国际医药卫生导报 ›› 2025, Vol. 31 ›› Issue (14): 2408-2413.DOI: 10.3760/cma.j.cn441417-20241209-14024

• 临床研究 • 上一篇    下一篇

基于多模态超声构建肝细胞癌微血管浸润的术前预测模型

纪莉 张姣 黎静 朱莉敏   

  1. 陕西省核工业二一五医院215功能科,咸阳 712000

  • 收稿日期:2024-12-09 出版日期:2025-07-01 发布日期:2025-08-06
  • 通讯作者: 张姣,Email:bf29186911@163.com
  • 基金资助:

    2024年陕西省科技计划(2024JC-YBMS-674)

Construction of a preoperative prediction model for microvascular infiltration in hepatocellular carcinoma based on multimodal ultrasound 

Ji Li, Zhang Jiao, Li Jing, Zhu Limin   

  1. 215 Functional Department, Shaanxi Nuclear Industry 215 Hospital, Xianyang 712000, China

  • Received:2024-12-09 Online:2025-07-01 Published:2025-08-06
  • Contact: Zhang Jiao, Email: bf29186911@163.com
  • Supported by:

    Plan of Science and Technology in Shaanxi in 2024 (2024JC-YBMS-674)

摘要:

目的 基于多模态超声构建肝细胞癌(hepatocellular carcinoma,HCC)微血管浸润(microvascular invasion,MVI)的术前预测模型。方法 选择2022年6月至2024年9月陕西省核工业二一五医院收治的HCC患者121例为研究对象。根据术前病理检查结果将其分为MVI发生组和未发生组。所有患者均完成二维超声、超声造影等多模态超声检查,并记录相关参数;查阅两组病历资料,对HCC微血管浸润可能影响因素进行分析;绘制列线图,构建HCC患者MVI的术前预测模型,并完成预测效能分析和验证。采用t检验和χ2检验进行统计分析。结果 121例HCC患者中,35例发生MVI,占28.93%。单因素及多因素分析结果表明,病灶长径和动脉相(arterial phase,AP)、门静脉相(portal phase,PP)、血管后相(posterior vascular phase,PVP)信号为HCC患者MVI发生的独立危险因素(均P<0.05)。构建模型公式:Logit(P)=12.151+4.280×病灶长径-AP×10.089-PP×6.231-PVP×4.710,利用该公式预测HCC患者术前MVI的发生;受试者操作特征曲线结果显示,发生组曲线下面积(AUC)0.971(95%CI 0.934~1.000),未发生组AUC 0.886(95%CI 0.751~1.000)。决策曲线分析结果显示,发生组阈概率在0~0.4和未发生组阈概率在0.1~1.0时,模型具有较高的正向净收益。临床影响曲线结果显示,发生组阈概率在0~0.4和未发生组阈概率在0~0.7时,该模型预测准确度高。结论 HCC患者MVI的发生常受到病灶长径、AP信号、PP信号及PVP信号的影响。基于多模态超声参数构建HCC患者MVI预测模型,具有较高的预测效能。

关键词: 肝细胞癌, 多模态超声, 微血管浸润, 术前预测模型, 预测效能

Abstract:

Objective To construct a preoperative prediction model of microvascular invasion (MVI) in hepatocellular carcinoma (HCC) by multimodal ultrasound. Methods A total of 121 patients with HCC treated at Shaanxi Nuclear Industry 215 Hospital from June 2022 to September 2024 were selected as the study objects, and were divided into an MVI occurrence group and a non-occurrence group according to the preoperative pathological examination results. All the patients completed multimodal ultrasound examinations, such as two-dimensional ultrasound and contrast-enhanced ultrasound, and the relevant parameters were recorded. Both groups' medical records were reviewed to analyze the possible influencing factors of MVI in HCC. The nomogram was drawn, and a preoperative prediction model of MVI in the patients was constructed; the predictive efficacy was analyzed and verified. t and χ2 tests were used for the statistical analysis. Results Among the 121 patients, 35 (28.93%) had MVI. The univariate and multivariate analyses results showed that lesion size and signals during the arterial phase (AP), portal phase (PP), and posterior vascular phase (PVP) were the independent risk factors for MVI in the patients (all P<0.05). The established model formula was as below: Logit (P)=12.151+4.280×lesion size-AP×10.089-PP×6.231-PVP×4.710, which was used to predict the occurrence of MVI in the patients before surgery. The receiver operating characteristic curve results showed that the area under the curve (AUC) of the occurrence group was 0.971 (95%CI 0.934-1.000), and that of the non-occurrence group 0.886 (95%CI 0.751-1.000). The results of decision curve showed that when the threshold probability of the occurrence group was 0-0.4 and the threshold probability of the non-occurrence group was 0.1-1.0, the model had a higher positive net profit. The clinical impact curve results showed that the prediction accuracy of the model was high when the threshold probability of the occurrence group was 0-0.4 and the threshold probability of the non-occurrence group was 0-0.7. Conclusions The occurrence of MVI in patients with HCC is often affected by lesion size and signals during AP, PP, and PVP. The prediction model of MVI in patients with HCC based on multimodal ultrasound parameters has high predictive efficacy.

Key words: Hepatocellular carcinoma,  , Multimodal ultrasound,  , Microvascular invasion,  , Preoperative prediction model,  , Predictive efficacy