Work

Imaging Biomarkers for Pulmonary Hypertension: AI‐Based Vascular Analysis Through Non‐Contrast Chest Computed Tomography Scans

Year: 2026
Type: article
Study design: Observational Study
Abstract: ABSTRACT Pulmonary Hypertension (PH) is a progressive, debilitating condition characterized by elevated pressures in the pulmonary vasculature. Accurate diagnosis and appropriate classification are essential for effective management of this disorder. This study evaluated artificial intelligence (AI)‐based artery‐vein phenotyping analysis (AVX) of non‐contrast chest computed tomography (CT) scans as a means to improve diagnostic precision. A retrospective cohort study was conducted that enrolled patients who underwent non‐contrast chest CT and right heart catheterization within a 6‐month period. AVX provided segmentation and quantification of pulmonary arterial and venous trees, enabling an examination of vascular volume distribution. Radiological biomarkers were examined in the context of both PH classification and hemodynamic profiles. Significant differences in pulmonary vascular volumes were observed between three of the PH diagnostic groups (Group 1, Group 2, and Group 3, N = 163 patients). Patients stratified by hemodynamic profiles (pre‐, isolated post‐, and combined pre‐ and post‐capillary) also showed distinct patterns of arterial and venous volumes. Several CT‐derived vascular biomarkers were correlated with invasive hemodynamic measurements. Collectively, our findings suggest that the use of AI‐based artery‐vein quantification tools may provide complementary information for characterizing pulmonary vascular diseases. The results of principal component analysis highlight the value of this comprehensive approach. However, given the study's monocentric nature, the omission of PH Groups 4 and 5, and the lack of a healthy control cohort, larger‐scale validation studies will be needed to confirm these findings.
Language: English
Field: Medicine
Open Access status: gold
Estimated APC paid: $2,310
OpenAlex ID: W7225883158
FWCI:
Cites: 42
Cited by:
Related to: 10