Feng Jia-Wei, Zheng Feng, Liu Shui-Qing, Qi Gao-Feng, Ye Xin, Ye Jing, Jiang Yong
Department of thyroid surgery, The Third Affiliated Hospital of Soochow University, Changzhou First People's Hospital, Changzhou, Jiangsu, China; Department of General Surgery, Sir Run Run Shaw Hospital, School of Medicine, Zhejiang University, Hangzhou, Zhejiang, China.
Department of Critical Care Medicine, The Third Affiliated Hospital of Soochow University, Changzhou First People's Hospital, Changzhou, Jiangsu, China.
Acad Radiol. 2025 Mar;32(3):1360-1372. doi: 10.1016/j.acra.2024.10.001. Epub 2024 Oct 22.
The study aimed to analyze the patterns and frequency of Level V lymph node metastasis (LNM) in papillary thyroid carcinoma (PTC), identify its risk factors, and construct predictive models for assessment.
We conducted a retrospective analysis of 325 PTC patients who underwent thyroidectomy and therapeutic unilateral bilateral modified radical neck dissection from October 2020 to January 2023. Patients were randomly allocated into a training cohort (70%) and a validation cohort (30%). The radiomics signature model was developed using ultrasound images, applying the minimum Redundancy-Maximum Relevance and Least Absolute Shrinkage and Selection Operator regression to extract high-throughput quantitative features. Concurrently, the clinic signature model was formulated based on significant clinical factors associated with Level V LNM. Both models were independently translated into nomograms for ease of clinical use.
The radiomics signature model, without the inclusion of clinical factors, showed high discriminative power with an area under the curve (AUC) of 0.933 in the training cohort and 0.912 in the validation cohort. Conversely, the clinic signature model, composed of tumor margin, simultaneous metastasis, and high-volume lateral LNM, achieved an AUC of 0.749 in the training cohort. The radiomics signature model exhibited superior performance in sensitivity, specificity, positive predictive value, negative predictive value across both cohorts. Decision curve analysis demonstrated the clinical utility of the radiomics signature model, indicating its potential to guide more precise treatment decisions.
The radiomics signature model outperformed the clinic signature model in predicting Level V LNM in PTC patients. The radiomics signature model, available as a nomogram, offers a promising tool for preoperative assessment, with the potential to refine clinical decision-making and individualize treatment strategies for PTC patients with potential Level V LNM.
本研究旨在分析甲状腺乳头状癌（PTC）中Ⅴ级淋巴结转移（LNM）的模式和频率，确定其危险因素，并构建预测模型以进行评估。
我们对2020年10月至2023年1月期间接受甲状腺切除术和治疗性单侧或双侧改良根治性颈清扫术的325例PTC患者进行了回顾性分析。患者被随机分为训练队列（70%）和验证队列（30%）。使用超声图像开发了放射组学特征模型，应用最小冗余-最大相关性和最小绝对收缩与选择算子回归来提取高通量定量特征。同时，基于与Ⅴ级LNM相关的显著临床因素构建了临床特征模型。两个模型都独立转化为列线图以便于临床使用。
在不纳入临床因素的情况下，放射组学特征模型显示出较高的判别能力，训练队列中的曲线下面积（AUC）为0.933，验证队列中的AUC为0.912。相反，由肿瘤边缘、同时转移和高容量侧方LNM组成的临床特征模型在训练队列中的AUC为0.749。放射组学特征模型在两个队列中的敏感性、特异性、阳性预测值、阴性预测值方面均表现出卓越性能。决策曲线分析证明了放射组学特征模型的临床实用性，表明其有潜力指导更精确的治疗决策。
在预测PTC患者的Ⅴ级LNM方面，放射组学特征模型优于临床特征模型。作为列线图的放射组学特征模型为术前评估提供了一个有前景的工具，有可能优化临床决策并为有潜在Ⅴ级LNM的PTC患者制定个体化治疗策略。
