5/13/2021 Huan Song
Varun Kelkar is a fourth-year Ph.D. candidate in the Computational Imaging Science Laboratory led by bioengineering department head, Mark. A. Anastasio. Kelkar has received a fellowship through the Oak Ridge Institute and the U.S. Food and Drug Administration (FDA) to research new tools to better evaluate machine learning-based medical imaging reconstruction methods.
Written by Huan Song
Varun Kelkar is a fourth-year Ph.D. candidate from the department of electrical and computer engineering working in the Computational Imaging Science Laboratory led by bioengineering department head, Mark. A. Anastasio. Kelkar has received a fellowship through the Oak Ridge Institute and the U.S. Food and Drug Administration (FDA) to research new tools to better evaluate machine learning-based medical imaging reconstruction methods.
His project primarily works with the Center for Devices and Radiological Health (CDRH) within the FDA. The CDRH creates regulations for medical devices and innovations used on the human body to protect the public and consumers.
"The goal of this project is to gather first-hand experience with deep learning-based image reconstruction algorithms, characterize the properties of deep learning reconstructed images and develop objective performance assessment methodologies for this category of reconstruction technology," said Anastasio. Kelkar's research in evaluating medical imaging reconstruction methods will inform future regulations around medical imaging devices for diagnostic and treatment purposes.
Originally from India, Kelkar graduated from the Indian Institute of Technology, Madras with an undergraduate degree in engineering physics which combined physics and electrical engineering. There, he developed an interest in the fields of wave physics and imaging. "The reason I became interested in this field is the unifying nature of a lot of different types of imaging systems," said Kelkar.
Although there might be many different medical imaging modalities, among several, there is a common thread of using waves to acquire information about objects, and reconstructing digital images from these measurements. For example, in x-ray tomography, which is one out of many medical imaging techniques, x-rays go through parts of the body at different angles. Diagnostic imaging devices collect the signature of x-rays leaving the body at different orientations and a computational model reconstructs these signatures to create images showing the inside of a human body.
Kelkar said, "many different types of medical imaging modalities such as ultrasound, MRI, CT, optical imaging, etc. have similar physics and mathematical structures, which leads to very similar signal processing and data processing techniques used to create images."
Researchers are now using machine learning to aid the process of image formation. However, machine learning techniques are still at an early stage where they may not necessarily produce a diagnostically useful image. The Computational Imaging Science Laboratory is designing evaluation tools for machine learning-based image reconstruction methods to gauge whether image features are correct and useful for diagnostic imaging services.
"Illinois is very interdisciplinary and I was able to take courses in signal processing, machine learning, optics and electromagnetics," he said. "It gave me a really broad and also significantly deep understanding of different aspects of imaging science and related areas". Kelkar also believes that the research conducted at Illinois is a testament to interdisciplinary collaborations. "Dr. Anastasio comes from a similar interdisciplinary background to mine and I really enjoy discussing research with him," he said.
About ORISE
The Oak Ridge Institute for Science and Education (ORISE) is a U.S. Department of Energy (DOE) asset that is dedicated to enabling critical scientific, research, and health initiatives of the department and its laboratory system by providing world class expertise in STEM workforce development, scientific and technical reviews, and the evaluation of radiation exposure and environmental contamination.