基于影像组学的煤工尘肺结节分析联合临床因素预测肺癌发生的价值

孟珊; 惠东明; 王坤; 李志超

doi:10.16766/j.cnki.issn.1674-4152.002800

基于影像组学的煤工尘肺结节分析联合临床因素预测肺癌发生的价值

doi: 10.16766/j.cnki.issn.1674-4152.002800

孟珊¹,
惠东明¹,
王坤²,
李志超^2, ,

1.
重庆西区医院放射科, 重庆 400050
2.
重庆市九龙坡区第二人民医院放射科, 重庆 400050

基金项目:

重庆市科卫联合医学科研项目 2022MSXM140

详细信息

通讯作者:
李志超，E-mail：31956839@qq.com

中图分类号: R814.42 R135.2
计量
- 文章访问数: 420
- HTML全文浏览量: 185
- PDF下载量: 26
- 被引次数: 0
出版历程
- 收稿日期: 2022-10-21
- 网络出版日期: 2023-04-07

Value of analysis of coal workers' pneumoconiosis nodules based on radiomics combined with clinical factors in predicting the occurrence of lung cancer

1.
Department of Radiology, Chongqing Western Hospital, Chongqing 400050, China

摘要

摘要: 目的研究影像组学、临床因素与煤工尘肺结节恶变间的关系，构建煤工尘肺结节恶变的最佳预测模型。方法收集2015年1月—2019年6月在重庆市九龙坡区第二人民医院就诊的425例煤工尘肺患者的临床资料及共628个结节CT图像，以7 : 3的比例随机分成训练集和验证集，每组数据内分为恶变组和未恶变组。对结节进行基本影像特征判读并勾画感兴趣区，提取影像组学特征。利用影像组学关键特征构建Radscore公式, 通过logistic回归对临床特征、影像组学特征、临床特征和Radscore建立预测模型并使用AUC及Delong检验比较模型的预测效能。结果 3年随访过程中病理证实恶性结节54枚，直径为(1.70±0.63)cm，未发生恶变结节574枚，直径为(1.68±0.76)cm，2组直径比较差异无统计学意义(t=0.468，P=0.642)；2组接触煤尘工龄、家族史、吸烟史、饮酒史、毛刺征与空泡征比较，差异有统计学意义(均P＜0.05)；年龄、结核史、COPD病史比较，差异无统计学意义(均P＞0.05)。LASSO筛选出11个影像组学特征。在验证集中混合模型取得最佳效果，AUC为0.895, 影像组学模型和临床模型的AUC分别为0.671、0.654。Delong检验显示模型差异有统计学意义(P＜0.05)。结论接触煤尘工龄、家族史、吸烟史、饮酒史、毛刺征、空泡征及影像组学特征与煤工尘肺结节恶变存在着一定的相关性，可用来构建预测模型。
- 煤工尘肺 /
- 肺结节 /
- 影像组学 /
- 肺癌
Abstract: Objective To investigate the relationship between radiomics, clinical factors and malignant transformation of coal workers' pneumoconiosis nodules, establishing the best prediction model for malignant transformation of coal workers' pneumoconiosis nodules. Methods The clinical data of 425 cases of coal workers' pneumoconiosis treated in the Second People ' s Hospital of Jiulongpo district from January 2015 to June 2019 and CT imaging data of 628 pneumoconiosis nodules were collected. They were randomly divided into training set and verification cohorts at a ratio of 7 : 3. Each group of data was divided into malignant group and non-malignant group. The basic image features of the nodule were interpreted and the region of interest (ROI) was delineated. The Radscore calculation formula was constructed using the key features of radiomics. The predictive models were established base on clinical features, radiomics features, and clinical features combined with Radscore through logistic regression, and AUC and Delong were used to compare the prediction efficiency of models. Results During the 3-year follow-up, 54 malignant nodules were confirmed by pathology, with a diameter of (1.70±0.63) cm, 574 nodules without malignant change, with a diameter of (1.68±0.76) cm. There was no significant difference in the diameter between the two groups (t=0.468, P=0.642). Statistically significant differences were existed between the two groups in terms of length of service exposed to coal dust, family history, smoking history, drinking history and image features of spicule sign and air sign (all P < 0.05). There was no significant difference in age, tuberculosis history and chronic obstructive pulmonary disease (COPD) history (all P>0.05). LASSO screened 11 imaging features. The mixed model had the best effect in the validation set, with an AUC of 0.895. AUC of radiomics model and clinical model were 0.671 and 0.654 respectively. Delong test showed that the model difference was statistically significant (P < 0.05). Conclusion There is a certain correlation between the length of service exposed to coal dust, family history, smoking history, drinking history, burr sign, air sign, radiomics features and the malignant change of coal worker pneumoconiosis nodule, which can be used to build a model to predict the malignant change of coal worker pneumoconiosis nodule.
- Coal workers' pneumoconiosis /
- Pulmonary nodule /
- Radiomics /
- Lung cancer

