Magnetic Resonance Imaging (MRI) is a vital medical imaging technique. However, limitations in hardware and acquisition time often result in lower-resolution images (e.g. at 1.5T or with specific fast scanning protocols). This project aims to develop a superresolution algorithm capable of constructing higher-resolution MRI images (e.g. 3T or the full resolution from a fast scan) from lower resolution ones.

By simulating lower resolution images from high resolution images using suitable k-space (similar to Fourier space) representations and simulations of the imaging sequences, we can construct a dataset having only high resolution images available. This dataset can be used to train and validate a machine learning approach to map lower resolutions to higher resolutions. This project involves developing the k-space simulation technique (for a particular protocol, etc), devising a superresolution machine learning technique, and evaluating and improving its performance.

For this project you need some understanding of MRI techniques and imaging protocols, mathematical and programming skills, in particular in machine learning (using Python or Julia and a suitable deep learning framework such as tensorflow, torch, jax, lux).

Ideally, the code would be made available under the AGPL v3 or compatible license to integrate with our other code.