Graph Inference Towards ICD Coding

TL;DR

This paper introduces LabGraph, a novel graph-based framework for automated ICD coding that improves prediction accuracy and robustness by reformulating the task as graph generation and employing advanced learning techniques.

Contribution

The paper presents LabGraph, a unified graph generation framework that incorporates adversarial domain adaptation, reinforcement learning, and a label graph discriminator for improved ICD coding.

Findings

Outperforms previous methods on benchmark datasets

Achieves higher micro-F1, micro-AUC, and P@K scores

Demonstrates robustness and better generalization

Abstract

Automated ICD coding involves assigning standardized diagnostic codes to clinical narratives. The vast label space and extreme class imbalance continue to challenge precise prediction. To address these issues, LabGraph is introduced -- a unified framework that reformulates ICD coding as a graph generation task. By combining adversarial domain adaptation, graph-based reinforcement learning, and perturbation regularization, LabGraph effectively enhances model robustness and generalization. In addition, a label graph discriminator dynamically evaluates each generated code, providing adaptive reward feedback during training. Experiments on benchmark datasets demonstrate that LabGraph consistently outperforms previous approaches on micro-F1, micro-AUC, and P@K.