-
肺癌是全球癌症相关死亡的主要原因[1],非小细胞肺癌(nonsmall-cell lung cancer, NSCLC)最常见的基因突变是表皮生长因子受体(epidermal growth factor receptor,EGFR)突变[2]。与野生型EGFR病人相比,酪氨酸激酶抑制剂(tyrosine kinase inhibitors, TKIs)对EGFR突变病人有着更高的应答率[3],能够提高病人的无进展生存期[4],改善生活质量,因此TKIs被美国国家综合癌症网络(national comprehensive cancer network, NCCN)推荐为NSCLC的一线治疗方法[5]。活检组织病理学标本最常用于检测EGFR的突变状态,但是由于肿瘤的异质性、取样的偏差、耗时的程序、有创并且增加了癌症转移的风险[6],一定程度上限制了其在EGFR突变检测中的应用。影像组学是从影像图像中高通量地提取大量影像信息,将视觉影像信息转化为深层次的定量影像特征来进行量化研究,不仅有效地解决了肿瘤异质性难以定量评估的问题,而且无创,可以重复进行[7]。影像组学在肺部病变的定性、肺癌的分级与分期、肺癌的疗效评估和预后预测等方面已有较为广泛的研究[8-9],本文旨在探究影像组学联合临床特征在预测肺腺癌EGFR突变状态中的价值。
-
无吸烟史、毛刺征、GGO以及胸膜牵拉与肺腺癌EGFR突变状态显著相关(P<0.05)(见表 1)。
项目 训练组 验证组 EGFR(-)(n=25) EGFR(+)(n=49) χ2 P EGFR(-)(n=23) EGFR(+)(n=28) χ2 P 年龄/岁 63.24±12.89 60.47±8.27 0.98* >0.05 63.21±10.31 62.75±9.39 0.17* >0.05 性别 男
女12 (48.0)
13 (52.0)18 (36.7)
31 (63.3)0.87 >0.05 15 (65.2)
8 (34.8)6 (21.4)
22 (78.6)10.00# <0.01 吸烟史 有
无5 (20.0)
20 (80.0)23 (46.9)
26 (53.1)5.11 <0.05 5 (21.7)
18 (78.3)15 (53.6)
13 (46.4)5.36 <0.05 大小/cm 2.80(1.35-5.40) 3.00(2.10-4.00) 0.532△ >0.05 3.27±1.85 3.34±1.67 0.14* >0.05 形态 规则
不规则16 (64.0)
9 (36.0)26 (53.1)
23 (46.9)0.81 >0.05 7 (30.4)
16 (69.6)14 (50.0)
14 (50.0)2.00 >0.05 边界 清晰
模糊24 (96.0)
1 (4.0)42 (85.7)
7 (14.3)0.91# >0.05 4 (17.4)
19 (82.6)1(3.6)
27 (96.4)1.39 >0.05 分叶征 有
无18 (72.0)
7 (28.0)41 (83.7)
8 (16.3)1.40 >0.05 18(78.3)
5 (21.7)23 (82.1)
5 (17.9)0.01 >0.05 毛刺征 有
无7 (28.0)
18 (72.0)26 (53.1)
23 (46.9)4.21 <0.05 7 (30.4)
16(69.6)17 (60.7)
11 (39.3)4.65 <0.05 病灶密度 磨玻璃
实性13 (52.0)
12 (48.0)12 (24.5)
37 (75.5)5.60 <0.05 14 (60.9)
9 (39.1)7 (25.0)
21 (75.0)6.71 <0.01 晕征 有
无5 (20.0)
20 (80.0)14 (28.6)
35 (71.4)0.64 >0.05 10 (43.5)
13 (56.5)8 (28.6)
20 (71.4)1.23 >0.05 钙化 有
无3 (12.0)
22 (88.0)12 (24.5)
37 (75.5)1.60 >0.05 2 (8.7)
21 (91.3)5 (17.9)
23 (82.1)0.29# >0.05 瘤内坏死 有
无4 (16.0)
21 (84.0)15 (30.6)
34 (69.4)1.17 >0.05 6 (26.1)
17 (73.9)6 (21.4)
22 (78.6)0.15 >0.05 空泡征 有
无6 (24.0)
19 (76.0)12 (24.5)
37 (75.5)0.00 >0.05 4 (17.4)
19 (82.6)5 (17.9)
23 (82.1)0.11# >0.05 空洞征 有
无1 (4.0)
24 (96.0)1 (2.0)
48 (98.0)— 1.00▼ 1 (4.3)
22 (95.7)3 (10.7)
25 (89.3)0.10# >0.05 空气支气管征 有
无5(20.0)
20 (80.0)4 (8.2)
45 (91.8)1.20# >0.05 4(17.4)
19 (82.6)5 (17.9)
23 (82.1)0.11# >0.05 周围性肺气肿 有
无1 (4.0)
24(96.0)2(4.1)
47(95.9)0.37# >0.05 1(4.3)
22(95.7)3(10.7)
25(89.3)0.10# >0.05 外周纤维化 有
无5(20.0)
20(80.0)19 (38.8)
30(61.2)2.66 >0.05 4 (17.4)
