Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Structural Graph Neural Networks with Anatomical Priors for Explainable Chest X-ray Diagnosis

Khaled Berkani

TL;DR

This work introduces a structural graph reasoning framework that embeds anatomical priors into patch-level graphs derived from CNN features to enable explainable chest X-ray diagnosis. A novel structural GNN propagates information via spatial displacement-aware message passing, producing intrinsic node-level explanations while delivering strong graph-level diagnostic performance. The approach jointly supports lesion-aware node predictions and global reasoning, achieving superior accuracy and AUC with intrinsic interpretability, demonstrated on ChestX-ray8-like data. By directly encoding spatial anatomy into the graph reasoning layer, the method offers a principled, domain-agnostic substrate for structure-aware and explainable AI across vision tasks.

Abstract

We present a structural graph reasoning framework that incorporates explicit anatomical priors for explainable vision-based diagnosis. Convolutional feature maps are reinterpreted as patch-level graphs, where nodes encode both appearance and spatial coordinates, and edges reflect local structural adjacency. Unlike conventional graph neural networks that rely on generic message passing, we introduce a custom structural propagation mechanism that explicitly models relative spatial relations as part of the reasoning process. This design enables the graph to act as an inductive bias for structured inference rather than a passive relational representation. The proposed model jointly supports node-level lesion-aware predictions and graph-level diagnostic reasoning, yielding intrinsic explainability through learned node importance scores without relying on post-hoc visualization techniques. We demonstrate the approach through a chest X-ray case study, illustrating how structural priors guide relational reasoning and improve interpretability. While evaluated in a medical imaging context, the framework is domain-agnostic and aligns with the broader vision of graph-based reasoning across artificial intelligence systems. This work contributes to the growing body of research exploring graphs as computational substrates for structure-aware and explainable learning.

Structural Graph Neural Networks with Anatomical Priors for Explainable Chest X-ray Diagnosis

TL;DR

This work introduces a structural graph reasoning framework that embeds anatomical priors into patch-level graphs derived from CNN features to enable explainable chest X-ray diagnosis. A novel structural GNN propagates information via spatial displacement-aware message passing, producing intrinsic node-level explanations while delivering strong graph-level diagnostic performance. The approach jointly supports lesion-aware node predictions and global reasoning, achieving superior accuracy and AUC with intrinsic interpretability, demonstrated on ChestX-ray8-like data. By directly encoding spatial anatomy into the graph reasoning layer, the method offers a principled, domain-agnostic substrate for structure-aware and explainable AI across vision tasks.

Abstract

We present a structural graph reasoning framework that incorporates explicit anatomical priors for explainable vision-based diagnosis. Convolutional feature maps are reinterpreted as patch-level graphs, where nodes encode both appearance and spatial coordinates, and edges reflect local structural adjacency. Unlike conventional graph neural networks that rely on generic message passing, we introduce a custom structural propagation mechanism that explicitly models relative spatial relations as part of the reasoning process. This design enables the graph to act as an inductive bias for structured inference rather than a passive relational representation. The proposed model jointly supports node-level lesion-aware predictions and graph-level diagnostic reasoning, yielding intrinsic explainability through learned node importance scores without relying on post-hoc visualization techniques. We demonstrate the approach through a chest X-ray case study, illustrating how structural priors guide relational reasoning and improve interpretability. While evaluated in a medical imaging context, the framework is domain-agnostic and aligns with the broader vision of graph-based reasoning across artificial intelligence systems. This work contributes to the growing body of research exploring graphs as computational substrates for structure-aware and explainable learning.
Paper Structure (22 sections, 5 equations, 3 figures, 3 tables)

This paper contains 22 sections, 5 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Proposed Method
  4. Overview
  5. Patch-level Graph Construction
  6. Structural GNN Layer
  7. Node-level and Graph-level Prediction
  8. Explainability
  9. End-to-End Architecture
  10. Experimental Results
  11. Dataset and Preprocessing
  12. Implementation Details
  13. Evaluation Metrics
  14. Node-level Performance
  15. Graph-level Performance
  16. ...and 7 more sections

Figures (3)

  • Figure 1: Confusion matrix for graph-level diagnosis (Structural GNN).
  • Figure 2: ROC Curve for Graph-level Diagnosis
  • Figure 3: Node importance heatmap