Master & bachelor projects
Denoising of complex-valued diffusion MRI images
Diffusion MRI is a powerful tool to quantify biological tissue microstructure in vivo and non-invasively. However, diffusion MRI signal analysis is notoriously hampered by low signal-to-noise ratio (SNR) in heavily diffusion-weighted images where the signal is substantially attenuated. Among the denoising techniques proposed over the years, Marchenko-Pastur Principle Component Analysis (MP-PCA) denoising has been the most promising (Veraart et al., NeuroImage 2016; Moeller et al., NeuroImage 2021). In order to meet the underlying assumption of Gaussian noise and avoid interference from Rician bias, diffusion MRI data are best denoised in complex-valued space rather than magnitude space. Phase maps from diffusion weighted datasets are however heavily affected by fluctuations due to the diffusion-weighting itself.
In this project, we propose to develop a robust pipeline for pre-processing the magnitude and phase images to enable reliable MP-PCA denoising of complex-valued diffusion MRI data, and thus provide a dramatic boost in SNR. The test data for this project will then be used to quantify brain gray matter microstructure using a new biophysical model (NEXI, Jelescu et al., NeuroImage 2022) with high accuracy and precision of parameter estimates.
Profile: MSc or BSc students in physics, engineering, computer science or mathematics. Flexible duration aligned with your University and degree requirements from a research project.