MDDI 演講稿 · 2025-03-19
陳杰豪高階政務部長在 AI 健康世界峰會 2025 上的開幕致辭
要點
- • 新加坡醫療行業 AI 早期採用——放射科最早用 AI 檢測影像異常；近期 SingHealth「Note Buddy」即時多語種問診轉寫。
- • 「個性化干預」案例：SGH 的 AI2D（個性化抗生素處方，降低耐藥風險）；AI Singapore 的 CURATE.AI（個性化化療劑量）。
- • 「上游預防」案例：aiTriage（僅靠 ECG + 手機 App 預測 3 / 30 天心臟事件風險，7 分鐘分診）；CARES²（術後併發症預測）；NUS × CMU × FriendsLearn 在新加坡建「數字疫苗」精準預防卓越中心。
- • 新啟動「SingHealth Duke-NUS AI in Medicine Institute」（AIMI）——四大優先：前沿 AI 研究（基礎模型 + 智慧體 AI）、AI 素養、研究到落地的橋樑、倫理治理。
- • $1.2 億新元投入「AI for Science」計劃——支援 AI 與科研的融合（含生物醫學）。Synapxe「HEALIX」平臺整合公立醫院資料基礎設施。
- • 今天見證 SingHealth × CHAI（Coalition for Health AI，美國非營利，成員含 Google、Johns Hopkins、梅奧診所、微軟、斯坦福醫學院）簽署 MOU——共建 AI 醫療指南、出版物與政策建議。
完整譯文（繁體中文）
MDDI 英文原文譯文 · 翻譯日期: 2026-05-03
本文已從早期版本的網站遷移過來——格式可能有不一致之處。
SingHealth 集團執行長 Ng Wai Hoe 教授，
Duke-NUS 醫學院研究高階副院長 Patrick Tan 副教授，
A*STAR 生物醫學研究理事會助理執行長 Lisa Ooi 博士，
「Coalition for Health in AI」執行長 Brian Anderson 博士，
各位嘉賓、代表、
來自世界各地的朋友們：
早安。我很高興出席第三屆「AI Health World Summit」（AI 健康世界峰會）。
我向 SingHealth Duke-NUS 學術醫學中心與 A*STAR 致敬——他們把這一跨邊界、跨學科、跨產業的社群召集到一起。
今天在場的我們都共享同一個目標——改造我們社群的醫療——並都相信 AI 在實現這一目標中扮演重要角色。
新加坡一直是「用技術克服多種約束」的堅定信仰者——土地、勞動力、小型市場。
在這一點上——AI 是一項有重大潛能的技術。
我們的醫療行業很早就擁抱了 AI。
a. 比如——我們的放射科醫生是「用 AI 檢測醫學影像異常」最早的採用者之一。
生成式 AI 進一步打開了「臨床效率」的新可能。
a. 一個例子是去年 8 月推出的 SingHealth「Note Buddy」工具。
b. 這一工具即時轉寫並總結問診——並支援多種語言。
c. 它減少了醫生的行政工作量——讓他們與病人有更多有意義的互動。
重要的是——AI 也讓我們能在「規模化、低成本」的同時，對醫療干預採取「更個性化」的方法。
a. 一個例子是新加坡總醫院（SGH）的「AI2D」（Augmented Intelligence in Infectious Diseases）模型——個性化為患者開抗生素，降低抗生素耐藥的風險。
b. 另一個例子是 AI Singapore 的「CURATE.AI」——為個體患者最佳化化療劑量。
c. 不再只把醫療干預建立在「群體級臨床試驗」之上——這些早期努力有可能通過讓決策反映個人獨特的「基因、生活方式、病史」相關的生物標誌物反應——改善干預結果。
AI 也增強了我們的預測能力——讓醫生能把更多幹預「向上遊推進」——朝預防性照護方向走。
a. 比如——SGH、Duke-NUS、DxD Hub 共同開發的心臟風險工具「aiTriage」——能預測病人在未來 3 天或 30 天內出現重大心臟事件的風險。
i. 僅用一臺 ECG 裝置與手機 App——醫護人員就能識別高危患者並在 7 分鐘內做出分診——顯著提高心臟驟停的存活率。
b. 這種早期檢測能力在圍手術期醫學中也很有價值——「CARES 2」就是這樣的手術風險模型。
i. 由 SGH 與新加坡國立大學（NUS）共同開發——能預測術後併發症的可能性、並提示是否需要再入院或入 ICU。這類預測讓早期干預成為可能。
AI 給了我們「再向上游推進」的工具——比如——如何及早塑造行為、在症狀出現之前大幅降低疾病風險。
a. 一個例子是——NUS 的醫學院與計算學院、卡內基梅隆大學（CMU）的 Heinz 資訊系統與公共政策學院——以及作為技術發明者與轉化夥伴的 FriendsLearn Inc 之間的合作。