HTML全文

图 1 LASSO的参数曲线路线图及提取的特征

注：A为参数曲线路线图。B为提取的特征图，其中a为original_shape_MinorAxisLength；b为original_shape_SurfaceVolumeRatio；c为original_firstorder_RootMeanSquared；d为original_glcm_Id；e为original_glrlm_ShortRunEmphasis；f为original_ngtdm_Strength；g为log-sigma-1-0-mm-3D_firstorder_10Percentile；h为log-sigma-1-0-mm-3D_glszm_GrayLevelNonUniformity；i为log-sigma-1-0-mm-3D_glszm_LargeAreaHighGrayLevelEmphasis；j为log-sigma-1-0-mm-3D_glszm_LowGrayLevelZoneEmphasis；k为log-sigma-1-0-mm-3D_gldm_LargeDependenceEmphasis。

Figure 1. Parameter curve roadmap and extracted features of LASSO

下载: 全尺寸图片幻灯片

图 2 3种模型的ROC曲线图

注：A示临床模型在训练组和验证组中的ROC曲线；B示影像组学模型在训练组和验证组中的ROC曲线；C示混合模型在训练组和验证组中的ROC曲线；D示3个模型在验证组中的ROC曲线。

Figure 2. ROC Curve of three models

下载: 全尺寸图片幻灯片

表 1 煤工尘肺患者恶变结节与未恶变结节之间的临床因素比较

Table 1. Analysis of clinical factors between malignant and non-malignant nodules in coal workers' pneumoconiosis patients

项目	恶变结节(n=54)	未恶变结节(n=574)	统计量	P值
年龄(x±s，岁)	57.6±12.4	58.3±10.0	0.468^a	0.640
结核(例)			0.277^b	0.599
伴结核	26	255
不伴结核	28	319
COPD病史(例)			0.151^b	0.698
伴COPD	29	324
不伴COPD	25	250
接触煤尘工龄(x±s，年)	16.76±3.86	7.86±2.34	24.941^a	＜0.001
尘肺分期(例)			6.455^c	0.040
Ⅰ期	8	168
Ⅱ期	34	268
Ⅲ期	12	138
结节直径(x±s，cm)	1.70±0.63	1.68±0.76	0.154^a	0.877
家族史(例)			6.777^b	0.009
有	32	235
无	22	339
吸烟史(例)			9.451^b	0.002
有	34	237
无	20	337
饮酒史(例)			7.646^b	0.006
有	34	249
无	20	325
毛刺征(例)			7.373^b	0.007
阳性	30	211
阴性	24	363
空泡征(例)			19.181^b	0.001
阳性	34	190
阴性	20	384
胸膜凹陷征(例)			0.729^b	0.393
阳性	24	290
阴性	30	284
空气充气征(例)			1.984^b	0.159
阳性	12	180
阴性	42	392
钙化(例)			2.890^b	0.089
阳性	44	405
阴性	10	169
平均CT值(x±s，Hu)	46.61±7.38	45.10±6.76	1.553^a	0.121
注：^a为t值，^b为χ²值，^c为H值。

下载: 导出CSV

表 2 CWP患者发生肺癌的多因素logistic回归分析

Table 2. Multivariate logistic regression analysis of lung cancer in patients with CWP

