Deep Learning-based MRI reconstruction

Presenter: Jon Andre Ottesen, PhD student at CRAI, Division of Radiology and Nuclear Medicine, Department of Physics and Computational Radiology, Oslo University Hospital, Oslo, Norway

Abstract: MRI data acquisition is an inherently slow process due to fundamental physical constraints that limit the signal (k-space) acquisition rate. This can lead to prolonged MRI sequences, during which the patient must remain still to achieve diagnostic-quality images. Traditionally, acceleration methods have entailed k-space/signal subsampling, followed by image reconstruction to reduce aliasing artifacts. In recent years, deep learning and convolutional neural networks have shown great promise as an alternative framework for MRI reconstruction to further accelerate scans beyond that of more traditional reconstruction methods.

Deep learning-based MRI reconstruction aims to learn to map a subsampled MR image to its fully sampled counterpart. One of the methods is a so-called cascading network that alternates between a U-Net-like architecture and data consistency. Here we aim to investigate how to improve MRI reconstruction frameworks through alterations to the cascading network formula, and show how it can be used on prospective data.

‍

In compliance with GDPR consent requirements, presentations given in a Visual Intelligence context may be recorded with the consent of the speaker. All recordings are edited to remove all faces, names and voices of other participants. Questions and comments by the audience will hence be removed and will not appear in the recording.  With the freely given consent from the speaker, recorded presentation may be posted on the Visual Intelligence YouTube channel.

‍