Workshop - BICV

Recent advances in computer vision have improved medical image analysis, enabling computer-assisted diagnosis and supporting medical intervention. However, biomedical imaging still presents important challenges due to strong anatomical and structural priors, limited annotations, and high variability between patients and institutions.

These issues are even more evident in ultra high-resolution data such as histopathology, which contains an enormous amount of fine-grained visual detail, going from subcellular structures to tissue architecture. In these cases, relevant diagnostic information can be subtle, spatially sparse, and can easily be lost without appropriate modeling strategies.

For this reason, new approaches focus on developing robust systems that can handle data heterogeneity, limited supervision, and the complexity of human morphology. These methods aim to preserve meaningful biological structure while learning generalizable representations. This workshop focuses on biomedical computer vision problems in these settings, with particular attention to human morphology across scales, from cellular level to organ level, and to emerging solutions designed to address these challenges.

Topics of interest include, but are not limited to:

• Vision Transformers and foundation models for high resolution biomedical imaging
• Multi-scale and hierarchical modeling of human morphology
• Self-supervised, contrastive, and weakly supervised learning
• Imaging techniques for diagnosis, prognosis, and risk analysis
• Domain generalization and adaptation across diverse clinical datasets
• Data efficient learning in low annotation regimes
• Synthetic data generation and augmentation for biomedical applications

More broadly, the workshop aims to connect advances in computer vision with real clinical needs, encouraging research that can support reliable and scalable solutions for medical practice.