基于CT图像的肺结核病灶治愈状态判定深度学习模型的建立

doi:10.19982/j.issn.1000-6621.20230457

中国防痨杂志 ›› 2024, Vol. 46 ›› Issue (3): 272-278.doi: 10.19982/j.issn.1000-6621.20230457

基于CT图像的肺结核病灶治愈状态判定深度学习模型的建立

秦李祎¹, 吕平欣², 郭琳³, 钱令军³, 肖谦³, 杨阳⁴, 尚园园⁵^,⁶, 贾俊楠¹, 初乃惠⁵, 刘远明³(), 李卫民¹()

¹首都医科大学附属北京胸科医院/国家结核病临床实验室/耐药结核病研究北京市重点实验室,北京 101149
²北京老年医院影像科,北京 100095
³深圳市智影医疗科技有限公司,深圳 518109
⁴首都医科大学附属北京胸科医院影像科,北京 101149
⁵首都医科大学附属北京胸科医院结核一科,北京 101149
⁶首都医科大学附属北京友谊医院老年科,北京 100050

收稿日期:2023-12-25 出版日期:2024-03-10 发布日期:2024-03-05
通信作者: 刘远明,Email: f.lure@hotmail.com;李卫民,Email: lwm_18@aliyun.com
作者简介:注:吕平欣与秦李祎对本研究具有同等贡献,为并列第一作者
基金资助:
国家自然科学基金(82373641);深圳市科技计划资助项目(KQTD2017033110081833);广州市基础研究计划市校(院)企联合资助项目(2023A03J0536)

Deep learning to determine the healing status of pulmonary tuberculosis lesions on CT images

Qin Liyi¹, Lyu Pingxin², Guo Lin³, Qian Lingjun³, Xiao Qian³, Yang Yang⁴, Shang Yuanyuan⁵^,⁶, Jia Junnan¹, Chu Naihui⁵, Liu Yuanming³(), Li Weimin¹()

¹National Clinical Laboratory on Tuberculosis/Beijing Key Laboratory of Drug Resistance Tuberculosis Research/Beijing Chest Hospital, Capital Medical University, Beijing 101149, China
²Department of Radiology, Beijing Geriatric Hospital, Beijing 100095, China
³Shenzhen Smart Imaging Healthcare Co.,Ltd, Shenzhen 518109, China
⁴Department of Radiology, Beijing Chest Hospital, Capital Medical University, Beijing 101149, China
⁵The First Department of Tuberculosis, Beijing Chest Hospital, Capital Medical University, Beijing 101149, China
⁶Department of Geriatrics, Beijing Friendship Hospital, Capital Medical University, Beijing 100050, China

Received:2023-12-25 Online:2024-03-10 Published:2024-03-05
Contact: Liu Yuanming,Email: f.lure@hotmail.com; Li Weimin,Email: lwm_18@aliyun.com
Supported by:
National Natural Science Foundation of China(82373641);Shenzhen Science and Technology Program(KQTD2017033110081833);Guangzhou Basic Research Program City School (Institute) Enterprise Joint Funding Project(2023A03J0536)

摘要/Abstract

摘要：

目的: 基于CT影像构建深度学习模型判定肺结核病灶的活动性。方法: 回顾性纳入2018年12月至2020年12月首都医科大学附属北京胸科医院就诊的具有治疗前、中和后时间点的CT影像资料的肺结核治愈患者(102例),按照8∶2的比例将病灶随机分为训练集和测试集。另外,于2021年10月至2022年12月在同一家医院前瞻性纳入肺结核治愈患者(72例),在治疗前、中和后时间点纳入CT资料作为独立验证集。通过迁移学习方式进行深度学习模型构建;采用掩膜区域卷积神经网络(Mask R-CNN)架构实现病灶自动分割及活动性判定。基于三维病灶标签进行模型训练,通过计算测试集受试者工作特性(ROC)曲线下面积(AUC)、敏感度、特异度,并与独立验证集比较,评估模型对肺结核病灶活动性的判定效能。结果: 回顾性队列共纳入符合标准的肺结核治愈患者102例,共收集到770份CT影像资料;332个病灶为活动性,464个病灶为非活动性。前瞻性队列纳入肺结核治愈患者72例,共收集到540份CT影像资料。基于迁移学习的Mask R-CNN深度学习模型计算,测试集的AUC为87.5%,敏感度为85.7%,特异度为78.6%;独立验证集的AUC为79.9%,敏感度为78.7%,特异度为75.0%。结论: 基于迁移学习的Mask R-CNN深度学习模型在小样本量肺结核病灶活动性预测中展现出一定潜力,可以为快速、自动的临床决策提供科学参考。

关键词: 结核, 肺, 体层摄影术, X线计算机, 模型, 结构, 人工智能

Abstract:

Objective: To construct a deep learning model based on CT images for activity assessment of pulmonary tuberculosis lesions. Methods: A retrospective cohort of 102 cured pulmonary tuberculosis patients at Beijing Chest Hospital, Capital Medical University between December 2018 and December 2020 was included, CT data were collected before, during, and after treatment. Lesions were randomly divided into training and test sets with an 8∶2 ratio. Additionally, a prospective cohort of 72 cured pulmonary tuberculosis patients was enrolled between October 2021 and December 2022, CT datasets were collected for an independent validation set. A deep learning model was constructed through transfer learning using the Mask R-CNN architecture to achieve automatic lesion segmentation and activity determination. The model was trained based on three-dimensional lesion labels from the training set, and its performance in determining the activity of pulmonary tuberculosis lesions was evaluated in the test set and independent validation set by calculating the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. Results: A retrospective cohort of 102 cured pulmonary tuberculosis patients who met the criteria was included, and a total of 770 CT imaging data were collected; 332 lesions were active, and 464 lesions were inactive. A prospective cohort of 72 cured patients with pulmonary tuberculosis was included, and a total of 540 CT imaging data were collected. The transfer learning-based Mask R-CNN deep learning model achieved an AUC of 87.5%, sensitivity of 85.7%, and specificity of 78.6% in the test set. In the independent validation set, the model obtained an AUC of 79.9%, sensitivity of 78.7%, and specificity of 75.0%. Conclusion: The transfer learning-based Mask R-CNN deep learning model has shown promising potential in predicting the activity of small-scale pulmonary tuberculosis lesions, could offer valuable scientific insights for rapid and automatic clinical decision-making.

Key words: Tuberculosis, pulmonary, Tomography, X-ray computed, Models, structural, Artificial intelligence

中图分类号:

R521
R81

秦李祎, 吕平欣, 郭琳, 钱令军, 肖谦, 杨阳, 尚园园, 贾俊楠, 初乃惠, 刘远明, 李卫民. 基于CT图像的肺结核病灶治愈状态判定深度学习模型的建立[J]. 中国防痨杂志, 2024, 46(3): 272-278. doi: 10.19982/j.issn.1000-6621.20230457

Qin Liyi, Lyu Pingxin, Guo Lin, Qian Lingjun, Xiao Qian, Yang Yang, Shang Yuanyuan, Jia Junnan, Chu Naihui, Liu Yuanming, Li Weimin. Deep learning to determine the healing status of pulmonary tuberculosis lesions on CT images[J]. Chinese Journal of Antituberculosis, 2024, 46(3): 272-278. doi: 10.19982/j.issn.1000-6621.20230457

