ZHANG Ruixiang, a master’s student in the IIST program at the Graduate School of Computer and Information Sciences, received the Best Presentation Award at the 9th International Conference on Artificial Intelligence and Big Data (ICAIBD 2026), held in China in May 2026.

Abstract—Tongue image analysis is important for computer-aided diagnosis in Traditional Chinese Medicine. However, existing deep learning methods often fail to distinguish tongue body features from tongue coating features and are sensitive to illumination variation and specular reflection, which distort color appearance and obscure coating-related cues. To address these issues, this paper proposes a region-guided and illumination-aware framework for multi-attribute tongue image analysis. Fuzzy C-Means is used to generate soft masks for tongue body and coating regions, which are incorporated into region-guided feature aggregation to promote region-specific representation learning. Specular reflection suppression and auxiliary illumination-level supervision are further introduced to enhance robustness under varying imaging conditions. Experiments on an expert-annotated tongue image dataset show that the proposed framework consistently outperforms baseline variants, especially for tongue color, coating color, and coating thickness recognition. These results demonstrate the effectiveness of jointly modeling regional ambiguity and illumination-related interference.

For more information:
https://cis.hosei.ac.jp/news/2026/07/13/17808/