基于超声影像组学和临床特征构建的联合模型诊断早期卵巢癌的临床价值  

Clinical value of a combined model based on ultrasound radiomics and clinical features in the diagnosis of early ovarian cancer

作　　者：何丽英 昌禹豪 马强 魏伟 闫娜 朱菲菲 江峰 HE Liying;CHANG Yuhao;MA Qiang;WEI Wei;YAN Na;ZHU Feifei;JIANG feng(Department of Ultrasound,the First Affiliated Hospital of Wannan Medical College,Yijishang Hospital,Wuhu 241000,China;Department of Ultrasound,Hangzhou First People’s Hospital,Hangzhou 310000,China)

机构地区：[1]皖南医学院第一附属医院弋矶山医院超声科,安徽芜湖241000 [2]杭州第一人民医院超声科,浙江杭州310000

出　　处：《临床超声医学杂志》2025年第1期39-47,共9页Journal of Clinical Ultrasound in Medicine

基　　金：皖南医学院中青年科研基金项目(WK2023ZQNZ29)。

摘　　要：目的基于超声影像组学和临床特征构建联合模型,探讨其诊断早期卵巢癌的临床价值。方法选取2012年1月至2023年1月于我院接受手术治疗的卵巢肿瘤患者272例,按7∶3比例随机分为训练集190例和内部验证集82例;另选2023年2月至2024年5月杭州市第一人民医院收治的卵巢肿瘤患者80例作为外部验证集。所有患者均行常规超声检查获得肿瘤位置、最大径、形态、外轮廓分界、Adler血流分级、肿瘤内部成分构成、肿瘤实性成分最大径、肿瘤内部乳头数量及腹水情况;收集患者年龄、绝经情况及血清肿瘤标志物。根据术后病理结果分为早期卵巢癌组和良性组,比较各数据集中两组上述参数的差异。基于训练集患者的术前二维超声灰阶图像提取超声影像组学特征,采用贪婪递归特征去除策略和最小绝对收缩和选择算子回归分析降维,保留系数非零的最优特征构建影像组学评分(Radscore)。采用单因素和多因素Logistic回归筛选诊断早期卵巢癌的独立影响因素,分别构建临床模型、联合模型并绘制列线图可视化。采用受试者工作特征(ROC)曲线、校准曲线、临床决策曲线评估各模型的区分度、校准度及临床适用性,并进行内部和外部验证。结果训练集中两组人附睾蛋白4(HE4)、肿瘤实性成分最大径、肿瘤最大径、肿瘤内部成分构成、Adler血流分级、腹水比较差异均有统计学意义(均P<0.05);内部验证集中两组HE4、糖类抗原125(CA125)、肿瘤实性成分最大径、肿瘤最大径、Adler血流分级、腹水比较差异均有统计学意义(均P<0.05);外部验证集中两组HE4、CA125、肿瘤最大径、Adler血流分级比较差异均有统计学意义(均P<0.05)。共提取1050个超声影像组学特征,经筛选及降维后保留13个系数非零的最优特征并计算Radscore即为影像组学模型。根据单因素和多因素Logistic回归分析结果,纳入Adler血流分级�Objective To establish a combined model based on ultrasound radiomics and clinical features,and to explore the clinical value of the model in the diagnosis of early ovarian cancer.Methods A total of 272 patients with ovarian tumors who received surgical treatment in our hospital from January 2012 to January 2023 were selected,and they were randomly divided into 190 cases in training set and 82 cases in internal verification set according to the ratio of 7∶3.While 80 patients with ovarian tumor in Hangzhou First People’s Hospital from February 2023 to May 2024 were selected as independent external validation set.All patients underwent routine ultrasound examination to obtain tumor location,maximum diameter,morphology,outer contour boundary,Adler blood flow grading,tumor internal composition,maximum diameter of solid tumor components,number of papillae within the tumor,and presence of ascites.The age,menopausal status,and serum tumor marker levels of patient were collected.According to the postoperative pathological results,the patients were divided into early ovarian cancer group and benign group,and the differences of above parameters between the two groups in each dataset were compared.The ultrasound radiomics features were extracted based on preoperative two-dimensional ultrasound grayscale images of patients in the training set,the greedy recursive feature removal strategy and the minimum absolute shrinkage and selection operator regression analysis were used to reduce the dimensionality,and the optimal features with non-zero coefficients were retained to construct the radiomics score(Radscore).Single factor and multiple factor Logistic regression were used to screen the independent influencing factors in the diagnosis of early ovarian cancer.A clinical model and a combined model were constructed,and visualized with Nomogram.The discrimination,calibration and clinical applicability of each model were evaluated by receiver operating characteristic(ROC)curve,calibration curve and clinical decision curve,and

关 键 词：超声检查 影像组学 卵巢癌 早期 鉴别诊断 

分 类 号：R445.1[医药卫生—影像医学与核医学] R737.31[医药卫生—诊断学]