The dual-energy CT imaging model can predict the expression of Ki-67 in gastric stromal tumors before operation
-
摘要:
目的 探讨基于双能量CT联合影像组学模型评估胃间质瘤(GST)Ki-67表达水平的应用价值。 方法 回顾性收集盐城市第一人民医院2021年11月~2023年9月行双能量CT增强扫描并经手术病理及免疫组化确诊的GST患者105例。按照7：3的比例随机分为训练组(n=74)及测试组(n=31)根据术后免疫组化结果再分为Ki-67高表达组及Ki-67低表达组。记录所有患者的一般临床特点，分析肿瘤的常规CT特征，于静脉期图像测量、计算病灶双能量CT定量参数、提取影像组学特征，利用单因素分析及LASSO算法对上述特征进行筛选，使用Logistic回归分别构建常规CT征象模型、双能量CT模型、影像组学模型及联合模型。采用ROC曲线下面积对各模型诊断效能进行比较。使用DeLong检验比较各模型间曲线下面积的差异。 结果 肿瘤最大径、标准化碘浓度、能谱曲线斜率及6个影像组学特征在两组间的差异有统计学意义(P < 0.05)，联合模型为最佳模型，具有最高的预测效能。联合模型与其他3个模型间的差异均有统计学意义(P < 0.05)，其余各模型间差异无统计学意义(P > 0.05)。 结论 基于双能量CT联合影像组学模型在评估GST Ki-67表达水平方面具有一定的临床价值。 Abstract:Objective To explore the application value of radiomics model based on dual-energy CT to predict the Ki-67 expression level of gastric stromal tumor (GST). Methods Retrospective analysis of 105 cases of GST who underwent dual-energy CT enhanced scanning and were diagnosed by surgical pathology and immunohistochemistry at the First People's Hospital of Yancheng City from November 2021 to September 2023. All cases were divided into training group (n=74) and test group (n=31) in a 7:3 ratio, and divided into Ki-67 high expression group and Ki-67 low expression group according to the postoperative immunohistochemistry results. General clinical characteristics of all patients were recorded, and conventional CT signs of the tumors were analyzed, quantitative dual-energy CT parameters were measured and calculated on venous phase, and imaging omics features were extracted. The above features were screened using univariate analysis and LASSO algorithm, and Logistic regression was used to establish a conventional CT sign model, a dual-energy CT model, an imaging histology model, and a combined model. The diagnostic efficacy of the models was compared using the ROC curve and AUC. DeLong test was used to compare the differences of each AUC. Results The differences in maximum tumor diameter, normalized iodine concentration, K and six imaging omics features between the two groups were statistically significant (P < 0.05). The combined model was the best model with the highest predictive efficacy. The differences between the combined model and the other three models were statistically significant (P < 0.05), while the differences between the other models were not statistically significant (P > 0.05). Conclusion The radiomics model based on dual-energy CT has clinical value in predicting Ki-67 expression levels in GST. -
Key words:
- gastric stromal tumor /
- dual-energy CT /
- radiomics /
- Ki-67
-
图 1 双能量CT参数
Figure 1. Dual-energy CT parameters. A-B: iodograms and energy profile plots of patients with Ki-67 high-expression GST; C-D: iodograms and energy profile plots of patients with Ki-67 low-expression GST.
图 2 影像组学特征筛选
Figure 2. Screening of imaging histologic features. A: distribution of regression coefficients; B: 10-fold cross-validation plot; C: screened imaging histologic features.
图 3 各模型的诊断效能比较
Figure 3. Comparison of the diagnostic efficacy of the models. A: ROC for each model in the training group; B: ROC for each model in the testing group.