19(82.6)7(25.0)
21(75.0)0.10# >0.05 胸膜牵拉 有
无10(40.0)
15(60.0)33(67.3)
16(32.7)5.09 <0.05 8(34.8)
15(65.2)22(78.6)
6(21.4)10.00 <0.01 胸膜接触 有
无11(44.0)
14(56.0)23(46.9)
26(53.1)0.06 >0.05 13(56.5)
10(43.5)10(35.7)
18(64.3)2.21 >0.05 胸膜增厚 有
无11(44.0)
14(56.0)20(40.8)
29(59.2)0.07 >0.05 7(30.4)
16(69.6)12(42.9)
16(57.1)0.83 >0.05 胸腔积液 有
无4(16.0)
21(84.0)11(22.4)
38(77.6)0.43 >0.05 5(21.7)
18(78.3)7(25.0)
21(75.0)0.07 >0.05 肺门/纵隔淋巴肿大 有
无10(40.0)
15(60.0)13(26.5)
36(73.5)1.40 >0.05 8(34.8)
15(65.2)11(39.3)
17(60.7)0.11 >0.05 *示t值;△示Z值;#示矫正χ2值;▼示确切概率法 表 1 肺腺癌病人的临床因素与CT征象统计[n; 构成比(%)]
-
共提取936个影像组学特征,经降维最终选取了6个与EGFR突变显著相关的影像组学特征,将其按照权重系数由高到低排列(见图 3)。
-
临床模型由吸烟史、毛刺征、GGO及胸膜牵拉构成;影像组学模型由提取的6个影像组学特征构成;联合模型由吸烟史、毛刺征、磨玻璃密度、胸膜牵拉以及6个影像组学特征构成。联合模型在训练组和验证组中的AUC均高于影像组学模型和临床模型(见表 2),联合模型比影像组学模型和临床模型具有更高的预测效能(见图 4)。
模型 AUC(95%CI) 训练组 验证组 临床模型 0.749(0.653~0.843) 0.753(0.612~0.863) 影像组学模型 0.818(0.711~0.898) 0.797(0.661~0.896) 联合模型 0.860(0.760~0.930) 0.855(0.728~0.938) 表 2 3种模型在训练组和验证组中的AUC
CT影像组学联合临床特征在预测肺腺癌EGFR突变中的价值
Value of the CT radiomics combined with clinical features in the prediction of EGFR mutation in lung adenocarcinoma
-
摘要:
目的探究CT影像组学联合临床特征对肺腺癌EGFR突变状态的预测效能。 方法对125例肺腺癌病人进行回顾性研究,分成训练组(n=74)与验证组(n=51),基于CT成像提取影像组学特征;采用支持向量机(SVM)分类器,分别构建临床模型、影像组学模型以及联合模型;受试者工作特征曲线(ROC)及曲线下面积(AUC)用于评价模型的预测效能。 结果临床模型、影像组学模型以及联合模型在训练组中的AUC分别为0.749(0.653~0.843)、0.818(0.711~0.898)、0.860(0.760~0.930),在验证组中的AUC分别为0.753(0.612~0.863)、0.797(0.661~0.896)、0.855(0.728~0.938)。 结论对于肺腺癌EGFR突变状态的预测,CT影像组学特征优于临床因素与CT征象,当影像组学结合临床因素与CT征象,能进一步提高预测效能。 -
关键词:
- 肺腺癌 /
- 表皮生长因子受体突变 /
- 影像组学
Abstract:ObjectiveTo explore the predictive efficacy of CT radiomics combined with clinical features in predicting EGFR mutation in lung adenocarcinoma. MethodsThe clinical data of 125 patients with lung adenocarcinoma were retrospectively analyzed, the patients were divided into the training group(n=74) and verification group(n=51).The radiomics features were extracted based on CT radiomics.The support vector machine(SVM) classifier was used to construct the clinical model, radiomics model and joint model, respectively.The receiver operating characteristic curve(ROC) and area under the curve(AUC) were used to evaluate the predictive efficacy of model. ResultsThe AUC of clinical model, radiomics model and joint model in training group were 0.749(0.653-0.843), 0.818(0.711-0.898) and 