文献速递腹部MR

基于钆塞酸二钠增强MRI诊断亚厘米肝细胞癌的新型决策树算法:多中心开发与验证

A Novel Diagnostic Algorithm for Subcentimeter Hepatocellular Carcinoma Utilizing Gd-EOB-DTPA-Enhanced MRI: Multicenter Development and Validation.
PubMed 文献 · Acad Radiol · 2025-09-25

该多中心回顾性诊断准确性研究纳入419例≤1.0 cm局灶性肝结节患者,利用训练集225例、内部测试集96例及两个独立中心外部验证集98例,开发并验证整合扩散受限、非环状动脉期高强化、移行期低信号及轻中度T2WI高信号四种MRI特征的决策树算法,其敏感度90.0%-93.3%、特异度89.1%-94.7%,显著优于LI-RADS LR-4、改良LR-4及JSH标准。

要点

  • 该决策树算法在训练集、内部测试集和外部验证集的准确率分别为93.8%、90.6%和90.8%。
  • 20%-23.3%的亚厘米肝细胞癌无动脉期高强化,但算法对此类病例的准确率仍达80.0%-87.5%。
  • 算法性能显著优于LI-RADS LR-4(P<0.001)、改良LR-4(P=0.002)和JSH标准(P=0.005)。

核心结论

  • 该算法降低了对动脉期高强化的依赖,整合高收益MRI特征并简化流程,在多中心队列中准确率高,支持其临床转化用于早期肝细胞癌检测。

研究设计与样本

  • 多中心回顾性诊断准确性研究,纳入419例≤1.0 cm局灶性肝结节患者,训练集225例和内部测试集96例来自单中心,外部验证集98例来自两个独立中心,采用多变量逻辑回归和分类回归树(CART)建模,参考标准未在原文摘要中明确说明。

关键影像表现/征象

  • 算法整合四种MRI特征:扩散受限、非环状动脉期高强化、移行期低信号、轻中度T2WI高信号。
  • 20%-23.3%的亚厘米肝细胞癌表现为动脉期高强化缺失。
  • 对于动脉期高强化缺失的亚厘米肝细胞癌,算法诊断准确率为80.0%-87.5%。

诊断标准与关键数值

  • 训练集:敏感度92.9%(104/112),特异度94.7%(107/113),准确率93.8%。
  • 内部测试集:敏感度90.0%(36/40),特异度91.1%(51/56),准确率90.6%。
  • 外部验证集:敏感度93.3%(40/43),特异度89.1%(49/55),准确率90.8%。
  • 与LI-RADS LR-4比较P<0.001,与改良LR-4比较P=0.002,与JSH标准比较P=0.005。

临床应用与阅片要点

  • 对于≤1.0 cm的局灶性肝结节,可综合评估扩散受限、非环状动脉期高强化、移行期低信号及轻中度T2WI高信号。
  • 无动脉期高强化的亚厘米肝细胞癌仍占20%-23.3%,算法对此类病例准确率达80.0%-87.5%,阅片时不应仅因缺乏动脉期高强化而排除诊断。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

英文原文(摘要)· 点击展开

OBJECTIVES: To develop and validate a novel decision-tree algorithm using Gd-EOB-DTPA-enhanced MRI for diagnosing subcentimeter hepatocellular carcinoma (scHCC ≤1.0 cm), addressing limitations of current guidelines in detecting lesions lacking arterial phase hyperenhancement (APHE).

METHODS: This multicenter retrospective diagnostic accuracy study analyzed 419 patients with focal liver nodules ≤ 1.0 cm, utilizing training set (n=225) and internal test set (n=96) from a single center, and an external validation set (n=98) from two independent centers. Multivariable logistic regression and classification and regression tree (CART) modeling integrated four MRI features: restricted diffusion, non-rim APHE, transitional phase hypointensity, and mild-moderate T2WI hyperintensity. Performance was compared to LI-RADS LR-4, modified LR-4, and Japan Society of Hepatology (JSH) criteria.

RESULTS: A total of 419 patients (mean age 53±12 years, 333 men) were analyzed. The algorithm achieved: Training set: Sensitivity 92.9% (104/112), specificity 94.7% (107/113), accuracy 93.8%. Internal test set: Sensitivity 90.0% (36/40), specificity 91.1% (51/56), accuracy 90.6%. External validation: Sensitivity 93.3% (40/43), specificity 89.1% (49/55), accuracy 90.8%. It significantly outperformed existing criteria (P < 0.001 vs. LR-4; P = 0.002 vs. modified LR-4; P = 0.005 vs. JSH). A critical finding was that 20%-23.3% of scHCCs demonstrated the absence of APHE, yet these cases achieved diagnostic accuracy rates of 80.0%-87.5%.

CONCLUSION: This algorithm reduces dependency on APHE, integrates high-yield MRI features, and streamlines protocols. Its high accuracy across multicenter cohorts supports immediate clinical translation for early HCC detection.

来源:Acad Radiol,查看原文。版权归原作者所有。

引用

  1. Zhang J, Lu Z, Shen H, et al. A Novel Diagnostic Algorithm for Subcentimeter Hepatocellular Carcinoma Utilizing Gd-EOB-DTPA-Enhanced MRI: Multicenter Development and Validation. Acad Radiol. 2025. PMID: 41006083. doi:10.1016/j.acra.2025.09.006

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。