文献速递腹部技术与安全PETCT

基于PET原发灶的影像组学特征对传统代谢参数及传统危险因素在胃癌术前淋巴结转移预测中的增量价值

Incremental value of PET primary lesion-based radiomics signature to conventional metabolic parameters and traditional risk factors for preoperative prediction of lymph node metastases in gastric cancer.
PubMed 文献 · Abdom Radiol (NY) · 2022-11-23

该回顾性研究纳入127例术前行PET/CT的胃癌患者,构建基于PET原发灶的影像组学特征(Rad-score),并与PET常规代谢参数及传统危险因素比较。结果显示Rad-score是淋巴结转移的独立危险因素(OR 6.38),将其加入常规参数和危险因素后AUC由0.751提升至0.882(P=0.006),NRI和IDI均为0.293。

要点

  • Rad-score在淋巴结转移组高于非转移组(0.35 vs -0.61,P<0.001)。
  • Rad-score是胃癌淋巴结转移的独立危险因素(OR 6.38,95%CI 2.73-14.91)。
  • 加入Rad-score后预测淋巴结转移的AUC从0.751升至0.882(P=0.006)。

核心结论

  • 在胃癌患者中,基于PET原发灶的影像组学特征与淋巴结转移显著相关,并可在PET常规代谢参数和传统危险因素之上提高对淋巴结转移的预测能力,可能为胃癌的个体化诊疗提供增量价值。

研究设计与样本

  • 回顾性研究,纳入127例术前行PET/CT的胃癌患者,收集基本临床数据和PET常规代谢参数;采用LASSO logistic回归构建影像组学特征;以术后组织学结果为参考标准将患者分为淋巴结转移组和非淋巴结转移组;采用ROC曲线评估Rad-score的判别能力,计算IDI和NRI确认增量价值;采用Bootstrap重采样(500次)进行内部验证。

关键影像表现/征象

  • 淋巴结转移组的Rad-score高于非淋巴结转移组(0.35 vs -0.61,P<0.001)。

诊断标准与关键数值

  • Rad-score预测淋巴结转移的独立危险因素OR为6.38(95%CI 2.73-14.91,P<0.0001)。
  • 传统危险因素模型的AUC为0.651,加入PET常规参数后AUC为0.751(P=0.02)。
  • 在PET常规代谢参数和传统危险指标基础上加入Rad-score后AUC为0.882,较0.751显著提高(P=0.006)。
  • Bootstrap重采样(500次)内部验证的95%CI为0.802-0.948,敏感度为89.5%,阳性预测值为93.5%。
  • 加入Rad-score后NRI为0.293(P=0.0040),IDI为0.293(P=0.0045)。

临床应用与阅片要点

  • 作者认为基于PET原发灶的影像组学特征可提高对胃癌淋巴结转移的预测能力,可能为个体化诊疗提供增量价值。

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

BACKGROUND: Precise preoperative prediction of lymph node metastasis (LNM) is crucial for optimal diagnosis and treatment in patients with gastric cancer (GC), in which existing imaging methods have certain limitations. We hypothesized that PET primary lesion-based radiomics signature could provide incremental value to conventional metabolic parameters and traditional risk indicators in predicting LNM in patients with GC.

METHODS: This retrospective study was performed in 127 patients with GC who underwent preoperative PET/CT. Basic clinical data and PET conventional metabolic parameters were collected. Radiomics signature was constructed by the least absolute shrinkage and selection operator algorithm (LASSO) logistic regression. Based on the postoperative histological results, the patients were divided into LNM group and non-lymph node metastasis (NLNM) group. Receiver-operating characteristic (ROC) was used to evaluate the discriminatory ability of Radiomics score (Rad-score) for predicting LNM and determine whether adding Rad-score to PET conventional metabolic parameters and traditional risk factors could improve the predictive value in LNM. The Integrated discrimination improvement (IDI) and net reclassification improvement (NRI) were calculated to further confirm the incremental value of Rad-score for predicting LNM in GC.

RESULTS: The LNM group had higher Rad-score than NLNM group [(0.35 (-0.13-0.85) vs. -0.61 (-1.92-0.18), P < 0 .001)]. After adjusted for gender, age, BMI, and FBG, multivariable logistic regression analysis illustrated that Rad-score (OR: 6.38, 95% CI: 2.73-14.91, P < 0.0001) was independent risk factors for LNM in GC. Adding PET conventional parameters to traditional risk factors increased the predictive value of LNM in GC (AUC 0.751 vs 0.651, P = 0.02). Additional inclusion of Rad-score to conventional metabolic parameters and traditional risk indicators significantly improved the AUC (0.882 vs 0.751; P = 0.006). Bootstrap resampling (times = 500) was used for internal verification, 95% confidence interval (CI) was 0.802-0.948, with the sensitivity equaled to 89.5%, and positive predictive value (PPV) was 93.5%. When Rad-score was added to conventional metabolic parameters and traditional risk indicators, net reclassification improvement (NRI) was 0.293 (P = 0.0040) and integrated discrimination improvement (IDI) was 0.293 (P = 0.0045).

CONCLUSION: In GC patients, PET Radiomics signature of the primary lesion-based was significantly associated with LNM and could improve the prediction of LNM above PET conventional metabolic parameters and traditional risk factors, which could provide incremental value for individual diagnosis and treatment of GC.

来源:Abdom Radiol (NY),查看原文。版权归原作者所有。

引用

  1. Xue XQ, Yu WJ, Shao XL, et al. Incremental value of PET primary lesion-based radiomics signature to conventional metabolic parameters and traditional risk factors for preoperative prediction of lymph node metastases in gastric cancer. Abdom Radiol (NY). 2022. PMID: 36418614. doi:10.1007/s00261-022-03738-4

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。