Integration of Multi-Source Medical Data for Medical Diagnosis Question Answering.

Area of Science:

• Artificial Intelligence in Medicine
• Clinical Informatics
• Medical Data Science

Background:

• Current medical question answering systems often rely on single data sources, limiting comprehensive diagnostic capabilities.
• Intelligent healthcare requires integrating diverse data like lab tests, radiology, and pathology for accurate medical diagnosis.
• Existing methods fall short in providing the holistic analysis needed for advanced clinical decision-making.

Purpose of the Study:

• To introduce the Medical Diagnosis Question Answering (MedDQA) task, focusing on multi-source data integration.
• To develop and evaluate a novel system for answering medical diagnosis questions using heterogeneous patient data.
• To establish a robust baseline for advanced medical question answering systems in intelligent healthcare.

Main Methods:

• Developed a large-scale dataset of multi-source diagnostic information from 250,917 patients.
• Utilized a large-scale model to construct question-and-answer pairs for the MedDQA task.
• Proposed a Medical Multi-Agent (MMA) system employing specialized agents for processing various data modalities into a uniform text output.

Main Results:

• The MMA system demonstrated enhanced capabilities in handling multi-source medical data.
• Experimental results showed significant improvements in medical diagnosis accuracy.
• The proposed system establishes a strong foundation for future research in medical question answering.

Conclusions:

• The MedDQA task and the MMA system represent a significant advancement in leveraging multi-source data for medical diagnosis.
• Integrating diverse data modalities through multi-agent systems is crucial for the progression of intelligent healthcare.
• The MMA system offers a scalable and effective approach to complex medical question answering challenges.