摘要
目的 基于计算机肺部感染辅助诊断系统,验证人工智能(AI)技术在浸润性肺结核病灶检测及评估中应用的准确性及可行性。方法 回顾性搜集2020年1月至2022年7月延安市第二人民医院确诊的初次就诊的浸润性肺结核患者120例,对肺部病变范围和病变征象进行半定量评分。将各肺叶的病变范围和各病变征象分数相加,得到最终的分数。采用Pearson或Spearman相关分析全肺及各肺叶视觉评分(病变范围得分、病变严重程度得分、各征象得分)与CT定量指标[病灶体积(LeV)、病灶占肺体积的百分比(LeV%)及病灶质量(LM)]之间的相关性。结果 全肺病变范围得分、病变严重程度得分与全肺定量CT指标LeV、LeV%、LM呈高度相关(r=0.783~0.826,P<0.001);各肺叶LeV%与人工评估的病变比例得分、各肺叶LM与人工评估的病变严重程度得分均呈高度相关(r=0.761~0.913,P<0.001)。结论 AI获得的全肺及各肺叶CT定量指标与传统视觉评分具有高度相关性,证实了其在浸润性肺结核客观影像学定量评估中具有准确性及可行性。
Objective Verification of artificial intelligence(artificial)based on computer aided diagnosis system of pulmonary infection.The accuracy and feasibility of intelligence,AI technique in the detection and evaluation of invasive pulmonary tuberculosis.Methods A total of 120 patients with invasive pulmonary tuberculosis diagnosed for the first time in Yan'an second people's Hospital from January 2020 to July 2022 were collected retrospectively.The lesion range of each lung lobe and the lesion sign score were added to get the final score.Pearson or Spearman correlation analysis was used to analyze the correlation between the visual score of the whole lung and each lobe(lesion range score,lesion severity score,sign score)and CT quantitative index[lesion volume(lesion Volume,LeV/ml),percentage of lesion to lung volume(per⁃centage of lesion,LeV%)and lesion mass(lesion mass,LM/g)].Results The whole lung lesion range score and lesion severity score were highly correlated with the whole lung quantitative CT index LeV,LeV%and LM(r=0.783-0.826,P<0.001),and the LeV%score of each lobe was highly correlated with the lesion proportion score and the LM score of each lobe with the lesion severity score evaluated by manual(r=0.761-0.913,P<0.001).Conclusion The CT quan⁃titative indexes of wh1ole lung and each lobe obtained by AI are highly correlated with the traditional visual score,which proves that it is accurate and feasible in the objective imaging quantitative evaluation of invasive pulmonary tuberculosis.
作者
巩盼
李星宇
来坤
阮梦雨
沙文霞
李建龙
黄晓旗
郭佑民
王改莲
GONG Pan;LI Xingyu;LAI Kun(Department of Radiology,Yanan University Affiliated Hospital,Yanan,Shaanxi Province 716000,P.R.China)
出处
《临床放射学杂志》
北大核心
2024年第4期567-573,共7页
Journal of Clinical Radiology
基金
陕西省卫生健康科研项目(编号:2022B008)
大学生创新创业训练计划创新训练项目(编号:D2022138)。
关键词
肺结核
定量CT
视觉评估
人工智能
计算机辅助诊断
Pulmonary tuberculosis
Quantitative CT
Visual assessment
Artificial intelligence
Computer⁃aided diagnosis
作者简介
通讯作者:王改莲。
