Exploring the applicability of machine learning based artificial intelligence in the analysis of cardiovascular imaging

Worldwide, the prevalence of cardiovascular diseases has doubled, demanding new diagnostic tools. Artificial intelligence, especially machine learning and deep learning, offers innovative possibilities for medical research. Despite historical challenges, such as a lack of data, these techniques have potential for cardiovascular research. This thesis explores the application of machine learning and deep learning in cardiology, focusing on automation and decision support in cardiovascular imaging.
Part I of this thesis focuses on automating cardiovascular MRI analysis. A deep learning model was developed to analyze the ascending aorta in cardiovascular MRI images. The model's results were used to investigate connections between genetic material and aortic properties, and between aortic properties and cardiovascular diseases and mortality. A second model was developed to select MRI images suitable for analyzing the pulmonary artery.
Part II focuses on decision support in nuclear cardiovascular imaging. A first machine learning model was developed to predict myocardial ischemia based on CTA variables. In addition, a deep neural network was used to identify reduced oxygen supply through the arteries supplying oxygen-rich blood to the heart and cardiovascular risk features using PET images.
This thesis successfully explores the possibilities of machine learning and deep learning in cardiovascular research, with a focus on automated analysis and decision support.