摘要
目的由于DWI-Alberta卒中项目早期CT评分(DWI-ASPECTS)与临床医师评分在急性缺血卒中患者应用中均存在缺陷,文中旨在采用自动化评分(eDWI-ASPECTS)软件,并与不同年资神经放射科医师评估作对比,验证eDWI-ASPECTS的评估效能。方法回顾性分析2010年7月至2019年12月东部战区总医院神经内科309例入院并被诊断为大脑中动脉(Middle Cerebral Artery,MCA)供血区缺血性卒中的患者。使用基于机器学习的eDWI-ASPECTS进行评分,以及两名高年资和低年资(分别具有15年、5年神经影像方面的临床经验)的神经放射科医师分别进行评分。根据梗死发生的部位分为3个组:皮质组,即梗死在皮层区域(岛叶及M1-M6);深部组,即梗死部位在白质区及基底节区(尾状核、豆状核、内囊);混合组,即累及皮层、皮层下及深部区域。采用组内相关系数(ICC)、Cohen′s Kappa系数分别评估总体、医师之间以及医师与eDWI-ASPECTS之间的一致性,并将ASPECTS二分类后进一步评估医师与eDWI-ASPECTS的一致性。结果对于总体数据,低年资与高年资医师之间、低年资医师与eDWI-ASPECTS、高年资与eDWI-ASPECTS一致性较好(ICC=0.94、0.91、0.93)。皮质组一致性最好(P<0.05),而深部组及混合组的一致性下降,混合组一致性最差(P<0.05)。二分法评估结果:eDWI-ASPECTS与低年资医师(kappa=0.71,P<0.05)、eDWI-ASPECTS与高年资医师(kappa=0.76,P<0.05)、医师之间(kappa=0.79,P<0.05)显示一致性均达到较好水平。结论eDWI-ASPECTS与医师一致性较好,为急性缺血性卒中患者的诊断和治疗提供了标准化的评估工具;但是,不同梗死部位会影响到医师与eDWI-ASPECTS的一致性。
Objective DWI-Aspects,derived from the Alberta Stroke Program Early CT Aspects(CT-Aspects),is a semi-quantitative tool for early ischemic changes.However,the consistency of scores among clinicians remains low.This study aims to evaluate the consistency of MRI diffusion-weighted imaging(DWI)-ASPECTS evaluated by physicians with different seniority,and to further compare it with eDWI-Aspects.Methods A retrospective analysis was conducted on 309 patients admitted to the Department of Neurology of the General Hospital of the Eastern Theater Command from July 2010 to December 2019 who were diagnosed with ischemic stroke in the Middle Cerebral Artery(MCA)blood supply region.The EDWI-Aspects evaluation software based on machine learning and two neuroradiologists with high and low experience were used to evaluate.There were three groups according to the site of infarct:Cortical group,i.e.,infarcts in cortical areas(insula and M1-M6);Deep group,i.e.,infarcts in the white matter region and basal ganglia region(caudate nucleus,lentiform nucleus,and internal capsule);Mixed group,i.e.,cortical,subcortical,and deep areas were involved using intra-group correlation coefficient(ICC).Cohen's Kappa coefficient evaluates the consistency of the physicians in different groups and that between physicians and EDWI-Aspects,respectively,and further evaluates the consistency between physicians and EDWI-Aspects after the dichotomy of ASPECTS(0-6 vs 7-10).Results For the overall data,there was good consistency between the physicians of low and high seniority,low seniority and eDWI-Aspects,and high seniority and eDWI-Aspects(ICC=0.94,0.91,0.93).The Cortical group showed the best consistency(ICC=0.79~0.86,P<0.05),while that of the deep group and mixed group decreased,and the mixed group performed the worst,showing moderate to good consistency(ICC=0.62-0.76,P<0.05).Results of dichotomous evaluation:Kappa values of eDWI-Aspects and physicians with low seniority(Kappa=0.71,P<0.05),eDWI-Aspects and physicians with high seniority(Kappa=0.76,P<0.05),and among physicians(Kappa=0.79,P<0.05)all reached a good level of consistency.Conclusion eDWI-Aspects is consistent with physicians and provides a standardized assessment tool for the diagnosis and treatment of acute ischemic stroke patients.However,different infarct sites will affect the consistency between physicians and EDWI-Aspects.
作者
苏晓芹
程晓青
周长圣
董铮
施佳倩
刘权慧
潘成伟
李秀丽
俞益洲
张志强
SU Xiao-qin;CHENG Xiao-qing;ZHOU Chang-sheng;DONG Zheng;SHI Jia-qian;LIU Quan-hui;PAN Cheng-wei;LI Xiu-li;YU Yi-zhou;ZHANG Zhi-qiang(Department of Medical Imaging,Jinling Hospital,Nanjing University School of Medicine/General Hospital of Eastern Theater Command,PLA,Nanjing 210002,Jiangsu,China;Department of Medical Imaging,Xuzhou Medical University,Xuzhou 221004,Jiangsu,China;Deep Wise AI Lab,Beijing 100080,China)
出处
《医学研究生学报》
CAS
北大核心
2021年第2期177-181,共5页
Journal of Medical Postgraduates
