[2602.23509] SegReg: Latent Space Regularization for Improved Medical Image Segmentation

Electrical Engineering and Systems Science > Image and Video Processing arXiv:2602.23509 (eess) [Submitted on 26 Feb 2026] Title:SegReg: Latent Space Regularization for Improved Medical Image Segmentation Authors:Puru Vaish, Amin Ranem, Felix Meister, Tobias Heimann, Christoph Brune, Jelmer M. Wolterink View a PDF of the paper titled SegReg: Latent Space Regularization for Improved Medical Image Segmentation, by Puru Vaish and 4 other authors View PDF Abstract:Medical image segmentation models are typically optimised with voxel-wise losses that constrain predictions only in the output space. This leaves latent feature representations largely unconstrained, potentially limiting generalisation. We propose {SegReg}, a latent-space regularisation framework that operates on feature maps of U-Net models to encourage structured embeddings while remaining fully compatible with standard segmentation losses. Integrated with the nnU-Net framework, we evaluate SegReg on prostate, cardiac, and hippocampus segmentation and demonstrate consistent improvements in domain generalisation. Furthermore, we show that explicit latent regularisation improves continual learning by reducing task drift and enhancing forward transfer across sequential tasks without adding memory or any extra parameters. These results highlight latent-space regularisation as a practical approach for building more generalisable and continual-learning-ready models. Comments: Subjects: Image and Video Processing (eess.IV...