Biological Spatial Priors Regularize Foundation Model Representations for Cross-Site MSI Generalization in Colorectal Cancer

• URL: http://arxiv.org/abs/2605.02660v1
• Date: Mon, 04 May 2026 14:39:40 GMT
• Title: Biological Spatial Priors Regularize Foundation Model Representations for Cross-Site MSI Generalization in Colorectal Cancer
• Abstract summary: We introduce a biologically motivated spatial prior based on peripheral distance encoding.<n>We evaluate six foundation model and MIL aggregator combinations as a reference, then assess the effect of each spatial prior.<n>Results suggest biologically grounded spatial priors act as regularizers that reduce reliance on site-specific imaging patterns.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Predicting microsatellite instability (MSI) status from routine hematoxylin and eosin (H&E) whole slide images (WSIs) offers a practical alternative to molecular testing, but models trained at one institution tend to generalize poorly to slides acquired at a different site. Foundation model representations, despite their generality, still encode site-specific texture alongside the conserved biological morphology underlying MSI. We investigate whether tile-level spatial priors derived from known MSI histology can guide these representations toward more site-invariant features. We introduce a biologically motivated spatial prior based on peripheral distance encoding, reflecting the Crohn's-like peripheral lymphocytic reaction at the tumor invasive margin, and evaluate a secondary local immune neighborhood encoding reflecting the lymphocyte-to-tumor ratio in each tile's immediate spatial neighborhood. Both priors are injected into a TransMIL aggregator before self-attention, allowing the transformer to integrate spatial biological context with UNI2-h or Virchow2 features across all attention layers. We evaluate six foundation model and MIL aggregator combinations as a reference, then assess the effect of each spatial prior. Training on TCGA-COAD (137 slides) and evaluating externally on TCGA-READ (50 slides) without retraining, peripheral distance encoding achieves MSI AUC 0.959 +/- 0.012 on COAD and MSS specificity 1.000 on READ, compared to 0.957 and 0.939 for the strongest reference configuration. Local immune neighborhood encoding achieves comparable internal AUC but lower cross-site specificity, suggesting margin proximity encodes a more site-invariant biological signal than local immune density. Results suggest biologically grounded spatial priors act as regularizers that reduce reliance on site-specific imaging patterns.

