摘要
目的:基于常规实验室指标,验证人工智能(Artificial intelligence,AI)在浸润性肺结核临床评估中的价值.方法:回顾性收集2021年1月-2022年1月在延安市第二人民医院确诊的96例浸润性肺结核患者的临床资料、实验室指标及胸部CT,并对胸部CT上病变范围和病变征象进行半定量评分,计算出病变比例得分、病变严重程度得分.用AI获取肺结核患者的定量CT指标(病灶体积(LeV,mL)、病灶占双肺体积的比例(LeV％)及病灶质量(LM,g)),采用Pearson或Spearman检验分析人工视觉评分、定量CT指标与实验室指标之间的相关性,并绘制定量CT指标与实验室指标、人工视觉评分与实验室指标的相关性分析图.结果:人工视觉评分(全肺病变比例得分、病变严重程度得分)(r=0.225～0.497,P＜0.001)、CT定量指标(LeV、LeV％、LM)(r=0.290～0.576,P＜0.001)与实验室指标 WBC、NEU、LYM、ALB、PAB、ESR、A/G、MLR、NLR、PLR 呈轻-中度相关.结论:AI 定量分析的定量CT指标、人工视觉评分与实验室指标的定量指标呈轻-中度相关,证实了定量CT在浸润性肺结核的临床诊断中具有可行性.
Abstract
Objective:To verify the value of artificial intelligence(AI)in the clinical evaluation of invasive pulmonary tu-berculosis based on conventional laboratory indicators.Methods:The clinical data,laboratory indexes and chest CT of 96 pa-tients with invasive pulmonary tuberculosis diagnosed in Yan'an Second People's Hospital from January 2021 to January 2022 were collected retrospectively.The lesion range and signs on chest CT were scored semi-quantitatively,and the lesion propor-tion score and lesion severity score were calculated.The quantitative CT indexes of pulmonary tuberculosis patients(LeV,mL),percentage of focus to lung volume(LeV％)and lesion mass(LM,g)were obtained by AI.The correlation between artificial vi-sion score,quantitative CT index and laboratory index were analyzed by Pearson or Spearman test,and the correlation diagram between quantitative CT index and laboratory index,artificial vision score and laboratory index were drawn.Results:Artificial vision score(whole lung lesion proportion score,lesion severity score)(r=0.225～0.497,P＜0.001)and CT quantitative index(LeV,LeV％,LM)(r=0.290～0.576,P＜0.001)were mildly to moderately correlated with laboratory indexes WBC,NEU,LYM,ALB,PAB,ESR,A/G,MLR,NLR,PLR.Conclusion:The quantitative CT index and artificial vision score of AI quantitative analy-sis are mildly to moderately correlated with those of laboratory indexes,which confirms the feasibility of quantitative CT in the clinical diagnosis of invasive pulmonary tuberculosis.
基金项目
大学生创新创业训练计划创新训练项目(D2022138)
陕西省卫生健康科研项目(2022B008)
维普
万方数据