Han Shuai, Liu Yu, Li Xiaohang, Jiang Xiao, Li Baifeng, Zhang Chengshuo, Zhang Jialin
Department of Hepatobiliary Surgery, The First Hospital of China Medical University, Shenyang, China.
Department of Radiology, The First Hospital of China Medical University, Shenyang, China.
Front Oncol. 2022 Mar 25;12:800449. doi: 10.3389/fonc.2022.800449. eCollection 2022.
The purpose of this study was to develop and validate a preoperative nomogram of differentiating benign and malignant gallbladder polypoid lesions (GPs) combining clinical and radiomics features.
The clinical and imaging data of 195 GPs patients which were confirmed by pathology from April 2014 to May 2021 were reviewed. All patients were randomly divided into the training and testing cohorts. Radiomics features based on 3 sequences of contrast-enhanced computed tomography were extracted by the Pyradiomics package in python, and the nomogram further combined with clinical parameters was established by multiple logistic regression. The performance of the nomogram was evaluated by discrimination and calibration.
Among 195 GPs patients, 132 patients were pathologically benign, and 63 patients were malignant. To differentiate benign and malignant GPs, the combined model achieved an area under the curve (AUC) of 0.950 as compared to the radiomics model and clinical model with AUC of 0.929 and 0.925 in the training cohort, respectively. Further validation showed that the combined model contributes to better sensitivity and specificity in the training and testing cohorts by the same cutoff value, although the clinical model had an AUC of 0.943, which was higher than 0.942 of the combined model in the testing cohort.
This study develops a nomogram based on the clinical and radiomics features for the highly effective differentiation and prediction of benign and malignant GPs before surgery.
本研究旨在开发并验证一种结合临床和影像组学特征的术前列线图，用于鉴别胆囊息肉样病变（GPs）的良恶性。
回顾性分析2014年4月至2021年5月间195例经病理证实的GPs患者的临床和影像数据。所有患者被随机分为训练组和测试组。通过Python中的Pyradiomics软件包提取基于3个序列的增强CT的影像组学特征，并通过多元逻辑回归建立进一步结合临床参数的列线图。通过区分度和校准来评估列线图的性能。
195例GPs患者中，132例病理结果为良性，63例为恶性。在训练组中，为鉴别GPs的良恶性，联合模型的曲线下面积（AUC）为0.950，而影像组学模型和临床模型的AUC分别为0.929和0.925。进一步验证表明，尽管在测试组中临床模型的AUC为0.943，高于联合模型的0.942，但联合模型在训练组和测试组中以相同的截断值具有更好的敏感性和特异性。
本研究基于临床和影像组学特征开发了一种列线图，用于术前高效鉴别和预测GPs的良恶性。
