MDDI 演讲稿 · 2025-05-27
Rahayu Mahzam 政务次长在 AI 健康研讨会 2025 上的闭幕致辞
Closing Address by MOS Rahayu Mahzam at AI in Health Symposium 2025
要点
- • 新加坡医疗 AI 三原则：「解决真实问题」+「负责任地规模化」+「广泛协作」。
- • 新加坡总医院 + GovTech 合作的 Proph Abby——AI 助理把抗生素处方对指南的合规率从 80% 提至 95%、复杂病例时间节省最高 90%。
- • BiliSG（SGH × SingHealth × Synapxe）让父母在家用智能手机给新生儿做黄疸筛查；MerMed-FM（SingHealth × A*STAR 合作的医疗基础模型）让小型机构也能基于 AI 起步。
- • 信任与治理：HSA 早在 2019 年发布 AI 医疗设备指南；2021 年 MOH+IHiS（现为 Synapxe）+ HSA 联合发布《医疗 AI 指南》；TRUST 健康数据交换平台支持匿名分析。
- • 当日见证两份合作备忘录：Enigma Health × Roche 把 AI 与制药能力结合服务临床试验；Enigma Health × ST Engineering 把医疗创新带给最懂患者需求的人。
完整译文（中文）
MDDI 英文原文译文 · 翻译日期：2026-05-03
瑞士驻新加坡大使 Frank Grütter 阁下，
各位嘉宾、产业领袖、医疗专业人士、女士与先生：
开场
我很高兴今天为「AI 在健康领域研讨会——以生成式 AI 规模化与可持续化医疗」做闭幕。
和其他国家一样——新加坡的医疗系统面对人口结构演变与生活方式变化带来的挑战。我们的人口在老龄化，慢性疾病的患病率在上升。这些挑战给医疗专业人士与医院基础设施带来压力——也呼唤新的医疗交付方式。
因此——借助技术「以有效且负责任的方式交付优质护理」对我们至关重要。AI 在「改造我们如何照护病人、支持医疗工作者、构建更具韧性的健康系统」上有巨大潜力。
我很高兴看到——今天的会议让大家有机会有意义地审视新加坡的医疗创新方法。让我把我们方法之下的几个关键考量讲清楚。
解决真实问题
第一——我们采取务实的方法，把 AI 与技术应用到医生与病人面对的真实问题上。我们走「证据为本、性价比为先」的路——看哪里能产生最大影响。
新加坡总医院（SGH）每年完成超过 10 万台手术，每一台都需要预防性抗生素。但选对抗生素并不简单——医生必须考虑手术类型、病人过敏史与疾病史。SGH 与 GovTech 的「开放政府产品」（Open Government Products）团队合作开发了「Proph Abby」——帮助医生在复杂病例中——按既定指南——开出正确的抗生素。这把 AI 直接嵌入到临床流程中——帮助医生更高效地交付护理。
在测试中——这款 AI 助理对既定指南的合规率达到 95%——较之前医生 80% 的合规率有显著改进。它在复杂病例上还能为医生节省最多 90% 的时间。这个例子说明——AI 同时改善了病人结果与护理效率。
我们也在把医疗带得离家庭与家人更近。黄疸影响超过半数的新生儿——这要求多次往返门诊——让脆弱的婴儿暴露于潜在感染——也让新家庭备受压力。「BiliSG」由 SGH、SingHealth Polyclinics、Synapxe 共同开发——让父母用智能手机 App 在家就能为新生儿筛查黄疸。BiliSG 的试点 App 达到了高度精准的灵敏度——并减少了不必要的门诊就诊。它正在儿科医院与综合诊所做进一步测试。
这些机会令人印象深刻——但需要被广泛获得。为医疗用例开发定制基础模型，需要大量数据集与资源。这可能形成一种「鸿沟」——只有大型、资金充裕的机构能从 AI 进展中获益，小医院与诊所被甩在后面。因此——我们也必须把 AI 能力的获取「民主化」。
想想 SingHealth 与 A*STAR 共同开发中的「MerMed-FM」基础模型。这个模型旨在以「比传统方法更少的训练数据」让 AI 开发更高效。如果成功——MerMED-FM 能让较小的医疗机构在不必从零开始的情况下构建 AI 应用。
负责任地规模化
但即便我们在解决痛点——也必须负责任地去做。
毕竟——生成式 AI 带来强大的新能力的同时——也带来关于隐私、安全与信任的新风险——我们必须回应。
「Enigma」是 Enigma Health 推出的医疗 AI 平台，自动化医院的行政工作。在 KK 妇女儿童医院与 PRISM¹（SingHealth Duke-NUS 精准医学研究所）开展的试点中——Enigma 把每份基因报告的撰写时间从 30 分钟缩短到几秒——一小时可处理 1,400 份，原本需要数周。重要的是——数据隐私得以保留。
「ELVF-FM」是 SingHealth 与 A*STAR 的研究合作——旨在帮助医生解读医学影像、验证临床结果。如果成功——这款工具会高亮其检测到的异常并解释自身推理——它有望克服许多今天 AI 系统的「黑箱」属性——让 AI 应用更安全。
在单一应用之外——我们还需要构建支撑创新的安全基础设施。比如健康研究——能帮助我们理解疾病、开发新治疗、规划健康项目、改进公共卫生政策——但必须以维护公众信任、保障个人隐私的方式进行。新加坡国家级的健康相关数据交换平台「TRUST」²（Trusted Research and Real World Data Utilisation and Sharing Tech）——通过安全基础设施与一套稳健框架，支持匿名化的大规模分析。研究在「受控访问 + 经过审核的输出」下进行——但不涉及个人识别符。
Enigma、ELVF-FM 与 TRUST——为我们展示了如何在「保持被信任」的同时驾驭 AI 的力量。
在医疗 AI 的采用中——「良好治理」与「技术进展」同等关键。没有清晰规则——公司不敢投资、医生不敢采纳新技术。
这就是为什么——新加坡卫生科学局（HSA）早在 2019 年就引入 AI 医疗设备指南、提供清晰监管路径。2021 年——卫生部、当时的「整合健康信息系统」（IHiS，现为 Synapxe）与 HSA——在此基础上联合发布《医疗 AI 指南》。
广泛协作
但即便是最精密的技术与最稳健的治理框架——也无法独立成功。医疗转型需要集体努力、共享专长。
没有一家机构能独自应对医疗 AI 的复杂性。我们必须跨行业、跨利益相关方协作——在公共与私营部门之间。
今天——我很高兴见证两份谅解备忘录的签署：一份是 Enigma Health 与罗氏（Roche），另一份是 Enigma Health 与新科工程（ST Engineering）。
这两份 MOU 体现了我们对医疗创新的协作式做法。
通过这次合作——Enigma Health 与罗氏将把 AI 能力与制药专长结合起来——加强临床试验与市场准入。这一合作将帮助创新治疗更高效地抵达病人。
Enigma Health 与新科工程的合作——会让医疗创新更易被「最懂病人需求」的人获得。
总体上——这些「解决真实问题、负责任规模化、广泛协作」的努力——让新加坡能借力 AI 增强医疗中的人类专长，从而为新加坡与新加坡人交付更好结果。
感谢各位为本次研讨会做出的贡献——感谢你们成为新加坡为后代构建更好医疗这一旅程的一部分。
我真心相信——携手同行——我们能为所有人构建更健康的未来。祝大家日安。
1 SingHealth Duke-NUS 精准医学研究所
