🎁 Wrapping up Medical Imaging with Deep Learning (MIDL) 2025 in Salt Lake City (July 9–11), and what an inspiring conference it was!
Our team from FAU Erlangen-Nürnberg, Mischa Dombrowski, Felix Nützel, and myself had the pleasure of presenting three full papers and one short paper, showcasing cutting-edge work in medical generative models, visual grounding, and trustworthy AI evaluation metrics.
🧠 A few highlights:
🔹 “Generate to Ground”: We showed how generative text-to-image diffusion models can outperform state-of-the-art discriminative methods in grounding medical phrases in images, doubling the accuracy of previous approaches! [Felix Nützel, Mischa Dombrowski, Bernhard Kainz]
🔹 “Can Diffusion Models Generalize?”: We introduced a new metric (t') for assessing privacy and fairness trade-offs in generative medical imaging models, which is key for secure synthetic data sharing. [Mischa Dombrowski, Bernhard Kainz]
🔹 “CRG Score”: A new clinically-aware metric for evaluating long radiology reports, addressing class imbalance and boosting interpretability. [Ibrahim Hamamci, Sezgin Er, Suprosanna Shit, Hadrien Reynaud, Bernhard Kainz, Bjoern Menze]
🔹 “Intelligent Lesion Selection”: A collaboration with Siemens Healthineers and Helsinki University Hospital on automated RECIST-like tracking of lung metastases in breast cancer. [Melika Qahqaie et al.]
💻 All projects were powered by Erlangen National High Performance Computing Center (NHR@FAU) new helma GPU cluster, demonstrating the impact of German supercomputing in AI for healthcare.
We also had the chance to connect with fantastic researchers and old friends Daniel Rueckert, Ipek Oguz, Mirabela Rusu, Ronald Summers, Tal Arbel, Mattias Paul Heinrich, Mathias Unberath, Alessa Hering, Tolga Tasdizen, and many more brilliant PhD students.
Huge thanks to the MIDL organizers for fostering such a vibrant, open scientific community.
🔗 Papers: https://lnkd.in/dzWazZPQ
#MIDL2025 #MedicalImaging #DeepLearning #FAU #NHRFAU #GenerativeAI #AIinMedicine #Radiology #DiffusionModels #PrivacyFairness #AITrustworthiness #ResearchImpact #FAUErlangen
#HTA