图/表 7

图1

表1

表2

表3

图2

图3

图4

参考文献 30

[1]	舒薇, 刘宇红. 世界卫生组织《2023年全球结核病报告》解读. 结核与肺部疾病杂志, 2024, 5(1): 15-19. doi:10.19983/j.issn.2096-8493.2024006.
[2]	邓国防, 路希维. 规范肺结核活动性的判断流程不断提高临床诊断与治疗水平. 中国防痨杂志, 2020, 42 (6): 541-543. doi:10.3969/j.issn.1000-6621.2020.06.002.
[3]	中国防痨协会. 非活动性肺结核诊断及预防发病专家共识. 结核与肺部疾病杂志, 2021, 2 (3): 197-201. doi:10.3969/j.issn.2096-8493.20210085.
[4]	国家感染性疾病临床医学研究中心, 深圳市第三人民医院,《中国防痨杂志》编辑委员会. 肺结核活动性判断规范及临床应用专家共识. 中国防痨杂志, 2020, 42 (4): 301-307. doi:10.3969/j.issn.1000-6621.2020.04.001.
[5]	Marais BJ, Graham SM. The Value of Chest Radiography in Tuberculosis Preventive Treatment Screening in Children and Adolescents. Am J Respir Crit Care Med, 2022, 206(7): 814-816. doi:10.1164/rccm.202205-1023ED.
[6]	Thiel BA, Bark CM, Nakibali JG, et al. Reader variability and validation of the Timika X-ray score during treatment of pulmonary tuberculosis. Int J Tuberc Lung Dis, 2016, 20(10): 1358-1363. doi:10.5588/ijtld.16.0186. pmid: 27725048
[7]	吕平欣, 李多, 骆宝建. CT评价肺结核病灶活动性的作用. 中华放射学杂志, 2013, 47(1): 87-90. doi:10.3760/cma.j.issn.1005-1201.2013.01.022.
[8]	Pinto LM, Pai M, Dheda K, et al. Scoring systems using chest radiographic features for the diagnosis of pulmonary tuberculosis in adults: a systematic review. Eur Respir J, 2013, 42(2): 480-494. doi:10.1183/09031936.00107412. pmid: 23222871
[9]	贺伟, 李成海, 谢汝明, 等. 活动性肺结核抗结核治疗后的HRCT影像动态分析及残余病变的临床意义. 中华实验和临床感染病杂志(电子版), 2015, 9(6): 38-42. doi:10.3877/cma.j.issn.1674-1358.2015.06.011.
[10]	Yoon SH, Kim YJ, Doh K, et al. Interobserver variability in Lung CT Screening Reporting and Data System categorisation in subsolid nodule-enriched lung cancer screening CTs. Eur Radiol, 2021, 31(9): 7184-7191. doi:10.1007/s00330-021-07800-5. pmid: 33733688
[11]	马利利, 饶会林, 唐飞, 等. 肺隐球菌病患者肺部CT影像学表现及临床病理特征分析. 临床军医杂志, 2022, 50(12): 1310-1312,1316. doi:10.16680/j.1671-3826.2022.12.30.
[12]	Cadena AM, Fortune SM, Flynn JL. Heterogeneity in tuberculosis. Nat Rev Immunol, 2017, 17(11): 691-702. doi:10.1038/nri.2017.69. pmid: 28736436
[13]	Moreno-Molina M, Shubladze N, Khurtsilava I, et al. Genomic analyses of Mycobacterium tuberculosis from human lung resections reveal a high frequency of polyclonal infections. Nat Commun, 2021, 12(1): 2716. doi:10.1038/s41467-021-22705-z. pmid: 33976135
[14]	Urbanowski ME, Ordonez AA, Ruiz-Bedoya CA, et al. Cavitary tuberculosis: the gateway of disease transmission. Lancet Infect Dis, 2020, 20(6): e117-e128. doi:10.1016/S1473-3099(20)30148-1. pmid: 32482293
[15]	Hosny A, Parmar C, Quackenbush J, et al. Artificial intelligence in radiology. Nat Rev Cancer, 2018, 18(8): 500-510. doi:10.1038/s41568-018-0016-5. pmid: 29777175
[16]	马依迪丽·尼加提, 米日古丽·达毛拉, 张斌, 等. 基于CT深度学习的活动性/非活动性肺结核分类模型构建及验证. 暨南大学学报(自然科学与医学版), 2023, 44(1): 69-77, 86. doi:10.11778/j.jdxb.20220268.
[17]	Yan C, Wang L, Lin J, et al. A fully automatic artificial intelligence-based CT image analysis system for accurate detection, diagnosis, and quantitative severity evaluation of pulmonary tuberculosis. Eur Radiol, 2022, 32(4): 2188-2199. doi:10.1007/s00330-021-08365-z.
[18]	Nijiati M, Zhou R, Damaola M, et al. Deep learning based CT images automatic analysis model for active/non-active pulmonary tuberculosis differential diagnosis. Front Mol Biosci, 2022, 9: 1086047. doi:10.3389/fmolb.2022.1086047.
[19]	Lee S, Yim JJ, Kwak N, et al. Deep Learning to Determine the Activity of Pulmonary Tuberculosis on Chest Radiographs. Radiology, 2021, 301(2): 435-342. doi:10.1148/radiol.2021210063.
[20]	He K, Gkioxari G, Dollar P, et al. Mask R-CNN. IEEE Trans Pattern Anal Mach Intell, 2020, 42(2): 386-397. doi:10.1109/TPAMI.2018.2844175.
[21]	中华人民共和国国家卫生和计划生育委员会. WS 288—2017 肺结核诊断. 2017-11-09.
[22]	吕平欣, 贺伟, 吕岩, 等. 基于影像学特征把握非活动性肺结核的影像学诊断要点. 中国防痨杂志, 2022, 44(4): 305-309. doi:10.19982/j.issn.1000-6621.20220002.
[23]	秦李祎, 吕岩, 杨阳, 等. 肺结核患者治愈后肺内残留的非活动性结核病灶的CT特征研究. 中国防痨杂志, 2022, 44(4): 336-342. doi:10.19982/j.issn.1000-6621.20210734.
[24]	Lewinsohn DM, Leonard MK, LoBue PA, et al. Official American Thoracic Society/Infectious Diseases Society of America/Centers for Disease Control and Prevention Clinical Practice Guidelines: Diagnosis of Tuberculosis in Adults and Children. Clin Infect Dis, 2017, 64(2): 111-115. doi:10.1093/cid/ciw778. pmid: 28052967
[25]	Ordonez AA, Tucker EW, Anderson CJ, et al. Visualizing the dynamics of tuberculosis pathology using molecular imaging. J Clin Invest, 2021, 131(5): e145107. doi:10.1172/JCI145107.
[26]	Im JG, Itoh H, Shim YS, et al. Pulmonary tuberculosis: CT findings-early active disease and sequential change with antituberculous therapy. Radiology, 1993, 186(3): 653-660. doi:10.1148/radiology.186.3.8430169. pmid: 8430169
[27]	Jiang H, Zhu C, Qin L, et al. Case report: A 9-year systema-tic treatment failure of a pulmonary tuberculosis patient. Front Public Health, 2022, 10: 966891. doi:10.3389/fpubh.2022.966891.
[28]	Tavaziva G, Harris M, Abidi SK, et al. Chest X-ray Analysis With Deep Learning-Based Software as a Triage Test for Pulmonary Tuberculosis: An Individual Patient Data Meta-Analysis of Diagnostic Accuracy. Clin Infect Dis, 2022, 74(8): 1390-1400. doi:10.1093/cid/ciab639.
[29]	Mahbub MK, Biswas M, Gaur L, et al. Deep features to detect pulmonary abnormalities in chest X-rays due to infectious diseaseX: Covid-19, pneumonia, and tuberculosis. Inf Sci (N Y), 2022, 592:389-401. doi:10.1016/j.ins.2022.01.062.
[30]	Hansun S, Argha A, Liaw ST, et al. Machine and Deep Learning for Tuberculosis Detection on Chest X-Rays: Systematic Literature Review. J Med Internet Res, 2023, 25: e43154. doi:10.2196/43154.

