Swift, Andy JKiely, David GAlandejani, Faisal Abdulwahab2023-06-052023-06-052022https://hdl.handle.net/20.500.14154/68276Pulmonary arterial hypertension (PAH) is a progressive, life-threatening condition characterised by pulmonary vascular resistance and associated with severe outcomes. According to the European Society of Cardiology and European Respiratory Society (ESC/ERS) guidelines, the risk assessment and prognosis of PAH are reliant on multiple investigations, including cardiac magnetic resonance imaging (MRI) and N-terminal prohormone brain natriuretic peptide (NT-proBNP). The overall aim of this body of work was to investigate the clinical benefit of cardiac MRI including artificial intelligence (AI) approaches, and NT-proBNP to assess the right atrium (RA) and right ventricle (RV) for severity assessment and prognosis in PAH. A cardiac MRI-based automated AI analysis of the RA and RV was developed, and the AI failure rate of the model was tested. Repeatability and agreement of contours of automated cardiac MRI analysis of the RA were evaluated. The prognostic significance of AI RV and AI RA area and their utility to risk stratify patients with PAH have been identified and compared with one another. The importance and relationship between automated cardiac MRI and NT-proBNP in PAH have been highlighted. The work has shown that cardiac MRI RV and RA area measurements can be fully automated using AI with a very low failure rate. The variability of AI-derived RA area measurements was lower than manual measurements in a scan-rescan cohort. Manual and automated RA area measurements moderately correlate with invasive haemodynamics. NT-proBNP showed a moderate correlation when compared to automated RV function. Measures of RV function and RA area have prognostic value; nonetheless, only measures of RV function but not ESC/ERS RA area thresholds identify patients at low-risk of 1-year mortality. Finally, the need for further work exploring larger cohorts with NT-proBNP and cardiac MRI measurements to investigate the incremental value of different approaches when assessing the right ventricle has been recommended.170enPulmonary arterial hypertensionRight atrial areaRight ventricularCardiac MRINT-proBNPClinical testingRepeatability assessmentMortality predictionRisk stratificationEuropean Society of Cardiology and European Respiratory Society guidelinesConvolutional neural networksArtificial intelligenceDeep learning trainingThesis