[1]耿然.基于深度学习的人工智能影像辅助诊断系统对肺结节的诊断效能评价[J].新乡医学院学报,2022,39(11):1031-1035.[doi:10.7683/xxyxyxb.2022.11.006]
　GENG Ran.Evaluation of efficiency of artificial intelligence image aided diagnosis system based on deep learning in the diagnosis of pulmonary nodules[J].Journal of Xinxiang Medical University,2022,39(11):1031-1035.[doi:10.7683/xxyxyxb.2022.11.006]
点击复制

基于深度学习的人工智能影像辅助诊断系统对肺结节的诊断效能评价
分享到：

《新乡医学院学报》[ISSN:1004-7239/CN:41-1186/R]

卷:
39
期数:
2022年11
页码:
1031-1035
栏目:
临床研究
出版日期:
2022-11-05

文章信息/Info

Title:
Evaluation of efficiency of artificial intelligence image aided diagnosis system based on deep learning in the diagnosis of pulmonary nodules
作者:
耿然
(洛阳市第一人民医院CT室,河南　洛阳　471002)
Author(s):
GENG Ran
（CT Room,Luoyang First People′s Hospital,Luoyang 471002,Henan Province,China）
关键词:
人工智能; 计算机断层扫描; 肺结节; 诊断效能; 特异度; 假阳性率
Keywords:
artificial intelligence; computed tomography; pulmonary nodules; diagnostic efficiency; specificity; false positive rate
分类号:
R563
DOI:
10.7683/xxyxyxb.2022.11.006
文献标志码:
A
摘要:
目的　探讨人工智能（AI）影像辅助诊断系统对肺结节的诊断效能。方法　回顾性分析2016年1月至2019年12月于洛阳市第一人民医院经术前计算机断层扫描（CT）检查及病理检查结果确诊的85例肺结节患者的胸部CT影像资料，分别采用AI影像辅助诊断系统阅片及影像医师常规人工阅片，比较2种阅片方法对肺结节的检出情况，并比较这2种阅片方法诊断肺恶性结节的灵敏度、特异度、假阳性率以及评定结节性质的特异度。结果　85例患者共检出真性结节290枚，其中直径<5 mm结节171枚，5 mm≤直径≤10 mm结节83枚，直径>10 mm结节36枚。171枚直径<5 mm 的小结节中，AI阅片检出162枚（94.74%），人工阅片检出78枚（45.61%），其中人工阅片检出而AI阅片未检出9枚（5.26%），人工阅片未检出而AI阅片检出93枚（54.38%），AI阅片检出率显著高于人工阅片检出率（ χ2=4.317，P<0.01）。83枚5 mm≤直径≤10 mm 的结节中，AI阅片检出83枚（100.00%），人工阅片检出45枚（54.21%），其中人工阅片检出而AI阅片未检出0枚（0.00%），人工阅片未检出而AI阅片检出38枚（45.78%），AI阅片检出率显著高于人工检出率（ χ2=1.744，P<0.01）。36枚直径>10 mm的结节中，AI阅片检出率和人工阅片检出率均为100.00%（36/36）。在检出的290枚真性结节中，AI阅片检出281枚，其中AI阅片评定为低危结节99枚、中低危结节52枚、中危结节38 枚、高危结节（恶性结节）92枚；人工阅片检出159枚，其中人工阅片检出评定为低危结节59枚、中低危结节24枚、中危结节21枚、高危结节（恶性结节）55枚。经病理证实的恶性结节共66枚，其中AI阅片判定正确62枚，4枚未检出,准确率为93.93%(62/66)；人工阅片判定正确52枚，14枚未检出，准确率为78.79%(52/66)；AI阅片判定准确率显著高于人工阅片判定（ χ2=3.216，P<0.05）。AI阅片对肺恶性结节检查的灵敏度、假阳性率显著高于人工阅片，特异度和假阴性率显著低于人工阅片（ χ2= 2.311、4.165、7.896、2.311，P<0.05）。 结论　AI影像辅助诊断系统阅片对肺结节有较高的检出率及灵敏度，但其特异度低于影像医师人工阅片，有一定的假阳性率。在临床工作中应采用AI影像辅助诊断系统阅片联合影像医师人工阅片的模式诊断肺结节，以有效提高对肺结节的早期发现及良恶性鉴别的准确度，减少漏诊率。
Abstract:
Objective　To investigate the diagnostic efficiency of artificial intelligence (AI) image aided diagnostic system in pulmonary nodules.Methods　The chest computed tomography(CT) imaging data of 85 patients with pulmonary nodules diagnosed by preoperative CT examination and pathological examination in Luoyang First People′s Hospital from January 2016 to December 2019 was retrospectively analyzed.The detection of pulmonary nodules in the images readed by the AI image aided diagnostic system and conventional manual film-reading by imaging physicians was compared;and the sensitivity,specificity,false positive rate and the specificity in evaluating the nature of nodules between the two film-reading methods were compared.Results　A total of 290 true nodules were detected in 85 patients,including 171 nodules with the diameter <5 mm,83 nodules with the diameter from 5 mm≤ to ≤10 mm,and 36 nodules with the diameter >10 mm.Of the 171 small nodules with the diameter of <5 mm,162 nodules (94.74%) were detected by AI film-reading method,and 78 nodules (45.61%) were detected by manual film-reading method;among them,9 nodules (5.26%) were detected by manual film-reading method but not by AI film-reading method,and 93 nodules (54.38%) were detected by AI film-reading method but not by manual film-reading method;the detection rate of nodules detected by AI film-reading method was significantly higher than that of manual film-reading method (χ2=4.317,P<0.01).Of the 83 nodules with the diameter from 5 mm≤ to ≤10 mm,83 nodules (100.00%) were detected by AI film-reading method,and 45 nodules (54.21%) were detected by manual film-reading method;among them,0 nodules(0.00%) was detected by manual film-reading method but not detected by AI film-reading method,and 38 nodules (45.78%) were detected by AI film-reading method but not by manual film-reading method;the detection rate of AI film-reading method was significantly higher than that of manual film-reading method ( χ2=1.744,P<0.01).The detection rate of AI film-reading method and manual film-reading method were 100.00% (36/36) in 36 nodules with the diameter>10 mm.Of the 290 true nodules,281 nodules were detected by AI film-reading method,including 99 low-risk nodules,52 medium low-risk nodules,38 medium-risk nodules,and 92 high-risk nodules (malignant nodules);159 nodules were detected by manual film-reading method,including 59 low-risk nodules,24 medium low-risk nodules,21 medium-risk nodules and 55 high-risk nodules (malignant nodules).There were 66 malignant nodules confirmed by pathology,of which 62 nodules were correctly confirmed by AI film-reading method,and 4 nodules were not detected by AI film-reading method,with an accuracy rate of 93.93% (62/66);there were 52 malignant nodules were correctly confirmed by manual film-reading method,and 14 nodules were not detected by manual film-reading method,with an accuracy rate of 78.79% (52/66);the accuracy rate of malignant nodules correctly confirmed by AI film-reading method was significantly higher than that of manual film-reading method (χ2=3.216,P<0.05).The sensitivity and false positive rate of AI film-reading method in the diagnosis of pulmonary malignant nodules were significantly higher than those of manual film-reading method,and the specificity and false negative rate were significantly lower than those of manual film-reading method (χ2=2.311,4.165,7.896,2.311;P<0.05).Conclusion　AI image aided diagnostic system has a high detection rate and sensitivity for pulmonary nodules,however,but its specificity is lower than that of manual film-reading method,with a certain false positive rate.In clinical work,the mode of AI image aided diagnosis system film-reading combined with manual film-reading should be applied to diagnose pulmonary nodules,so as to effectively improve the accuracy of early detection of pulmonary nodules and differential diagnosis between benign and malignant,and reduce the rate of missed diagnosis.

