人工智能对提高放射科住院医生诊断胸部肋骨骨折一致性的价值

doi:10.19723/j.issn.1671-167X.2023.04.017

摘要/Abstract

摘要：

目的: 探讨人工智能(artificial intelligence, AI)提高放射科住院医生对外伤性肋骨骨折病灶检出率和不同阅片者之间检出一致性的价值。方法: 回顾性收集来自吉林大学中日联谊医院(单位02)和上海交通大学医学院附属第九人民医院(单位03)的393例急诊胸部外伤患者胸部CT图像，三位影像学专家的阅片结果作为评估的参考标准。所有图像分配到三个单位：北京大学第一医院(单位01)、单位02和单位03，并随机分为A组和B组(A组包括197例患者，B组包括196例患者)。每个单位各由一位低年资放射科住院医生对每组数据进行肋骨骨折检出的试验阅片，试验阅片时每位医生针对同一组数据进行两次阅片，一次为医生独立阅片(简称“医生”)，另一次为医生在AI软件辅助下阅片(简称“医生+AI”)。比较“医生”和“医生+AI”对不同类型(错位型、隐匿型)肋骨骨折病灶的检出率，并评价阅片一致性。结果: “医生+AI”对错位型肋骨骨折和隐匿型肋骨骨折病灶检出率均高于“医生”(94.56% vs. 78.40%, 76.60% vs. 49.42%, P < 0.001)。“医生”阅片的Kappa系数除单位01和单位03之间稍大于0.4(一致性中等)外，单位01和单位02及单位02和单位03均小于0.4(一致性较差)，且Phi系数均小于0.6(中度相关)，而“医生+AI”阅片的Kappa系数和Phi系数均等于或大于0.6(一致性较好，相关较强)。结论: AI软件可自动检出可疑肋骨骨折病灶，帮助提高医生对骨折病灶的检出率，并提高不同阅片者之间的一致性。

关键词: 肋骨骨折, 电子计算机断层扫描成像, 胸部外伤, 人工智能

Abstract:

Objective: To explore the value of artificial intelligence (AI) in improving the detection rate of traumatic rib fractures by radiologist residents and the consistency among different readers. Methods: Chest CT images of 393 patients with acute trauma from China-Japan Union Hospital of Jilin University (hospital 02) and Shanghai Ninth People' s Hospital (hospital 03) were collected in this research. The consensus achieved by three radiology experts was regarded as the reference standard. All the images assigned to three hospitals: Peking University First Hospital (hospital 01), hospital 02 and hospital 03, and were then randomly divided into two groups (group A and group B: group A included 197 patients, and group B included 196 patients). Each group was read by one radiologist resident from each hospital for rib fracture detection. Each case was read twice by the same radiologist, with and without the assistance of the AI ["radiologist-only" reading and "radiologist + AI" reading]. The detection rates of different types of rib fractures (displaced fractures and occult fractures) were compared between "radiologist-only" reading and "radiologist + AI" reading. The consistencies of different radiologists with different reading methods were evaluated. Results: The detection rates of displaced rib fractures and occult rib fractures by "radiologist + AI" reading were significantly higher than those read by "radiologist-only" reading (94.56% vs. 78.40%, 76.60% vs. 49.42%, P < 0.001). For "radiologist-only reading", the Kappa coefficients of the radiologists between hospital 01 and hospital 03 were slightly greater than 0.4 (indicating moderate consistency), the coefficients of the radiologists between hospital 01/hospital 02 and hospital 02/hospital 03 were less than 0.4 (indicating poor consistency). The Phi coefficients of the radiologists among different hospitals were all less than 0.6 (indicating moderate correlation). With "radiologist + AI" reading, the Kappa and Phi coefficient among the radiologists in dif-ferent hospitals were greater than or equal to 0.6 (indicating good consistency and correlation). Conclusion: AI software can be used to automatically detect suspected rib fracture lesions, which helps to improve the detection rate of fracture lesions and the consistency among different readers.

Key words: Rib fracture, Computed tomography imaging, Chest trauma, Artificial intelligence

中图分类号:

R814.4

刘想,谢辉辉,许玉峰,张晓东,陶晓峰,柳林,王霄英. 人工智能对提高放射科住院医生诊断胸部肋骨骨折一致性的价值[J]. 北京大学学报（医学版）, 2023, 55(4): 670-675.

