Towards automation and quantification of reperfusion assessment in medical images of the brain, heart, and reconstructed gastric tube

Impaired perfusion can result in tissue necrosis, which, if it occurs in vital organs such as the brain or heart, can have fatal consequences. The continuous development of medical imaging technologies, characterized by enhanced resolution and faster imaging capabilities, enables the assessment of tissue perfusion through direct visual examination of medical images as well as indirect assessment via the analysis of contrast dynamics. Although visual perfusion assessment methods have been found to have prognostic value in clinical applications, they have several problems that need to be addressed: the predominant use of ordinal and low-resolution scales, dependence on observers, and time-consuming nature. This thesis explores multiple avenues to establish a foundation for automated and quantitative perfusion measures. We investigate the use of fluorescence imaging for intraoperative perfusion evaluation and develop a model to analyze fluorescent dynamics in gastric tube perfusion. Additionally, we evaluate an alternative method called the quantitative blush evaluator (QuBE) for myocardial perfusion assessment, highlighting the need for improvements. Furthermore, we develop a semi-automated quantitative brain perfusion assessment, the quantified TICI scale (qTICI), as an alternative to visual grading, demonstrating its association with functional outcome. Lastly, we develop a cerebral collateral score based on CT perfusion parameters (CTP-CS), showing its potential for assessing collateral capacity and predicting outcomes. The thesis concludes with a discussion on future prospects, including the integration of deep learning and hybrid systems for automated quantitative perfusion assessment.