| dc.description.abstract | Prostate cancer was in 2020 the most frequently diagnosed cancer in men in over half of the world’s countries, and the leading cause of cancer death among men in 48 countries. Medical imaging is one of the main pillars of comprehensive cancer care, where Magnetic Resonance Imaging (MRI) plays an increasingly important role in many steps of prostate cancer management.
The rapid advancements in medical technology have provided physicians with an unprecedented volume of data for making diagnoses and treatment decisions. Artificial intelligence (AI) offers the promise of rapidly analysing the vast amounts of medical images to assist radiologists. However, despite the numerous AI-based tools available in the market today, only a limited number of them are successfully integrated into clinical practice. One of the current major challenges for AI in medical image analysis is the lacking ability of a novel predictive or prognostic model to generalise to new data, which potentially can be solved by image standardisation.
An alternative to standardasing images is utilising quantitative imaging techniques, such as Magnetic Resonance Fingerprinting (MRF), as these are ideally independent of the type and set-up of the scanner. Quantitative MRF images also hold promise in differentiating healthy and cancerous tissue, as well as low and intermediate/high-grade cancers, thus aiding diagnostics. Improving spatial resolution and scan times are key obstacles to implementing MRF in routine use, where a challenge lies in reducing scan time while keeping artefacts to a minimum. In particular, the extreme undersampling factors used in radial MRF of the prostate lead to pronounced streak-like artefacts from blood flow in the femoral vessels.
The research precented in this thesis contributes to reduce the inter-scan variability of MR images of the prostate, with the long-term aim of improving the prostate cancer diagnostic pathway. An in-house developed method for standardasing qualitative 𝑇�2-weighted images is evaluated and extended. Its ability to replicate quantitative images is investigated in asymptomatic volunteers, and the standardisation performance assessed on a large, multicentre and multivendor dataset. Two alternatives to reduce or eliminate artefacts in radial MRF of the prostate are presented; through image reconstruction and pulse sequence modifications. Advances made in this thesis are expected to provide quantitative and standardised MR images of the prostate, instrumental for the development of diagnostic MRI. | en_US |
