Best Paper Award at the IEEE International Symposium on Biomedical Imaging (ISBI) 2026
A new paper by researchers at VinUniversity has been accepted to the 2026 IEEE International Symposium on Biomedical Imaging (ISBI), a premier conference dedicated to the mathematical, algorithmic, and computational aspects of biological and biomedical imaging. The study, titled “Synergizing Deep Learning and Biological Heuristics for Extreme Long-Tail White Blood Cell Classification,” by Duc T. Nguyen et al., also achieved a top-tier ranking in the ISBI 2026 WBCBench Challenge and was honored with the Best Paper Award.
The research addresses white blood cell (WBC) classification from microscopic images, an important task for leukemia screening and hematological analysis. This remains a major challenge for biomedical AI because real-world clinical datasets are often highly imbalanced, abnormal cell types are extremely rare, and image quality can be affected by noise and domain variation across different sources.
Fig. 1. Overview of the proposed three-stage hybrid framework. The pipeline seamlessly integrates generative domain restoration (Stage 1), dual-branch semantic feature extraction (Stage 2), and biological filtering (Stage 3) to achieve robust classification under extreme long-tail distributions.
To tackle these challenges, the team developed a robust three-stage hybrid framework. The first stage focuses on image restoration and standardization to reduce noise and recover important cellular details. The second stage combines a Swin Transformer ensemble with MedSigLIP-based contrastive representations to improve feature learning and enhance recognition of rare cell types. The final stage introduces a biologically inspired refinement strategy based on morphological characteristics of blood cells, helping the system better identify difficult and easily overlooked minority classes.
By integrating the representation power of deep learning with biologically meaningful domain knowledge, the proposed framework improves both rare-cell recognition and robustness in challenging real-world scenarios. This achievement highlights VinUniversity’s growing research strength in AI for biomedical imaging, precision medicine, and clinical diagnostic support.
