基于灰阶超声特征结合影像组学对甲状腺滤泡性肿瘤的鉴别诊断价值  

The Differential Diagnostic Value of Gray-Scale Ultrasound Features Combined with Radiomics for Follicular Thyroid Neoplasms

作　　者：陈志明[1] 武晓凤 王伟[1] 徐莉萍 金晨阳 张凯仁 董凤林[1] Chen Zhiming;Wu Xiaofeng;Wang Wei;Xu Liping;Jin Chenyang;Zhang Kairen;Dong Feng(Department of Ultrasound,The First Affiliated Hospital of Soochow University,Suzhou 215006,China)

机构地区：[1]苏州大学附属第一医院超声科,苏州市215006

出　　处：《中国超声医学杂志》2024年第6期605-608,共4页Chinese Journal of Ultrasound in Medicine

摘　　要：目的 探讨基于灰阶超声特征结合影像组学的不同分类器模型对甲状腺滤泡腺瘤与癌的鉴别诊断价值。方法 回顾性分析术后病理证实为甲状腺滤泡性肿瘤患者的154个结节的灰阶超声资料。随机将所有结节按7∶3分为训练集和验证集。提取并筛选最优特征,构建5种分类器模型。最后,通过受试者工作特征(ROC)曲线评估每个分类模型及联合灰阶超声特征(以下称联合模型)的诊断效能。结果 基于超声影像组学特征的分类模型中训练集和验证集曲线下面积(AUC)分别为:0.681~0.740、0.545~0.739;逻辑回归(LR)模型在训练集中表现最佳(AUC:0.740)。联合模型在训练集和验证集的AUC分别为:0.740~0.853、0.730~0.848;LR模型仍表现最佳(AUC:0.853)。联合模型的LR的AUC与仅基于超声影像组学的LR模型的AUC比较差异具有统计学意义(P<0.05)。结论 基于灰阶超声特征结合影像组学模型对甲状腺滤泡腺瘤与癌的鉴别诊断具有较高的价值。Objective To explore the diagnostic value of different classification models based on gray-scale ultra-sound combined with ultrasound radiomics in distinguishing follicular thyroid adenoma from follicular thyroid carcino-ma.Methods A retrospective analysis was conducted on a cohort of 154 thyroid follicular tumors,which were con-firmed through surgical and pathological examination.The clinical and gray-scale ultrasound data of all patients were collected.The nodules were divided into training and test sets using a random partitioning approach,with a ratio of 7:3.Optimal features were extracted and selected before constructing models utilizing five machine learning classifi-cation algorithms.The diagnostic performance of each model was then assessed by receiver operating characteristic(ROC)curve analysis.Results In the classification model based only on ultrasound radiomics features,the area under the ROC curve(AUC)of training set and test set were 0.681-0.740 and 0.545-0.739,respectively.The logistic regression(LR)model exhibited the strongest discriminative performance on the training set(AUC:0.740).The AUC of the training set and the test set in the combined model were 0.740-0.853 and 0.730-0.848,respectively.The LR model outperformed on training data,with an AUC of o.853.The comparison of AUC of LR based on com-bined model and LR model based solely on ultrasound radiomics has statistical significance(P<0.05).Conclusions The classifier models based on gray-scale ultrasound combined with ultrasound radiomics features have a high diagnos-tic value in distinguishing follicular thyroid adenomas from follicular thyroid carcinomas.

关 键 词：甲状腺滤泡性肿瘤 影像组学 灰阶超声 

分 类 号：R445.1[医药卫生—影像医学与核医学] R736.1[医药卫生—诊断学]