b. 他們將在新加坡建立一個「精準預防 AI 卓越中心」——通過我們所謂的「數字疫苗」。
c. 這些「數字疫苗」使用神經認知訓練、內隱學習、沉浸式遊戲——以安全有效的方式——為兒童培育健康行為——同時為神經發育與腸道菌群健康打底。
d. 這能在童年「習慣形成」的關鍵階段——在神經-生理層面引發並強化與「行為與選擇」相關的變化。我們期待跟蹤這種方法如何為後代改善「終身的健康與認知福祉」。
我們才剛剛觸及 AI 給醫療帶來的巨大可能性的表層。
a. 前沿專案正在改造我們「研究病理與設計醫療干預」的方式。
b. 比如——Google DeepMind 的新系統「AlphaProteo」——能為與癌症、糖尿病相關的蛋白生成新的「分子結合體」（molecule binders）——幫助調控關鍵的細胞過程。
c. 我們正在更接近「靶向曾被認為『不可成藥』的蛋白」——以及「曾被認為不可治癒的疾病」。
d. 藉助這些進展——也許有生之年我們能治癒更多種癌症！
這些進展的影響是深遠的。
同樣令人驚歎的是——它們出現於不同領域的交匯——電腦科學、分子生物學、臨床醫學等等。
這種「專長的匯流」提醒我們——最大的突破不是僅靠技術——而是把多元的頭腦帶到一起。
這種跨學科對話——必須從把生態的不同部分聚到一起開始。
a. 這就是為什麼——我們的國家健康科技機構 Synapxe 去年推出「HEALIX」分析平臺——以整合公立醫院的資料基礎設施。藉助一個統一、安全的 AI 開發平臺——HEALIX 讓基於「匿名化國家級資料」的分析專案成為可能。
我們也在彌合「研究」與「臨床落地」之間的鴻溝。在最近的國會預算辯論中——副總理王瑞傑談到了「雙語」人才的重要性——把 AI 研究與領域專長結合的研究者——就像今天在場的許多位。
對「應用 AI 研究」的聚焦至關重要。政府已為「AI for Science」計劃承諾投入 1.2 億新元——這一倡議支援 AI 與科研的整合——包括生物醫學與健康科學領域。
為持續把研究流向實踐——我們將繼續為「跨學科學習與應用」搭建平臺。
今天——我很高興啟動「SingHealth Duke-NUS AI in Medicine Institute」（AIMI，AI 在醫學領域研究院）。
a. 這個研究院將把 SingHealth 與 Duke-NUS 醫學院的專長彙集起來——四大優先：
i. 第一——領跑前沿 AI 技術研究——包括基礎模型與智慧體 AI。
ii. 第二——倡導 AI 素養——在醫療專業人士與公眾中培育「知情采用」的文化。
iii. 第三——彌合研究與真實世界應用之間的鴻溝——推動創新 AI 專案的全球商業化。
iv. 第四——主導制定穩健的倫理指南——強調負責任 AI、患者安全、資料隱私。
這一研究院的優先事項——也凸顯了我們前方的幾項挑戰。
上一次（2023 年）舉辦這一峰會時——生成式 AI 剛剛席捲世界。從那以後——全球對 AI 的視角已經成熟——對它的實際收益與風險有了更清晰的認識。
「信任」必須是我們的北極星。
信任一直是醫療的根基——我們的社群、我們的病人——把生命託付給我們。
但有了 AI——我們走在「醫療突破」與「風險」之間的細微界線上。從「演算法透明度」到「不具代表性的訓練資料」——種種挑戰可能損害 AI 方案的安全性，並侵蝕患者對醫療機構與專業人士的信任。
因此——即便我們在「實驗與創新」新技術——我們也不能丟掉「穩健的治理」與「以患者結果為中心」。
我們必須把發展過程中的關鍵利益相關方聚在一起——臨床醫生、研究者、決策者，重要的——還有患者及其家屬。
正是這種集體努力——構建出信任——協作與共同承諾會帶來持久的信心。這就是為什麼像今天這樣的峰會很關鍵。
本著這種精神——我也很高興見證 SingHealth 與「Coalition for Health AI」（CHAI）簽署 MOU。
a. CHAI 是一家位於美國的非營利組織——致力於在醫療中為「安全、有效、公平的 AI」建立指南與最佳實踐。
b. CHAI 把學術醫療體系、機構與專家實務者聚到一起——成員包括 Google、約翰霍普金斯大學、梅奧診所、微軟、斯坦福醫學院。
今天簽署的 MOU——是雙方共同承諾——提升醫療中「負責任 AI 使用」的認知。
a. 這將包括共同制定 AI 指南、出版物與政策建議——確保醫療 AI 使用持續保持倫理、透明，並聚焦於改造病人照護。
前方有許多激動人心的機會與可能——作為社群、作為社會、作為人類。但讓我們始終記住——AI 在醫療中成功的真正衡量——不是技術進步——而是它「改善生命」的能力。
祝大家峰會順利。
英文原文
MDDI 官網原始記錄 · 抓取日期: 2026-05-02
This article has been migrated from an earlier version of the site and may display formatting inconsistencies.
