基于多模态超声参数及临床指标的结直肠癌变诊断的预测模型  

Predictive model for colorectal cancer transformation diagnosis based on multimodal ultrasound parameters and clinical indicators

作　　者：孟祥安 郝永欣 骆允 李阳 韩小林 MENG Xiang-an;HAO Yong-xin;LUO Yun;LI Yang;HAN Xiao-lin(Department of Ultrasound,Hebei Petro China Central Hospital,Lang-fang 065000,China;Department of General Surgery,Hebei Petro China Central Hospital,Lang-fang 065000,China)

机构地区：[1]河北中石油中心医院超声医学科,河北廊坊065000 [2]河北中石油中心医院普通外科,河北廊坊065000

出　　处：《中国现代普通外科进展》2025年第1期17-22,共6页Chinese Journal of Current Advances in General Surgery

基　　金：廊坊市科技支撑计划项目(KYLL-2022-27)。

摘　　要：目的:构建基于多模态超声参数和临床指标的结直肠良恶性病灶预测模型,并评估预测模型效能。方法:收集2020年3月—2024年3月河北中石油中心医院治疗的结直肠占位病变198例患者的临床资料,依据病理学诊断结果分为良性病灶组(n=102)和恶性病灶组(n=96),所有患者均行多模态超声检查。采用多因素Logistic回归分析筛选结直肠癌变发生的影响因素;采用R4.2.3软件构建列线图预测模型,Hosmer-Lemeshow检验、校正曲线、ROC曲线、临床决策曲线进行验证。结果:良性病灶组与恶性病灶组内部回声、形态、血流信号、上升时间(RT)、造影剂快进快出、杨氏模量平均值(Emean)、年龄、便潜血阳性、息肉比较差异有统计学意义(P<0.05)。内部回声不均匀、形态不规则、血流信号丰富、造影剂快进快出、年龄≥60岁、便潜血阳性、息肉≥2个是结直肠癌变发生的独立危险因素(P<0.05),RT是结直肠癌变发生的保护因素(P<0.05)。列线图预测模型内部验证结果显示,Hosmer-Lemeshow检验χ^(2)=3.661,P=0.886,校正曲线显示实际概率与预测概率基本一致,ROC曲线的AUC为0.802(95%CI:0.732~0.871),提示该列线图预测模型校准度、区分度良好。在0.28~0.98的高风险阈值范围内,决策曲线在All线和None线上方,此范围内有较高的净收益,临床实用性较高。结论:内部回声、形态、血流信号、造影剂快进快出、年龄、便潜血阳性、息肉、RT是结直肠癌变发生的影响因素,以此为基础构建的列线图模型预测效能良好,为临床医师的早期干预提供参考。Objective:To construct a predictive model for benign and malignant colorectal lesions based on modal ultrasound parameters and clinical indicators,and evaluate the effectiveness of the predictive model.Methods:Clinical data of 198 patients with colorectal lesions treated in Hebei Petro China Central Hospital from March 2020 to March 2024 were recorded.According to pathological diagnosis,they were grouped into a benign lesion group of 102 cases and a malignant lesion group of 96 cases.All patients underwent multimodal ultrasound examination.Multivariate Logistic regression analysis was applied to screen the influencing factors of colorectal cancer progression.R software package was applied to build a nomogram prediction model.Hosmer-Lemeshow test,calibration curve,ROC curve,and clinical decision curve were used for validation.Results:There were statistically obvious differences in internal echo,morphology,blood flow signal,rise time(RT),contrast agent of"fast in and fast out",mean Young's modulus(Emean),age,positive fecal occult blood,and polyps between the benign and malignant lesion groups(P<0.05).Uneven internal echoes,irregular shapes,abundant blood flow signals,contrast agent of"fast in and fast out",age≥60 years,positive fecal occult blood,and polyps≥2 were independent risk factors for colorectal cancer(P<0.05),while RT is a protective factor for colorectal cancer(P<0.05).The internal validation results of the nomogram prediction model showed that the Hosmer Lemeshow test showed χ^(2)=3.661 and P=0.886.The calibration curve showed that the actual probability was basically consistent with the predicted probability,and the AUC of the ROC curve was 0.802(95%CI:0.732-0.871),indicating that the calibration and discrimination of the column chart prediction model were good.Within the high-risk threshold range of 0.28-0.98,the decision curve was above the All line and None line,indicating a high net benefit and clinical practicality.Conclusion:Internal echo,morphology,blood flow signal,contrast agent of"fast in

关 键 词：结直肠肿瘤 多模态超声 影响因素 预测模型 

分 类 号：R735.3[医药卫生—肿瘤]