Computational Methods for Biomedical Imaging
Name
gerlach-connormg-sm-eecs-2023-thesis.pdf
Description
Thesis PDF
Size
1.73 MB
Format
Adobe PDF
Checksum (MD5)
72eda476a27f792d9d6273de3a38abfc
Author(s)
Gerlach, Connor Michael
Advisor(s)
You, Sixian
Date Issued
September 2023
Publisher
Massachusetts Institute of Technology
Abstract
This work aims to survey and advance the state of the art in methods for biomedical imaging and disease diagnosis. We demonstrate the generation of non-diffracting beams using Lee Holography, and argue that the rich depth information made possible by these beams is well-suited for machine learning applications, where 2D images can contain 3D contextual information without the added computational overhead of performing 3D convolutions. We begin with a review of important non-diffracting beams in the existing literature, and proceed to discuss the necessary experimental design for their generation. We then demonstrate the experimental generation of these beams, including the novel generation of a rotating beam and needle beam via Lee holography. This is followed by the presentation and analysis of a particular semi-supervised machine learning method, contrastive learning, and a novel demonstration of how transfer learning can further improve the representations made by contrastive learning.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
