Four-dimensional magnetic resonance flow imaging (4D flow MRI) enables comprehensive time-resolved three-dimensional measurement of blood velocity, permitting retrospective flow visualization, quantification, and advanced hemodynamic analysis throughout the heart and great vessels. De-spite its considerable clinical and research potential, broader adoption of cardiovascular 4D flow MRI remains limited by technical challenges, including long scan times, reduced spatial and temporal resolution, phase offset errors, large data volumes, and time-consuming image processing workflows.

The overall aim of this thesis was to improve data quality, image processing, and hemodynamic analyses in cardiovascular 4D flow MRI, with the goal of facilitating robust and clinically applicable workflows.

The first part of this thesis focused on improving data quality. Clinically suit-able 4D flow MRI acquisitions were identified and compared, and demonstrated the importance of higher-order polynomial background phase offset correction for accurate flow quantification and improved data consistency. To address the limitations of polynomial fitting in regions with insufficient static tissue, particularly around the heart, a fully automatic deep learning-based method for background phase offset correction was developed, using static phantom measurements for training.

The subsequent part addressed the challenge of large-scale data processing through the development of a deep learning-based method for time-resolved cardiovascular segmentation of 4D flow MRI data. Compared with atlas-based approaches, this method substantially reduced computational burden and processing time, thereby facilitating efficient analyses in large cohorts.

Building on these methodological developments, the final part of this thesis expanded the clinical applicability of 4D flow MRI through novel hemodynamic analyses. An automated framework for the assessment of diastolic dysfunction was introduced, demonstrating the potential of 4D flow MRI for streamlined functional evaluation. In addition, a hemodynamic atlas-based framework for voxel-wise comparison of disease groups was proposed, over-coming challenges related to anatomical variability and enabling direct regional analyses across subjects.

Taken together, this thesis advances the reliability, automation, and analytical scope of cardiovascular 4D flow MRI. By improving data quality, simplifying image processing, and introducing new frameworks for clinically relevant hemodynamic assessments, the presented work contributes toward more robust workflows and supports the broader translational use of 4D flow MRI in cardiovascular disease.