Deng Yisen, Liu Zhan, Wang Xuming, Gao Xixi, Zhang Zhaohua, Zhang Dingkai, Xu Mingyuan, Chen Haijie, Fan Xueqiang, Yang Yuguang, Ye Zhidong, Liu Peng, Wen Jianyan
Department of Cardiovascular Surgery, Peking University China-Japan Friendship School of Clinical Medicine, 100191 Beijing, China.
Department of Cardiovascular Surgery, China-Japan Friendship Hospital, 100029 Beijing, China.
Rev Cardiovasc Med. 2023 Nov 23;24(11):327. doi: 10.31083/j.rcm2411327. eCollection 2023 Nov.
Postoperative new atrial fibrillation (POAF) is a commonly observed complication after off-pump coronary artery bypass surgery (OPCABG), and models based on radiomics features of epicardial adipose tissue (EAT) on non-enhanced computer tomography (CT) to predict the occurrence of POAF after OPCABG remains unclear. This study aims to establish and validate models based on radiomics signature to predict POAF after OPCABG.
Clinical characteristics, radiomics signature and features of non-enhanced CT images of 96 patients who underwent OPCABG were collected. The participants were divided into a training and a validation cohort randomly, with a ratio of 7:3. Clinical characteristics and EAT CT features with statistical significance in the multivariate logistic regression analysis were utilized to build the clinical model. The least absolute shrinkage and selection operator (LASSO) algorithm was used to identify significant radiomics features to establish the radiomics model. The combined model was constructed by integrating the clinical and radiomics models.
The area under the curve (AUC) of the clinical model in the training and validation cohorts were 0.761 (95% CI: 0.634-0.888) and 0.797 (95% CI: 0.587-1.000), respectively. The radiomics model showed better discrimination ability than the clinical model, with AUC of 0.884 (95% CI: 0.806-0.961) and 0.891 (95% CI: 0.772-1.000) respectively for the training and the validation cohort. The combined model performed best and exhibited the best predictive ability among the three models, with AUC of 0.922 (95% CI: 0.853-0.990) in the training cohort and 0.913 (95% CI: 0.798-1.000) in the validation cohort. The calibration curve demonstrated strong concordance between the predicted and actual observations in both cohorts. Furthermore, the Hosmer-Lemeshow test yielded value of 0.241 and 0.277 for the training and validation cohorts, respectively, indicating satisfactory calibration.
The superior performance of the combined model suggests that integrating of clinical characteristics, radiomics signature and features on non-enhanced CT images of EAT may enhance the accuracy of predicting POAF after OPCABG.
术后新发心房颤动（POAF）是非体外循环冠状动脉旁路移植术（OPCABG）后常见的并发症，基于非增强计算机断层扫描（CT）上心外膜脂肪组织（EAT）的放射组学特征预测OPCABG后POAF发生的模型尚不清楚。本研究旨在建立并验证基于放射组学特征的模型，以预测OPCABG后的POAF。
收集96例行OPCABG患者的临床特征、放射组学特征及非增强CT图像特征。参与者随机分为训练组和验证组，比例为7:3。将多因素逻辑回归分析中有统计学意义的临床特征和EAT CT特征用于构建临床模型。采用最小绝对收缩和选择算子（LASSO）算法识别显著的放射组学特征以建立放射组学模型。通过整合临床模型和放射组学模型构建联合模型。
临床模型在训练组和验证组的曲线下面积（AUC）分别为0.761（95%CI：0.634-0.888）和0.797（95%CI：0.587-1.000）。放射组学模型显示出比临床模型更好的区分能力，训练组和验证组的AUC分别为0.884（95%CI：0.806-0.961）和0.891（95%CI：0.772-1.000）。联合模型在三个模型中表现最佳，预测能力最强，训练组的AUC为0.922（95%CI：0.853-0.990），验证组的AUC为0.913（95%CI：0.798-1.000）。校准曲线显示两组中预测值与实际观察值之间具有高度一致性。此外，训练组和验证组的Hosmer-Lemeshow检验值分别为0.241和0.277，表明校准效果良好。
联合模型的卓越性能表明，整合临床特征、放射组学特征以及EAT的非增强CT图像特征可能会提高OPCABG后预测POAF的准确性。