Professor Ng Wai Hoe, Group CEO of SingHealth
Associate Professor Patrick Tan, Senior Vice-Dean of Research at Duke-NUS Medical School
Dr Lisa Ooi, Assistant Chief Executive of the Biomedical Research Council at A*STAR
Dr Brian Anderson, CEO of the Coalition for Health in AI
Distinguished speakers, delegates
Friends from around the world
A very good morning and it is my pleasure to join you today at the third AI Health World Summit.
I commend the SingHealth Duke-NUS Academic Medical Centre and A*STAR for convening this community from across borders, disciplines, and industries.
All of us here have the shared goal to transform healthcare in our communities and are joined in the belief that AI plays an important role to achieve this aim.
Singapore has always been a firm believer in using technology to overcome our many constraints – land, labour and small market size.
In this regard, AI is a technology with significant potential.
Our healthcare sector had embraced AI early on.
a. For example, our radiologists were amongst the earliest adopters in using AI to detect anomalies in medical scans.
Generative AI has further opened up new possibilities for clinical efficiency.
a. An example is SingHealth’s Note Buddy tool that was launched last August.
b. The tool transcribes and summarises consultations in real-time and in different languages.
c. It reduces the doctors’ administrative workload, freeing them to have more meaningful interactions with their patients.
Importantly, AI has also given us the capability to adopt a more personalised approach to medical intervention, at scale and at lower cost.
a. An example is the Singapore General Hospital’s AI2D (Augmented Intelligence in Infectious Diseases) model. The model personalises antibiotic prescriptions for patients, reducing the risk of antibiotic resistance.
b. Another example is AI Singapore’s CURATE.AI that optimises chemotherapy doses for individual patients.
c. Instead of basing medical interventions just on population-based clinical trials, these early efforts have the potential to improve intervention outcomes by enabling decisions to be informed by an individual’s unique responses to biomarkers related to genetics, lifestyle, and medical history.
AI also improves our predictive capabilities, allowing doctors to shift more interventions upstream towards preventive care.
a. For example, aiTriage, a cardiac risk tool developed by the Singapore General Hospital, Duke-NUS Medical School, and DxD Hub, can predict a patient’s risk of a major cardiac event in the next 3 or 30 days.
i. Using just an ECG device and mobile app, healthcare workers can identify at-risk patients and make triaging decisions in under 7 minutes, thereby significantly improving the chances of someone surviving a cardiac arrest.
b. These early detection capabilities are also valuable in peri-operative medicine. CARES 2 is one such surgical risk model.
i. Developed by the Singapore General Hospital and National University of Singapore, it can predict the likelihood of post-operation complications, and inform the need for re-admission or ICU admission. These predictions facilitate early interventions.
AI gives us the tools to go further upstream, for example looking at how we can shape behaviours early and reduce disease risks well before symptoms emerge.
a. An example is a partnership between the National University of Singapore's Schools of Medicine and Computing, and Carnegie Mellon University's Heinz College of Information Systems and Public Policy, along with FriendsLearn Inc – a technology inventor and translation partner.
b. They will establish a Centre of Excellence in Singapore for the use of AI in precision prevention, via what we call ‘digital vaccines’.
c. These digital vaccines use neurocognitive training, implicit learning, and immersive gaming to encourage healthier behaviours in children – safely and effectively priming neurodevelopment and gut biome health.
d. This induces and reinforces changes at the neuro-physiological level underlying behaviours and choices, during the critical habit-formation stage of childhood. We look forward to tracking how this method can improve lifelong health and cognitive wellbeing for generations to come.
