YEREVAN, Armenia — From July 15–17, a collaborative research paper by the American University of Armenia (AUA) Zaven P. and Sonia Akian College of Science and Engineering (CSE) and the Department of Computer Science at the University of Oxford was recognized as the Runner-Up for the Best Paper Award at the Medical Image Understanding and Analysis (MIUA) 2025 conference, held at the University of Leeds, United Kingdom.
The paper, titled “Semantic Segmentation with Spreading Scribbles,” was co-authored by AUA alumna and research associate, Yeva Gabrielyan (BSCS ’19, MSCIS ’22), AUA Associate Professor Dr. Varduhi Yeghiazaryan, and Dr. Irina Voiculescu, associate professor at the University of Oxford. The implementation and presentation of the project at the conference was made possible with the support of the Afeyan Family Foundation Seed Fund for Collaborative Grants, which aims to strengthen global research partnerships and advance academic excellence at AUA.
MIUA 2025 brought together over 150 participants from 23 countries, including researchers, clinicians, and industry professionals, to explore the latest advancements in medical image analysis, AI-assisted diagnostics, and personalized medicine. The program featured oral and poster sessions, a doctoral consortium and mentoring, as well as a vibrant panel discussion uniting experts from academia, clinical research, and industry. Attendees engaged in extensive networking opportunities and shared ongoing projects and visions for future collaboration.
“This work was our first attempt to contribute to scribble-supervised medical image segmentation, said Dr. Yeghiazaryan. “I am very glad that it received appreciation from the medical image analysis community. This encourages us to expand efforts in this direction.”
This is not the first collaboration of this team. In June, their joint research article, “Parallel Watershed Partitioning: GPU-Based Hierarchical Image Segmentation,” was accepted for publication in the Journal of Parallel and Distributed Computing, a high-ranking journal published by Elsevier. The current conference paper applies the algorithms from that article to improve medical image segmentation accuracy, even under annotation-scarce conditions using scribble e supervision.
“It is truly rewarding to see our work recognized at an international conference of this scale. Receiving the Runner-Up nomination demonstrates the impact of our research in making semantic segmentation more accessible for clinical use by reducing the effort required from clinicians,” remarked Gabrielyan. “This motivates us to continue advancing in this direction and improving segmentation quality while further minimizing the need for extensive human input.”
Medical image segmentation is the task of delineating a specific tissue/organ on a medical scan (left image). This is useful in the context of diagnosis, treatment planning, etc. Deep learning systems, in order to learn to solve this task, need vast amounts of training examples. If the training data contains only scribbles (middle image) instead of fully detailed segmentation masks, the task becomes significantly harder to learn. The approach proposed in the MIUA 2025 paper “spreads,” or propagates, the scribble labels across the image (right image) to boost the amount of training data. Experimental results show that classic deep learning architectures with simple loss functions, not specifically designed for scribble supervision, perform on par with the advanced models targeted for scribble supervision.
CSE faculty and alumni continue to strengthen international research collaborations, advancing innovation in their fields, and contributing results that carry influence on the global stage.
Founded in 1991, the American University of Armenia (AUA) is a private, independent university located in Yerevan, Armenia, affiliated with the University of California, and accredited by the WASC Senior College and University Commission in the United States. AUA provides local and international students with Western-style education through top-quality undergraduate and graduate degree and certificate programs, promotes research and innovation, encourages civic engagement and community service, and fosters democratic values.
