人工智能辅助诊断软件探究新型冠状病毒肺炎核酸转阴后的肺部CT表现及其与入院时临床分型的关系

Imaging manifestations of chest CT in patients with coronavirus disease 2019 after nucleic acid conversion and its relationship with clinical classification at admission based on artificial intelligence aided diagnosis software

目的 采用人工智能辅助诊断软件探究新型冠状病毒肺炎(COVID-19)患者核酸转阴后肺部CT的影像表现及其与入院时临床分型的关系。方法 回顾性分析2020年6月至2021年8月首都医科大学附属北京地坛医院经核酸检测确诊且收住入院的COVID-19患者529例,应用人工智能肺炎辅助诊断软件从患者肺部CT中提取病灶位置、病灶体积、病灶体积占比、病灶密度等指标,并从患者临床病例中获取入院时临床分型信息,入组患者357例,共有5种临床分型:无症状感染者(27例,7.6%)、轻型(80例,22.4%)、普通型(227例,63.6%)、重型(21例,5.9%)、危重型(2例,0.6%)。对于各个肺叶内的病灶情况进行统计描述,对患者左右肺病灶体积进行差异性分析;对患者入院时及核酸转阴后的病灶体积、病灶体积占比、病灶密度进行差异性分析;将患者核酸转阴后的病灶体积、病灶体积占比、病灶密度与入院时的临床分型进行相关性分析。结果 本组357例患者中,入院时肺部CT显示有病灶的患者有248例(69.5%),经治疗核酸转阴后,双肺仍有病灶者140例(39.2%)。病灶主要位于双肺下叶,其中右肺上叶有病灶者72例(51.4%),右肺中叶有病灶者42例(30.0%),右肺下叶有病灶者109例(77.9%),左肺上叶有病灶者70例(50.0%),左肺下叶有病灶者97例(69.3%)。在经治疗核酸转阴后肺内仍有病变的140例患者中,左肺有病灶者108例(77.1%),右肺有病灶者121例(86.4%)。右肺的病灶体积显著大于左肺,差异有统计学意义(Z=2.359,P<0.05);核酸转阴后病灶体积、病灶体积占比及病灶平均密度均明显低于入院时相应指标,差异均有统计学意义(Z=5.889、5.603、6.352,P<0.05)。病灶体积与病灶体积占比与入院时临床分型呈正相关(P<0.05),病灶密度与入院时临床分型无关(P>0.05)。结论 COVID-19患者核酸转阴后肺内仍可有残余病灶,且病灶体积及体积占比与患者入院时的临床分型呈正相关。使用人工智能辅助诊断软件可精准提取患者肺部病灶,对核酸转阴后的COVID-19患者的后续治疗提供必要的影像学指导。

Objective To investigate the imaging manifestations of chest CT in patients with coronavirus disease 2019(COVID-19) after nucleic acid conversion and its relationship with clinical classification at admission based on artificial intelligence aided diagnosis software.Methods A retrospective analysis was performed for 529 cases of COVID-19 admitted to Beijing Ditan Hospital,Capital Medical University,from June 2020 to August 2021 who were confirmed by nucleic acid testing,and artificial intelligence pneumonia assisted diagnosis software was used to extract lesion location,lesion volume,lesion volume proportion,lesion density and other indicators from chest CT,and obtained clinical classification information at admission from the clinical cases of patients,357 cases enrolled via natriuresis.Among them,5 clinical classifications were asymptomatic carriers(27 cases,7.6%),mild carriers(80 cases,22.4%),common carriers(227 cases,63.6%),severe carriers(21 cases,5.9%),and critically ill carriers(2 cases,0.6%).The lesions in each lung lobe were statistically described.The volume of left and right lung lesions was differentiated.The volume of lesions,the proportion of lesion volume and the density of lesions at the time of admission and after nucleic acid conversion were differentiated.The correlation between the volume of lesions,the proportion of lesion volume,the density of lesions and the clinical classification at admission were analyzed.Results Among the 357 cases in this group,248 cases(69.5%) showed lesions on lung CT upon admission.After treatment and nucleic acid turned negative,140 cases(39.2%) still had lesions in both lungs.And the lesions were mainly located in the lower lobes of both lungs,including 72 cases(51.4%) with lesions in the upper lobe of the right lung,42 cases(30.0%) with lesions in the middle lobe of the right lung,109 cases(77.9%) with lesions in the lower lobe of the right lung,70 cases(50.0%) with lesions in the upper lobe of the left lung,and 97 cases(69.3%) with lesions in the lower lobe of the left lung.Among the 140 patients who still had lesions in the lungs after treatment and nucleic acid turned negative,108 cases(77.1%) had lesions in the left lung and 121 cases(86.4%) had lesions in the right lung.The volume of lesions in the right lung was significantly larger than that in the left lung,and the difference was statistically significant(Z=2.359,P<0.05);the volume of lesions,the proportion of lesion volume and the average density of lesions after nucleic acid conversion were significantly lower than the corresponding indexes at admission(Z=5.889,5.603,6.352,P<0.05).The proportion of lesion volume and lesion volume was positively correlated with clinical classification at admission(P<0.05),and lesions were not related to clinical classification at admission(P>0.05).Conclusion Patients with COVID-19 may still have residual lesions in the lungs after nucleic acid conversion negative,and the volume and volume proportion of lesions are positively correlated with the clinical classification of patients at the time of admission.The use of artificial intelligence-assisted diagnosis software can extract the characteristics of patients' lung lesions,which can provide guidance for the follow-up treatment of patients with COVID-19 after nucleic acid conversion negative.