临床资料	回顾性肺结核治愈队列(102例)		前瞻性肺结核治愈队列(72例)		χ²值	P值
临床资料	例数	构成比(%)	例数	构成比(%)	χ²值	P值
性别					4.507	0.042
男性	53	52.0	49	68.1
女性	49	48.0	23	31.9
年龄组(岁)					0.666	0.415
≥30	56	54.9	44	61.1
<30	46	45.1	28	38.9
肺结核病史					17.858	<0.001
有	67	65.7	67	93.1
无	35	34.3	5	6.9
耐药情况					57.007	<0.001
敏感	10	9.8	0	0.0
单耐药	16	15.7	0	0.0
耐多药	64	62.7	35	48.6
多耐药	8	7.9	37	51.4
其他^a	4	3.9	0	0.0

采样数	回顾性肺结核治愈队列(770份)		前瞻性肺结核治愈队列(540份)		统计检验值	P值
采样数	例数	构成比(%)	例数	构成比(%)	统计检验值	P值
人均采样^a	7.0	6.0,8.8	7.0	6.0,9.0	Z=―0.360	0.719
治疗过程					χ²=26.114	<0.001
治疗前	105	13.6	94	17.4
治疗中	524	68.1	294	54.4
治疗后	141	18.3	152	28.1

数据集	CT图像数量		病灶标签三维感兴趣体积数量		曲线下面积 (%,95%CI值)	敏感度 (%)	特异度 (%)
数据集	治疗前	治疗后	活动性组	非活动性稳定组	曲线下面积 (%,95%CI值)	敏感度 (%)	特异度 (%)
训练集	85	113	260	365	92.4(90.1~94.7)	85.7	89.8
测试集	21	28	72	99	87.5(84.3~90.7)	85.7	78.6
验证集	94	152	313	493	79.9(74.4~84.7)	78.7	75.0

基于CT图像的肺结核病灶治愈状态判定深度学习模型的建立

Deep learning to determine the healing status of pulmonary tuberculosis lesions on CT images

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 30

相关文章 6

编辑推荐

Metrics

本文评价

友情链接

[1]	高珊, 聂文娟, 侯代伦, 初乃惠. 非结核分枝杆菌肺病影像学表现及应用人工智能新技术的研究进展[J]. 中国防痨杂志, 2024, 46(3): 362-366.
[2]	刘雪艳, 王芳, 李春华, 唐光孝, 郑娇凤, 王惠秋, 李玉蕊, 王佳男, 舒伟强, 吕圣秀. 基于深度学习的继发性肺结核CT辅助诊断模型构建及验证[J]. 中国防痨杂志, 2024, 46(3): 279-287.
[3]	李多, 吕平欣. 推动人工智能技术在肺结核影像领域的发展和运用[J]. 中国防痨杂志, 2024, 46(3): 253-259.
[4]	吴键, 侯代伦. 深度学习在肺结核影像诊断中的应用[J]. 中国防痨杂志, 2022, 44(1): 91-94.
[5]	王前, 何金戈, 李玉红, 陈明亭, 周林. 人工智能自动阅片技术用于HIV/AIDS人群结核病主动筛查效果的研究[J]. 中国防痨杂志, 2021, 43(6): 557-561.
[6]	曹盼,王斐,刘哲,刘锦程,梁矿立,袁吉欣,池峰,黄烨东,杨健. 特征金字塔网络在胸部X线摄影图像上筛检肺结核的价值[J]. 中国防痨杂志, 2019, 41(3): 288-293.