[Paper Review] A Graph Based and Patient Demographics Aware Dialogue System for Disease Diagnosis.

Xinyan Zhao, Liangwei Chen+1 more|arXiv (Cornell University)|Oct 21, 2020

Topic Modeling26 references4 citations

TL;DR

This paper proposes a graph-based, demographics-aware dialogue system for disease diagnosis using a bidirectional graph convolutional network (BG-DQN) that models symptom-disease relationships and integrates patient demographic attributes. The system improves diagnostic accuracy, reduces dialogue turns, and enhances discrimination between diseases with similar symptoms compared to existing methods.

ABSTRACT

A dialogue system for disease diagnosis aims at making a diagnosis by conversing with patients. Existing disease diagnosis dialogue systems highly rely on data-driven methods and statistical features, lacking profound comprehension of medical knowledge, such as symptom-disease relations. In addition, previous work pays less attention to demographic attributes of a patient, which are important factors in clinical diagnoses. To tackle these issues, this work presents a graph based and demographic attributes aware dialogue system for disease diagnosis. Specifically, we first build a weighted bidirectional graph based on clinical dialogues to depict the relationship between symptoms and diseases and then present a bidirectional graph based deep Q-network (BG-DQN) for dialogue management. By extending Graph Convolutional Network (GCN) to learn the embeddings of diseases and symptoms from both the structural and attribute information in the graph, BG-DQN could capture the relations between diseases and symptoms better. Moreover, BG-DQN also encodes the demographic attributes of a patient to assist the disease diagnosis process. Experimental results show that the proposed dialogue system outperforms several competitive methods in terms of diagnostic accuracy. More importantly, our method can complete the task with less dialogue turns and possesses better distinguishing capability on diseases with similar symptoms.

Motivation & Objective

• Address the lack of deep medical knowledge integration in existing data-driven disease diagnosis dialogue systems.
• Incorporate patient demographic attributes—often overlooked in prior work—into the diagnosis process.
• Improve the modeling of symptom-disease relationships using structural and attribute information from clinical dialogues.
• Reduce the number of dialogue turns required for accurate diagnosis while maintaining or improving diagnostic performance.
• Enhance the system's ability to distinguish between diseases that present with similar symptoms.

Proposed method

• Construct a weighted, bidirectional graph from clinical dialogues to represent symptom-disease relationships.
• Develop a bidirectional graph-based deep Q-network (BG-DQN) that leverages graph convolutional networks (GCN) to learn embeddings from both structural and attribute information in the graph.
• Extend GCN to encode both the topological structure of the symptom-disease graph and node-specific attributes such as symptoms and diseases.
• Integrate patient demographic attributes (e.g., age, gender) into the BG-DQN framework to inform diagnosis decisions.
• Train the BG-DQN using reinforcement learning to optimize dialogue policy for efficient and accurate diagnosis.
• Use the learned embeddings to guide dialogue management, enabling the system to select the most informative questions at each turn.

Experimental results

Research questions

• RQ1 Can a graph-based approach better model the complex relationships between symptoms and diseases compared to traditional statistical or embedding-based methods?
• RQ2 To what extent do patient demographic attributes improve diagnostic accuracy in a dialogue-based system?
• RQ3 Does the proposed BG-DQN reduce the number of dialogue turns required to reach a diagnosis while maintaining or improving accuracy?
• RQ4 How well does the system distinguish between diseases that share similar symptom profiles?
• RQ5 Can the integration of structural and attribute information in a graph network lead to more robust and explainable diagnosis decisions?

Key findings

• The proposed BG-DQN system achieves higher diagnostic accuracy than several competitive baseline methods.
• The system reduces the average number of dialogue turns required to reach a diagnosis, improving efficiency.
• The model demonstrates superior distinguishing capability between diseases with similar symptoms, reducing misdiagnosis risks.
• Incorporating demographic attributes leads to measurable improvements in diagnostic performance, especially for age- or gender-related conditions.
• The graph-based representation learning effectively captures symptom-disease relationships, outperforming models that rely solely on statistical or embedding-based features.