Search papers, labs, and topics across Lattice.
This paper introduces WSI-INR, a novel patch-free framework for lesion segmentation in whole-slide images (WSIs) that leverages Implicit Neural Representations (INRs) to model WSIs as continuous functions mapping spatial coordinates to tissue semantics. Multi-resolution hash grid encoding is used to achieve consistent feature representation across different resolutions, and a shared INR decoder captures general priors across different cases. Experiments demonstrate that WSI-INR maintains robust segmentation performance across resolutions, significantly outperforming patch-based U-Net and TransUNet, especially at lower resolutions.
Ditch the patches: Implicit Neural Representations can now segment highly variable pathological lesions in whole-slide images, even outperforming standard methods at lower resolutions.
Whole-slide images (WSIs) are fundamental for computational pathology, where accurate lesion segmentation is critical for clinical decision making. Existing methods partition WSIs into discrete patches, disrupting spatial continuity and treating multi-resolution views as independent samples, which leads to spatially fragmented segmentation and reduced robustness to resolution variations. To address the issues, we propose WSI-INR, a novel patch-free framework based on Implicit Neural Representations (INRs). WSI-INR models the WSI as a continuous implicit function mapping spatial coordinates directly to tissue semantics features, outputting segmentation results while preserving intrinsic spatial information across the entire slide. In the WSI-INR, we incorporate multi-resolution hash grid encoding to regard different resolution levels as varying sampling densities of the same continuous tissue, achieving a consistent feature representation across resolutions. In addition, by jointly training a shared INR decoder, WSI-INR can capture general priors across different cases. Experimental results showed that WSI-INR maintains robust segmentation performance across resolutions; at Base/4, our resolution-specific optimization improves Dice score by +26.11%, while U-Net and TransUNet decrease by 54.28% and 36.18%, respectively. Crucially, this work enables INRs to segment highly heterogeneous pathological lesions beyond structurally consistent anatomical tissues, offering a fresh perspective for pathological analysis.
