Synthetic data generation pipeline to effectively train deep learning augmented super-resolution ultrasound imaging
Katakam, Swathi
This item's files can only be accessed by the System Administrators group.
Permalink
https://hdl.handle.net/2142/129788
Description
Title
Synthetic data generation pipeline to effectively train deep learning augmented super-resolution ultrasound imaging
Author(s)
Katakam, Swathi
Issue Date
2025-05-09
Director of Research (if dissertation) or Advisor (if thesis)
Chen, Yun-sheng
Department of Study
Electrical & Computer Eng
Discipline
Electrical & Computer Engr
Degree Granting Institution
University of Illinois Urbana-Champaign
Degree Name
M.S.
Degree Level
Thesis
Keyword(s)
Super-resolution
Ultrasound
synthetic data
microscopy
Abstract
Super-resolution ultrasound imaging is an emerging ultrasound technique capable of imaging microvascular structures and flow in unprecedented detail. Advancements in imaging speed and resolution are required to enable efficient 2D and 3D SRUI for clinical and preclinical applications. Deep learning methods are used to accelerate the acquisition and processing of SRUI datasets towards real-time imaging, but the efficacy of these methods is reliant on large quantities of high quality data. Synthetic data, typically generated moving microbubbles quasi-randomly in a 2-dimensional image is sometimes used to train these algorithms, but this method does not reliably mimic the vascular network and flow, which is 3-dimensional. The need for high quality 3D datasets as the field of 3D imaging advances. To address the gap in data availability, we present a workflow to generate large quantities of synthetic SRUI data from micro-CT angiographies of the mouse cerebrovascular network. Using image processing techniques we generate a realistic vector field through which we propagate the microbubbles through the vasculature. These microbubbles are gathered into frames which are simulated using ultrasound simulation software and subsequently analyzed to produce super-resolution ultrasound images similar in structure and flow to in-vivo datasets. This workflow advances current methods for generating synthetic data in SRUI by simplifying the simulation of microbubble flow in high-resolution whole-brain volumes.
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.