We are scratching at the surface of the immense possibilities that AI brings to healthcare.
a. Cutting-edge projects are transforming the way we study pathology and design medical interventions.
b. Take for example Google Deepmind’s new AlphaProteo system. It can generate new molecule binders for proteins associated with cancer and diabetes, helping regulate critical cellular processes.
c. We are coming closer to targeting proteins once thought ‘undruggable’, and conditions thought incurable.
d. With these advances, we can possibly cure many more forms of cancer in our lifetime!
The implications of these advances are profound.
But equally remarkable is how they've emerged from the intersection of different fields – computer science, molecular biology, clinical medicine and more.
This convergence of expertise reminds us that our greatest breakthroughs come not from technology alone, but from bringing diverse minds together.
Such interdisciplinary conversations must begin by bringing together different parts of the ecosystem.
a. This is why our national healthtech agency, Synapxe, launched its HEALIX analytics platform last year to consolidate data infrastructure across our public hospitals. With a unified secure platform for AI development, HEALIX makes analytics projects with anonymised national data possible.
We are also bridging the gap between research and clinical implementation. At the recent Budget Debate in Parliament, DPM Heng spoke of the importance of ‘bilingual’ talent – researchers who combine AI research with domain expertise like many of you in the audience here today.
This focus on applied AI research is crucial. The Government has committed $120 million to the AI for Science programme. This initiative supports the integration of AI and scientific research, including in the biomedical and health sciences fields.
To sustain this flow of research to practice, we will continue investing in platforms for interdisciplinary learning and application.
Today, I am pleased to launch the SingHealth Duke-NUS AI in Medicine Institute, or AIMI.
a. This Institute will bring together expertise from SingHealth and Duke-NUS Medical School, with four priorities.
i. First, to spearhead research in frontier AI technologies, including foundation models and agentic AI.
ii. Second, to champion AI literacy, fostering a culture of informed adoption amongst healthcare professionals and the public alike.
iii. Third, to bridge the gap between research and real-world application, driving the global commercialisation of innovative AI projects.
iv. Fourth, to lead the development of robust ethical guidelines, which emphasise responsible AI, patient safety, and data privacy.
The priorities of this institute highlight some of the challenges that lie ahead of us.
When we last held this Summit in 2023, GenAI had just taken the world by storm. Since then, global perspectives towards AI have matured, with a clearer view of its actual benefits and risks.
Trust must be our North Star.
Trust has always underpinned healthcare; our communities, our patients entrust us with their lives.
But with AI, we tread a fine line between medical breakthroughs and risks. Challenges from algorithmic transparency to unrepresentative training data may undermine the safety of AI solutions, and erode patients’ trust in the healthcare institutions and professionals.
That is why we must not lose sight of robust governance and the focus on patient outcomes even as we experiment and innovate with new technologies.
We must bring together key stakeholders in the development process - clinicians, researchers, policymakers, and importantly patients and their families.
It is this collective effort that builds trust, where collaboration and shared commitment lead to lasting confidence. This is why efforts like today’s Summit are critical.
It is in this spirit that I am also heartened to witness the signing of an MOU between SingHealth and the Coalition for Health AI, or CHAI.
a. CHAI is a US-based non-profit organisation, dedicated to establishing guidelines and best practices for safe, effective, and equitable AI across healthcare.
b. CHAI brings together academic health systems, organisations, and expert practitioners, with members including Google, Johns Hopkins University, Mayo Clinic, Microsoft, and Stanford Medicine.
The MOU being signed today is a pledge of commitment by both organisations to raise awareness of responsible AI use in healthcare.
a. This will involve the co-creation of AI guidelines, publications, and policy recommendations, ensuring that the use of AI in healthcare remains ethical, transparent, and focused on transforming patient care.
There are many exciting opportunities and possibilities lie ahead of us – as a community, as society, as humanity. But let us always remember the true measure of AI’s success in healthcare is not the technical advancements but its ability to improve lives.
I wish everyone a fruitful Summit.