表 1 训练组及测试组患者一般临床特点及常规CT征象比较
Table 1. Comparison of general clinical characteristics and routine CT signs of patients in the training and test groups
Index Training group Test group Low expression(n=56) High expression(n=18) χ2/t P Low expression(n=19) High expression(n=12) χ2/t P Age (year, Mean±SD) 62.86±8.41 60.61±7.27 -1.016 0.313 63.42±6.75 65.33±9.63 0.651 0.520 Gender [n(%)] 3.609 0.057 1.777 0.183 Female 36(83.7) 7(16.3) 11(73.3) 4(26.7) Male 20(64.5) 11(35.5) 8(50.0) 8(50.0) Clinical symptom [n(%)] 0.147 0.701 0.176 0.675 No 19(73.1) 7(26.9) 5(55.6) 4(44.4) Yes 37(77.1) 11(22.9) 14(63.6) 8(36.4) Portion [n(%)] 0.342 0.559 1.052 0.305 Body 33(73.7) 12(26.7) 13(68.4) 6(31.6) Fundus 23(79.3) 6(20.7) 6(50.0) 6(50.0) Growth method [n(%)] 1.729 0.421 1.057 0.590 Intracavitary 26(74.3) 9(25.7) 11(68.8) 5(31.2) Extragastric 25(73.5) 9(26.5) 6(50.0) 6(50.0) Both 5(100.0) 0(0.0) 2(66.7) 1(33.3) Morphology [n(%)] 2.306 0.129 1.517 0.218 Orbicular 39(81.3) 9(18.8) 15(68.2) 7(31.8) Irregular 17(65.4) 9(34.6) 4(44.4) 5(55.6) Calcify [n(%)] 0.149 0.700 0.247 0.619 No 41(74.5) 14(25.5) 17(63.0) 10(37.0) Yes 15(78.9) 4(21.1) 2(50.0) 2(50.0) Necrotic [n(%)] 1.028 0.272 3.656 0.056 No 39(79.6) 10(20.4) 13(76.5) 4(23.5) Yes 17(68.0) 8(32.0) 6(42.9) 8(57.1) Ulcers [n(%)] 2.511 0.113 0.327 0.567 No 47(79.7) 12(20.3) 13(65.0) 7(35.0) Yes 9(60.0) 6(40.0) 6(54.5) 5(45.5) Maximum diameter (mm, Mean±SD) 32.46±20.07 65.50±59.18 2.255 0.037 31.00±16.53 54.42±18.45 3.688 0.001 
下载: 导出CSV
表 2 训练组及测试组双能量CT参数比较
Table 2. Comparison of dual-energy CT parameters in the training and test group (Mean±SD)
Index Training group Test group Low expression(n=56) High expression(n=18) χ2/t P Low expression(n=19) High expression(n=12) χ2/t P NIC (mg/mL) 25.13±3.26 29.51±4.35 2.478 0.016 27.98±5.35 32.55±3.73 4.024 0.001 K 1.87±0.65 2.59±1.04 2.736 0.012 2.37±0.59 3.37±1.06 2.985 0.009 NIC: Normalized iodine concentration; K: Slope of energy spectrum curve.
下载: 导出CSV
表 3 各模型在训练组及测试组诊断效能中的比较
Table 3. Comparison of the models in the training and test groups
Group AUC Sensitivity(%) Specificity(%) 95% CI Training group Conventional CT signs model 0.682 67 75 0.518-0.846 Dual-energy CT parameters model 0.707 72 75 0.543-0.871 Radiomics model 0.666 72 64 0.531-0.801 Combind model 0.794 61 84 0.669-0.918 Test group Conventional CT signs model 0.759 66 78 0.571-0.866 Dual-energy CT parameters model 0.798 75 77 0.629-0.868 Radiomics model 0.829 72 75 0.683-0.875 Combind model 0.947 82 79 0.772-0.967
下载: 导出CSV
-
[1] Dermawan JK, Rubin BP. Molecular pathogenesis of gastrointestinal stromal tumor: a paradigm for personalized medicine[J]. Annu Rev Pathol, 2022, 17: 323-44. doi: 10.1146/annurev-pathol-042220-021510 [2] Blay JY, Kang YK, Nishida T, et al. Gastrointestinal stromal tumours[J]. Nat Rev Dis Primers, 2021, 7(1): 22. doi: 10.1038/s41572-021-00254-5 [3] Gerdes J, Lemke H, Baisch H, et al. Cell cycle analysis of a cell proliferation- associated human nuclear antigen defined by the monoclonal antibody Ki-67[J]. J Immunol, 1984, 133(4): 1710-5. doi: 10.4049/jimmunol.133.4.1710 [4] Liu QX, Peng ZH, Shen LF, et al. Prognostic and clinicopathological value of ki-67 in melanoma: a meta-analysis[J]. Front Oncol, 2021, 11: 737760. doi: 10.3389/fonc.2021.737760 [5] Flores AR, Rêma A, Mesquita JR, et al. Vimentin and Ki-67 immunolabeling in canine gastric carcinomas and their prognostic value[J]. Vet Pathol, 2022, 59(6): 903-14. doi: 10.1177/03009858221117858 [6] Gown AM. The biomarker ki-67: promise, potential, and problems in breast cancer[J]. Appl Immunohistochem Mol Morphol, 2023, 31 (7): 478-84. doi: 10.1097/PAI.0000000000001087 [7] Kadado KJ, Abernathy OL, Salyers WJ, et al. Gastrointestinal stromal tumor and ki-67 as a prognostic indicator[J]. Cureus, 2022, 14(1): e20868. [8] Ma B, Huang XT, Zou GJ, et al. Relationship between Ki-67 and CD44 expression and microvascular formation in gastric stromal tumor tissues[J]. World J Clin Cases, 2022, 10(2): 469-76. doi: 10.12998/wjcc.v10.i2.469 [9] 陈泓臻, 邱惠芳. 