In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
will defend his proposal
From Medicine to Space Exploration, Intelligently Filling Gaps In Imaging Data
AbstractAlthough it was already possible to land on the Moon and assess biological tissue, the technology necessary to dive into practical deep learning research only became available recently. Deep learning research has crossed over from computer science and now spans multiple fields including medical imaging and space exploration. Deep learning algorithms are known to be both computational, expensive, and data hungry, often requiring thousands or millions of training data samples to reach acceptable performance. In the medical domain, obtaining valid imaging data in amounts large enough to train these algorithms can be difficult due to many factors-- including patient privacy, governmental regulations, and lack of variability. When talking about space exploration, there is usually no available data until we are able to perform a data collection mission, such as the NASA LRO. In both scenarios, a possible solution to the vacuum of available data is to generate or synthesize entirely new data. In this thesis, we propose novel deep learning methods to generate realistic imaging data by taking into account the physical properties of the imaging methodology to be synthesized. This allows us to train our models with less real data, while reaching acceptable levels of realism.
Date: Friday, May 29, 2020
Time: 11:00 AM - 12:00 PM
Place: Online Presentation - MS Teams
Advisor: Dr. Nikolaos Tsekos
Faculty, students, and the general public are invited.