变量	B	SE	Wald χ²	P值	OR值	95% CI
家族史	0.644	0.219	8.674	0.003	1.905	1.241-2.925
吸烟史	1.011	0.228	19.717	＜0.001	2.747	1.759-4.292
接触煤尘工龄	1.029	0.048	466.016	＜0.001	2.798	2.548-3.072
注：各变量赋值情况，发生恶变肺结节有=1，无=0；家族史，有=1，无=0；吸烟史，有=1，无=0；接触煤尘年龄以实际值赋值。

下载: 导出CSV

参考文献(22)

[1]	RONSMANS S, NEMERY B. Pneumoconiosis in coal miners: anthracosilicosis after all[J]. Ann Am Thorac Soc, 2022, 19(9): 1451-1452. doi: 10.1513/AnnalsATS.202206-528ED
[2]	SONG Y, SOUTHAM K, BEAMISH B B, et al. Effects of chemical composition on the lung cell response to coal particles: implications for coal workers' pneumoconiosis[J]. Respirology, 2022, 27(6): 447-454. doi: 10.1111/resp.14246
[3]	张建红, 化静. 煤工尘肺并发肺癌的临床分析[J]. 中国继续医学教育, 2021, 13(18): 130-133. https://www.cnki.com.cn/Article/CJFDTOTAL-JXUY202118036.htm ZHANG J H, HUA J. Clinical Analysis of Coal workers' Pneumoconiosis Complicated With Lung Cancer[J]. China Continuing Medical Education, 2021, 13(18): 130-133. https://www.cnki.com.cn/Article/CJFDTOTAL-JXUY202118036.htm
[4]	YANG X, DONG X, WANG J, et al. Computed tomography-based radiomics signature: a potential indicator of epidermal growth factor receptor mutation in pulmonary adenocarcinoma appearing as a subsolid nodule[J]. Oncologist, 2019, 24(11): e1156-e1164. doi: 10.1634/theoncologist.2018-0706
[5]	YANG R, HUI D M, LI X, et al. Prediction of single pulmonary nodule growth by CT radiomics and clinical features-a one-year follow-up study[J]. Front Oncol, 2022, 12: 1034817. DOI: 10.3389/fonc.2022.1034817.
[6]	梁如意, 董超前, 袁亮, 等. 煤矿粉尘对工人健康损害的流行病学研究进展[J]. 中华劳动卫生职业病杂志, 2022, 40(6): 476-480. LIANG R Y, DONG C Q, YUAN L, et al. Progress in the epidemiological studies on coal mine dust exposure with workers' health damage[J]. Chinese journal of industrial hygiene and occupational diseases, 2022, 40(6): 476-480.
[7]	ALMBERG K S, FRIEDMAN L S, R0SE C S, et al. Progression of coal workers' pneumoconiosis absent further exposure[J]. Occup Environ Med, 2020, 77(11): 748-751. doi: 10.1136/oemed-2020-106466
[8]	梁伟, 聂继盛. 同煤总医院煤工尘肺并发肺癌病例的危险因素、临床表现和CT征象[J]. 职业与健康, 2020, 36(17): 2321-2325. https://www.cnki.com.cn/Article/CJFDTOTAL-ZYJK202017007.htm LIANG W, NIE J S. Risk factors, clinical features, and CT characteristics of coal workers' pneumoconiosis complicated with lung cancer in Tongmei General Hospital[J]. Occupation and Health, 2020, 36(17): 2321-2325. https://www.cnki.com.cn/Article/CJFDTOTAL-ZYJK202017007.htm
[9]	GO L H T, COHEN R A. Coal workers' pneumoconiosis and other mining-related lung disease: new manifestations of illness in an age-old occupation[J]. Clin Chest Med, 2020, 41(4): 687-696. doi: 10.1016/j.ccm.2020.08.002
[10]	ALIF S M, SIM M R, HO C, et al. Cancer and mortality in coal mine workers: a systematic review and meta-analysis[J]. Occup Environ Med, 2022, 79(5): 347-357. doi: 10.1136/oemed-2021-107498
