SNIC
SUPR
SNIC SUPR
Super resolution MRI using machine learning
Dnr:

SNIC 2019/5-13

Type:

SNAC Small Compute

Principal Investigator:

Anders Garpebring

Affiliation:

UmeƄ universitet

Start Date:

2019-02-05

End Date:

2020-03-01

Primary Classification:

20603: Medical Image Processing

Webpage:

Allocation

Abstract

Magnetic resonance imaging (MRI) provides many benefits compared to other medical imaging modalities. For instance no ionizing radiation is needed and superior contrast in soft tissue can be achieved. However, a large downside with MRI is its long scan times combined with high price which leads to high costs per patient. Several methods has been presented over the years to speed up MRI, e.g. Echo Planar Imaging, parallel imaging and more recently compressed sensing. The current development and next step in improving the acquisition speed of MRI to use machine learning and in particular convolutional neural networks (CNNs). This projects aims at synthetically increase the resolution of MR images using CNNs and thereby reduce the time required for acquisition of high resolution images. In particular, this project will investigate the possibility to increase the slice resolution in multi-contrast imaging by learning how high resolution information can be transferred from high resolution images to low resolution images with different contrasts.