丁香园董事长李天天：以 “负责任创新” 探索医疗人工智能发展新路径

Core Viewpoint - The article emphasizes the concept of "responsible innovation" in the development of medical artificial intelligence (AI), highlighting the importance of data quality and risk management in the application of AI in healthcare [1][2]. Group 1: Responsible Innovation in Medical AI - The medical industry must prioritize patient safety and quality, ensuring that any technological path adheres to the principles of evidence-based medicine [2]. - The company positions its AI products as decision-support tools for medical professionals rather than replacements for clinical decision-making, aiming to reduce information retrieval costs and minimize errors [2]. Group 2: Data Quality and Governance - The company launched a clinical decision support tool called "Clinical Decision" in October 2025, which relies on high-quality evidence-based medical data and AI technology to provide intelligent diagnostic support [3]. - A strict data screening mechanism is in place, prioritizing authoritative clinical guidelines and high-level evidence to avoid biases and uncertainties [3]. - The company employs multi-layered data cleaning and structuring processes to clarify key concepts and risk warnings, preventing ambiguities that could lead to potential risks [3]. Group 3: Dynamic Updates and Peer Review - A dynamic update mechanism is established to keep pace with evolving medical knowledge, ensuring timely revisions of guidelines and removal of outdated information [4]. - The company incorporates peer review and multiple rounds of manual verification for critical content, focusing on safety boundaries and high-risk scenarios to meet professional consensus and quality standards [4]. Group 4: Practical Applications and Future Plans - The company prioritizes risk control and correction capabilities in its decision-support system, exposing potential risks before providing conclusions to avoid misleading doctors in complex situations [5]. - High-level evidence is emphasized in conclusion presentation, with clear indications of applicability, evidence sources, and uncertainty boundaries to respect clinical guidelines [6]. - The company plans to donate AI products and services to grassroots medical institutions and doctors, aiming to enhance decision-making quality and reduce systemic risks in under-resourced areas by 2026 [6][7]. Group 5: Future Directions - The company will continue to refine data governance and risk control mechanisms, cautiously advancing the application of medical AI to provide practical experiences for orderly development within a regulatory framework [7].