Editor's note:
In response to the Belt and Road Initiative, and with the aim of sharing expertise and addressing shared health challenges, our hospital has launched the Overseas Promotion Program for New Medical Technologies. If you are interested in any of the medical technologies covered under this program, please feel free to contact the International Cooperation Office at faowch@163.com.
KOM: A Multi-Agent Artificial Intelligence System Enabling Full-Process Knee Osteoarthritis (KOA) Health Management
Knee osteoarthritis (KOA) affects over 600 million people globally, leading to pain, loss of function, and disability. While personalized multidisciplinary care can slow disease progression and enhance quality of life, it requires considerable healthcare resources and is often inaccessible in resource-constrained settings. To address these challenges, we developed KOM (Knee Osteoarthritis Manager), the first clinically evaluated, fully integrated multi-agent system for managing KOA. KOM spans the entire clinical pathway, encompassing disease assessment, progression prediction, and treatment planning. By automating the evaluation of KOA, forecasting progression risks, and generating individualized treatment plans, KOM aligns seamlessly with clinical workflows. This approach significantly reduces the time and resources typically needed for personalized care, making it particularly valuable in areas with limited healthcare resources.
The KOM system integrates multiple AI technologies, including large language models (LLMs) for patient interaction and convolutional neural networks (CNNs) for radiograph analysis. This AI system offers three core agents: an Assessment Agent that collects patient data, analyzes radiographs, and generates evaluation reports; a Risk Agent that predicts the progression of the disease and identifies patient-specific risk factors; and a Therapy Agents Group that simulates multidisciplinary team discussions to generate evidence-based, personalized treatment plans. By automating key tasks such as patient data collection, radiograph analysis, and treatment planning, KOM ensures faster, more accurate, and personalized care.
In a randomized, three-arm simulation study, KOM outperformed advanced language models and doctoral trainees in several clinical tasks, including patient interaction, image analysis, risk prediction, and management planning. When KOM collaborated with trainees, it reduced diagnosis and planning time by 38.5% and provided more personalized treatment plans compared to either component working alone. This collaboration demonstrated the efficiency and enhanced clinical decision-making capabilities that KOM provides, highlighting its potential to optimize clinical workflows.
The innovations brought by KOM extend beyond KOA management, offering a scalable framework for managing other chronic degenerative diseases. By incorporating AI-driven automation, KOM ensures faster, more efficient treatment planning, making it an invaluable tool for healthcare professionals, particularly in areas with high disease prevalence or limited healthcare resources like Chengdu. KOM’s ability to predict disease progression, assess risks, and generate personalized treatment plans presents significant value for global health improvement. It provides a foundation for the development of AI-assisted management systems that can be tailored to other chronic conditions, offering a promising solution to the global healthcare challenge. Through its multi-agent framework, KOM empowers clinicians with intelligent decision-making tools, reducing the cognitive load on healthcare professionals and allowing them to focus on delivering high-quality, individualized care.
