检索历史 清除

目的:采用基于胸部CT图像的人工智能技术探讨慢性阻塞性肺疾病（COPD）患者肺血管重塑的规律。方法:该研究为横断面研究。回顾性分析2018年1月至2022年10月在山西白求恩医院接受胸部高分辨CT（HRCT）和肺功能检查（PFT）的稳定期COPD患者257例的临床及影像资料。另同期收集HRCT及PFT正常的健康人28名作为对照组。根据慢性阻塞性肺病全球倡议（GOLD）分级标准将COPD患者分为GOLD 1 级31例、GOLD 2级116例、GOLD 3级82例、GOLD 4级28例。采用FACT数字肺软件自动分割所有受检者CT图像的肺动脉及肺静脉，计算相关肺血管参数，包括肺总容积（TLV）、各级血管容积［横截面积小于5 mm 2（CSA <5）、5~10 mm 2（CSA 5~10）及大于10 mm 2（CSA >10）］、血管分支数量、血管密度（肺血管容积/TLV）等；计算所有受检者的肺气肿百分比（%LAA）及肺动脉直径/主动脉直径（PAD/AD）。多组间比较采用ANOVA或Kruskal-Wallis H检验，组内两两比较采用LSD检验或Bonferroni校正。使用Spearman相关性检验对对照组及COPD组的CT肺血管参数与肺功能指标、%LAA进行相关性分析。 结果:5组间年龄、体重指数、肺功能参数、%LAA及PAD/AD差异均有统计学意义（ P<0.001）。5组间总体肺血管密度参数差异均有统计学意义（ P<0.05）；5组间CSA <5、CSA 5~10、CSA >10的肺动脉密度参数差异均有统计学意义（ P<0.05），其中GOLD 1级CSA <5、CSA 5~10及CSA >10的肺动脉密度参数均高于对照组，而后随COPD严重程度增加呈下降趋势；5组间CSA <5、CSA 5~10、CSA >10的肺静脉密度参数差异均有统计学意义（ P<0.001），GOLD 1级患者的CSA 5~10肺静脉密度高于对照组，其余随COPD严重程度增加肺静脉密度表现为逐渐降低趋势；对照组、GOLD 1级、GOLD 2级、GOLD 3级、GOLD 4级患者的动脉、静脉血管分支数量/TLV均呈下降趋势（ P<0.001）。肺血管密度参数与PFT各参数均呈正相关（ r=0.138~0.510， P<0.05），与%LAA呈负相关（ r=-0.340~-0.671， P<0.001）；PAD/AD与PFT参数呈负相关（ r=-0.208~-0.286， P<0.001），与%LAA呈正相关（ r=0.131， P<0.05）。 结论:基于胸部CT图像的人工智能技术可以定量分析各种肺血管密度参数，能够揭示COPD患者肺血管重塑的变化规律。

Objective:To explore the pattern of pulmonary vascular remodeling in patients with chronic obstructive pulmonary disease (COPD) using artificial intelligence technology based on chest CT images.Methods:This was a cross-sectional study. The clinical and imaging data of 257 patients with stable COPD who underwent chest high resolution CT (HRCT) and pulmonary function tests (PFT) from January 2018 to October 2022 at Shanxi Bethune Hospital were retrospectively analyzed. In addition, 28 healthy individuals with normal HRCT and PFT were collected in the same period as a control group. According to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) grading criteria, COPD patients were classified into 31 cases of GOLD 1, 116 cases of GOLD 2, 82 cases of GOLD 3, and 28 cases of GOLD 4. FACT digital lung software was used to automatically segment the pulmonary arteries and pulmonary veins of all the cases, and to calculate the relevant pulmonary vascular parameters, including total lung volume (TLV), vessel volumes at all levels [cross-sectional area less than 5 mm 2 (CSA <5), between 5 and 10 mm 2 (CSA 5-10), and more than 10 mm 2 (CSA >10)], number of vascular branches, and vascular density (pulmonary vascular volume/TLV). Percentage of emphysema (%LAA) and pulmonary artery diameter/aortic diameter (PAD/AD) were calculated for all cases. ANOVA or Kruskal-Wallis H test was used for multiple intergroup comparisons, and LSD test or Bonferroni correction was used for within-group pairwise comparisons. Spearman correlation test was conducted to examine the relationship between CT pulmonary vascular parameters and pulmonary function parameters, as well as %LAA, in both the control group and the COPD group. Results:Differences in age, body mass index, pulmonary function parameters, %LAA and PAD/AD were statistically significant among the 5 groups ( P<0.001). Differences in overall pulmonary vascular density parameters were statistically significant among the 5 groups ( P<0.05). Differences in pulmonary arterial density parameters among the 5 groups with CSA <5, CSA 5-10, and CSA >10 were statistically significant ( P<0.05). The pulmonary arterial density values of GOLD 1 CSA <5, CSA 5-10 and CSA >10 were higher than those of the control group, and then showed a decreasing trend with the increase of COPD severity. The differences in pulmonary venous density parameters among the 5 groups with CSA< 5, CSA 5-10, and CSA >10 were statistically significant ( P<0.001), and the CSA 5-10 pulmonary venous density was higher in GOLD 1 patients than in the control group, and the remaining pulmonary venous densities showed a gradual decreasing trend with the increase in the severity of COPD. The number of arterial and venous vascular branches/TLV tended to decrease in the control group, GOLD 1, GOLD 2, GOLD 3, and GOLD 4 patients ( P<0.001). Pulmonary vascular density parameters were positively correlated with all PFT parameters ( r=0.138-0.510, P<0.05), and negatively correlated with %LAA ( r=-0.340--0.671, P<0.001); PAD/AD was negatively correlated with PFT parameters ( r=-0.208--0.286, P<0.001) and positively correlated with %LAA ( r=0.131, P<0.05). Conclusion:Various pulmonary vascular density parameters can be quantitatively analyzed by artificial intelligence technology based on chest CT images, which can reveal the changing pattern of pulmonary vascular remodeling in COPD patients.