文献速递腹部US

胰腺神经内分泌肿瘤:基于超声图像的影像组学模型预测肿瘤分级

Pancreatic neuroendocrine tumor: prediction of tumor grades by radiomics models based on ultrasound images.
PubMed 文献 · Br J Radiol · 2023-07-26

回顾性纳入64例经手术病理确诊的胰腺神经内分泌肿瘤患者,基于B型超声图像构建影像组学评分,用于预测G2/G3与G1的区分。训练队列AUC为0.844,测试队列AUC为0.833,训练/测试队列敏感度分别为0.750和0.786,特异度均为0.833。结论认为基于B型超声的影像组学数据有潜力预测胰腺神经内分泌肿瘤的病理分级。

要点

  • 基于B型超声图像的影像组学评分用于区分G2/G3与G1胰腺神经内分泌肿瘤,训练队列AUC为0.844,测试队列AUC为0.833。
  • 训练队列准确率81.8%、敏感度0.750、特异度0.833;测试队列准确率80.0%、敏感度0.786、特异度0.833。
  • 决策曲线分析显示影像组学评分具有临床获益。

核心结论

  • 基于B型超声图像构建的影像组学数据有潜力预测胰腺神经内分泌肿瘤患者的组织病理学肿瘤分级和Ki-67增殖指数。

研究设计与样本

  • 回顾性纳入64例经手术和组织病理学确诊的胰腺神经内分泌肿瘤患者(男34例、女30例,平均年龄52.4±12.2岁),分为训练队列44例和测试队列20例;采用最大相关最小冗余和最小绝对收缩与选择算子进行特征筛选,以受试者工作特征曲线分析评估模型性能。

诊断标准与关键数值

  • 按WHO 2017标准基于Ki-67增殖指数和有丝分裂活性将肿瘤分为G1、G2、G3级;最终纳入G1 18例、G2 35例、G3 11例;基于B型超声图像的影像组学评分预测G2/G3与G1,训练队列AUC 0.844、测试队列AUC 0.833;训练队列准确率81.8%、敏感度0.750、特异度0.833,测试队列准确率80.0%、敏感度0.786、特异度0.833。

临床应用与阅片要点

  • 作者认为基于B型超声图像的影像组学模型有潜力预测胰腺神经内分泌肿瘤的组织病理学分级和Ki-67增殖指数。

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

OBJECTIVE: We aimed to investigate whether the radiomics analysis based on B-mode ultrasound (BMUS) images could predict histopathological tumor grades in pancreatic neuroendocrine tumors (pNETs).

METHODS: A total of 64 patients with surgery and histopathologically confirmed pNETs were retrospectively included (34 male and 30 female, mean age 52.4 ± 12.2 years). Patients were divided into training cohort (n = 44) and validation cohort (n = 20). All pNETs were classified into Grade 1 (G1), Grade 2 (G2), and Grade 3 (G3) tumors based on the Ki-67 proliferation index and the mitotic activity according to WHO 2017 criteria. Maximum relevance minimum redundancy, least absolute shrinkage and selection operator were used for feature selection. Receiver operating characteristic curve analysis was used to evaluate the model performance.

RESULTS: Finally, 18 G1 pNETs, 35 G2 pNETs, and 11 G3 pNETs patients were included. The radiomic score derived from BMUS images to predict G2/G3 from G1 displayed a good performance with an area under the receiver operating characteristic curve of 0.844 in the training cohort, and 0.833 in the testing cohort. The radiomic score achieved an accuracy of 81.8% in the training cohort and 80.0% in the testing cohort, a sensitivity of 0.750 and 0.786, a specificity of 0.833 and 0.833 in the training/testing cohorts. Clinical benefit of the score also exhibited superior usefulness of the radiomic score, as shown by the decision curve analysis.

CONCLUSIONS: Radiomic data constructed from BMUS images have the potential for predicting histopathological tumor grades in patients with pNETs.

ADVANCES IN KNOWLEDGE: The radiomic model constructed from BMUS images has the potential for predicting histopathological tumor grades and Ki-67 proliferation indexes in patients with pNETs.

来源:Br J Radiol,查看原文。版权归原作者所有。

引用

  1. Dong, Yang, Tian, et al. Pancreatic neuroendocrine tumor: prediction of tumor grades by radiomics models based on ultrasound images. Br J Radiol. 2023. PMID: 37393539. doi:10.1259/bjr.20220783

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。