Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Lymphoid Infiltration Assessment of the Tumor Margins in H&E Slides

Zhuxian Guo, Amine Marzouki, Jean-François Emile, Henning Müller, Camille Kurtz, Nicolas Loménie

TL;DR

This work addresses the limitations of IHC-based lymphoid infiltration assessment at tumor margins by proposing an H&E-based pipeline. It combines a contextual-aware segmentation model trained on public lymphocyte datasets with a distance-transform approach to generate lymphoid infiltration curves, enabling margin-focused quantification in colorectal cancer. The method demonstrates cross-center generalization (Dice scores on external datasets) and shows that H&E-derived curves closely resemble IHC benchmarks, with a Turing-test–style clinical validation indicating potential utility as a supplementary tool in immunotherapy planning. The study advocates adopting Turing-test–inspired validation in medical AI and points to future work on clinically meaningful cut-offs and automated neoplastic tissue segmentation to further integrate H&E-based assessment into practice.

Abstract

Lymphoid infiltration at tumor margins is a key prognostic marker in solid tumors, playing a crucial role in guiding immunotherapy decisions. Current assessment methods, heavily reliant on immunohistochemistry (IHC), face challenges in tumor margin delineation and are affected by tissue preservation conditions. In contrast, we propose a Hematoxylin and Eosin (H&E) staining-based approach, underpinned by an advanced lymphocyte segmentation model trained on a public dataset for the precise detection of CD3+ and CD20+ lymphocytes. In our colorectal cancer study, we demonstrate that our H&E-based method offers a compelling alternative to traditional IHC, achieving comparable results in many cases. Our method's validity is further explored through a Turing test, involving blinded assessments by a pathologist of anonymized curves from H&E and IHC slides. This approach invites the medical community to consider Turing tests as a standard for evaluating medical applications involving expert human evaluation, thereby opening new avenues for enhancing cancer management and immunotherapy planning.

Lymphoid Infiltration Assessment of the Tumor Margins in H&E Slides

TL;DR

This work addresses the limitations of IHC-based lymphoid infiltration assessment at tumor margins by proposing an H&E-based pipeline. It combines a contextual-aware segmentation model trained on public lymphocyte datasets with a distance-transform approach to generate lymphoid infiltration curves, enabling margin-focused quantification in colorectal cancer. The method demonstrates cross-center generalization (Dice scores on external datasets) and shows that H&E-derived curves closely resemble IHC benchmarks, with a Turing-test–style clinical validation indicating potential utility as a supplementary tool in immunotherapy planning. The study advocates adopting Turing-test–inspired validation in medical AI and points to future work on clinically meaningful cut-offs and automated neoplastic tissue segmentation to further integrate H&E-based assessment into practice.

Abstract

Lymphoid infiltration at tumor margins is a key prognostic marker in solid tumors, playing a crucial role in guiding immunotherapy decisions. Current assessment methods, heavily reliant on immunohistochemistry (IHC), face challenges in tumor margin delineation and are affected by tissue preservation conditions. In contrast, we propose a Hematoxylin and Eosin (H&E) staining-based approach, underpinned by an advanced lymphocyte segmentation model trained on a public dataset for the precise detection of CD3+ and CD20+ lymphocytes. In our colorectal cancer study, we demonstrate that our H&E-based method offers a compelling alternative to traditional IHC, achieving comparable results in many cases. Our method's validity is further explored through a Turing test, involving blinded assessments by a pathologist of anonymized curves from H&E and IHC slides. This approach invites the medical community to consider Turing tests as a standard for evaluating medical applications involving expert human evaluation, thereby opening new avenues for enhancing cancer management and immunotherapy planning.
Paper Structure (11 sections, 4 figures, 2 tables)

This paper contains 11 sections, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Datasets
  3. Method
  4. Lymphocyte segmentation in H&E
  5. Distance transform
  6. Lymphoid infiltration curve
  7. Experimental study
  8. Evaluation of lymphocyte segmentation in H&E
  9. Evaluation of infiltration curve similarity
  10. Enhanced clinical validation: A Turing test approach
  11. Conclusion

Figures (4)

  • Figure 1: Workflow of the lymphoid infiltration assessment of the tumor margin. Annotations by an expert pathologist on the H&E (A) and its corresponding IHC (B) slides. (C), (D): Lymphocyte segmentation results on the paired slides (identified lymphocytes in green). Lymphocyte density curves as a function of distance from the tumor margin from H&E (E) and IHC (F) slides.
  • Figure 2: Selected segmentation results on NuCLS and BCa. The annotations in the ground truth images of BCa dataset were dilated for visualisation purpose.
  • Figure 3: A tertiary lymphoid structure (TLS, a lymphoid follicle, a cluster of CD20$^{+}$ cells surrounded by CD3$^{+}$ cells). For H&E images, both CD3$^{+}$ and CD20$^{+}$ will be identified (middle). The CD3$^{+}$ in its corresponding IHC lymphocyte cluster (right).
  • Figure 4: Tumor cell in necrotic areas within large tumor glands and lymphocytes appear similar. Using a contextual aware neural network, a U-Net architecture with frozen Mix Vision Transformer encoder (UNet (MixVit-B1)), has better performance in distinguishing necrotic cells and lymphocytes.