Xiang LIU,Hui-hui XIE,Yu-feng XU,Xiao-dong ZHANG,Xiao-feng TAO,Lin LIU,Xiao-ying WANG. Value of artificial intelligence in the improvement of diagnostic consistency of radiology residents[J]. Journal of Peking University (Health Sciences), 2023, 55(4): 670-675.

图/表 6

表1

表2

图1

图2

表3

表4

参考文献 23

1	Hashmi ZG , Kaji AH , Nathens AB .Practical guide to surgical data sets: National trauma data bank (NTDB)[J].JAMA Surg,2018,153(9):852-853. doi: 10.1001/jamasurg.2018.0483
2	Peek J , Beks RB , Hietbrink F , et al.Epidemiology and outcome of rib fractures: A nationwide study in the Netherlands[J].Eur J Trauma Emerg Surg,2022,48(1):265-271. doi: 10.1007/s00068-020-01412-2
3	Weikert T , Noordtzij LA , Bremerich J , et al.Assessment of a deep learning algorithm for the detection of rib fractures on whole-body trauma computed tomography[J].Korean J Radiol,2020,21(7):891-899. doi: 10.3348/kjr.2019.0653
4	Pishbin E , Ahmadi K , Foogardi M , et al.Comparison of ultrasonography and radiography in diagnosis of rib fractures[J].Chin J Traumatol,2017,20(4):226-228. doi: 10.1016/j.cjtee.2016.04.010
5	Bizimungu R , Sergio A , Baumann BM , et al.Thoracic spine fracture in the panscan era[J].Ann Emerg Med,2020,76(2):143-148. doi: 10.1016/j.annemergmed.2019.11.017
6	Dennis BM , Bellister SA , Guillamondegui OD .Thoracic trauma[J].Surg Clin North Am,2017,97(5):1047-1064. doi: 10.1016/j.suc.2017.06.009
7	Cho SH , Sung YM , Kim MS .Missed rib fractures on evaluation of initial chest CT for trauma patients: Pattern analysis and diagnostic value of coronal multiplanar reconstruction images with multidetector row CT[J].Br J Radiol,2012,85(1018):e845-e850. doi: 10.1259/bjr/28575455
8	Singleton JM , Bilello LA , Canham LS , et al.Chest computed tomography imaging utility for radiographically occult rib fractures in elderly fall-injured patients[J].J Trauma Acute Care Surg,2019,86(5):838-843. doi: 10.1097/TA.0000000000002208
9	Talbot BS , Gange CP Jr , Chaturvedi A , et al.Traumatic rib injury: Patterns, imaging pitfalls, complications, and treatment[J].Radiographics,2017,37(2):628-651. doi: 10.1148/rg.2017160100
10	Murphy CE , Raja AS , Baumann BM , et al.Rib fracture diagnosis in the panscan era[J].Ann Emerg Med,2017,70(6):904-909. doi: 10.1016/j.annemergmed.2017.04.011
11	Langdorf MI , Medak AJ , Hendey GW , et al.Prevalence and clinical import of thoracic injury identified by chest computed tomography but not chest radiography in blunt trauma: Multicenter prospective cohort study[J].Ann Emerg Med,2015,66(6):589-600. doi: 10.1016/j.annemergmed.2015.06.003
12	Kim J , Kim S , Kim YJ , et al.Quantitative measurement method for possible rib fractures in chest radiographs[J].Healthc Inform Res,2013,19(3):196-204. doi: 10.4258/hir.2013.19.3.196
13	He J , Baxter SL , Xu J , et al.The practical implementation of artificial intelligence technologies in medicine[J].Nat Med,2019,25(1):30-36. doi: 10.1038/s41591-018-0307-0
14	Gao XW , Qian Y .Prediction of multidrug-resistant TB from CT pulmonary images based on deep learning techniques[J].Mol Pharm,2018,15(10):4326-4335. doi: 10.1021/acs.molpharmaceut.7b00875
15	Onishi Y , Teramoto A , Tsujimoto M , et al.Automated pulmonary nodule classification in computed tomography images using a deep convolutional neural network trained by generative adversarial networks[J].Biomed Res Int,2019,2019,6051939.
16	刘想, 谢辉辉, 许玉峰, 等.AI软件自动检出胸部CT图像上肋骨骨折的诊断效能研究[J].临床放射学杂志,2021,40(7):1369-1374.
17	Zhou QQ , Wang J , Tang W , et al.Automatic detection and classification of rib fractures on thoracic CT using convolutional neural network: Accuracy and feasibility[J].Korean J Radiol,2020,21(7):869-879. doi: 10.3348/kjr.2019.0651
18	张武平, 王春风, 王琳, 等.人工智能辅助在提高CT肋骨骨折检出率作用分析[J].影像研究与医学应用,2020,4(13):211-212.
19	杨超朋, 赵俊彦, 何光龙, 等.基于深度学习的人体肋骨骨折智能检测技术[J].刑事技术,2021,46(2):134-139.
20	周清清, 王佳硕, 唐雯, 等.基于卷积神经网络成人肋骨骨折CT自动检测和分类的应用研究[J].影像诊断与介入放射学,2020,29(1):27-31.
21	Chapman BC , Herbert B , Rodil M , et al.RibScore: A novel radiographic score based on fracture pattern that predicts pneumonia, respiratory failure, and tracheostomy[J].J Trauma Acute Care Surg,2016,80(1):95-101. doi: 10.1097/TA.0000000000000867
22	王建林, 李晓芬, 费峻眙.多层螺旋CT "一站式"图像后处理模式在肋骨骨折诊断以及司法鉴定中的应用探讨[J].现代医用影像学,2018,27(8):2755-2756.
23	羊鸿钧.多层螺旋CT三维重建在肋骨骨折诊断及法医临床鉴定中的应用[J].法制博览,2020,(8):155-156.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

