We are happy to share that we have three papers accepted to the Information Processing in Medical Imaging Conference (IPMI 2025).
- Skelite: Compact Neural Networks for Efficient Iterative Skeletonization
Luis D. Reyes Vargas, Martin J. Menten, Johannes C. Paetzold, Nassir Navab and Mohammad Farid Azampour
Skeletonization creates thin representations that capture geometry and topology, crucial for preserving connectivity in curvilinear structures like vessels in medical imaging; however, existing differentiable methods trade off between efficiency and accuracy. This work introduces a novel, learnable iterative skeletonization framework that leverages synthetic data, augmentation, and model distillation to produce thin, connected skeletons with 100× speedup over traditional methods while maintaining high accuracy and generalizing well across domains. - BioSonix: Can Physics-based Sonification Perceptualize Tissue Deformations From Tool Interactions?
Veronica Ruozzi, Sasan Matinfar, Laura Schütz, Benedikt Wiestler, Alberto Redaelli, Emiliano Votta, Nassir Navab
This paper introduces BioSonix, a mixed reality framework that enhances surgical tool navigation by sonifying tool-tissue interactions. Using biomechanical simulations, it maps tissue displacements to auditory cues, encoding properties like stiffness and density. User studies with clinical and biomedical experts demonstrated its effectiveness in improving task accuracy and intuitive perception of interactions. - Hierarchical Neural Cellular Automata for Lightweight Microscopy Image Classification
Chen Yang*, Michael Deutges*, Nassir Navab, Carsten Marr+, Ario Sadafi+
This paper introduces a hierarchical Neural Cellular Automata (NCA) model for lightweight microscopy image classification, capturing both local and global features efficiently. It outperforms existing NCA-based methods on six datasets. NCA are efficient architectures requiring fewer parameters, making them suitable for resource-constrained environments.
Congrats to all the authors!