第一部分 基于健康成人胸部平扫CT图像自动化三维定量分析肺内血管体积

目的：基于健康受试者的胸部平扫CT图像，在肺窗及纵隔窗上自动分割并定量测量肺内血管体积（intrapulmonary vascular volume, IPVV）并比较其差异。

方法：本研究回顾性收集了2023年1月至2023年11月期间，我院常规体检时胸部平扫CT结果正常的258例受试者，其中男性116例（45%），女性142例（55%），年龄范围21-83岁。每位受试者的肺窗及纵隔窗的影像数据采用自动化算法测量IPVV，包括总肺内血管体积（total intrapulmonary vascular volume, TIPVV）、肺内动脉血管体积（intrapulmonary arterial vascular volume, IPVVa）、肺内静脉血管体积（intrapulmonary venous vascular volume, IPVVv）及五组不同血管直径（0.8 mm<血管直径≤1.6 mm组，1.6 mm<血管直径≤2.4 mm组，2.4 mm<血管直径≤3.2 mm组，3.2 mm<血管直径≤4.0 mm组，血管直径>4.0 mm组）的IPVV。比较肺窗及纵隔窗测量的IPVV的差异。

结果：基于肺窗测量的TIPVV、IPVVa、IPVVv分别为120.42 (105.19, 138.08) mL·m^-2、51.28 (46.48, 57.54) mL·m^-2、69.08 (57.74, 81.12) mL·m^-2，基于纵隔窗测量的TIPVV、IPVVa、IPVVv分别为102.33 (92.82, 112.23) mL·m^-2、44.67 (40.83, 49.46) mL·m^-2、57.10 (51.25, 64.28) mL·m^-2。基于肺窗测量的TIPVV、IPVVa、IPVVv均显著大于基于纵隔窗测量的结果，两组间具有显著的统计学差异（P均<0.001）。基于肺窗测量的不同血管直径的肺小血管的肺内动脉血管体积及肺内静脉血管体积亦均显著大于基于纵隔窗测量的结果（P均<0.01）。在血管直径0.8 mm-1.6 mm组肺动脉和肺静脉的IPVV、以及血管直径2.4 mm-3.2 mm组肺静脉的IPVV中，男性与女性间没有统计学差异（P>0.05）；在血管直径1.6 mm-2.4 mm组的肺动脉和肺静脉中，女性的IPVV明显大于男性（P均<0.01）；在其余血管直径组中，男性的IPVV则大于女性（P均<0.05）。

结论：计算机辅助肺血管分割算法能够有效识别、分割胸部平扫CT的肺内血管，并能够自动测量IPVV。基于胸部平扫CT肺窗测量的IPVV均大于基于纵隔窗测量的IPVV。计算机辅助肺血管分割算法有望在胸部平扫CT上分割异常肺血管，使在胸部平扫CT上识别肺血管异常成为可能。

第二部分 肺血管受累血管炎患者的肺内血管体积的自动化三维定量分析

目的：比较计算机断层扫描肺动脉造影（computed tomography pulmonary angiography, CTPA）显示肺血管受累血管炎患者与CTPA阴性受试者测量的肺内血管体积（intrapulmonary vascular volume, IPVV）的差异。

方法：本研究纳入了2019年3月至2024年11月期间207例我院诊断为血管炎且CTPA显示肺血管受累血管炎患者的影像数据以及2019年2月至2025年2月期间于我院CTPA检查为阴性的202名受试者的影像数据。采用计算机辅助肺血管分割算法测量每位肺血管受累血管炎患者及CTPA阴性受试者的IPVV，定量评估包括总肺内血管体积（total intrapulmonary vascular volume, TIPVV）、肺内动脉血管体积（intrapulmonary arterial vascular volume, IPVVa）、肺内静脉血管体积（intrapulmonary venous vascular volume, IPVVv），以及五组不同血管直径（0.8 mm<血管直径≤1.6 mm组，1.6 mm<血管直径≤2.4 mm组，2.4 mm<血管直径≤3.2 mm组，3.2 mm<血管直径≤4.0 mm组，血管直径>4.0 mm组）的IPVV；并对CTPA阴性受试者与肺血管受累血管炎患者的整个肺部、肺内动脉、肺内静脉及不同血管直径分组的IPVV进行了比较。

