Uncertainty-Aware Clinical Knowledge Graphs for Chest X-Ray Device Reasoning

This framework constructs a clinical knowledge graph for chest radiograph analysis, materializing over 914,000 B7 evidence nodes and 884,000 relationships. It tracks device tip estimates, placement probabilities, and provenance, offering granular uncertainty metrics that standard segmentation or labeling models lack.