醫療人工智慧應用爭議與法制規範課題

人工智慧技術改變包括醫療在內、幾近所有產業的固有面貌。研究指出診斷支持、疾病預防及影像診斷是醫療人工智慧最具潛力的三個應用領域，然而人工智慧技術於醫療領域的運用亦衍生若干過往新興課題。面對此一嶄新科技帶來的挑戰與規範課題，在病患可得直接感知的外部應用部分，最受矚目者當為「醫療器材軟體」（SaMD），而現時規範趨勢大抵以衡平SaMD發展並確保其安全無虞為目標，如美國2021年提出的SaMD行動計畫，係立於產品生命週期之基礎提出SaMD規範重點方針。而不易為外界感知的醫療人工智慧內在應用，亦即機器學習／演算法所產生之課題，除演算法可能造成醫療專業知識水平下降外，演算法本身欠缺透明度及存在偏見／歧視可能，無疑是醫療人工智慧發展上最受質疑之處。主要國家提出的規範機制，無論是建議性文件或實質性立法推動，逐步聚焦如何避免人工智慧應用出現歧視性結果，並確保充足的透明度與獨立性。針對國內醫療人工智慧的應用推動與相關之法制建構，本文建議可優先聚焦：1、建立醫療領域「人工智慧風險評估機制」；2、建立醫療人工智慧應用行為之「問責機制」；3、建立醫療人工智慧系統「持續檢視機制」；及4、研議必要的醫療人工智慧倫理規範。此外，亦應同步確保資料品質及醫療主管機關在立法與執法層面之充分參與，亦至為重要。

關鍵字

人工智慧；醫療；醫療器材軟體；演算法；風險評估；問責

並列摘要

Artificial Intelligence (AI) technology is being introduced and applied in almost every industry. PwC identified three areas with the biggest AI potential for medical, including supporting diagnosis, early identification of potential pandemics and imaging diagnostics. However, the application of AI technology in the medical field has also caused lots of new issues. In the external application area, the focus is on Software as Medical Device (SaMD). The current regulatory trend is to emphasize equalizing the development of SaMD and ensuring its security. The Action Plan proposed by the United States in 2021 is based on the product-life-cycle concept and proposes the regulatory focus direction of SaMD. Machine Learning/Algorithm is an indispensable intrinsic application of medical AI. Studies have shown that algorithms may contribute to a decline in medical expertise, other possible problems include a lack of transparency and the possibility of discrimination. It is necessary to construct supervision leg al system toward application of medical AI. This article holds that the following matters should be given priority: 1. Establish ＂AI Risk-Assessment mechanism＂ in the medical field; 2. Establish ＂Accountability mechanism＂ for medical AI application behavior; 3. Establish ＂continuous inspection mechanism＂ for medical AI Systems; 4. Necessary Ethical regulation of medical AI.