cs.CVJan 13, 2026

ISLA: Ensemble learning and domain adaptation for generalizable acute ischemic stroke lesion segmentation in diffusion MRI with clinical validation

Authors: Vincent Roca, Hakim Daouairi, Sevak Yepremian, Hilde Henon, Nicolas Martinage, Laurent Puy, Grégory Kuchcinski, Martin Bretzner, 

Organizations: Univ. Lille, CNRS, Inserm, CHU Lille, Institut Pasteur de Lille, US 41 - UAR 2014 - PLBS, F-59000 Lille, France · CHU Lille, Département de Neurologie vasculaire, F-59037 Lille, France · Univ. Lille, Inserm, CHU Lille, U1172 - Lille Neurosciences & Cognition, F-59037 Lille, France · CHU Lille, Département de Neuroradiologie, F-59037 Lille, France · 

Abstract

Accurate segmentation of acute ischemic stroke (AIS) lesions on MRI is essential for clinical decision-making. Deep learning methods offer a promising approach to automating this task. However, they often generalize poorly to unseen clinical sites, and few AIS lesion segmentation methods have been externally validated. In this study, we propose ISLA\mathit{ISLA} (Ischemic Stroke Lesion Analyzer), a framework for AIS lesion segmentation built on the design principles of nnU-Net\mathit{nnU\text{-}Net}. ISLA\mathit{ISLA} combines attention modules, deep supervision, and ensemble learning across different input resolutions, convolutional block types, and loss functions. We also introduce ISLA-MT\mathit{ISLA\text{-}MT}, a variant that leverages unlabeled target-domain images through semi-supervised learning. We trained both approaches on the SOOP and ISLES 2022 datasets and evaluated on a held-out internal validation set of 253 cases, as well as on 590 cases from an independent multicenter clinical cohort. We compared them with three state-of-the-art pre-trained models and with nnU-Net\mathit{nnU\text{-}Net} trained on the same data. ISLA\mathit{ISLA} outperformed the pre-trained models on the internal validation set, while both ISLA\mathit{ISLA} and ISLA-MT\mathit{ISLA\text{-}MT} outperformed them on the external cohort. Although nnU-Net\mathit{nnU\text{-}Net} performed slightly better than ISLA\mathit{ISLA} internally, both ISLA\mathit{ISLA} variants demonstrated better generalization to the external cohort. Furthermore, ISLA-MT\mathit{ISLA\text{-}MT} significantly improved performance over ISLA\mathit{ISLA} on the external cohort. The proposed models performed particularly well on low-contrast lesions, and qualitative error analyses suggested greater robustness to acquisition artifacts. Both models are fast (∼\sim 5 s per case), and their code and weights are publicly available at https://github.com/RocaVincent/isla_segmentation.

Figures & tables

Appendix figures & tables14 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Deep Learning Segmentation of Diffusion-Weighted MRI Acute Ischaemic Stroke: A Pragmatic Evaluation Across Three Datasets

    Aug 26, 2026Atle Bjørnerud, Till Schellhorn, Thor H. Skattør +4Medical Image SegmentationU-Net

  2. Automated Lesion Segmentation of Stroke MRI Using nnU-Net: A Comprehensive External Validation Across Acute and Chronic Lesions

    Jan 13, 2026Tammar Truzman, Matthew A. Lambon Ralph, Ajay D. HalaiMedical Image SegmentationConvolutional Neural Networks

  3. Native-Space 3D CarveMix for Multi-Site T1w Stroke Segmentation

    Aug 24, 2026Dexter Wen Jie Teo, Kumaradevan PunithakumarMedical Image Segmentation3D Medical Image Segmentation