AI-derived pulmonary artery and vein volumes from CTPA increased significantly with height, weight, body mass index, and body surface area (all p < 0.001), with distinct sex and age interactions.
Observational (n=376)
Yes
Accounting for anthropometrics, sex, and age at diagnosis is essential for accurately translating AI-quantified pulmonary blood volumes from CTPA into predictive models for pulmonary hypertension.
p-value: p=<0.001
Introduction The influence of anthropometrics, sex, and age at diagnosis on artificial intelligence (AI)–derived pulmonary blood volumes (PBV) from computed tomography pulmonary angiography (CTPA) remains poorly characterized. These physiological and biological determinants may affect PBV-based pulmonary hypertension (PH) prediction models. Methods An AI-based segmentation model quantified pulmonary artery and vein volumes in a secondary CTPA analysis from the Cambridge PH Registry. PBV were modelled as functions of anthropometrics, sex, and age at diagnosis, adjusting for pulmonary vascular resistance (PVR), PH diagnostic category (group 1, 2, 3 or 4 PH, or no PH), and number of cardiac comorbidities. The impact of PBV normalization strategies, including anthropometrics, on sex-related differences was assessed. Multivariable linear regression evaluated associations between PBV and invasively measured PVR and cardiac output (CO), and the incremental predictive value of anthropometrics. Results 376 patients (median age 60 years; 57% female) investigated with right heart catheter were included: 120 pulmonary arterial hypertension, 30 group 2 PH, 79 group 3 PH, 102 chronic thromboembolic PH, and 45 without PH. Pulmonary artery volume increased with height, weight, body mass index (BMI), and body surface area (BSA) (all p 0.001), with a stronger height-related effect in males. Older diagnosis age associated with larger pulmonary artery volume, particularly those with higher weight, BMI, and BSA (all p 0.001). Pulmonary vein volume also increased with anthropometrics, but with older age disproportionately increasing volumes in females with higher weight, BMI, and BSA. PBV normalization to anthropometrics (height, weight, and BSA) attenuated sex-related differences ( p 0.001). In hemodynamic models, larger pulmonary artery volume and lower pulmonary vein volume were independently associated with higher PVR, while larger pulmonary vein volume was associated with higher CO (all p 0.001). Inclusion of anthropometrics, particularly BSA, significantly improved prediction of PVR ( F = 12.8, p 0.001) and CO (F = 23.8, p 0.001). Conclusion AI-derived PBV from CTPA are shaped by complex interactions between anthropometrics, sex, and diagnosis age, with distinct effects on pulmonary arterial and venous compartments. Accounting for these determinants is essential for translating AI-quantified PBV into clinically intuitive PH prediction or phenotyping models.
Ghani et al. (Mon,) conducted a observational in Pulmonary hypertension (n=376). Anthropometrics, sex, and age at diagnosis was evaluated on Pulmonary artery and vein volumes quantified by AI from CTPA (p=<0.001). AI-derived pulmonary artery and vein volumes from CTPA increased significantly with height, weight, body mass index, and body surface area (all p < 0.001), with distinct sex and age interactions.