0.860(0.760-0.930), respectively.The AUC of clinical model, radiomics model and joint model in verification group were 0.753(0.612-0.863), 0.797(0.661-0.896) and 0.855(0.728-0.938), respectively. ConclusionsFor the prediction of EGFR mutation status in lung adenocarcinoma, the CT radiomics features are superior to clinical factors and CT signs.The radiomics combined with clinical factors and CT signs can further improve the prediction efficiency. -
表 1 肺腺癌病人的临床因素与CT征象统计[n; 构成比(%)]
项目 训练组 验证组 EGFR(-)(n=25) EGFR(+)(n=49) χ2 P EGFR(-)(n=23) EGFR(+)(n=28) χ2 P 年龄/岁 63.24±12.89 60.47±8.27 0.98* >0.05 63.21±10.31 62.75±9.39 0.17* >0.05 性别 男
女12 (48.0)
13 (52.0)18 (36.7)
31 (63.3)0.87 >0.05 15 (65.2)
8 (34.8)6 (21.4)
22 (78.6)10.00# <0.01 吸烟史 有
无5 (20.0)
20 (80.0)23 (46.9)
26 (53.1)5.11 <0.05 5 (21.7)
18 (78.3)15 (53.6)
13 (46.4)5.36 <0.05 大小/cm 2.80(1.35-5.40) 3.00(2.10-4.00) 0.532△ >0.05 3.27±1.85 3.34±1.67 0.14* >0.05 形态 规则
不规则16 (64.0)
9 (36.0)26 (53.1)
23 (46.9)0.81 >0.05 7 (30.4)
16 (69.6)14 (50.0)
14 (50.0)2.00 >0.05 边界 清晰
模糊24 (96.0)
1 (4.0)42 (85.7)
7 (14.3)0.91# >0.05 4 (17.4)
19 (82.6)1(3.6)
27 (96.4)1.39 >0.05 分叶征 有
无18 (72.0)
7 (28.0)41 (83.7)
8 (16.3)1.40 >0.05 18(78.3)
5 (21.7)23 (82.1)
5 (17.9)0.01 >0.05 毛刺征 有
无7 (28.0)
18 (72.0)26 (53.1)
23 (46.9)4.21 <0.05 7 (30.4)
16(69.6)17 (60.7)
11 (39.3)4.65 <0.05 病灶密度 磨玻璃
实性13 (52.0)
12 (48.0)12 (24.5)
37 (75.5)5.60 <0.05 14 (60.9)
9 (39.1)7 (25.0)
21 (75.0)6.71 <0.01 晕征 有
无5 (20.0)
20 (80.0)14 (28.6)
35 (71.4)0.64 >0.05 10 (43.5)
13 (56.5)8 (28.6)
20 (71.4)1.23 >0.05 钙化 有
无3 (12.0)
22 (88.0)12 (24.5)
37 (75.5)1.60 >0.05 2 (8.7)
21 (91.3)5 (17.9)
23 (82.1)0.29# >0.05 瘤内坏死 有
无4 (16.0)
21 (84.0)15 (30.6)
34 (69.4)1.17 >0.05 6 (26.1)
17 (73.9)6 (21.4)
22 (78.6)0.15 >0.05 空泡征 有
无6 (24.0)
19 (76.0)12 (24.5)
37 (75.5)0.00 >0.05 4 (17.4)
19 (82.6)5 (17.9)
23 (82.1)0.11# >0.05 空洞征 有
无1 (4.0)
24 (96.0)1 (2.0)
48 (98.0)— 1.00▼ 1 (4.3)
22 (95.7)3 (10.7)
25 (89.3)0.10# >0.05 空气支气管征 有
无5(20.0)
20 (80.0)4 (8.2)
45 (91.8)1.20# >0.05 4(17.4)
19 (82.6)5 (17.9)
23 (82.1)0.11# >0.05 周围性肺气肿 有
无1 (4.0)
24(96.0)2(4.1)
47(95.9)0.37# >0.05 1(4.3)
22(95.7)3(10.7)
25(89.3)0.10# >0.05 外周纤维化 有
无5(20.0)
20(80.0)19 (38.8)
30(61.2)2.66 >0.05 4 (17.4)
19(82.6)7(25.0)
21(75.0)0.10# >0.05 胸膜牵拉 有
无10(40.0)
15(60.0)33(67.3)
16(32.7)5.09 <0.05 8(34.8)
15(65.2)22(78.6)
6(21.4)10.00 <0.01 胸膜接触 有
无11(44.0)
