mlsanityData to 14 Oct 2026

Margin-Consistent Deep Subtyping of Invasive Lung Adenocarcinoma via Perturbation Fidelity in Whole-Slide Image Analysis

arXiv 2603.06650 ↗First posted 27 Feb 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Whole-slide image classification for invasive lung adenocarcinoma subtyping remains vulnerable to real-world imaging perturbations that undermine model reliability at the decision boundary. We propose a margin consistency framework evaluated on 203,226 patches from 143 whole-slide images spanning five adenocarcinoma subtypes in the BMIRDS-LUAD dataset. By combining attention-weighted patch aggregation with margin-aware training, our approach achieves robust feature-logit space alignment measured by Kendall correlations of 0.88 during training and 0.64 during validation. Contrastive regularization, while effective at improving class separation, tends to over-cluster features and suppress fine-grained morphological variation; to counteract this, we introduce Perturbation Fidelity (PF) scoring, which imposes structured perturbations through Bayesian-optimized parameters. Vision Transformer-Large achieves 95.20 +/- 4.65% accuracy, representing a 40% error reduction from the 92.00 +/- 5.36% baseline, while ResNet101 with an attention mechanism reaches 95.89 +/- 5.37% from 91.73 +/- 9.23%, a 50% error reduction. All five subtypes exceed an area under the receiver operating characteristic curve (AUC) of 0.99. On the WSSS4LUAD external benchmark, ResNet50 with an attention mechanism attains 80.1% accuracy, demonstrating cross-institutional generalizability despite approximately 15-20% domain-shift-related degradation and identifying opportunities for future adaptation research.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 19 of its 33 references that are in the map. First posted on arXiv 27 Feb 2026.

Builds on

19 of 33 references.

References · 19

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 12,049
  3. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,513
  4. 2016Towards Evaluating the Robustness of Neural Networkscited by 2,998
  5. 2015DeepFool: a simple and accurate method to fool deep neural networkscited by 1,318
  6. 2020Prevalence of Neural Collapse during the terminal phase of deep learning trainingcited by 719
  7. 2020Overfitting in adversarially robust deep learningcited by 506
  8. 2020Geometry-aware Instance-reweighted Adversarial Trainingcited by 198
  9. 2018Large Margin Deep Networks for Classificationcited by 112
  10. 2021Probabilistic Margins for Instance Reweighting in Adversarial Trainingcited by 57
  11. 2019Pathologist-level classification of histologic patterns on resected lung adenocarcinoma slides with deep neural networkscited by 43
  12. 2021Multi-Layer Pseudo-Supervision for Histopathology Tissue Semantic Segmentation using Patch-level Classification Labelscited by 43
7 more
  1. 2022WSSS4LUAD: Grand Challenge on Weakly-supervised Tissue Semantic Segmentation for Lung Adenocarcinomacited by 36
  2. 2023Exploring and Exploiting Decision Boundary Dynamics for Adversarial Robustnesscited by 36
  3. 2020Understanding the Decision Boundary of Deep Neural Networks: An Empirical Studycited by 20
  4. 2024Detecting Brittle Decisions for Free: Leveraging Margin Consistency in Deep Robust Classifierscited by 10
  5. 2024Double InfoGAN for Contrastive Analysiscited by 7
  6. 2024Dynamic Label Adversarial Training for Deep Learning Robustness Against Adversarial Attackscited by 4
  7. 2024Semi-Supervised Contrastive VAE for Disentanglement of Digital Pathology Imagescited by 1

Built on by

Not cited yet.