Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Employing Graph Representations for Cell-level Characterization of Melanoma MELC Samples

Luis Carlos Rivera Monroy, Leonhard Rist, Martin Eberhardt, Christian Ostalecki, Andreas Baur, Julio Vera, Katharina Breininger, Andreas Maier

TL;DR

This work describes a pipeline that uses suspected melanoma samples that have been characterized using Multi-Epitope-Ligand Cartography (MELC), which achieves classification accuracy of 87 %, outperforming existing approaches by 10 %.

Abstract

Histopathology imaging is crucial for the diagnosis and treatment of skin diseases. For this reason, computer-assisted approaches have gained popularity and shown promising results in tasks such as segmentation and classification of skin disorders. However, collecting essential data and sufficiently high-quality annotations is a challenge. This work describes a pipeline that uses suspected melanoma samples that have been characterized using Multi-Epitope-Ligand Cartography (MELC). This cellular-level tissue characterisation is then represented as a graph and used to train a graph neural network. This imaging technology, combined with the methodology proposed in this work, achieves a classification accuracy of 87%, outperforming existing approaches by 10%.

Employing Graph Representations for Cell-level Characterization of Melanoma MELC Samples

TL;DR

This work describes a pipeline that uses suspected melanoma samples that have been characterized using Multi-Epitope-Ligand Cartography (MELC), which achieves classification accuracy of 87 %, outperforming existing approaches by 10 %.

Abstract

Histopathology imaging is crucial for the diagnosis and treatment of skin diseases. For this reason, computer-assisted approaches have gained popularity and shown promising results in tasks such as segmentation and classification of skin disorders. However, collecting essential data and sufficiently high-quality annotations is a challenge. This work describes a pipeline that uses suspected melanoma samples that have been characterized using Multi-Epitope-Ligand Cartography (MELC). This cellular-level tissue characterisation is then represented as a graph and used to train a graph neural network. This imaging technology, combined with the methodology proposed in this work, achieves a classification accuracy of 87%, outperforming existing approaches by 10%.
Paper Structure (12 sections, 2 figures, 1 table)

This paper contains 12 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Dataset
  4. Graph Representation
  5. Dimensionality Reduction
  6. Graph Neural Network (GNN)
  7. Experimental Setup and Metrics
  8. Experimental Results
  9. Discussion
  10. Conclusions and Future Work
  11. Compliance with ethical standards
  12. Acknowledgments

Figures (2)

  • Figure 1: Overview of the proposed methodology. In the first place, the segmentation of cell instances (a), and second the representation of the stain profiles as tabular data (b). Thirdly, the graph representation of the cells (c) and, finally, the node classification was trained using a graph convolutional neural network (d). The outcome is a binary classification for cell instances on the tissue sample (e).
  • Figure 2: Melanoma MELC sample. Tissue stained to the five cell sub-types targets and example of staining agent specific for each type. Cell nuclei (Propidium iodide), Immune cells (CD43), Vascular cells (Collagen IV), Epithelial cells (Phospho Connexin), Melanoma associated stain agents (CD63), and the manual annotation of the melanoma tumour provided by the specialist.