14(56.0)23(46.9)
26(53.1)0.06 >0.05 13(56.5)
10(43.5)10(35.7)
18(64.3)2.21 >0.05 胸膜增厚 有
无11(44.0)
14(56.0)20(40.8)
29(59.2)0.07 >0.05 7(30.4)
16(69.6)12(42.9)
16(57.1)0.83 >0.05 胸腔积液 有
无4(16.0)
21(84.0)11(22.4)
38(77.6)0.43 >0.05 5(21.7)
18(78.3)7(25.0)
21(75.0)0.07 >0.05 肺门/纵隔淋巴肿大 有
无10(40.0)
15(60.0)13(26.5)
36(73.5)1.40 >0.05 8(34.8)
15(65.2)11(39.3)
17(60.7)0.11 >0.05 *示t值;△示Z值;#示矫正χ2值;▼示确切概率法 表 2 3种模型在训练组和验证组中的AUC
模型 AUC(95%CI) 训练组 验证组 临床模型 0.749(0.653~0.843) 0.753(0.612~0.863) 影像组学模型 0.818(0.711~0.898) 0.797(0.661~0.896) 联合模型 0.860(0.760~0.930) 0.855(0.728~0.938) -
[1] SIEGELl RL, MILLER KD, JEMAL A. Cancer statistics, 2018[J]. CA Cancer J Clin, 2018, 68(1): 7. doi: 10.3322/caac.21442 [2] JIA Y, YUN CH, PARK E, et al. Overcoming EGFR(T790M) and EGFR(C797S) resistance with mutant-selective allosteric inhibitors[J]. Nature, 2016, 534(7605): 129. doi: 10.1038/nature17960 [3] ETTINGER DS, WOOD DE, AISNER DL, et al. Non-small cell lung cancer, version 5.2017, NCCN clinical guidelines in Oncology[J]. J Natl Compr Canc Netw, 2017, 15(4): 504. doi: 10.6004/jnccn.2017.0050 [4] RIELY GJ, PAO W, PHAM D, et al. Clinical course of patients with non-small cell lung cancer and epidermal growth factor receptor exon 19 and exon 21 mutations treated with gefitinib or erlotinib[J]. Clin Cancer Res, 2006, 12(3): 839. doi: 10.1158/1078-0432.CCR-05-1846 [5] SORIA JC, OHE Y, VANSTEENKISTE J, et al. Osimertinib in untreated EGFR-Mutated advanced non-small-cell lung cancer[J]. N Engl J Med, 2018, 378(2): 113. doi: 10.1056/NEJMoa1713137 [6] LOUGHRAN CF, KEELING CR. Seeding of tumour cells following breast biopsy: a literature review[J]. Br J Radiol, 2011, 84(1006): 869. doi: 10.1259/bjr/77245199 [7] LEE G, LEE HY, PARK H, et al. Radiomics and its emerging role in lung cancer research, imaging biomarkers and clinical management: State of the art[J]. Eur J Radiol, 2017, 86(16): 297. [8] 张静, 武志峰, 鄂林宁. 肺结节(≤ 2 cm)及其周围组织的影像组学特征在其良恶性鉴别中的价值[J]. 中国临床医学影像杂志, 2020, 31(7): 478. [9] 单文莉, 柏根基, 孔丹, 等. 基于常规CT图像应用影像组学模型预测浸润性肺腺癌不同分化程度的价值[J]. 临床放射学杂志, 2020, 39(6): 1095. [10] ZHANG L, CHEN B, LIU X, et al. Quantitative biomarkers for prediction of epidermal growth factor receptor mutation in non-small cell lung cancer[J]. Transl Oncol, 2018, 11(1): 94. doi: 10.1016/j.tranon.2017.10.012 [11] LIU G, XU Z, GE Y, et al. 3D radiomics predicts EGFR mutation, exon-19 deletion and exon-21 L858R mutation in lung adenocarcinoma[J]. Transl Lung Cancer Res, 2020, 9(4): 1212. doi: 10.21037/tlcr-20-122 [12] DIGUMARTHY SR, PADOLE AM, GULLO RL, et al. Can CT radiomic analysis in NSCLC predict histology and EGFR mutation status?