基于CT影像组学及临床多因素回归分析预测肝细胞肝癌TACE术后中重度腹痛的模型构建

Prediction Model of Moderate and Severe Abdominal Pain after TACE for Hepatocellular Carcinoma Based on CT Imaging and Clinical Multivariate Regression Analysis

目的探讨肝细胞癌(HCC)病人肝动脉插管化疗栓塞(TACE)术后急性严重腹痛发生危险因素,并分别构建影像组学模型、Logistics回归模型。方法回顾性分析128例经病理证实为肝细胞肝癌并实行TACE术的患者,分别收集其影像学资料及临床资料。按照8:2分成训练集及验证集,按照BPI评分标准,“0~3”分无或轻度疼为阴性组,共计57例,“4~10”分中重度腹痛分为阳性组,共计71例,比较阳性组与阴性组患者的临床及影像资料,采用影像组学及多因素Logistics回归分析筛选危险因素,并建立预测模型;受试者工作(ROC)曲线分析预测效能。结果影像组学模型:对动脉期、门脉期、延迟期、动脉期+门脉期、动脉期+延迟期、门脉期+延迟期及动脉期+门脉期+延迟期的模型进行分析,结果显示动脉期+门脉期+延迟期的联合组学模型的训练集AUC值为0.79,验证集AUC值为0.78,其效能高于其它六组组学模型。临床因素Logistics回归分析显示TACE术史、TACE术后腹痛病史、肿瘤距肝包膜距离是否>1cm、肿瘤最大直径是否>5cm、肿瘤数目是否>2、有无血管侵犯及TACE术式是其独立危险因素,临床因素Logistics回归分析筛选独立因素建立的预测模型AUC值为0.85。结论影像组学中动脉期+门脉期+延迟期的联合模型对于预测肝细胞肝癌TACE术后腹痛的发生有一定的临床价值,为以后影像组学对于肝癌TACE术后腹痛预测研究提供一定的价值,临床多因素回归分析模型的可以提供更多的临床因素,具有较好的效能,组学模型与临床多因素回归模型两者之间互有补充,建立预测模型,为临床提供更好的个性化疼痛管理,减轻病人的痛苦。

Objective To investigate the risk factors for acute severe abdominal pain after hepatocellular carcinoma(HCC)patients with hepatocellular carcinoma(HCC)after hepatic artery catheter chemoembolization(TACE),and to construct radiomics models and logistic regression models,respectively.Methods A retrospective analysis was performed for 128 patients with pathologically confirmed hepatocellular carcinoma and TACE surgery,and their imaging data and clinical data were collected.According to the 8:2 divided into training set and verification set,according to the BPI scoring standard,"0~3"score no or mild pain was negative group,a total of 57 cases,"4~10"moderate and severe abdominal pain was divided into positive group,a total of 71 cases,compared the clinical and imaging data of patients with moderate and severe pain with no or mild pain,radiomics and multivariate logistic regression analysis were used to screen risk factors,and a predictive model was established.Subject-operated(ROC)curve analysis predicts potency.Results The radiomics model:The models of arterial phase,portal phase,delayed phase,arterial phase+portal phase,arterial phase+delay period,portal phase+delay period and arterial phase+portal phase+delay period were analyzed,and the results showed that the AUC value of the training set of the combined omics model of arterial phase+portal phase+delay period was 0.79 and the AUC value of the validation set was 0.78,which was higher than that of the other six omics models.Logistic regression analysis of clinical factors showed the history of TACE,the history of abdominal pain after TACE,whether the distance between the tumor and the liver capsule was>1cm,whether the maximum diameter of the tumor was>5cm,and whether the number of tumors was>2.Whether there is vascular invasion is its independent risk factor,and the AUC value of the prediction model established by clinical factor logistic regression analysis to screen independent factors is 0.85.Conclusion The combined model of arterial phase+portal phase+delayed phase in radiomics has certain clinical value for predicting the occurrence of abdominal pain after TACE surgery for hepatocellular carcinoma,which provides certain value for future radiomics for the prediction of abdominal pain after TACE surgery for hepatocellular carcinoma,and the clinical multivariate regression analysis model can provide more clinical factors and has good efficacy,and the omics model and the clinical multivariate regression model complement each other,and the prediction model is established to provide better personalized pain management for the clinic.Alleviates the patient's suffering.