Non-invasive characterisation of tissue microstructure from MRI using Deep Learning: applications to brain cancer

点击翻译按钮获取中文摘要

Non-invasive and in vivo characterisation of brain tissue microstructures is of utmost importance to medicine, neuroscience, and basic biological research. If available, it would allow us to study not only healthy tissue development but also disease stage and progression, helping specialists to determine the optimal treatment. To this end, researchers have combined magnetic resonance imaging (MRI) with biophysical models to estimate tissue characteristics at the cellular level. However, these models are based on assumptions on the size and shape of cellular components to make the problem mathematically tractable, leading to unwanted errors and degeneracy. Addressing this problem, the supervisory team is exploring a potentially disruptive methodology borrowed from materials science. It consists of measuring statistical descriptors (SDs) of tissue microstructure using signals from the MRI scanner, from which histology-like representations may be reconstructed. These SDs have the advantage of describing the statistical nature of tissue components without relying on prior assumptions on cell shapes and arrangements, with huge potential to depict microarchitectures in the living body. Nevertheless, initial experiments were unstable and computationally demanding, lasting for days even in modern computers. This PhD project will focus on developing a solution to the problem by introducing machine learning (ML) approaches: first, to generate fast tissue microstructure reconstructions based on MRI-based SDs; and second, to perform quick simulations of MRI signals for any given microstructure for optimising the MRI acquisition (i.e., maximising accuracy and precision of SDs inference). Convolutional Neural Networks will be developed to solve these issues due to their flexibility and accuracy. The framework's potential will be illustrated in the study of brain cancer microstructures, with the aim of providing unique diagnostic information. Synthetic datasets representing brain cancer tissues at different stages will be utilised to train and test the algorithms. Experiments using physical phantoms will be also performed, building the grounds for potential testing on participants towards the end of the project. Measurements will be carried out in the Connectom MRI scanner, a unique facility designed to characterise tissue microstructure available at the research centre.