊断:基于超声可解释性机器学习模型
Differential diagnosis of benign and malignant thyroid nodules:based on an interpretable ultrasound machine learning model
摘要
Abstract
Objective To investigate the innovation and effectiveness of two-dimensional ultrasonography and shear wave elastography(SWE)combined with the XGBoost machine learning model in the differential diagnosis of benign and malignant thyroid nodules.Methods 2D-ultrasound images and SWE measurements were analyzed in 156 patients with thyroid nodules(209 nodules)from the North District of the First Affiliated Hospital of Anhui Medical University from May 2021 to September 2022 with pathology as the gold standard.A machine learning model based on two-dimensional ultrasonography and SWE was developed using the XGBoost algorithm.The feature importance was assessed using the Shapley additive interpretation method.ROC curves were plotted,and the AUC was calculated to assess the performance of the XGBoost model and SWE.Additionally,decision curve analysis and calibration curves were used to evaluate the application value and diagnostic efficacy of the XGBoost model.Results The AUC,accuracy,sensitivity,specificity,positive predictive value,and negative predictive value of the XGBoost model in the diagnosis of benign and malignant thyroid nodules were 0.890,0.776,89.6%,65.7%,83.3%,76.7%in the training cohort and 0.913,0.788,92.7%,64.9%,82.9%,82.8%in the validation cohort,respectively.Decision curve analysis and calibration curve analysis showed that the XGBoost model showed good clinical application value in the diagnosis of benign and malignant thyroid nodules,as well as high accuracy and reliability.Conclusion The XGBoost machine learning model based on two-dimensional ultrasound features and SWE has important application value in the differential diagnosis of benign and malignant thyroid nodules and provides a new and efficient tool for clinical decision-making.关键词
弹性成像/机器学习/鉴别诊断/甲状腺结节Key words
elastography/machine learning/differential diagnosis/thyroid nodule引用本文复制引用
陈冬冬,解翔,詹小林,虞红珍,周燕,陈芳..甲状腺结节良恶性鉴别诊断:基于超声可解释性机器学习模型[J].分子影像学杂志,2024,47(12):1290-1297,8.基金项目
安徽医科大学校科研基金项目(2021xkj174) （2021xkj174）
