Abstract
Perivascular spaces (PVS), also known as Virchow–Robin spaces, are key imaging biomarkers for diagnosing neurological disorders. Magnetic resonance imaging (MRI) can reliably depict these tiny structures, providing a basis for quantitative analysis. However, in clinical practice, accurate segmentation of PVS is highly challenging due to their small size, sparse distribution, and complex morphology. To address this challenge, we propose a vision–language model (VLM)-based framework for precise PVS delineation. Specifically, we introduce a Hierarchical Textual Conditioning mechanism to adaptively fuse global semantic priors with local morphological descriptions, improving the localization of sparsely distributed PVS. Furthermore, we design a Dual-Domain Cooperative Attention module to amplify responses to small PVS structures and improve the completeness of small-object segmentation. In addition, we propose an Uncertainty-guided Boundary–Shape Consistency module that explicitly models feature and boundary uncertainty via a learnable Gaussian membership function, thereby enabling better adaptation to diverse PVS morphological variants and improving contour stability. Experimental validation on an independent T1-weighted MRI dataset demonstrates that our method outperforms existing state-of-the-art approaches.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/SWITCH_021.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=4pauh3hFmn
BibTex
@InProceedings{XuWan_UncertaintyGuided_MICCAISAT2026,
author = { Xu, Wangyang AND Chen, Tao AND Wang, Zihan AND Long, Xi AND Breeuwer, Marcel AND Zinger, Sveta AND Huang, Peiyu AND Zhang, Jiong},
title = { { Uncertainty-Guided Hierarchical Textual Modeling for Perivascular Space Segmentation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17255},
month = {pending},
page = {pending}
}