EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Adaptive multi-stage domain unlearning for white-matter lesion segmentation

Domen Preložnik, Žiga Špiclin

Frontiers in Medicine · 31.07.2026 · 13

DOI: 10.3389/fmed.2026.1875760 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Medicine, online erschienen am 31.07.2026. Verfasst von Domen Preložnik und 1 weiteren Autor:innen. Beteiligte Einrichtung: Faculty of Electrical Engineering, University of Ljubljana, Ljubljana, Slovenia. Die Arbeit nennt 61 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: The Slovenian Research and Innovation Agency.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Introduction

Inter-scanner variability in magnetic resonance imaging (MRI) adversely affects the diagnostic and prognostic quality of scans and necessitates the development of models that are robust to domain shift arising from the unseen scanner data. A review of recent advances in domain adaptation and domain generalization showed that the efficacy of strategies involving modifications or constraints on the latent space appears to be contingent upon the level and/or depth of supervision during model training.

Methods

We propose an adaptive multi-stage domain unlearning (ADMU) technique to improve robustness to unseen scanner domains. Building on the state-of-the-art segmentation framework nnU-Net, we employ deep supervision at deep encoder stages by applying domain classifier unlearning, sequentially across these stages to reduce domain-discriminative latent features. Following the self-configurable approach of nnU-Net, the auxiliary feedback loop implements an adaptive backpropagation schedule for unlearning. Experiments were conducted on four public datasets (one for training, three for testing) to benchmark white-matter lesion segmentation methods. Five benchmark models and/or strategies, spanning passive to active domain-robust training strategies, were tested, and five state-of-the-art methods were compared.

Results

AMDU demonstrated consistent, robust and improved cross-dataset segmentation performance on three test sets versus baseline nn-Unet variants. The advantage of AMDU was in enhanced lesion sensitivity with balanced false detections, resulting in good overall segmentation quality, as measured by segmentation overlap and relative lesion volume error. Compared to continuous domain unlearning, the adaptive scheduling balanced the adverse impact of unlearning onto the main segmentation task. Intensity-based preprocessing was found to be detrimental to segmentation performance.

Discussion

The proposed AMDU strategy was shown to be complementary to data augmentation. Demonstrated for white-matter lesion segmentation it relied only the FLAIR modality, simplifying preprocessing to spatial normalization to brain atlas, with no intensity harmonization, for best cross-dataset segmentation performance. The source code is available at https://github.com/Pubec/nnunetv2-unlearning.

Schlagwörter

Stichwörter der Autor:innen

adaptive domain unlearningcomparative evaluationdomain generalizationimage segmentationopen-source and reproduciblescanner variability

Autor:innen und Institutionen

  • Faculty of Electrical Engineering, University of Ljubljana, Ljubljana, Slovenia.

Förderung

  • The Slovenian Research and Innovation Agency

Zitieren

APA

Preložnik, D., & Špiclin, Ž. (2026). Adaptive multi-stage domain unlearning for white-matter lesion segmentation. Frontiers in Medicine, 13. https://doi.org/10.3389/fmed.2026.1875760

Vancouver

Preložnik D, Špiclin Ž. Adaptive multi-stage domain unlearning for white-matter lesion segmentation. Frontiers in Medicine. 2026;13. doi:10.3389/fmed.2026.1875760

BibTeX
@article{Prelonik2026,
  author = {Preložnik, Domen and Špiclin, Žiga},
  title = {Adaptive multi-stage domain unlearning for white-matter lesion segmentation},
  journal = {Frontiers in Medicine},
  year = {2026},
  volume = {13},
  doi = {10.3389/fmed.2026.1875760}
}
Kostenlos starten →