April 4, 2024
Left: Two frames from an endoscopy video. Center: The resulting depth estimation for each frame using a common method, with incompleteness (white regions) and errors in the reconstruction. Right: A more complete depth estimation using Sengupta’s method that reasons about reflection of light.
Assistant Professor Roni Sengupta was recognized with the Trailblazer R21 Award by the National Institute of Biomedical Imaging and Bioengineering (NIBIB) in the National Institutes of Health (NIH).
The Trailblazer R21 Award supports new and early-stage investigators pursuing research programs that are of high interest to the National Institute of Biomedical Imaging and Bioengineering, at the interface of life sciences with engineering and the physical sciences. The award will provide Sengupta with $400,000 in direct funding over the three-year funding period and roughly $580,000 in total funding.
Sengupta’s project seeks to develop a 3D modeling system that can revolutionize endoscopy by providing accurate reconstructions of internal organs. Current methods struggle to accurately model or even capture these organs because their structures are not rigid and may not be marked by distinct geometric figures, and they are covered with a layer of mucus that can reflect the light from the endoscope. Sengupta will use new deep-neural-network-based computer vision algorithms to reconstruct the shape of rigid organs based on the way light reflects off of their mucus layer. For non-rigid organs, she will develop an additional system to extract the organ’s cyclical change in shape over time.
Sengupta hopes that the project will improve diagnosis accuracy and patient comfort during endoscopic procedures, as well as enable related tasks like semi-autonomous maneuvering of the endoscope, better and more accurate visualization, and guidance back to regions that were missed.
The project is a collaboration between Sengupta and co-investigators Marc Niethammer in the Department of Computer Science and Sarah McGill and Carlton Zdanski in the School of Medicine. This project is also inspired by Dr. Stephen M. Pizer’s research on 3D colon modeling and blind spot detection from colonoscopy, currently funded by the Olympus Corporation.
For more information about the Trailblazer R21 Award, visit the award website.