JMIR Publications

ABSTRACT

Background:

Knowledge discovery from treatment data records from prestigious Chinese physicians is a dramatic challenge in the applications of artificial intelligence models to the research of Traditional Chinese Medicine (TCM).

Objective:

This paper aims to construct a TCM knowledge graph from prestigious Chinese physicians and apply it to decision-making assistant of TCM diagnosis and treatment.

Methods:

A new framework leveraging a representation learning method for TCM knowledge graph construction and application is designed. A Transformer-based Contextualized Knowledge Graph Embedding (CoKE) model is applied to knowledge graph representation learning and knowledge distillation. Automatic identification and expansion of multi-hop relations are integrated with the CoKE model as a pipeline. Based on the framework, a TCM knowledge graph, containing 598,82 entities including diseases, symptoms, examinations, drugs, etc., 17 relations, and 604,700 triples, is constructed. The framework is validated through a link predication task.

Results:

Experiments show that the framework outperforms a set of baseline models in the link prediction task using standard metrics MRR and Hits@N. The knowledge embedding multi-tagged TCM discriminative diagnosis metrics also indicates the improvement of our framework compared with the baseline models

Conclusions:

Experiments show that the clinical knowledge graph representation learning and application framework is effective for knowledge discovery and decision-making assistance in diagnosis and treatment . Our framework shows superiority of application prospects in tasks such as knowledge graph fused multi-modal information diagnosis, knowledge graph embedding based text classification and knowledge inference based medical question answering.