Rodrigo Andujar Lugo

Magnetic Resonance Imaging (MRI) is a critical diagnostic tool in medical imaging, where the accuracy of the images heavily depends on the patient remaining still during the scan. Motion artifacts significantly degrade the image quality, particularly in 2D acquisitions due to their inherent limitations, such as thick slices and gaps between slices, which worsen the artifacts by missing crucial information. While several techniques have been developed to mitigate these effects in 3D acquisitions, 2D acquisitions remain challenging.

This thesis aims to develop a novel deep learning model specifically designed for motion correction and reconstruction in 2D MRI acquisitions based on SAMER [1]. The proposed model will focus on implicitly filling in the missing gaps and correcting motion artifacts by learning from a vast dataset of MRI images with varying degrees of motion artifacts. By training the model on this data, it will be able to reconstruct high quality images from 2D acquisitions, thereby improving the diagnostic accuracy and reliability of MRI scans.

[1] Polak D, Splitthoff DN, Clifford B, et al. Scout accelerated motion estimation and reduction (SAMER). Magn Reson Med. 2021; 87: 163–178. https://doi.org/10.1002/mrm.28971