胃肠道间质瘤患者Ki-67和钙激活通道蛋白1表达与预后的关系[J]. 浙江医学, 2022, 44(23): 2526-30. doi: 10.12056/j.issn.1006-2785.2022.44.23.2022-1785 [10] 陈长波, 马俊, 兰小惠. 胃肠道间质瘤CT诊断及与病理肿瘤恶性潜能相关性分析[J]. 中国CT和MRI杂志, 2022, 20(6): 156-8. doi: 10.3969/j.issn.1672-5131.2022.06.051 [11] 郭建榕, 蔡美花, 吴立萍. 老年胃肠道间质瘤患者临床诊断中多层螺旋CT应用的准确性[J]. 中国老年学杂志, 2021, 41(19): 4206-9. doi: 10.3969/j.issn.1005-9202.2021.19.015 [12] 陈馨, 陈晓, 刘永康, 等. DCE-CT鉴别诊断胃异位胰腺和间质瘤的价值[J]. 放射学实践, 2020, 35(8): 1048-53. [13] 张于凤, 李辉, 李晶晶, 等. 能谱CT多模态参数联合临床参数对胃间质瘤Ki-67的预测价值[J]. 分子影像学杂志, 2022, 45(5): 688-692. doi: 10.12122/j.issn.1674-4500.2022.05.11 [14] 王学东, 刘爱连, 田士峰, 等. 单源双能CT能谱分析定量参数预测胃间质瘤Ki-67表达的可行性研究[J]. 放射学实践, 2020, 35(7): 893-9. [15] 原士超, 赵庆, 范红燕. 双能CT对胃间质瘤中Ki-67表达的预测价值及其与病理参数的关系[J]. 临床荟萃, 2021, 36(11): 1013-8. doi: 10.3969/j.issn.1004-583X.2021.11.011 [16] Liu Y, He CY, Fang WD, et al. Prediction of Ki-67 expression in gastrointestinal stromal tumors using radiomics of plain and multiphase contrast-enhanced CT[J]. Eur Radiol, 2023, 33(11): 7609-17. doi: 10.1007/s00330-023-09727-5 [17] Zhang QW, Gao YJ, Zhang RY, et al. Personalized CT-based radiomics nomogram preoperative predicting Ki-67 expression in gastrointestinal stromal tumors: a multicenter development and validation cohort[J]. Clin Transl Med, 2020, 9(1): 12. doi: 10.1186/s40169-020-0263-4 [18] 刘芸, 李传明, 曾平, 等. 基于动脉期增强CT影像组学模型鉴别小肠间质瘤组织中Ki-67表达状态的应用[J]. 胃肠病学和肝病学杂志, 2023, 32(4): 405-8. doi: 10.3969/j.issn.1006-5709.2023.04.009 [19] 李根, 刘琨, 于海韵, 等. 基于增强CT的深度学习模型预测胃肠道间质瘤Ki-67表达的双中心研究[J]. 国际医学放射学杂志, 2024, 47 (2): 172-7. [20] Cai W, Guo K, Chen YX, et al. Sub-regional CT radiomics for the prediction of ki-67 proliferation index in gastrointestinal stromal tumors: a multi-center study[J]. Acad Radiol, 2024: S1076-S6332 (24)00421-5. [21] Zhao YL, Feng MB, Wang MH, et al. CT radiomics for the preoperative prediction of Ki67 index in gastrointestinal stromal tumors: a multi-center study[J]. Front Oncol, 2021, 11: 689136. doi: 10.3389/fonc.2021.689136 [22] Xie ZH, Suo ST, Zhang W, et al. Prediction of high Ki-67 proliferation index of gastrointestinal stromal tumors based on CT at non-contrast-enhanced and different contrast-enhanced phases [J]. Eur Radiol, 2024, 34(4): 2223-32. [23] 李晶晶, 张梦琪, 刘焱. 能谱CT影像组学及CT征象对GIST危险度分级探讨[J]. 中国CT和MRI杂志, 2024, 22(7): 156-9. [24] Liang YM, Li XH, Li WM, et al. Prognostic significance of PTEN, Ki-67 and CD44s expression patterns in gastrointestinal stromal tumors[J]. World J Gastroenterol, 2012, 18(14): 1664-71. doi: 10.3748/wjg.v18.i14.1664 [25] Zhang P, Tao KX. Pay attention to misdiagnosis and differential diagnosis of gastric gastrointestinal stromal tumor[J]. Chin J Gastrointest Surg, 2021, 24(9): 758-61. [26] Shibata T, Koyama K, Yamada H, et al. The association between the growth pattern and malignancy in the formation of ulceration in gastric gastrointestinal stromal tumor[J]. Intern Med, 2023, 62(17): 2451-5. doi: 10.2169/internalmedicine.0983-22 [27] 杨采薇, 刘曦娇, 魏毅, 等. CT影像组学联合传统影像特征对胃肠道间质瘤Ki-67指数表达的预测价值[J]. 放射学实践, 2022, 37(9): 1068-73. [28] 李卫卫. CT影像组学在预测胃间质瘤KIT基因突变及Ki-67表达中的应用研究[D]. 郑州: 郑州大学, 2021. [29] Rizzo S, Botta F, Raimondi S, et al. Radiomics: the facts and the challenges of image analysis[J]. Eur Radiol Exp, 2018, 2(1): 36. doi: 10.1186/s41747-018-0068-z [30] 贾济波, 刘原庆, 冯飞文, 等. CT影像组学列线图模型预测胃肠道间质瘤Ki-67表达的临床研究[J]. 临床放射学杂志, 2023, 42(5): 783-8. -
点击查看大图
图(3) / 表(3)
计量
- 文章访问数: 46
- HTML全文浏览量: 17
- PDF下载量: 6
- 被引次数: 0
