Page 88 - 南京医科大学自然版
P. 88
南京医科大学学报(自然科学版) 第46卷第10期
·1518 · Journal of Nanjing Medical University(Natural Sciences) 2026年10月
·临床研究·
双能CT碘图影像组学特征对甲状腺乳头状癌侧颈区淋巴结转
移的诊断价值
周 雨,周 燕,许泳康,旷欣汝,沈美萍,斯 岩,王 勇,苏国义,吴飞云 *
南京医科大学第一附属医院放射科,江苏 南京 210029
[摘 要] 目的:探讨双能CT(dual energy CT,DECT)碘图影像组学特征对甲状腺乳头状癌(papillary thyroid cancer,PTC)侧颈
区淋巴结转移(lateral cervical lymph node metastasis,LLNM)的诊断价值。方法:回顾性分析2022年6月—2024年6月于南京医
科大学第一附属医院术前行DECT检查并经病理证实的200例PTC患者临床和影像资料,共纳入777枚淋巴结(其中379枚转
移淋巴结,398枚非转移淋巴结)。将患者按7∶3的比例随机分配到训练集与验证集中。纳入淋巴结相关的CT图像特征参数:
大小、形状、边缘、强化程度、是否合并钙化及囊变、是否结外侵犯。纳入淋巴结DECT定量参数:动脉期和静脉期的碘浓度、标
准化碘浓度、有效原子序数、标准化有效原子序数和能谱曲线斜率。同时从动脉期和静脉期碘图中手动勾画感兴趣区,提取影
像组学特征。结果:除结外侵犯外,转移淋巴结与非转移淋巴结的所有图像特征及DECT定量参数差异均有统计学意义(P均
< 0.05)。与传统影像特征模型[训练集曲线下面积(area under curve,AUC)=0.803,95%CI:0.765~0.841;验证集 AUC=0.804,95%
CI:0.748~0.861]相比,DECT定量参数模型(训练集AUC=0.939,95%CI:0.918~0.960;验证集 AUC=0.933,95%CI:0.901~0.965)、影
像组学模型(训练集 AUC=0.977,95%CI:0.966~0.988;验证集 AUC=0.931,95%CI:0.900~0.962)和联合模型(训练集 AUC=
0.977,95%CI:0.965~0.988;验证集 AUC=0.938,95%CI:0.908~0.968)对 LLNM 的诊断效能均更优(P 均<0.05)。训练集中影像
组学模型诊断效能优于DECT定量参数模型(P < 0.01),影像组学模型与联合模型效能相仿(P = 0.950)。结论:DECT碘图影
像组学特征对PTC患者LLNM具有较高的诊断价值。
[关键词] 甲状腺乳头状癌;双能CT;淋巴结转移;影像组学
[中图分类号] R736.1;R730.44 [文献标志码] A [文章编号] 1007⁃4368(2026)10⁃1518⁃10
doi:10.7655/NYDXBNSN251409
Value of dual⁃energy CT iodine map derived radiomics features in diagnosing the lateral
cervical lymph node metastasis in papillary thyroid cancer
Zhou Yu,Zhou Yan,Xu Yongkang,Kuang Xinru,Shen Meiping,Si Yan,Wang Yong,Su Guoyi,Wu Feiyun *
Department of Radiology,the First Affiliated Hospital of Nanjing Medical University,Nanjing 210029,China
[Abstract] Objective:To investigate the diagnostic value of radiomics features derived from iodine map of dual energy CT(DECT)
in diagnosing lateral cervical lymph node metastasis(LLNM)in patients of papillary thyroid cancer(PTC). Methods:A retrospective
analysis was performed on the clinical and imaging data of 200 patients with PTC who underwent preoperative DECT at the First
Affiliated Hospital of Nanjing Medical University between June 2022 and June 2024,all with pathological confirmation. A total of 777
lymph nodes were included,comprising 379 metastatic lymph nodes and 398 non⁃metastatic lymph nodes. Patients were randomly
allocated into the training and validation sets at a ratio of 7∶3. Clinical and CT features of lymph nodes,including size,shape,margin,
enhancement degree,calcification,cystic change,and extranodal extension were included. Quantitative DECT parameters including
iodine concentration,normalized IC,effective atomic number,normalized Zeff and slope of energy spectrum curve in arterial and
venous phase. Radiomics features were manually extracted from arterial and venous phase iodine maps by sketching region of interest.
Results:All clinical and DECT features showed significant differences between LLNM and non ⁃ LLNM(all P < 0.05)except for
extranodal extension. Compared with the conventional imaging model[training set,area under curve(AUC)=0.803,95% CI:0.765-
[基金项目] 国家自然科学基金(82171928,82471980);江苏省自然科学基金(BK20241983)
通信作者(Corresponding author),E⁃mail:wfy_njmu@163.com(ORCID:0000⁃0002⁃0343⁃0458)
∗