Related papers

• Bridging Vision Foundation Model Priors with CLIP for Spatial-aware Few-shot Anomaly Detection in Medical Images [48.646448077284276]
  We propose Spatial-FAD, a framework for effective few-shot medical anomaly detection.<n>It improves lesion localization by combining VFM spatial priors with CLIP semantics.<n>Our method achieves an average improvement of over 11.4% in Dice score and 1.8% in AUC.
  arXiv  Detail & Related papers  (2026-09-11T05:21:01Z)
• Conserved Immune Topology Improves Pathology Foundation Model Generalization for Cross-Cancer MSI-H Prediction [0.0]
  We propose Conserved Immune Top (CIT), a lightweight spatial representation for cross-cancer MSI-H prediction.<n>CIT uses unsupervised clustering to identify immune-associated tiles, then encodes lymphoid structures, peritumoral immune reactions, multi-scale tumor-infiltrating lymphocyte density.<n>Results suggest that CIT provides potentially an organ-invariant representation for MSI-H prediction, supporting cross-cancer generalization of pathology foundation models.
  arXiv  Detail & Related papers  (2026-09-04T14:17:03Z)
• Disentangling Co-Occurring Retinal Pathologies with Saliency-Guided Sparse Expert Routing [3.332219204498619]
  We propose a novel architecture that resolves this via sparse conditional computation.<n>We show that expert routing aligns with localized lesions and geometrically maps co-occurring cases.
  arXiv  Detail & Related papers  (2026-08-10T15:43:43Z)
• Rethinking Attention Locality in Spiking Transformers [50.00501275379363]
  Spiking Transformers provide a promising paradigm for efficient visual processing with spike-driven computation.<n>Their Softmax-free Spiking Self-Attention (SSA) struggles to establish spatially localized token interactions.<n>We propose Spatially Contiguous Local Attention with Boundary Continuity Pathway (SCLA-BCP)<n>Our approach improves mAP@50 by up to 9.50% on COCO on 2017 and mIoU by up to 3.42% on ADE20K.
  arXiv  Detail & Related papers  (2026-08-09T07:36:17Z)
• Predicting Immune Biomarkers with MultiModal Mixture-of-Expert Pathology Foundation Models Empowers Precision Oncology [47.79180201495836]
  MixTIME is a multimodal foundation model that integrates pathology foundation models trained across distinct modalities.<n>It predicts multiplex immunofluorescence (mIF) protein expression from hematoxylin and eosin whole-slide images.<n>MixTIME achieves state-of-the-art performance across 17 protein markers as measured by correlation metrics.
  arXiv  Detail & Related papers  (2026-06-16T16:22:42Z)
• A physics-informed foundation model for quantitative diffusion MRI [46.74037857287099]
  A physics-informed generative microstructure network (PIGMENT) learns a universal generative prior to human brain microstructure and adapts it zero-shot to each participant's measured data to recover subject-specific maps.<n>PIGMENT estimates demonstrated strong biological validity, preserving submillimeter cortical microarchitectural patterns and early-childhood white matter developmental trajectories from 10-fold accelerated scans.<n>These results establish PIGMENT as a physics-informed foundation model that extends quantitative diffusion MRI into regimes traditionally too sparse, heterogeneous, or clinically constrained for reliable analysis.
  arXiv  Detail & Related papers  (2026-05-29T07:26:35Z)
• FLORO: A Multimodal Geospatial Foundation Model for Ecological Remote Sensing Across Sensors and Scales [65.4821703903285]
  FLORO is a multimodal foundation model designed to learn transferable representations from a small but highly diverse remote sensing corpus.<n> FLORO is pretrained using masked autoencoding on a heterogeneous combination of Sentinel-1, Sentinel-2, SkySAT imagery, elevation, and UAV-derived data.<n>We evaluated FLORO on the PANGAEA benchmark under a frozen-encoder protocol across scene classification, segmentation, and regression tasks.
  arXiv  Detail & Related papers  (2026-05-27T08:55:54Z)
• PC-MIL: Decoupling Feature Resolution from Supervision Scale in Whole-Slide Learning [15.567271338271032]
  Whole-slide image (WSI) classification in computational pathology is commonly formulated as slide-level Multiple Instance Learning (MIL) with a single global bag representation.<n>We propose Progressive-Context MIL (PC-MIL), a framework that treats the spatial extent of supervision as a first-class design dimension.
  arXiv  Detail & Related papers  (2026-04-13T22:19:28Z)
• Learning to Synergize Semantic and Geometric Priors for Limited-Data Wheat Disease Segmentation [51.06355116973389]
  We propose a framework that treats wheat disease segmentation as a coupled task of disease-specific semantic perception and disease boundary localization.<n>Our core insight is that pretrained DINOv2 provides robust category-aware semantic priors to handle appearance shifts.<n>We show that SGPer consistently achieves state-of-the-art performance on wheat disease and organ segmentation benchmarks.
  arXiv  Detail & Related papers  (2026-04-07T04:19:39Z)
• A Unified Framework for Joint Detection of Lacunes and Enlarged Perivascular Spaces [3.9313804276175506]
  Cerebral small vessel disease (CSVD) markers, specifically enlarged perivascular spaces (EPVS) and lacunae, present a unique challenge in medical image analysis.<n>We propose a morphologydecoupled framework where Zero- Gated CrossTask Attention exploits dense EPVS context to guide sparse lacune detection.
  arXiv  Detail & Related papers  (2026-03-04T16:30:46Z)
• A Tumor Aware DenseNet Swin Hybrid Learning with Boosted and Hierarchical Feature Spaces for Large-Scale Brain MRI Classification [0.6991281327290525]
  Two tumor aware experimental setups are introduced to address class-specific diagnostic challenges.<n>DenseNet is customized at the input level to match MRI spatial characteristics.<n>Swin_t models global tumor morphology, effectively suppressing false negatives in meningioma and pituitary tumor classification.
  arXiv  Detail & Related papers  (2026-01-26T10:14:57Z)
• SAGE-FM: A lightweight and interpretable spatial transcriptomics foundation model [2.060288975192133]
  We introduce SAGE-FM, a lightweight spatial transcriptomics foundation model based on graph convolutional networks (GCNs)<n>Trained on 416 human Visium samples spanning 15 organs, SAGE-FM learns spatially coherent embeddings that robustly recover masked genes.<n>Results demonstrate that simple, parameter-efficient GCNs can serve as biologically interpretable and spatially aware foundation models for large-scale spatial transcriptomics.
  arXiv  Detail & Related papers  (2026-01-21T22:22:38Z)
• Learning to Factorize and Adapt: A Versatile Approach Toward Universal Spatio-Temporal Foundation Models [42.152122602443164]
  We present FactoST-v2, an enhanced factorized framework for universal temporal learning.<n>We show that FactoST-v2 achieves state-of-the-art accuracy with linear efficiency.<n>This factorized paradigm offers a practical, scalable path toward truly universal STFMs.
  arXiv  Detail & Related papers  (2026-01-17T15:20:08Z)
• An Interpretable Implicit-Based Approach for Modeling Local Spatial Effects: A Case Study of Global Gross Primary Productivity [9.352810748734157]
  In Earth sciences, unobserved factors exhibit non-stationary distributions, causing the relationships between features and targets to display spatial heterogeneity.<n>In geographic machine learning tasks, conventional statistical learning methods often struggle to capture spatial heterogeneity.<n>We propose a novel perspective - that is, simultaneously modeling common features across different locations alongside spatial differences using deep neural networks.
  arXiv  Detail & Related papers  (2025-02-10T05:44:54Z)
• CIMIL-CRC: a clinically-informed multiple instance learning framework for patient-level colorectal cancer molecular subtypes classification from H\&E stained images [42.771819949806655]
  We introduce CIMIL-CRC', a framework that solves the MSI/MSS MIL problem by efficiently combining a pre-trained feature extraction model with principal component analysis (PCA) to aggregate information from all patches. We assessed our CIMIL-CRC method using the average area under the curve (AUC) from a 5-fold cross-validation experimental setup for model development on the TCGA-CRC-DX cohort.
  arXiv  Detail & Related papers  (2024-01-29T12:56:11Z)
• Corneal endothelium assessment in specular microscopy images with Fuchs' dystrophy via deep regression of signed distance maps [48.498376125522114]
  This paper proposes a UNet-based segmentation approach that requires minimal post-processing. It achieves reliable CE morphometric assessment and guttae identification across all degrees of Fuchs' dystrophy.
  arXiv  Detail & Related papers  (2022-10-13T15:34:20Z)
• Site Generalization: Stroke Lesion Segmentation on Magnetic Resonance Images from Unseen Sites [40.32385363670918]
  Strokes are the main cause of various cerebrovascular diseases. Deep learning-based models have been proposed for this task, but generalizing these models to unseen sites is difficult. We propose a U-net--based segmentation network termed SG-Net to improve unseen site generalization for stroke lesion segmentation on MR images.
  arXiv  Detail & Related papers  (2022-05-09T14:33:06Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.