结果：肺血管受累血管炎患者及CTPA阴性的受试者间的TIPVV及IPVVv没有明显的统计学差异（P>0.05）。CTPA阴性受试者测量的IPVVa为47.79 (42.48, 54.10) mL·m^-2，肺血管受累血管炎患者测量的IPVVa为44.86 (39.33, 52.58) mL·m^-2，肺血管受累血管炎患者的IPVVa明显小于CTPA阴性受试者，且具有显著的统计学差异（P<0.01）。在血管直径0.8 mm–1.6 mm组中，肺血管受累血管炎患者的IPVVa为4.71 (3.93, 5.75) mL·m^-2，CTPA阴性受试者的IPVVa为5.14 (4.15, 6.03) mL·m^-2；在血管直径2.4 mm–3.2 mm组中，肺血管受累血管炎患者的IPVVa和CTPA阴性受试者的IPVVa分别为11.61 (10.18, 14.03) mL·m^-2、12.93 (11.00, 15.03) mL·m^-2。在血管直径0.8 mm–1.6 mm组和2.4 mm–3.2 mm组肺动脉中，肺血管受累血管炎患者的IPVVa小于CTPA阴性受试者，且具有统计学差异（P<0.05）。在血管直径1.6 mm–2.4 mm组中，肺血管受累血管炎患者的IPVVv为9.94 (8.08, 11.74) mL·m^-2，CTPA阴性受试者的IPVVv为9.08 (7.58, 11.75) mL·m^-2；在血管直径3.2 mm–4.0 mm组中，肺血管受累血管炎患者的IPVVv和CTPA阴性受试者的IPVVv分别为10.16 (8.07, 12.27) mL·m^-2、8.85 (7.34, 10.89) mL·m^-2。在血管直径1.6 mm–2.4 mm组和3.2 mm–4.0 mm组肺静脉中，肺血管受累血管炎患者的IPVVv大于CTPA阴性的受试者，且具有统计学差异（P<0.05）。

结论：计算机辅助肺血管分割算法自动测量IPVV可量化评价血管炎累及肺小血管的情况，肺血管受累血管炎患者的IPVVa明显小于CTPA阴性受试者。计算机辅助肺血管分割算法自动测量IPVV可用于肺血管受累血管炎患者的量化管理。

Part 1. Automated 3D Quantitative Analysis of Intrapulmonary Vascular Volume on Noncontrast CT in Healthy Individuals

Objective: To compare the differences in automated segmentation and quantitative measurements of intrapulmonary vascular volume (IPVV) between lung and mediastinal windows in healthy individuals based on chest computed tomography (CT) plain scans.

Materials and methods: In total, 258 individuals (aged 21–83 years) who underwent routine physical examinations and had normal chest CT scans from January to November 2023 were retrospectively enrolled. Among them, 116 were male (45%) and 142 were female (55%). For each healthy individual, a pulmonary vascular segmentation algorithm was employed to automatically extract the total intrapulmonary vascular volume (TIPVV), intrapulmonary arterial vascular volume (IPVVa), intrapulmonary venous vascular volume (IPVVv) based on the lung window and mediastinal window. Additionally, IPVVs were analyzed within five specific vessel diameter groups: 0.8 mm -1.6 mm, 1.6 mm-2.4 mm, 2.4 mm-3.2 mm, 3.2 mm-4.0 mm, and >4.0 mm. The differences in IPVVs extracted were then compared based on the lung window versus those extracted based on the mediastinal window.

Results: The TIPVV, IPVVa, IPVVv measured from the lung window were 120.42 (105.19, 138.08) mL·m^-2, 51.28 (46.48, 57.54) mL·m^-2 and 69.08 (57.74, 81.12) mL·m^-2, respectively, while those measured from the mediastinal window were 102.33 (92.82, 112.23) mL·m^-2, 44.67 (40.83, 49.46) mL·m^-2 and 57.10 (51.25, 64.28) mL·m^-2, respectively. The TIPVV, IPVVa, IPVVv measured from the lung window were significantly higher than those from the mediastinal window, with statistically significant differences (all P<0.001). Similarly, the IPVVa and IPVVv of small pulmonary vessels categorized by different diameters extracted from the lung window were significantly higher than those extracted from the mediastinal window (all P<0.01). No significant sex-based differences in IPVV were observed for pulmonary arteries and veins with diameters between 0.8 mm and 1.6 mm, as well as pulmonary veins with diameters between 2.4 mm and 3.2 mm (P>0.05). However, for pulmonary arteries and veins with diameters between 1.6 mm and 2.4 mm, females had significantly higher IPVVs than males (all P<0.01). In all other cases, IPVVs were larger in males than in females (all P<0.05).

