[Defense] Upscaling Low Quality MRI Images using Deep Learning
Friday, November 11, 2022
10:00 am - 11:00 am
In
Partial
Fulfillment
of
the
Requirements
for
the
Degree
of
Doctor
of
Philosophy
Rishabh
Sharma
will
defend
his
proposal
Upscaling
Low
Quality
MRI
Images
using
Deep
Learning
Abstract
MR scans of low gamma X-nuclei, low concentration metabolites, or standard imaging at very low field entail a challenging tradeoff between resolution, signal-to-noise and acquisition duration. Deep learning (DL) techniques, can potentially be used for improving such 鈥渓ow quality鈥 (LQ) images. In this dissertation I will develop and investigate different deep learning methods for upscaling LQ MRI under different acquisition scenarios that occur in clinical settings. These scenarios include K-Space truncated images, compressed sensing images, and other low quality MRI acquisitions. Collected images will be upscaled and evaluated under two acquisition conditions with same-subject High Quality Complementary Priors (HQCP) scenario, a LQ and a high quality (HQ) image are collected and both LQ and HQ will be input to deep learning methods and No Complementary Priors (NoCP) scenario, where only the LQ images are collected and used as sole input to deep learning methods. Augmentation technique will be introduced to create synthetic dataset to perform these studies as this study requires paired dataset of LQ and HQ images, which is not available in public domain. The deep learning methods will be in upscaling 1/8, 1/4, and 1/2 undersampled images for both scenarios. Statistical models will be used to evaluate the significance of deep learning models, acquisition matrix, acquisition scenario, and type of ground truth, using mixed effects model to correctly account for repeated measures in the study.
Friday,
November
11,
2022
10:00AM
-
11:00AM
CT
Online
via
聽
Dr. Nikolaos V. Tsekos, dissertation advisor
Faculty, students and the general public are invited.
