T03: BVM2025 Hands-on Tutorial on
Implicit Neural Representations (INRs) in Medical Imaging
Complementary to the recent success in population-based training of deep neural networks for medical image analysis tasks there has been an exiting development and advancement of optimisation driven approaches using Implicit Neural Representations (INR). INRs, which model e.g. mappings between input coordinates and output gray values as continuous functions, are promising in various vision tasks including motion estimation, image reconstruction, denoising, superresolution, and compression. Over the last few years, those ideas have been transferred into the medical imaging field, where they offer unique advantages in adapting to task- and instance-specific challenges that have thus far prevented a wider spread adoption. Our tutorial will provide both a comprehensive overview lecture of INR research in the medical imaging domain as well as practical demonstrations and hands-on learning for interested students and researchers with little or no prior knowledge of the topic.
The Notebook used during the session can be accessed and edited through Colab: 
The solution can be found on GitHub: 
Also check out our INR Playground on Hugging face Spaces: 
Talks
| Time | Topic | Contents |
|---|---|---|
| 14:00 - 14:10 | Welcome & Overview | Introduction of speakers and organisational tutorial information |
| 14:10 - 14:30 | Introduction to INR | Basic conceptual and mathematical foundations on INRs and associated activation functions |
| 14:30 - 16:30 | Hands-On (incl. individual breaks) | Assisted programming of INR solutions for reconstruction, denoising and image registration in Jupyter notebooks |
| 16:30 - 17:10 | Advanced Topic Mini-Lectures (4x10min) | 10 Minute Mini-Lecutures on challenging MRI reconstruction, rendering INRs (NERF), implicit neural image registration and multi-instance INR generalization |
| 17:10 - 17:30 | Evaluation | Discussion / Q&A |
Organizers
Universität zu Lübeck
Mattias Paul Heinrich is an Associate Professor at the Institue of Medical Informatics in Lübeck, Germany, and leads a research group focusing on diverse topics in the field of Medical Deep Learning and has acquired third-party funding for various projects. His research expertise lies in the fields of data analysis, machine learning and medical imaging. Together with collaborators from UK, France and Germany within academia and industry, Mattias has published more than 50 peer-reviewed articles with high citation impact and won several international awards.
Universität zu Lübeck
Fenja is a research assistant and Ph.D. candidate at the Medical Deep Learning Lab in the Institute of Medical Informatics in Lübeck. She has a mathematical background with a M.Sc. in Computational Life Science. Her research is focused around medical image registration, leveraging a combination of population-learned models with instance-based optimization.
Universität zu Lübeck
Ziad Al-Haj Hemidi is a Ph.D. student at the Institute of Medical Informatics, University of Lübeck. He earned his B.Sc. and M.Sc. in Medical Computer Science there (2019, 2022), focusing on MRI motion correction. Since 2022, he has been researching medical imaging in Prof. Mattias Heinrich’s Medical Deep Learning group, optimizing cardiac MRI reconstruction and motion artifact reduction for the MEDICARE project. His interests include MR imaging, k-space artifact reduction, fast reconstruction, deep learning, and implicit neural representations.
Universität zu Lübeck
Christoph is a research assistant and Ph.D. candidate at the Institute of Medical Informatics at the lab of Mattias Heinrich in Lübeck, Germany. He holds an M.S. degree in Medical Engineering Science and works on improving ultrasound image analysis and reconstruction, especially for pediatric patients. He is further actively involved in research related to medical image registration and has a faible for Implicit Neural Representations.