Umeå University's logo

Background: Medical imaging is increasingly used to inform on the clinical decision-making in prostate cancer (PCa). However, the ways in which tumour pathology is reflected in imaging remains poorly understood. The aim of this thesis was to provide insights into the associations between image characteristics and histopathological features that can be leveraged for improving radiotherapy. 

Methods: A pipeline of registration algorithms were developed to align a gold standard histopathological reference and in vivo imaging. We investigated the ability of image summary measures to discriminate between histological grades of PCa, and examined how detectability varied across lesions characterized by grades and by combined markers of cellular proliferation and differentiation. Finally, we conducted an in silico evaluation of a radiotherapy treatment protocol, based on the ongoing HYPO-RT-PC-boost phase II trial (NCT06220435). 

Results: The registration pipeline provided the means to investigate associations between imaging characteristics and histopathological features. We demonstrated that image measures derived from in vivo imaging can distinguish between lower- and higher-grade PCa, using partially discriminative cut-off values. Further, we showed that many detected lesions were both high-grade and had a higher-risk profile, characterized by high proliferation and low differentiation. Undetected lesions were more often lower-grade, but did not predominantly exhibit the low-risk combination of low proliferation and high differentiation. Furthermore, we showed that image summary measures can distinguish between higher- and lower-risk lesions, suggesting further prognostic potential of the imaging modalities. By incorporating multiple observer delineations of the visible tumour, the results of the in silico evaluation indicate that radiation oncologists delineate different tumour volumes, but they could all still obtain good coverage in sites containing more aggressive disease. These results provide a rationale for prioritizing sensitive structures during treatment planning.

Conclusion: Medical imaging modalities hold untapped potential to inform the clinical decision making. However, the inability to identify tumour on medical imaging does not necessarily translate to inadequate dose coverage in radiotherapy. The inherent complexity of generating large-scale datasets with co-registered imaging and histopathology limits generalizability and underscores the importance of interstudy harmonization in imaging protocols and histopathological evaluation.