Effect of AI-CAD assisting doctors with different seniority in CT image interpretation to predict the enlargement of hematoma in early stage of cerebral hemorrhage
10.20039/j.cnki.1007-3949.2024.05.009
- VernacularTitle:AI-CAD辅助不同年资医生CT影像判读预测脑出血早期血肿扩大的效果分析
- Author:
Wei WEI
1
;
Wenjing FAN
;
Xin CHEN
;
Zheming ZHANG
;
Guoliang LI
;
Dong CHEN
Author Information
1. 大连理工大学附属中心医院神经外科,辽宁省大连市 116033
- Keywords:
artificial intelligence;
cerebral hemorrhage;
hematoma enlargement
- From:
Chinese Journal of Arteriosclerosis
2024;32(5):429-436
- CountryChina
- Language:Chinese
-
Abstract:
Aim To investigate the effect of artificial intelligence(AI)assisting doctors with different seniority in predicting the enlargement of hematoma in the early stage of cerebral hemorrhage.Methods A total of 108 patients diagnosed with cerebral hemorrhage in Central Hospital Affiliated to Dalian University of Technology were retrospectively collected.CT images at admission and 24 hours after admission were collected.DICOM images obtained from plain CT scan were input into AI-CAD model developed by Biomind in collaboration with Temple of Heaven.A total of 9 doctors of different senior-level were selected in neurosurgery department of our hospital.Firstly,independent prediction was applied in the patients and then the study predicted whether patients would delelop hematoma enlargement within 24 hours combined with the results of auxiliary AI.The accuracy of independent prediction of doctors with different seniority and assisted AI prediction of aneurysm stability was calculated respectively.McNemar of paired samples was used to test the significance of difference between independent prediction coincidence rate and assisted AI prediction accuracy among differ-ent doctors.Results The accuracy of high,middle and low seniority doctors independently predicting the early expan-sion of cerebral hemorrhage was 58.95％,50.62％and 38.89％,respectively,and the accuracy of prediction was signifi-cantly improved after assisted AI(P＜0.001),the highest increase rate was low seniority doctors(25.92％),followed by middle seniority doctors(19.75％)and high seniority doctors(11.73％).The ability of senior physicians to independ-ently predict the expansion of intracerebral hemorrhage was strongest in patients and non-patients,with sensitivity of 18.75％(95％CI:9.44％～33.10％)and specificity of 65.94％(95％CI:59.98％～71.45％).The sensitivity of middle seniority doctors was 16.67％(95％CI:7.97％～30.76％),the specificity was 56.52％(95％CI:50.44％～62.42％),and the sensitivity of low seniority doctors was 8.33％(95％CI:2.70％～20.87％),the specificity was 44.20％(95％CI:38.29％～50.28％).However,after AI assisted the prediction of senior doctors,the sensitivity and specificity of each seniority group of doctors increased.The sensitivity of high seniority doctors was 60.42％(95％CI:45.29％～73.88％),the specificity was 72.46％(95％CI:66.72％～77.57％),the sensitivity of middle seniority doc-tors was 64.58％(95％CI:49.40％～77.45％),the specificity was 71.38％(95％CI:65.59％～76.56％),and the sensitivity of low seniority doctors was 68.75％(95％CI:53.60％～80.91％),the specificity was 64.13％(95％CI:58.13％～69.73％).Conclusion AI-CAD assisted doctors with high,middle and low seniority can improve the ac-curacy of predicting the enlargement of hematoma in early stage of cerebral hemorrhage,especially the ability of doctors with low seniority to find patients can be significantly improved,which can make up for the lack of work experience of doc-tors with low seniority to a certain extent.
