[PDF]

Quantifying human brain microstructure with self-supervised machine learning and MRI


Abdul Rauf Bhatti

05/09/2024

Supervised by Paddy Slator; Moderated by Charith Perera

Magnetic resonance imaging (MRI) plays a key role in diagnosis and monitoring of many diseases. Many emerging MRI techniques use machine learning to process MRI data into microstructural maps – maps of small-scale tissue structure and function, e.g., axon diameter, cell size, and blood flow. Self-supervised machine learning is emerging as a promising technique for producing such maps[1,2]. The aim of this project is to implement a self-supervised machine learning processing pipeline for a novel brain microstructure mapping technique – Soma and Neurite Density Imaging (SANDI) and to apply the pipeline to MRI data acquired at the world-leading Cardiff University Brain Research Imaging Centre (CUBRIC). The pipeline would be valuable contribution that would have immediate utility to numerous medical imaging researchers in Cardiff and beyond. For the duration of the project, the student would have the opportunity to interact with a research group with CUBRIC. Requirements: python. This is a research-focused project that seeks to create new knowledge and would need an appropriately qualified and motivated student. Co-supervised by Dr Marco Palombo. References: 1. Barbieri et al. Deep learning how to fit an intravoxel incoherent motion model to diffusion-weighted MRI. Magnetic Resonance in Medicine 2019. https://doi.org/10.1002/mrm.27910, 2.https://github.com/sebbarb/deep_ivim, 3.Palombo et al. SANDI: A compartment-based model for non-invasive apparent soma and neurite imaging by diffusion MRI. NeuroImage 2020. https://doi.org/10.1016/j.neuroimage.2020.116835


Final Report (05/09/2024) [Zip Archive]

Publication Form