Medical Image Computing (MIC) @DKFZ German Cancer Research Center

🚀 𝗝𝗼𝗶𝗻 𝘂𝘀 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗛𝗜𝗗𝗔 𝗟𝗲𝗰𝘁𝘂𝗿𝗲 𝗦𝗲𝗿𝗶𝗲𝘀 𝗼𝗻 𝗔𝗜 Artificial Intelligence is transforming industries across the board—from revolutionizing healthcare to optimizing public services. As part of our monthly HIDA lecture Series on AI, we bring together experts to explore AI’s impact in different fields. Here’s what’s coming up next: 📅 01.04.2025, 11:00 am 𝗚𝗲𝗻𝗔𝗜 𝗶𝗻 𝘁𝗵𝗲 𝗣𝘂𝗯𝗹𝗶𝗰 𝗦𝗲𝗰𝘁𝗼𝗿: 𝗧𝗵𝗲 𝗻𝗼𝘄 & 𝗻𝗲𝘅𝘁 𝗼𝗳 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 https://lnkd.in/e6CzPQjE 📅 21.05.2025, 02:00 pm 𝗔𝗜, 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝘆 𝗚𝗮𝗽𝘀, 𝗮𝗻𝗱 𝗔𝘀𝘆𝗺𝗺𝗲𝘁𝗿𝗶𝗲𝘀 𝗯𝗲𝘁𝘄𝗲𝗲𝗻 𝗖𝗿𝗲𝗱𝗶𝘁 𝗮𝗻𝗱 𝗕𝗹𝗮𝗺𝗲 https://lnkd.in/e_hY6Jii 📅 12.06.2025, 10:00 am 𝗖𝗹𝗶𝗻𝗶𝗰𝗮𝗹 𝗟𝗮𝗿𝗴𝗲 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 𝗶𝗻 𝗛𝗲𝗮𝗹𝘁𝗵𝗰𝗮𝗿𝗲 – 𝗔𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀, 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀, 𝗮𝗻𝗱 𝗢𝗽𝗽𝗼𝗿𝘁𝘂𝗻𝗶𝘁𝗶𝗲𝘀 https://lnkd.in/epPKgs26 👉 Missed a few lectures? No worries! We've got them recorded for you. Watch now: https://lnkd.in/ebaDuQvY

nnInteractive: A Breakthrough in AI-Driven 3D Interactive Segmentation! 🧠✨ If you’re working in medical imaging, AI-driven segmentation, or 3D reconstruction, nnInteractive is something you need to check out. I just published a Medium article covering: • Why this is a game-changer for AI-powered 3D interactive segmentation • How to get started with nnInteractive right now • How to interact with its backend to build your own applications & plugins This tool is unlocking new levels of precision, efficiency, and interactivity, making AI-powered segmentation more accessible than ever. Curious? Dive into the full breakdown here: 👉 https://lnkd.in/dY6MdkN6 Let me know your thoughts! Have you explored nnInteractive yet? 🚀🔍 #AI #3DImaging #MedicalAI #DeepLearning #nnInteractive #Segmentation #TechInnovation

Are you on-FHIR 🔥 🙃 ? Then join us as medical informatics expert at the DKFZ!

Exciting to see the Helmholtz Foundation Model Initiative (HFMI) gaining traction! 🚀 With the Human Radiome Project, we at MIC are proud to contribute our expertise in medical AI and large-scale multimodal learning. The Human Radiome is set to become a foundational AI model for medical imaging, integrating vast amounts of MRI and CT data to advance our understanding of human anatomy and pathology. By linking radiological data with other modalities like genomics and clinical records, we aim to drive personalized medicine and more intuitive AI-assisted decision-making in healthcare. Looking forward to shaping the future of AI-driven medical science together! 💪

🚀 Proud to introduce 𝗻𝗻𝗜𝗻𝘁𝗲𝗿𝗮𝗰𝘁𝗶𝘃𝗲 – a breakthrough in 3D Interactive Segmentation! 🎉 After months of work, I’m beyond excited to share nnInteractive, our next-generation 3D segmentation framework that redefines interactive AI for medical imaging. Unlike existing tools, nnInteractive fully embraces volumetric segmentation, offering unmatched usability and flexibility: ✅ Supports diverse interactions – Points, scribbles, bounding boxes, and our novel lasso prompt ✅ Positive & negative prompting – Gain full control over segmentation refinement ✅ From simple 2D inputs to full 3D segmentation – 1D points or 2D prompts seamlessly create volumetric masks ✅ Trained on 120+ datasets across CT, MRI, PET, 3D Microscopy, and more ✅ Integrated into real-world tools – Available in Napari & MITK Interacting with 3D data has never been this user-friendly! Whether you’re a researcher, radiologist, or AI enthusiast, nnInteractive makes segmentation faster, smarter, and more intuitive than ever. A huge thanks to my amazing co-authors Fabian Isensee Lars Krämer and everyone who contributed to this project! Excited to see how the community adopts and builds on it. 📄 Paper: https://lnkd.in/e3KdvmiM 🛠 Try it out: - Napari Plugin: https://lnkd.in/eayfGHqJ - MITK Plugin: https://lnkd.in/eYnUwiaT Let me know your thoughts and feedback! 🚀 Stefan Dinkelacker Ashis Ravindran Florian Stritzke Benjamin Hamm Tassilo Wald Moritz Langenberg Constantin Ulrich Jonathan Deissler Ralf Floca Klaus Maier-Hein #MedicalImaging #DeepLearning #AI #3DSegmentation #Napari #MITK #nnInteractive #Radiology #Microscopy