[J]. Medicine, 2019, 98(1): e13963. doi: 10.1097/MD.0000000000013963 [13] GOH V, GANESHAN B, NATHAN P, et al. Assessment of response to tyrosine kinase inhibitors in metastatic renal cell cancer: CT texture as a predictive biomarker[J]. Radiology, 2011, 261(1): 165. doi: 10.1148/radiol.11110264 [14] LIU Q, SUN D, LI N, et al. Predicting EGFR mutation subtypes in lung adenocarcinoma using 18F-FDG PET/CT radiomic features[J]. Transl Lung Cancer Res, 2020, 9(3): 549. doi: 10.21037/tlcr.2020.04.17 [15] MEI D, LUO Y, WANG Y, et al. CT texture analysis of lung adenocarcinoma: can Radiomic features be surrogate biomarkers for EGFR mutation statuses[J]. Cancer Imaging, 2018, 18(1): 52. doi: 10.1186/s40644-018-0184-2 [16] RIZZO S, RAIMONDI S, DE JONG EEC, et al. Genomics of non-small cell lung cancer(NSCLC): association between CT-based imaging features and EGFR and K-RAS mutations in 122 patients-An external validation[J]. Eur J Radiol, 2019, 110(1): 148. [17] LIU Y, KIM J, BALAGURUNATHAN Y, et al. Radiomic features are associated with EGFR mutation status in lung adenocarcinomas[J]. Clin Lung Cancer, 2016, 17(5): 441. doi: 10.1016/j.cllc.2016.02.001 [18] DANG Y, WANG R, QIAN K, et al. Clinical and radiological predictors of epidermal growth factor receptor mutation in nonsmall cell lung cancer[J]. J Appl Clin Med Phys, 2021, 22(1): 271. doi: 10.1002/acm2.13107 [19] WANG H, GUO H, WANG Z, et al. The diagnostic value of quantitative CT analysis of ground-glass volume percentage in differentiating epidermal growth factor receptor mutation and subtypes in lung adenocarcinoma[J]. Biomed Res Int, 20192019: 1. [20] LEE HJ, KIM YT, KANG CH, et al. Epidermal growth factor receptor mutation in lung adenocarcinomas: relationship with CT characteristics and histologic subtypes[J]. Radiology, 2013, 268(1): 254. doi: 10.1148/radiol.13112553 [21] SUGANO M, SHUMIZU K, NAKANO T, et al. Correlation between computed tomography findings and epidermal growth factor receptor and KRAS gene mutations in patients with pulmonary adenocarcinoma[J]. Oncol Rep, 2011, 26(5): 1205.