人工智能对肺结核患者病变检出及定性诊断价值研究

新发传染病电子杂志 ›› 2018, Vol. 3 ›› Issue (4): 214-217.

人工智能对肺结核患者病变检出及定性诊断价值研究

闫明艳¹, 陈根铭², 赖超¹, 陈智红¹, 祝君兰¹, 李娇¹, 董淑雯¹, 成官迅¹

1.北京大学深圳医院医学影像科,广东深圳 518036;
2.深圳市宝安区中心医院影像科,广东深圳 518102

收稿日期:2018-09-25 出版日期:2018-11-30 发布日期:2020-07-09
通讯作者: 成官迅,Email：chengguanxun@outlook.com
基金资助:
深圳市宝安区科技计划社会公益项目(2015313）

Clinical value of Artificial Intelligence in detecting and analyzing lesions in patients with pulmonary tuberculosis

YAN Ming-yan¹, Chen Gen-ming², LAI Chao¹, CHEN Zhi-hong¹, ZHU Jun-lan¹, LI Jiao¹, DONG Shu-wen¹, CHENG Guan-xun¹

1. Medical Imaging Department, Peking University Shenzhen Hospital,Guangdong Shenzhen 518036,China;
2. Department of Imaging, Central Hospital of Baoan District, Shenzhen,Guangdong Shenzhen 518102,China

Received:2018-09-25 Online:2018-11-30 Published:2020-07-09

摘要/Abstract

摘要： 目的对比肺结节检测分析系统和放射科医师对肺结核患者病变的检出情况,探讨人工智能（Artificial Intelligence,AI）对肺结核患者病变检出的价值。方法收集北京大学深圳医院2016年12月至2018年3月行胸部高分辨CT平扫检查的32例肺结核病例并分为A、B两组,分别评价两组的CT图像。A组采用σ-Discover/Lung肺结节检测分析系统0.2版本,分析获得病变检测结果;B组为2名胸部影像医师分析并获得病变检测结果。统计分析A、B两组对肺结核患者病变的检出情况,回顾性分析CT图像,经2名经验丰富的胸部影像科医师阅片,剔除考虑干酪性肺炎等大片状阴影的病例,采用σ-Discover/Lung肺结节检测分析系统分析病变良恶性概率,以大于50%作为高恶性判断标准。结果本研究共纳入32例肺结核患者,共413处病灶。A组一共检出380处病灶,其中误检10处,漏检43处;B组一共检出407处病灶,其中无误检,漏检6处病变。A组对肺结核患者病变的检出率为89.83%,误检率为2.63%,漏诊率为10.17%;B组对病变的检出率为98.55%,误检率为0%,漏诊率为1.42%。肺结节检测分析系统判断一共217处病变恶性概率大于50%,57.1%的肺良性病变分析为高恶性概率。结论肺结节检测分析系统对肺结核患者病变的检出具有较好的能力,但检出能力不如放射科医师,漏诊率较高,且目前对肺结核患者病变的良恶性分析准确率较低。

关键词: 人工智能, 肺结核, 检出率, 漏诊率, 肺结节

Abstract: Objective To compare the detection abilities of pulmonary nodule detection and analysis system and radiologists in the diagnosis of pulmonary tuberculosis lesions. To explore the contribution of artificial intelligence to the detection of lesions in tuberculosis patients. Methods A retrospective study was conducted to analyze the computed tomography images of tuberculosis patients from December 2016 to March 2018 in Peking University Shenzhen hospital. Firstly, CT images were retrospectively analyzed by two methods and divided into two groups, A and B. Group A was analyzed by σ-Discover/Lung pulmonary nodule detection and analysis system 1.0. Group B was analyzed by two chest radiologists. The abilities of lesions detection in tuberculosis patients in groups A and B were compared. Secondly, after reading by two experienced chest radiologists, patients with malignant pulmonary nodules or masses were excluded, and σ-Discover/Lung pulmonary nodule detection and analysis system was used to analyze the probabilities of malignant lesions (more than 50% as a high malignant criterion). Results A total of 32 patients with pulmonary tuberculosis were enrolled in the study, with a total of 413 lesions. A total of 380 lesions were detected in group A, including 10 misdetections and 43 missed examinations. A total of 407 lesions were detected in group B and with no false detections and 6 missing lesions. The detection rate of lesions in group A was 89.83%, the rate of false detection was 2.63%, and the rate of missing diagnosis was 10.17%. The detection rate of lesions in group B was 98.55%, the rate of false detection was 0%, and the rate of missing diagnosis was 1.42%. The pulmonary nodule detection and analysis system judged that the total malignant probability of 217 lesions was greater than 50%, and 57.1% of benign lung lesions were analyzed for high malignant probability. Conclusion sThe pulmonary nodule detection and analysis system has a good ability to detect the lesions of pulmonary tuberculosis patients, but the detection ability is not as good as radiologists. The rate of missing diagnosis is high, and the current accuracy of diagnosis of benign lesions of tuberculosis patients is low.