参考文献/References:

［1］　中华医学会呼吸病学分会肺癌学组，中国肺癌放置联盟专家组.肺结节诊治中国专家共识（2018年版）［J］.中华结核和呼吸杂志，2018，41(10)：763-771.
LUNG CANCER GROUP OF RESPIRATORY BRANCH OF CHINESE MEDICAL ASSOCIATION,EXPERT GROUP OF CHINA LUNG CANCER PREVENTION AND CONTROL ALLIANCE.Consensus of Chinese expert on diagnosis and treatment of pulmonary nodules(2018 edition)［J］.Chin J Tuberc Respir Dis,2018,41(10)：763-771.
［2］　李欣菱,郭芳芳,周振,等.基于深度学习的人工智能胸部CT肺结节检测效能评估［J］.中国肺癌杂志,2019,22(6):336-340.
LI X L,GUO F F,ZHOU Z,et al.Performance of deep-learning-based artificial intelligence on detection of pulmonary nodules in chest CT［J］.Chin J Lung Cancer,2019,22(6):336-340.
［3］　刘娜,赵正凯,邹佳瑜,等.基于人工智能的胸部CT肺结节检出及良恶性诊断效能评估［J］.CT理论与应用研究,2021,30(6):709-715.
LIU N,ZHAO Z K,ZOU J Y,et al.Evaluation of detection and diagnostic efficiency of pulmonary nodules by chest CT based on artificial intelligence［J］.Comput Tomogr Theory Applic,2021,30(6):709-715.
［4］　宋振春,于铁链.基于人工智能肺结节CT特征性诊断优势分析［J］.影像研究与医学应用,2020,4(20):67-68.
SONG Z C,YU T L.Advantage analysis of CT characteristic diagnosis of pulmonary nodules based on artificial intelligence［J］.J Imaging Res Med Appl,2020,4(20):67-68.
［5］　CUI S J,MING S,LIN Y,et al.Development and clinical application of deep learning model for lung nodules screening on CT images［J］.Scientific Reports,2020,10(1):13657.
［6］　WILSHIRE C L,LOUIE B E,MANNING K A,et al.Radiologic evaluation of small lepidic adenocarcinomas to guide decision making in surgical resection［J］.ANN Thorac Surg,2015,100(3):979-988.
［7］　WANG Y,WU B,ZHANG N,et al.Research progress of computer aided diagnosis system for pulmonary nodules in CT images［J］.J Xray Sci Technol,2020,28(1):1-16.
［8］　WANG X Q,MAO K M,WANG L Z,et al.An appraisal of lung nodules automatic classification algorithms for CT images［J］.Sensors (Basel),2019,19(1):194.
［9］　REVEL M P,MANNES I,BENZAKOUN J,et al.Subsolid lung nodule classification:a CT criterion for improving interobserver agreement［J］.Radiology,2018,286(1):316-325.
［10］　MACMAHON H,NAIDICH D P,GOO J M,et al.Guidelines for management of incidental pulmonary nodules detected on CT images:from the Fleischner society 2017［J］.Radiology,2017,284(1):228-243.
［11］　TRAVIS W D,BRAMBILLA E,NOGUCHI M,et al.International association for the study of lung cancer/american thoracic society/european respiratory society international multidisciplinary classification of lung adenocarcinoma［J］.J Thorac Oncol,2011,6(2):244-285.

更新日期/Last Update: 2022-11-05