Logo: to the web site of Uppsala University

Whole-body positron emission tomography/computed tomography (PET/CT) imaging is primarily used for cancer patients and provides comprehensive information about disease extent and overall patient health. However, the manual analysis and interpretation of these images is often time-consuming and complex, necessitating computer-aided methods to maximize data utilization. This thesis presents the development, evaluation, and investigation of applications of an image registration method aimed at advancing visualization and analysis of standard-of-care whole-body PET/CT images. Paper I describes an image registration method that aligns whole-body PET/CT images by incorporating both low-level features from the CT scan and high-level features from tissue segmentation masks. This method effectively aligned major organs and anatomical structures within reasonable computation times for both within- and between-subject registration, as demonstrated in lymphoma and head and neck cancer patients. Paper II applied this image registration technique to track lesions in metastatic breast cancer patients by comparing pre- and post-treatment images. High performance in lesion tracking confirmed the method’s potential utility for automated treatment response evaluation. Paper III investigated longitudinal image features extracted through within-subject registration of whole-body [¹⁸F]fluorodeoxyglucose PET/CT images in patients with locally advanced non-small cell lung cancer. These features were shown to have prognostic value, with whole-body image analysis providing improved survival prediction compared to chest-focused analysis in certain cases. Paper IV explored spatial tumor heterogeneity across patients with lung cancer, lymphoma, and melanoma, and showed the feasibility of using whole-body group-wise analysis to visualize tumor characteristics. These results collectively demonstrate the utility of whole-body PET/CT image registration-based analysis across various cancer types. The developed image registration method enables detailed comparative analysis of whole-body PET/CT images for longitudinal studies and cohort-based characterization of disease patterns. Its versatility in cancer image analysis suggests that its application has promising potential to improve cancer care, including staging, treatment planning, and patient outcomes.