Abstract
<b>Introduction:</b> CMR measurements are powerful prognostic markers in PAH. We developed a deep learning model to automatically generate biventricular contours in CMR and validated its result in a prospective cohort of PAH patients. <b>Methods:</b> The CMR contouring model was developed in a retrospective cohort of 400 patients with suspected pulmonary hypertension. A prospective validation cohort of 66 PAH patients was recruited to compare automatic and manual CMR measurements. Agreement in CMR results was assessed using intraclass correlation coefficient (ICC) and Bland-Altman plots. Additionally, 33 patients had a right heart catheter (RHC) and CMR within 24 hours to compare RHC-thermodilution and CMR derived stroke volume (SV). <b>Results:</b> Automatic and manual values showed excellent agreement for all prognostic CMR measurements. The ICC for right ventricular (RV) end-systolic volume was 0.94, 95% CI [0.88, 0.97], for RV ejection fraction 0.71 [0.24, 0.86] and for left ventricular (LV) end-diastolic volume 0.91 [0.72, 0.96]. Bland-Altman plots showed strong agreement between manual and automatic measurements. Both automatic and manual CMR measurement correlated strongly with RHC derived SV; 0.70 [0.44, 0.85] and 0.72 [0.49, 0.87], respectively (figure 1). <b>Conclusion:</b> In a prospective PAH cohort, fully automatic CMR assessments were reliable and corresponded accurately to invasive hemodynamics
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call