We travelled to MICCAI 2025

We travelled to Daejeon, South Korea, to present 3 MICCAI main conference papers and a workshop paper.

This year, we presented the following papers at MICCAI.

Main conference papers:

• Prototype-Based Multiple Instance Learning for Gigapixel Whole Slide Image Classification by Susu Sun. Susu developed an inherently interpretable MIL classifier for histopathology images that operates on concepts automatically discovered using a sparse autoencoder.
  
• Mind the Detail: Uncovering Clinically Relevant Image Details in Accelerated MRI with Semantically Diverse Reconstructions by Jan Nikolas Morshuis. Nikolas developed a method based on adversarial perturbations that enables MR reconstruction techniques to generate semantically diverse reconstructions, helping to reveal rare pathologies that might otherwise be missed.
  
• Subgroup Performance Analysis in Hidden Stratifications by Alceu Bissoto and Lisa Koch from the MLM Lab in Bern. In this joint work, led by our colleagues at the University of Bern, we investigated whether subgroup discovery can serve as a tool for performance validation and monitoring in deep learning based medical diagnosis systems.

UNSURE workshop paper:

• Is Uncertainty Quantification a Viable Alternative to Learned Deferral? by Anna M. Wundram. In this paper, Anna investigated whether uncertainty quantification methods can be used to effectively defer medical decisions to human experts in both in-domain and out-of-domain settings.

Outside of the conference, we also got to enjoy Korean culture including amazing Korean food. It was also a very special MICCAI for me, because I got to travel to Korea together with my family.

Images