Conclusion: The computer-aided pulmonary vascular segmentation algorithm can effectively identify and segment intrapulmonary vessels in chest CT plain scans and automatically measure IPVV. The IPVV measured based on the lung window is higher than that measured based on the mediastinal window. The computer-aided pulmonary vascular segmentation algorithm has the potential to segment abnormal pulmonary vasculature on chest CT plain scans, making it possible to identify pulmonary vascular abnormalities on chest CT plain scans. 

Part 2. Automated 3D Quantitative Analysis of Intrapulmonary Vascular Volume in Vasculitis Patients with Pulmonary Vascular Involvement

Objective: To compare the differences in intrapulmonary vascular volume (IPVV) measured using computed tomography pulmonary angiography (CTPA) between vasculitis patients with pulmonary vascular involvement and subjects with negative CTPA findings.

Materials and methods: This study included imaging data from 207 patients diagnosed with vasculitis with pulmonary vascular involvement based on CTPA between March 2019 and November 2024, as well as 202 subjects who underwent CTPA between February 2019 and February 2025 and had negative findings. A computer-aided pulmonary vascular segmentation algorithm was employed to automatically measure the IPVV for each patient and subject. Quantitative assessments were conducted on the total intrapulmonary vascular volume (TIPVV), intrapulmonary arterial vascular volume (IPVVa), intrapulmonary venous vascular volume (IPVVv). Additionally, IPVVs were analyzed within five specific vessel diameter groups: 0.8 mm-1.6 mm, 1.6 mm-2.4 mm, 2.4 mm-3.2 mm, 3.2 mm-4.0 mm, and >4.0 mm. Comparisons of IPVVs were performed between CTPA-negative subjects and vasculitis patients with pulmonary vascular involvement for the entire lung, intrapulmonary arteries, intrapulmonary veins and different vessel diameter groups .  

Results: TIPVV and IPVVv showed no significant differences between CTPA-negative subjects and vasculitis patients with pulmonary vascular involvement (P>0.05). The IPVVa measured in CTPA-negative subjects was 47.79 (42.48, 54.10) mL·m^-2, while that in vasculitis patients with pulmonary vascular involvement was 44.86 (39.33, 52.58) mL·m^-2. The IPVVa in vasculitis patients with pulmonary vascular involvement was significantly lower than that in CTPA-negative subjects, with a statistically significant difference (P<0.01). In pulmonary arteries with diameters of 0.8 mm-1.6 mm, the IPVVa measured in vasculitis patients with pulmonary vascular involvement was 4.71 (3.93, 5.75) mL·m^-2, while that in CTPA-negative subjects was 5.14 (4.15, 6.03) mL·m^-2. In pulmonary arteries with diameters of 2.4 mm-3.2 mm, the IPVVa measured in vasculitis patients with pulmonary vascular involvement and CTPA-negative subjects were 11.61 (10.18, 14.03) mL·m^-2 and 12.93 (11.00, 15.03) mL·m^-2. In pulmonary arteries with diameters of 0.8 mm-1.6 mm and 2.4 mm-3.2 mm, the IPVVa in vasculitis patients with pulmonary vascular involvement was lower than that in CTPA-negative subjects, with a statistically significant difference (P<0.05). In pulmonary veins with diameters of 1.6 mm-2.4 mm, the IPVVv measured in vasculitis patients with pulmonary vascular involvement was 9.94 (8.08, 11.74) mL·m^-2, while that in CTPA-negative subjects was 9.08 (7.58, 11.75) mL·m^-2. In pulmonary veins with diameters of 3.2 mm-4.0 mm, the IPVVv measured in vasculitis patients with pulmonary vascular involvement and CTPA-negative subjects were 10.16 (8.07, 12.27) mL·m^-2 and 8.85 (7.34, 10.89) mL·m^-2. In pulmonary veins with diameters of 1.6 mm-2.4 mm and 3.2 mm-4.0 mm, the IPVVv in vasculitis patients with pulmonary vascular involvement was higher than that in CTPA-negative subjects, with a statistically significant difference (P<0.05).

Conclusion: The computer-aided pulmonary vascular segmentation algorithm can automatically measure IPVV, providing a quantitative evaluation of the condition of small pulmonary vessels affected by vasculitis. The IPVVa in vasculitis patients with pulmonary vascular involvement was significantly lower than that in CTPA-negative subjects. The automated measurement of IPVV using this algorithm can be utilized for the quantitative management of vasculitis patients with pulmonary vascular involvement.