人工智能定量分析在CTD-ILD诊断和分级中的应用研究  

Artificial intelligence quantitative evaluation in diagnosis and grading of CTD-ILD

作　　者：储晨 方宇 孙艺 王春[3] 隗英 石峰 辛小燕[1] 赵盛楠[3] CHU Chen;FANG Yu;SUN Yi;WANG Chun;WEI Ying;SHI Feng;XIN Xiaoyan;ZHAO Shengnan(Department of Radiology,Nanjing Drum Tower Hospital,The Affiliated Hospital of Nanjing University,Nanjing 210008,China;Department of Radiology,Nanjing Drum Tower Hospital Clinical College of Nanjing University of Chinese Medicine,Nanjing 210008,China;Department of Rheumatology,Nanjing Drum Tower Hospital,The Affiliated Hospital of Nanjing University,Nanjing 210008,China;Department of Research and Development,Shanghai United Imaging Intelligence Co.,Ltd.,701 Yunjin Rd,Xuhui District,Shanghai 200232,China)

机构地区：[1]南京大学医学院附属鼓楼医院医学影像科,江苏南京210008 [2]南京中医药大学鼓楼临床医学院医学影像科,江苏南京210008 [3]南京大学医学院附属鼓楼医院风湿免疫科,江苏南京210008 [4]上海联影智能医疗科技有限公司研发部,上海200232

出　　处：《医学影像学杂志》2024年第8期54-57,61,共5页Journal of Medical Imaging

基　　金：国家自然科学基金资助项目(编号:81871282);江苏省“双创博士”人才引进计划项目(编号:JSSCBS20211495)。

摘　　要：目的 探讨人工智能定量分析在结缔组织病相关间质性肺疾病(CTD-ILD)诊断和分级中的应用。方法 选取CTD-ILD 128例患者,所有患者均进行CT扫描和肺功能检查,分为轻度组和重度组。独立样本t检验和ROC分析鉴别轻度组和重度组。方差分析和LSD检验鉴别各组病变成分。Spearman秩和检验比较各参数与肺功能等级的相关性。结果 重度组肺体积均显著小于轻度组;重度组肺病变体积和百分比均显著大于轻度组(P≤0.001)。ROC曲线显示肺病变体积和百分比指标具有较高诊断价值(AUC值均>0.700)。各组病变成分之间均具有显著性差异(P<0.001)。各参数与肺功能等级均具有相关性。结论 人工智能在CTD-ILD患者的定量分析中具有一定的优势,对患者的诊断和分级能够提供价值。Objective To explore the value of artificial intelligence in quantitative of connective tissue disease-associated in-terstitial lung disease(CTD-ILD).Methods A total of 128 CTD-ILD patients were prospectively enrolled and divided into mild and severe groups.The independent sample t test and ROC analysis were used to distinguish the parameters of pneumonia analysis in mild and severe groups.The analysis of variance and LSD test were used to compare the lesion components.The spear-man rank analysis was used to compare the correlation between all parameters and pulmonary function grades.Results The lung volume of severe groups was significantly higher than mild groups(P value were≤0.001).The ROC curves showed that the volume and percentage indexes of the lung were of high diagnostic value(AUCs>0.700).There were significantly differences be-tween different lesion components.The parameters were correlated with the pulmonary function grades.Conclusion Artificial intelligence has advantages in the quantitative analysis of CTD-ILD patients and can provide value for the diagnosis and grading of patients.

关 键 词：间质性肺疾病 结缔组织病 人工智能 体层摄影术 X线计算机 

分 类 号：R593.2[医药卫生—内科学] R814.42[医药卫生—临床医学]