Enhanced Retinal Blood Vessel Image Segmentation Using U-Net and Denoising Diffusion Probabilistic Models by Dr. Olubunmi Sule
Join us Tuesday, November 11 at 11:00 AM in Manc 229! Dr. Olubunmi Sule will talk about her research titled Enhanced Retinal Blood Vessel Image Segmentation Using U-Net and Denoising Diffusion Probabilistic Models.
Abstract: Accurate segmentation of retinal blood vessels is crucial for diagnosing and monitoring various retinal diseases. In this study, we first employed a U-Net model to segment retinal blood vessels from retinal images using the DRIVE dataset, achieving a binary IoU of 0.7531 and F1 score of 0.8378. To further improve the segmentation performance, we leveraged Denoising Diffusion Probabilistic Models (DDPMs), which model complex probability distributions and handle uncertainty. By learning to reverse a gradual noise corruption process, a U-Net-based denoising network iteratively refines the segmentation mask, starting from pure noise. Our results demonstrate significant improvements in segmentation accuracy, with a mean IoU of 0.9981, F1
Olubunmi O. Sule is a Visiting Assistant Professor in the Department of Computer Science at Wake Forest University. She earned her PhD in computer science at University of Kwazulu-Natal (UKZN), Durban, South Africa, in 2022, and a Post Graduate Program in Cybersecurity at The University of Texas at Austin, in 2024.
Her research interests include machine learning, computer vision, image processing, pattern recognition, medical image analysis and cybersecurity. She is focused on applying deep learning frameworks to medical images for early detection of abnormality and analysis. She has published her research articles in top-tier journals and conferences.
She was a recipient of the First Prize Award for Oral Presentation at the Postgraduate Research and Innovation Symposium, University of KwaZulu-Natal, in 2020 and the Honourable Mention Award for Poster Presentation at the Deep Learning IndabaX International Conference (South Africa Chapter) in 2019. She participates in reviewing papers for top journals.
As a dedicated teacher-scholar with over some years of experience, she has taught and tutored students different computer courses in Nigeria, South Africa, and USA.
All are welcome!