Borderless Ideas, Festive Science: Our Invitation to the 4th IDEA Lab Symposium in Erlangen
At the beginning of the winter season, xAILab Bamberg had the pleasure of participating in the 4th Borderless IDEA Lab Symposium, hosted in Erlangen by the IDEA Lab at FAU Erlangen-Nürnberg. Held in the newly inaugurated AIBE@FAU building, the symposium brought together researchers from across institutions and career stages to share ideas, challenge perspectives, and strengthen collaborations in medical AI; true to its “borderless” spirit.
We were delighted to contribute with two short talks that reflected both the depth and the diversity of research at xAILab.
Our PhD candidate Francesco Di Salvo presented his recent MICCAI work, “Embedding-Based Federated Data Sharing via Differentially Private Conditional VAEs.” Building on the growing need for privacy-preserving collaboration in medical imaging, Francesco outlined how compact, foundation-model-based embeddings combined with differential privacy can enable efficient and secure data sharing across institutions. His talk sparked lively discussion around federated learning, generative models, and practical pathways toward trustworthy medical AI.
Sebastian Doerrich complemented this with a more personal research talk, reflecting on his scientific journey so far and outlining his broader research direction. By connecting methodological developments with long-term research vision, his contribution resonated particularly well with early-career researchers navigating their own paths in academia.
A special highlight for our group was the official introduction of our new PhD student, Hanh Huyen My Nguyen, who recently joined xAILab Bamberg. In her introduction, she shared her academic background and the research questions she is eager to pursue, marking the beginning of her doctoral journey within both the xAILab and the wider IDEA Lab network.
Beyond our own contributions, the symposium offered a rich and varied scientific program. Extended talks from IDEA Lab members Jiayi Wang on Organ-wise Volume-Language Pretraining and Felix Nützel on Guided Residual Adapters with Sample-wise Partitioning for Diffusion Models showcased innovative approaches at the intersection of representation learning and generative modeling. Sandro Sage presented his Master’s thesis on Diffusion Models in K-Space, highlighting the growing role of diffusion-based methods in medical image reconstruction.
The program was further enriched by contributions from outside the host lab, including Manuel Milling (Technical University of Munich) on speech processing, Qingjie Meng (University of Birmingham) on Population-Scale Cardiac Digital Twins, and Jeremy Tan (ETH Zurich) on Synchronized Segmentation. Together, these talks painted a broad picture of current advances in AI for medicine, spanning imaging, multimodal learning, and large-scale physiological modeling.
What truly set the symposium apart, however, was its atmosphere: open, welcoming, and highly interactive. Discussions flowed easily across disciplines and career stages, reinforcing the sense of a shared scientific mission.
To round off the event, we had the pleasure of introducing our international colleagues to Franconia in Christmas time. A joint visit to the Nürnberger Christkindlesmarkt, one of the most famous Christmas markets in the world, provided a festive backdrop for continuing conversations: this time over mulled wine, gingerbread, and regional specialties.
And what's next?
Our participation in the 4th Borderless IDEA Lab Symposium reaffirmed the value of close-knit, interdisciplinary exchange alongside large international conferences. With the end of the year approaching, it is also the perfect time to reflect on the past year, celebrate our achievements, and recharge our batteries - so that we can start the new year energized, carrying forward the new connections, ideas, and collaborations sparked in Erlangen. We return to Bamberg inspired by the science, motivated by fresh perspectives, and excited to further strengthen our collaboration with the IDEA Lab and the broader BAIOSPHERE research community.
We thank the organizers, in particular Johanna P. Müller and Professor Bernhard Kainz for their warm hospitality and look forward to many more borderless ideas, scientific and otherwise.
