Excited to share our work on Multi-contrast Laser Endoscopy (MLE), published in npj Imaging.
Challenge: Evaluating new contrast methods through size-constrained endoscopes without disrupting clinical workflow.
Our solution: don’t replace the colonoscope — retrofit it.
(1/8)⬇️
In collaboration with @SNgamruengphong from @HopkinsGIHep, we imaged 31 colorectal polyps during colonoscopy procedures. Compared with conventional imaging, MLE demonstrated a 3× improvement in lesion contrast and a 5× increase in color contrast.
(7/8)⬇️
Welcome to our new Malone postdoctoral fellows! @tbobrow1, @jie_gao_056, and @DanielleRipsman comprise the latest cohort selected for the program. Learn more about their research here: https://t.co/XIOciNOPJc
@mode7studio@HopkinsEngineer@njdurr@mode7studio VR-Caps is a great rendered dataset. What makes C3VD new is that it uses real videos with paired ground truth. There are pros and cons for both rendered and C3VD data, so researchers should validate using both types. See our paper for a list of public datasets.
Applying #ComputerVision to colonoscopies has been a struggle for researchers. Now, thanks to a new "ground truth" dataset from @njdurr and team, researchers around the world are training and testing AI to estimate colon geometry and improve screening. https://t.co/vEa2Pwo0Sj
Thanks to PI @njdurr, co-authors @mayankgolhar & Rohan Vijayan, and our funding sponsors @NSF@NIBIBgov @OlympusMedUS that supported this work.
C3VD is available now at https://t.co/nAEWZSMIET
Publication: https://t.co/QXNgNw1crA
@JHUBME@HopkinsEngineer
(7/7)⬇️
Colonoscopy is an important application of computer vision. Testing new algorithms has been challenging w/o any datasets having real endoscope frames & ground truth depth. So we created C3VD - a colonoscopy 3D video dataset out in @ElsevierConnect Medical Image Analysis.
(1/7)⬇️
C3VD also includes four simulated screening colonoscopies performed by Dr. Venkata Akshintala of @HopkinsGIHep. These Videos come with 3D ground truth models and coarse ground truth camera poses.
(6/7)⬇️