Prediction of pathological complete response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer using contrast-enhanced ultrasound radiomics
Prediction of pathological complete response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer using contrast-enhanced ultrasound radiomics
秦琼 1吴玉泉 1文荣 1白秀梅 1高瑞智 1林雅丹 1吕佳忆 1何云 1杨红1
扫码查看
点击上方二维码区域，可以放大扫码查看
作者信息
1. 广西医科大学第一附属医院超声医学科，南宁 530021
折叠
摘要
目的 评估基于超声造影的影像组学模型对局部进展期直肠癌(LARC)患者新辅助放化疗(nCRT)后病理完全缓解(pCR)的预测效能。 方法 本研究回顾性纳入2018年4月至2023年4月在广西医科大学第一附属医院接受nCRT后行全直肠系膜切除的106例LARC患者，以6∶4随机划分为训练集63例(pCR者14例)和验证集43例(pCR者12例)。基于PyRadiomics从超声造影图像肿瘤感兴趣区域提取影像学特征。采用类内相关系数、Mann-Whitney U检验、最小绝对收缩和选择算子算法对特征进行降维。最后选取7个与pCR相关的影像学特征，基于R语言使用弹性网络回归构建超声造影影像组学模型，并与临床特征融合构建一个联合模型。采用ROC曲线下面积(AUC)评估模型的诊断效能。 结果 训练集中，超声造影影像组学模型的AUC为0.695(95%CI＝0.532～0.859)，联合模型的AUC为0.726(95%CI＝0.584～0.868)。验证集中，超声造影影像组学模型的AUC为0.763(95%CI＝0.625～0.902)，联合模型的AUC为0.790(95%CI＝0.653～0.928)。单因素及多因素逻辑回归分析均表明CA199(P<0.05)和超声造影影像组学评分(P<0.001)可作为LARC患者nCRT后pCR的独立预测因子。 结论 超声造影影像组学评分对LARC患者nCRT后是否达到pCR具有一定预测价值，有可能为预测LARC患者nCRT后达到pCR提供一种无创影像学生物标志物。 Objective To evaluate the diagnostic performance of radiomics model based on contrast-enhanced ultrasound(CEUS) in predicting pathological complete response(pCR) after neoadjuvant chemoradiotherapy(nCRT) in patients with locally advanced rectal cancer(LARC). Methods One hundred and six patients with LARC who underwent total mesorectal excision after nCRT between April 2018 and April 2023 in the First Affiliated Hospital of Guangxi Medical University were retrospectively included, the patients were randomly divided into a training set of 63(14 pCR patients) and a validation set of 43(12 pCR patients) in a 6∶4 ratios. Radiomics features were extracted from the tumors′ region of interest of CEUS images based on PyRadiomics. Intra-class correlation coefficient(ICC), Mann-Whitney U test, and least absolute shrinkage and selection operator(LASSO) algorithms were used to reduce features dimension. Finally, 7 radiomics features relevanted to pCR were selected to construct an ultrasomics model using elastic network regression, based on the R language. A combined model was constructed by jointing clinical feature. The performance of the models was assessed with the area under the ROC curve(AUC). Results The AUC of the ultrasomics model and the combined model was 0.695(95%CI=0.532-0.859) and 0.726(95%CI=0.584-0.868) respectively in the training set. The AUC of the ultrasomics model and the combined model was 0.763(95%CI=0.625-0.902) and 0.790(95%CI=0.653-0.928) respectively in the validation set. Both univariate and multivariate Logistic regression analyses showed that CA199(P<0.05) and ultrasomics score(P<0.001) could be an independent predictor of pCR after nCRT in patients with LARC. Conclusions The CEUS-based radiomics scores has certain predictive value for whether LARC patients achieve pCR after nCRT, and may provide a non-invasive imaging biomarker for predicting LARC patients achieve pCR after nCRT.
Abstract
Objective To evaluate the diagnostic performance of radiomics model based on contrast-enhanced ultrasound(CEUS) in predicting pathological complete response(pCR) after neoadjuvant chemoradiotherapy(nCRT) in patients with locally advanced rectal cancer(LARC). Methods One hundred and six patients with LARC who underwent total mesorectal excision after nCRT between April 2018 and April 2023 in the First Affiliated Hospital of Guangxi Medical University were retrospectively included, the patients were randomly divided into a training set of 63(14 pCR patients) and a validation set of 43(12 pCR patients) in a 6∶4 ratios. Radiomics features were extracted from the tumors′ region of interest of CEUS images based on PyRadiomics. Intra-class correlation coefficient(ICC), Mann-Whitney U test, and least absolute shrinkage and selection operator(LASSO) algorithms were used to reduce features dimension. Finally, 7 radiomics features relevanted to pCR were selected to construct an ultrasomics model using elastic network regression, based on the R language. A combined model was constructed by jointing clinical feature. The performance of the models was assessed with the area under the ROC curve(AUC). Results The AUC of the ultrasomics model and the combined model was 0.695(95%CI=0.532-0.859) and 0.726(95%CI=0.584-0.868) respectively in the training set. The AUC of the ultrasomics model and the combined model was 0.763(95%CI=0.625-0.902) and 0.790(95%CI=0.653-0.928) respectively in the validation set. Both univariate and multivariate Logistic regression analyses showed that CA199(P<0.05) and ultrasomics score(P<0.001) could be an independent predictor of pCR after nCRT in patients with LARC. Conclusions The CEUS-based radiomics scores has certain predictive value for whether LARC patients achieve pCR after nCRT, and may provide a non-invasive imaging biomarker for predicting LARC patients achieve pCR after nCRT.
万方数据