[11]	ZWANENBURG A, VALLIERES M, Abdalah M A, et al. The image biomarker standardization initiative: standardized quantitative radiomics for high-throughput image-based phenotyping[J]. Radiology, 2020, 295(2): 328-338. doi: 10.1148/radiol.2020191145
[12]	王海波, 崔薇, 杨玮丽. 基于多序列MRI影像组学预测早期宫颈癌淋巴血管侵犯的研究[J]. 中华全科医学, 2021, 19(12): 2088-2092. doi: 10.16766/j.cnki.issn.1674-4152.002244 WANG H B, WEI W, YANG W L. Multi-sequence MRI-based radiomics predicting lymph-vascular space invasion in early-stage cervical cancer[J]. Chinese Journal of General Practice, 2021, 19(12): 2088-2092. doi: 10.16766/j.cnki.issn.1674-4152.002244
[13]	YANG L, GU D S, WEI J W, et al. A radiomics nomogram for preoperative prediction of microvascular invasion in hepatocellular carcinoma[J]. Liver Cancer, 2019, 8(5): 373-386. doi: 10.1159/000494099
[14]	SCHNIERING J, MACIUKIEWICZ M, GABRYS H S, et al. Computed tomography-based radiomics decodes prognostic and molecular differences in interstitial lung disease related to systemic sclerosis[J]. Eur Respir J, 2022, 59(5): 2004503. DOI: 10.1183/13993003.04503-2020.
[15]	AUJAY G, ETCHEGARAY C, BLANC J F, et al. Comparison of MRI-based response criteria and radiomics for the prediction of early response to transarterial radioembolization in patients with hepatocellular carcinoma[J]. Diagn Interv Imaging, 2022, 103(7-8): 360-366. doi: 10.1016/j.diii.2022.01.009
[16]	ZHANG R P, ZHU L, ZHENG T, et al. Potential feature exploration and model development based on 18F-FDG PET/CT images for differentiating benign and malignant lung lesions[J]. Eur J Radiol, 2019, 121: 108735. DOI: 10.1016/j.ejrad.2019.108735.
[17]	GRANATA V, FUSCO R, DE MUZIO F, et al. Contrast MR-based radiomics and machine learning analysis to assess clinical outcomes following liver resection in colorectal liver metastases: a preliminary study[J]. Cancers(Basel), 2022, 14(5): 1110.
[18]	LI X, GUINDANI M, NG C S, et al. Spatial bayesian modeling of GLCM with application to malignant lesion characterization[J]. J Appl Stat, 2018, 46(2): 230-246.
[19]	MAYERHOEFER M E, MATERKA A, LANGS G, et al. Introduction to radiomics[J]. J Nucl Med, 2020, 61(4): 488-495.
[20]	WENG Q, ZHOU L, WANG H, et al. A radiomics model for determining the invasiveness of solitary pulmonary nodules that manifest as part solid nodules[J]. Clin Radiol, 2019, 74(12): 933 943.
[21]	赵诗雨, 何平, 杨成新, 等. 5 272名煤矿接尘工人肺量计数据分析[J]. 中华劳动卫生职业病杂志, 2021, 39(7): 546-549. YANG S Y, HE P, YANG C X, et al. Analysis of spirometer data of 5272 coal dust-exposed miners[J]. Chinese Journal of Industrial Hygiene and Occupational Diseases, 2021, 39(7): 546-549.
[22]	秦身钧, 陆青锋, 吴士豪, 等. 重庆中梁山晚二叠世煤有机地球化学特征[J]. 煤炭学报, 2018, 43(7): 1973-1982. https://www.cnki.com.cn/Article/CJFDTOTAL-MTXB201807021.htm QIN S J, LU Q F, WU S H, et al. Organic geochemistry of the Late Permian Coal from the Zhongliangshan mine, Chongqing[J]. Journal of China Coal Society, 2018, 43(7): 1973-1982. https://www.cnki.com.cn/Article/CJFDTOTAL-MTXB201807021.htm