Towards Knowledge-Infused Automated Disease Diagnosis Assistant

Mohit Tomar,Abhisek Tiwari,Sriparna Saha

2024-05-18

Abstract:With the advancement of internet communication and telemedicine, people are increasingly turning to the web for various healthcare activities. With an ever-increasing number of diseases and symptoms, diagnosing patients becomes challenging. In this work, we build a diagnosis assistant to assist doctors, which identifies diseases based on patient-doctor interaction. During diagnosis, doctors utilize both symptomatology knowledge and diagnostic experience to identify diseases accurately and efficiently. Inspired by this, we investigate the role of medical knowledge in disease diagnosis through doctor-patient interaction. We propose a two-channel, knowledge-infused, discourse-aware disease diagnosis model (KI-DDI), where the first channel encodes patient-doctor communication using a transformer-based encoder, while the other creates an embedding of symptom-disease using a graph attention network (GAT). In the next stage, the conversation and knowledge graph embeddings are infused together and fed to a deep neural network for disease identification. Furthermore, we first develop an empathetic conversational medical corpus comprising conversations between patients and doctors, annotated with intent and symptoms information. The proposed model demonstrates a significant improvement over the existing state-of-the-art models, establishing the crucial roles of (a) a doctor's effort for additional symptom extraction (in addition to patient self-report) and (b) infusing medical knowledge in identifying diseases effectively. Many times, patients also show their medical conditions, which acts as crucial evidence in diagnosis. Therefore, integrating visual sensory information would represent an effective avenue for enhancing the capabilities of diagnostic assistants.

Artificial Intelligence,Computation and Language

What problem does this paper attempt to address?

This paper aims to address issues in automatic disease diagnosis. With the development of internet communication and remote healthcare, more and more people are seeking medical services online. However, the increasing number of diseases and symptoms has made diagnosis more challenging. The paper proposes a disease diagnosis model called Knowledge-Integrated, Context-Aware Disease Diagnosis Model (KI-DDI), which identifies diseases through interactions between doctors and patients and integrates medical knowledge. The model consists of two channels: one channel uses Transformer to encode the communication between patients and doctors, and the other channel uses Graph Attention Network (GAT) to create symptom-disease embeddings. Subsequently, the dialogue and knowledge graph embeddings are fused and fed into a deep neural network for disease identification. Additionally, the paper constructs an empathetic medical dialogue corpus containing intent and symptom information.

The research finds that relying solely on patient self-reports is insufficient for accurate disease diagnosis, and the experience and medical knowledge of doctors are crucial for diagnosis. Therefore, the paper proposes an external knowledge graph that emphasizes the importance of self-reports in dialogues. Through experiments, the model demonstrates significant improvements over the existing state-of-the-art models, validating the effectiveness of knowledge integration and additional symptom extraction in disease identification. The paper also suggests the possibility of enhancing the capabilities of diagnostic assistants through the integration of visual sensory information in the future.

Furthermore, the paper creates an empathetic medical dialogue dataset for annotating attention graphs and symptom information, aiming to facilitate dialogue understanding and coherent response generation. Research questions include whether patient self-reports are sufficient for accurate diagnosis, how medical knowledge graphs influence diagnostic performance, and how knowledge integration mechanisms affect the efficiency of disease diagnosis.