Key words: Artificial Intelligence, Tuberculosis, Detection rate, The rate of missed diagnosis, Pulmonary nodule

闫明艳, 陈根铭, 赖超, 陈智红, 祝君兰, 李娇, 董淑雯, 成官迅. 人工智能对肺结核患者病变检出及定性诊断价值研究[J]. 新发传染病电子杂志, 2018, 3(4): 214-217.

YAN Ming-yan, Chen Gen-ming, LAI Chao, CHEN Zhi-hong, ZHU Jun-lan, LI Jiao, DONG Shu-wen, CHENG Guan-xun. Clinical value of Artificial Intelligence in detecting and analyzing lesions in patients with pulmonary tuberculosis[J]. Electronic Journal of Emerging Infectious Diseases, 2018, 3(4): 214-217.

参考文献

[1] World Health Organization.Global tuberculosis report 2018[EB/OL].[2018-0-18]. http://www.who.int/tb/publications/global_report/en/ .
[2] SINGH A K, GUPTA U D.Animal models of tuberculosis: Lesson learnt[J]. The Indian journal of medical research, 2018, 147(5): 456-63.
[3] FAZAL M I, PATEL M E, TYE J, et al.The past, present and future role of artificial intelligence in imaging[J]. Eur J Radiol, 2018, 105: 246-250.
[4] NISHIKAWA R M.Current status and future directions of computer-aided diagnosis in mammography[J]. Computerized medical imaging and graphics : the official journal of the Computerized Medical Imaging Society, 2007, 31(4-5): 224-235.
[5] KLIGERMAN S, CAI L, WHITE C S.The effect of computer-aided detection on radiologist performance in the detection of lung cancers previously missed on a chest radiograph[J]. Journal of thoracic imaging, 2013, 28(4): 244-252.
[6] MASSALHA S, CLARKIN O, THORNHILL R, et al.Decision Support Tools, Systems, and Artificial Intelligence in Cardiac Imaging[J]. The Canadian journal of cardiology, 2018, 34(7): 827-838.
[7] XIONG Y, BA X, HOU A, et al.Automatic detection of mycobacterium tuberculosis using artificial intelligence[J]. Journal of thoracic disease, 2018, 10(3): 1936-1940.
[8] Y.M.LURE F, JAEGER S, ANTANI S, et al. 自动化显微镜检测和数字化胸片诊断系统在肺结核筛查中的应用[J]. 新发传染病电子杂志, 2017, 2(1): 5-9.
[9] LAKHANI P, SUNDARAM B.Deep Learning at Chest Radiography: Automated Classification of Pulmonary Tuberculosis by Using Convolutional Neural Networks[J]. Radiology, 2017, 284(2): 574-582.
[10] KIM D H, WIT H, THURSTON M.Artificial intelligence in the diagnosis of Parkinson's disease from ioflupane-123 single-photon emission computed tomography dopamine transporter scans using transfer learning[J]. Nuclear Medicine Communications, 2018, 39(10):887-893.
[11] CHUNG S W, HAN S S, LEE J W, et al.Automated detection and classification of the proximal humerus fracture by using deep learning algorithm[J]. Acta orthopaedica, 2018, 89(4): 468-473.
[12] BARBIERI C, MOLINA M, PONCE P, et al.An international observational study suggests that artificial intelligence for clinical decision support optimizes anemia management in hemodialysis patients[J]. Kidney International, 2016, 90(2): 422-429.
[13] PARK W J, PARK J-B.History and application of artificial neural networks in dentistry[J]. European Journal of Dentistry, 2018, 12(4): 594-601.
[14] HOSNY A, PARMAR C, QUACKENBUSH J, et al.Artificial intelligence in radiology[J]. Nature reviews Cancer, 2018, 18(8): 500-510.
[15] PESAPANE F, CODARI M, SARDANELLI F.Artificial intelligence in medical imaging: threat or opportunity? Radiologists again at the forefront of innovation in medicine[J]. European Radiology Experimental, 2018, 2(1): 35.