| 陆海琴,冯学梅,周嘉鑫,段丽瑞,朱力.基于人工智能CT定量分析观察慢性阻塞性肺疾病小气道结构的性别差异[J].中国医学影像技术,2025,41(12):2006~2010 |
| 基于人工智能CT定量分析观察慢性阻塞性肺疾病小气道结构的性别差异 |
| Artificial intelligence based CT quantitative analysis of gender differences of small airway structures in chronic obstructive pulmonary disease |
| 投稿时间：2025-04-08 修订日期：2025-10-20 |
| DOI：10.13929/j.issn.1003-3289.2025.12.015 |
| 中文关键词: 肺疾病,慢性阻塞性 体层摄影术,X线计算机 性别 |
| 英文关键词:pulmonary disease, chronic obstructive tomography, X-ray computed sexuality |
| 基金项目:宁夏回族自治区中央引导地方科技发展资金项目(2023FRD05010)。 |
|
| 摘要点击次数: 91 |
| 全文下载次数: 20 |
| 中文摘要: |
| 目的 基于人工智能CT定量分析观察慢性阻塞性肺疾病(COPD)小气道结构的性别差异。方法 依据GOLD分级和性别对633例稳定期COPD患者及445名健康对照进行分层匹配抽样,设立正常组和GOLD Ⅰ、Ⅱ、Ⅲ+Ⅳ级组(各含男、女性亚组各30人),共纳入240人。以AVIEW分析系统对胸部高分辨率CT进行自动测量,提取全肺特征、血管及气道特征;以随机森林法筛选关键CT特征,并以偏最小二乘回归分析特征与肺功能指标的相关性,选出复相关系数(R)＞0.6者;采用双因素方差分析评估性别与疾病分级的交互作用。结果 4组内男、女性亚组间年龄、体质量指数差异均无统计学意义(P均＞0.05)。随机森林分析选出的关键CT特征包括全肺容积、肺气肿指数(LAA%)、第6~8级气道腔体积及第5~9级气道壁厚度(WA%),其中7项特征与肺功能指标的R＞0.6,且在不同GOLD分级组间差异均有统计学意义(F=11.350~55.680,P均＜0.01)。性别为全肺容积和第6级气道腔体积的独立影响因素(F=77.100、32.100,P均＜0.01),男性上述参数值均高于女性(Cohens d值＞0);性别×分组间均无明显交互作用(P均＞0.05)。效应量分析显示,正常组性别差异最为显著,全肺容积和第6级气道腔体积均为大效应量(Cohens d值分别为1.428和0.828);GOLD Ⅲ+Ⅳ级组性别差异减弱,但全肺容积仍为大效应量(Cohens d值为0.884),女性第6~8级气道WA%多大于男性(Cohens d值＜0)。结论 正常人及不同严重程度COPD患者胸部CT定量特征均存在显著性别差异:男性全肺容积与大气道管腔均更大,而女性小气道壁均更厚。 |
| 英文摘要: |
| Objective To observe gender differences of small airway structures in chronic obstructive pulmonary disease (COPD) using artificial intelligence (AI) based CT quantitative analysis. Methods Totally 633 stable COPD patients and 445 healthy controls were retrospectively enrolled. Normal group and GOLD grade Ⅰ, Ⅱ, Ⅲ+Ⅳ groups were established using stratified matched sampling, each contained 30 male (male subgroup) and 30 female (female subgroup). AVIEW analysis system was used to automatically measure and extract lung features in high-resolution CT images, including vascular features and airway features. Random forest method was used to screen key CT features, and partial least squares regression was used to analyze the correlations between features and lung function indicators. Features with a complex correlation coefficient (R) greater than 0.6 were selected, and two factor analysis of variance was performed to evaluate the interaction between gender and disease grading. Results No significant difference of age nor body mass index was found between subgroups in all 4 groups (all P＞0.05). The key CT features screened by random forest method included total lung capacity, emphysema index (LAA%), lumen volume of airways at grades 6—8, and wall area percentage (WA%) of airways at grades 5—9. Partial least squares regression showed R＞0.6 between 7 features and lung function indicators, and significant differences of these features were found among GOLD groups (F=11.350—55.680, all P＜0.01). Gender was the independent impact factor of total lung volume and airway lumen volume at grade 6 (F=77.100, 32.100, both P＜0.01), and the latter two were higher in males than in females (Cohen's d value＞0), while no significant gender×group interaction was observed (all P＞0.05). Effect size analysis indicated the most pronounced gender differences in normal group, with large effect sizes for total lung volume and airway lumen volume at grade 6 (Cohen's d values of 1.428 and 0.828, respectively). In GOLD Ⅲ+Ⅳ group, gender differences diminished but total lung volume remained a large effect size (Cohen's d value=0.884), and WA% of airways at grades 6—8 in females were higher than in males (Cohen's d value＜0). Conclusion Significant gender differences of quantitative characteristics existed in chest CT in normal subjects and COPD patients with different severities. Men had larger total lung volume and large airway lumen, while women had thicker small airway walls. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
