Avis Co., Ltd. (CEO Daehong Lee), a company specializing in AI-based pathology analysis solutions, announced on the 12th that it had won the Certificate of Merit for the Best Poster Award at the 2025 regular academic conference of the Korean Society for Artificial Intelligence in Medicine (KoSAIM).
This award is a result of recognition of the technical achievements of the joint research project, ‘Improving the Accuracy of IHC (Immunohistochemistry) Quantitative Analysis Using a Next-Generation AI Model Based on the Transformer Model’ by the Avis AI Research Team and Professor Yosep Jeong of the Department of Pathology at Catholic University of Korea.
IHC staining is an essential pathological test for cancer diagnosis, but accurately identifying and quantifying tumor cells among the numerous cells on a slide has proven challenging. Existing CNN-based AI models have limitations in distinguishing between cancer and non-cancerous cells, resulting in low quantitative analysis accuracy and reliability.
To improve this, the Avis research team developed an AI model based on the Vision Transformer (ViT) architecture. This model clearly distinguishes individual cell nuclei as tumor cells and non-tumor cells even in complex tissue environments, significantly improving the accuracy of quantitative analysis.
The research team validated the performance of quantitative analysis of the Ki-67 marker on 239 gastric cancer tissue samples. The Avis model achieved results virtually identical to expert-level results, outperforming the "free-hand" analysis method used by pathologists by a mere 1.2 percentage points. It also demonstrated superior performance compared to other commercial AI solutions.
Lee Dae-hong, CEO of Avis, said, “This award is a meaningful achievement that recognizes Avis’s next-generation AI technology from the most prestigious medical artificial intelligence society in Korea,” and added, “We will continue research and development to overcome the limitations of existing technologies and provide reliable solutions in actual pathology diagnosis sites through our self-developed pathology AI specialized architecture.”
- See more related articles
You must be logged in to post a comment.