Parameters	Brilliance 64	Brilliance 256	Lightspeed 16	NeuViz 128	Aquilion ONE
Scan length/mm	500	500	500	500	400-500
Tube voltage/kV	120	120	120	120	120
Tube current/mA	345	194-492	210	273-580	281-440
Slice thickness/mm	1	1	1.25	1	1
Image matrix	512×512	512×512	512×512	512×512	512×512
Kernal	Bone	Bone	Bone	Bone	Bone

Types of rib fractures	Index	Radiologist-only, n (%)	Radiologist + AI, n (%)	P value
Displaced fracture	Detected	461(78.40)	556(94.56)	< 0.001
	Not detected	127(21.60)	32(5.44)
	Total	588(100.00)	588(100.00)
Occult fracture	Detected	642(49.42)	995(76.60)	< 0.001
	Not detected	657(50.58)	304(23.40)
	Total	1 299(100.00)	1 299(100.00)
All fractures	Detected	1 103(58.45)	1 551(82.19)	< 0.001
	Not detected	784(41.55)	336(17.81)
	Total	1 887(100.00)	1 887(100.00)

Items	χ²	P value	Phi coefficient	Degree of correlation	Kappa (95%CI)	Degree of consistency
A1-A2	38.586	< 0.001	0.402	Moderate	0.371 (0.250, 0.493)	Poor
A1-A3	61.843	< 0.001	0.509	Moderate	0.491 (0.366, 0.616)	Moderate
A2-A3	23.119	< 0.001	0.311	Moderate	0.308 (0.182, 0.435)	Poor
B1-B2	52.185	< 0.001	0.366	Moderate	0.365 (0.273, 0.457)	Poor
B1-B3	67.168	< 0.001	0.415	Moderate	0.415 (0.325, 0.505)	Moderate
B2-B3	58.210	< 0.001	0.386	Moderate	0.385 (0.294, 0.477)	Poor

Items	χ²	P value	Phi coefficient	Degree of correlation	Kappa (95%CI)	Degree of consistency
A1-A2	68.209	< 0.001	0.618	Good	0.601 (0.458, 0.744)	Good
A1-A3	85.393	< 0.001	0.707	Good	0.698 (0.564, 0.832)	Good
A2-A3	102.028	< 0.001	0.653	Good	0.652 (0.520, 0.783)	Good
B1-B2	144.135	< 0.001	0.608	Good	0.596 (0.497, 0.695)	Good
B1-B3	177.701	< 0.001	0.675	Good	0.668 (0.577, 0.759)	Good
B2-B3	232.594	< 0.001	0.772	Good	0.771 (0.687, 0.855)	Excellent

[1]	孙玉春,郭雨晴,陈虎,邓珂慧,李伟伟. 口腔精准仿生修复技术的自主创新研发与转化[J]. 北京大学学报（医学版）, 2022, 54(1): 7-12.
[2]	郭辅政,朱凤雪,邓玖旭,杜哲,赵秀娟. 严重多发伤患者机械通气的危险因素分析[J]. 北京大学学报（医学版）, 2020, 52(4): 738-742.