Accelerating MRI reconstructions using AI

Magnetic resonance imaging (MRI) is one of the most important imaging techniques when it comes to diagnostics. The clinical MR scans can take up to an hour, which might result in a tedious procedure for the patient. To speed up the scan, one often measures a fraction of the full data matrix. Advanced signal processing techniques are then used to recover the missing information, such that the final reconstructed images are artifact-free. This comes at a cost, however, as the image reconstruction latency increases as the reconstruction problem complexity increases. Therefore, there is a need to accelerate the algorithms. Artificial intelligence (AI) will play a crucial role in this.

Assignment

The aim of this project is to accelerate MR image reconstruction using deep learning techniques. The student will design a tailored deep learning network architecture, work on an efficient training algorithm and construct a training data set. Image reconstructions and testing will involve real MRI data. This project is in close collaboration with Philips and the Leiden University Medical Center.

Requirements

We are looking for an enthusiastic and motivated student with experience in deep learning. Knowledge of numerical linear algebra and signal processing is a pre, and an interest in image reconstruction problems is appreciated. Programming skills in Python are required. The expected project duration is 9 months.

Contact

Last modified: 2022-06-15