Beyond Occlusion: The BioMedAI Research Team to Present at ISBI 2025
The paper "Beyond Occlusion: In Search for Near Real-Time Explainability of CNN-Based Prostate Cancer Classification" has been accepted for presentation at the prestigious 2025 IEEE International Symposium on Biomedical Imaging (ISBI 2025).
This conference, taking place April 14–17, 2025, in Houston, TX, USA, is one of the leading events in biomedical imaging, bringing together cutting-edge advancements in medical image analysis, artificial intelligence, and computational imaging.
The authors Martin Krebs, Jan Obdržálek, Vít Musil, and Tomáš Brázdil from the BioMedAI project at the Faculty of Informatics, Masaryk University, focus on explainability in convolutional neural networks (CNNs) for prostate cancer classification in digital histopathology. Their study introduces an innovative approach that replaces traditional occlusion-based explainability methods, achieving a tenfold speedup without compromising output quality. This acceleration significantly enhances model development and fine-tuning, bringing AI-assisted prostate cancer diagnostics closer to clinical application.
At ISBI 2025, the paper will be presented by Jan Obdržálek and Martin Krebs. Congratulations to the authors on this achievement!
Original source: https://biomedai.muni.cz/news
- Web address
- Attachments
