Healie: A Personalized Healthcare Content Generation System Based on Knowledge Graphs and LLMs

Healie (Health Information Enhancement) is an innovative medical information system led by Dr. Christine Kakalou, integrating knowledge graphs, cognitive factors, social determinants of health (SDOH), and large language models (LLMs). Its core goal is to improve patients' health literacy through technological means, bridge the medical information gap, and shift towards a patient-centered direction in medical AI development.

In today's medical environment, patients face dual challenges of information overload and comprehension barriers: medical literature, diagnostic reports, and treatment plans are filled with professional terminology, making it difficult for ordinary patients to accurately grasp their meaning. Healie attempts to solve this problem through intelligent content generation technology.

Knowledge Graph-Driven

The system uses a medical knowledge graph as its core data layer, structurally representing concepts and relationships such as diseases, symptoms, treatments, and drugs, with advantages of interpretability, updatability, and multi-source integration.

LLM Integration

The knowledge graph provides a structured factual foundation, while LLMs are responsible for converting it into easy-to-understand natural language explanations, balancing content accuracy and expressive flexibility.

Cognitive Factor Modeling

It incorporates users' health literacy levels, information processing preferences, cognitive load tolerance, prior knowledge backgrounds, etc., to generate content that adapts to users' comprehension abilities.

Healie focuses on the impact of SDOH on health, including socioeconomic status, educational background, living environment, social support networks, and access to medical resources. The system incorporates these factors into personalized content generation: for example, patients with limited economic means are prioritized to receive cost-effective solutions, and patients in remote areas are recommended medical resources with high accessibility.

• Patient Education: Help understand diagnostic results, treatment plans, and medication instructions, reduce anxiety, and improve treatment adherence.
• Health Literacy Improvement: Dynamically adjust content complexity for users of different levels to gradually enhance their health knowledge reserves.
• Chronic Disease Management: Provide continuous health guidance (lifestyle recommendations, symptom monitoring reminders, etc.).
• Medical Decision Support: Assist in understanding the pros and cons of treatment options to support informed decision-making.

• Balancing Accuracy and Readability: A layered content generation strategy—first ensure medical facts are accurate, then optimize language through multiple rounds of LLM processing.
• Coordinating Personalization and Generalization: Use user profile modeling to balance group commonalities and individual characteristics.
• Privacy Protection and Data Utilization: Local deployment + differential privacy technology to achieve personalized services while protecting privacy.

Healie represents an important direction for healthcare AI to shift from doctor-centered to patient-centered, aligning with modern medical concepts. Its integrated architecture of knowledge graphs and LLMs provides a reference model for AI applications in other vertical fields.

Healie is expected to further expand its capabilities:

• Visual content generation (medical image interpretation and annotation)
• Voice interaction (lowering the threshold for use)
• Real-time health monitoring (integrating wearable device data)
• Multilingual support (serving global patients) This demonstrates the potential of AI in the medical field and the exploratory value of responsible development.