This research project aims to develop and validate novel techniques utilizing radiomics in repeatable and reproducible manner, and subsequent machine learning techniques for non-invasive characterization of tissue microstructure using magnetic resonance imaging (MRI). Usefulness of the developed methodologies are demonstrated in practice in two fields of application: Prostate cancer and neuroscience. We develop novel machine learning methods specifically tailored for fast prostate MRI acquisition, and evaluate conventional and novel MRI sequences. In brain, we evaluate repeatability together with potential biases due to MRI artefacts to find optimal MRI acquisition and post-processing protocols, subject drop-out rate, and accuracy/precision for discovery of novel neurodevelopmental findings.

2021

2024