Thanks to our awesome collaborators! Preeti Verma, Yuwen Chen, Hanxue Gu, Haoyu Dong, Yaqian Chen, Andrew Marshall, Lidia Garrucho Moras, Kaisar Kushibar, Daniel M. Lang, Gene S. Kim, Lars Grimm, John Lewin, James S. Duncan, @ja-schnabel.bsky.social, Oliver Diaz, Karim Lekadir and Maciej Mazurowski.

As part of this paper, we also present the most extensive framework for the meta-evaluation of medical image similarity metrics to date, available at github.com/mazurowski-l...!

FRD improves on other metrics (FID, RadiologyFID, KID, etc.) in many applications, including: OOD detection, image-to-image translation/image generation evaluation, correlation with expert-perceived image quality, compute stability+speed, and sensitivity to adversarial attacks + image corruptions.

FRD quantifies images via hundreds of standardized, interpretable radiomic features, rather than learned embeddings, which we find brings several advantages due to better and more robustly characterizing anatomical features (particularly those related to downstream tasks).

Thanks for reading! Co-authors: Zafer Yildiz, Qihang Li, Yaqian Chen, Haoyu Dong, Hanxue Gu and Maciej Mazurowski

With a novel dynamic "short-long" memory, it outperforms SAM 2 and other models by >7% DSC on average, in segmenting various organs, bones, and muscles across modalities! Crucially, it exhibits better robustness to the problem of "over-propagation" of annotations through slices.

Sadly, I’m unable to attend the conference next week, but I wanted to share the paper with those interested, and I am happy to answer any questions! (3/3)

I also found that the representation intrinsic dim. peaks consistently earlier in medical image models compared to natural image models, pointing to a difference in the abstractness of task-relevant features between these domains. (2/3)