🚀 We are stoked to present nnInteractive: The nnU-Net moment for 3D Interactive Segmentation! 🎉 3D medical image segmentation is crucial for clinical and research applications, yet existing interactive segmentation tools, like Meta's SAM, struggle with volumetric consistency, usability, and interactivity. Developed by Fabian Isensee Maximilian R. and Lars Krämer we present a game-changer in AI-driven 3D segmentation: ✅ Supports diverse prompts: Points, scribbles, bounding boxes, and a novel lasso prompt ✅ Seamless 3D segmentation from intuitive 2D interactions ✅ Trained on 120+ datasets (CT, MRI, PET, 3D Microscopy, etc.) for unmatched generalization ✅ Integrated into widely-used imaging platforms: Try it today in Napari or MITK! Check out the details here: 📃 Paper: https://lnkd.in/eQsR8GF9 ⚒️ Napari Plugin: https://lnkd.in/eNgciStG ⚒️ MITK Plugin: https://lnkd.in/emMXnm7A This isn't just another segmentation tool—it sets a new state-of-the-art for AI-driven volumetric annotation. Try it out and let us know how it transforms your workflow! 🚀 A huge thanks to the team: Fabian Isensee Maximilian R. Lars Krämer Stefan Dinkelacker Ashis Ravindran Florian Stritzke Benjamin Hamm Tassilo Wald Moritz Langenberg Constantin Ulrich Jonathan Deissler Ralf Floca Klaus Maier-Hein #MedicalImaging #DeepLearning #AI #3DSegmentation #Napari #MITK #nnInteractive #Radiology #Microscopy

🌵 Back from #WACV2025 in Tucson, Arizona! I had the opportunity to present our paper on predictive imaging biomarker discovery, where we explore how image features that help guide treatment decisions can be identified using deep-learning-based CATE estimators for image inputs. We also propose an evaluation framework to assess their performance. 📄 Paper: "Enhancing Predictive Imaging Biomarker Discovery Through Treatment Effect Analysis" https://lnkd.in/ektgmsZZ 💻 Code: https://lnkd.in/ey9WysHy A huge thank you to Lukas Klein, Jens Petersen, Philipp Vollmuth, Paul F. Jaeger and Klaus Maier-Hein for the great collaboration and to Medical Image Computing (MIC) @DKFZ German Cancer Research Center for making this possible! Also, I'm very grateful for all the wonderful and brilliant people I had the chance to meet at the conference! 🏜️ #WACV #MedicalImaging #ComputerVision #CausalInference

Congrats, Fabian, on this well-deserved recognition! 🥳 Your work on AI-driven medical image analysis, especially #nnUNet, has had a major impact on the field. We are proud to have you in our division!

Our Talks at BVM 2025 – Insights into AI and Medical Image Analysis As part of our contributions to BVM - German Conference on Medical Image Computing, we are excited to highlight two talks from our division. These presentations will dive into image analysis validation and a revisited perspective on nnU-Net, providing valuable insights into AI-driven medical imaging research. Metrics Reloaded: Recommendations and Online Toolkit for Image Analysis Validation Speaker: Emre Kavur Date & Time: Monday, March 10, 15:25 PM (V16) Paper: https://lnkd.in/ghi787j7  Website: https://lnkd.in/eZNtNqkF nnU-Net Revisited Speaker: Constantin Ulrich Date & Time: Tuesday, March 11, 15:20 PM (V29) Paper: https://lnkd.in/eRHk-mmC GitHub: https://lnkd.in/eq_AcFj6 These sessions will provide an opportunity to engage in discussions on evaluation metrics, deep learning architectures, and their applications in medical imaging. We look forward to insightful conversations with the community. See you at BVM 2025!

👀 Exploring Our Poster Contributions at BVM 2025! Following our hands-on Kaapana demo at BVM - German Conference on Medical Image Computing, we are excited to highlight another key aspect of our participation: our poster presentations. Our division is contributing research across a range of topics, from precision ICU resource planning to federated learning security and longitudinal segmentation techniques. If you are attending BVM, be sure to visit our posters and engage with the authors. Here’s a quick overview of our contributions: P11 – Robin Peretzke and Maximilian Fischer: Precision ICU Resource Planning: A Multimodal Model for Brain Surgery Outcomes https://lnkd.in/eCDpYaWc P14 - Markus Bujotzek: Real-world Federated Learning in Radiology https://lnkd.in/eBtyjsnr P20 – Yannick Kirchhoff and Maximilian R.: Skeleton Recall Loss for Connectivity Conserving and Resource Efficient Segmentation of Thin Tubular Structures https://lnkd.in/eyvzk-rG P21 – Constantin Ulrich: A Unified Framework for Foreground and Anonymization Area Segmentation in CT and MRI Data https://lnkd.in/ereVmHHm P31 – Santhosh Parampottupadam: Client Security Alone Fails in Federated Learning: 2D and 3D Attack Insights https://lnkd.in/eNRar7NU P41 – Yannick Kirchhoff and Maximilian R.: Longitudinal Segmentation of MS Lesions via Temporal Difference Weighting https://lnkd.in/eQQ-BgUQ S020 - Maximilian Fischer: Learned Image Compression for HE-stained Histopathological Images via Stain Deconvolution https://lnkd.in/ek8bXiQT These posters represent a diverse set of challenges and solutions in medical image analysis, machine learning, and AI-driven healthcare research. We look forward to insightful discussions at the conference. Stay tuned for more updates on our contributions to BVM 2025!