Artificial intelligence-driven quantitative analysis of CT morphological differences between chronic thromboembolic pulmonary hypertension and chronic thromboembolic disease.

Abstract

Background: The morphological differences in the pulmonary vascular tree between chronic thromboembolic pulmonary hypertension (CTEPH) and chronic thromboembolic disease (CTED) are not yet fully understood. This study aimed to use artificial intelligence (AI) segmentation technology to identify morphological markers that can be used to differentiate CTEPH from CTED using computed tomography pulmonary angiography (CTPA).

Methods: We conducted a retrospective cohort study with consecutive patients diagnosed with CTEPH, CTED, and control subjects at the China-Japan Friendship Hospital from January 2019 to October 2023. The study involved the automatic quantification of the pulmonary blood volume (BV), tortuosity, and fractal dimension (FD) from CTPA images using an AI workstation. These morphological metrics were compared among the three groups using the Kruskal-Wallis test. Correlations between these metrics and the hemodynamic parameters were evaluated using Spearman's rank correlation coefficients. Additionally, a receiver operating characteristic (ROC) curve analysis was conducted to assess the discriminative ability of pulmonary artery tortuosity to differentiate between each pair of groups.

Results: A total of 190 participants [57 years, interquartile range (IQR), 49-65 years, 97 men], including 116 CTEPH patients, 54 CTED patients, and 20 controls, were enrolled in this study. Pulmonary artery tortuosity in the control, CTED, and CTEPH groups showed a progressively increasing trend [1.07 (IQR, 1.06-1.10) vs. 1.10 (IQR, 1.07-1.14) vs. 1.14 (IQR, 1.10-1.18), P<0.01]. The area under the curve (AUC) values of pulmonary arterial tortuosity for differentiating between the CTEPH patients and controls, CTED patients and controls, and CTEPH patients and CTED patients were 0.859, 0.712, and 0.663, respectively. There was a positive correlation between pulmonary artery tortuosity and mean pulmonary arterial pressure (mPAP) (r=0.44, P<0.01), and pulmonary vascular resistance (PVR) (r=0.47, P<0.01). Additionally, the volume of the small- and medium-sized pulmonary arteries was significantly higher in the CTED patients than the CTEPH patients (P<0.01). The pulmonary arterial FD among the three groups was comparable (P=0.36).

Conclusions: Pulmonary arterial tortuosity on CTPA had auxiliary diagnostic value in differentiating between CTEPH patients and controls, but its value in differentiating between CTED and CTEPH patients requires further study. The reduced volume of small- and medium-sized pulmonary arteries in CTEPH patients could indicate impaired pulmonary hemodynamics.