2 Trusted Research and Real World Data Utilisation and Sharing Tech（受信任的研究与真实世界数据利用与共享平台）
英文原文
MDDI 官网原始记录 · 抓取日期：2026-05-02
His Excellency Mr Frank Grütter, Ambassador of Switzerland to Singapore
Distinguished guests, industry leaders, healthcare professionals, ladies and gentlemen.
Opening
It is my pleasure to close today’s AI in Health symposium on "Scaling and Sustaining Healthcare with GenAI".
Like other countries, Singapore’s healthcare system faces challenges arising from the evolving demography and changing lifestyles. Our population is aging and the prevalence of chronic illnesses is on the rise. These challenges create pressures on our healthcare professionals and hospital infrastructure. These pressures demand new ways of delivering care.
It is therefore important that we harness technology to deliver quality care in a way that is effective and responsible. AI offers tremendous potential to transform the way we care for patients, support our healthcare workers and create more resilient health systems.
I am heartened to note that the session today has allowed for a meaningful examination of Singapore’s approach to healthcare innovation. Let me articulate some key considerations underlying our approach.
Solve real problems
First, we take a practical approach and apply AI and technology to solve real-world problems for clinicians and patients. We take an evidence-based, cost-effective approach and look at areas where we can make the most impact.
Singapore General Hospital (SGH) performs over 100,000 surgeries annually, each requiring prophylactic antibiotics. But choosing the right antibiotic is complex and takes time. Doctors must consider the type of surgery, patient allergies and medical conditions. Working with the Open Government Products team at GovTech, SGH developed Proph Abby to help doctors prescribe the right antibiotics for complex cases, against established guidelines. This directly embeds AI into clinical flows to help clinicians deliver care more efficiently.
In testing, the AI assistant achieved 95% compliance with established guidelines: a significant improvement over the previous 80% compliance rate by doctors. It also saves doctors up to 90% of time on complex cases. This exemplifies how AI improves both patient outcomes and care efficiency.
We are also bringing healthcare closer to homes and families. Jaundice affects more than half of newborn babies. This requires multiple trips to the clinic, exposing vulnerable infants to potential infections and creating stress for new families. BiliSG, developed by SGH, SingHealth Polyclinics and Synapxe, helps parents screen their newborns for jaundice using a smartphone application, from the comfort of their homes. BiliSG’s pilot app achieved highly precise sensitivity and reduced unnecessary clinic visits. It is now undergoing further tests in paediatric hospitals and polyclinics.
These opportunities, while impressive, need to be made widely available. Developing customised foundational models for medical use cases requires massive datasets and significant resources. This can create a divide where only large, well-funded institutions can benefit from AI advances, while smaller hospitals and clinics are left behind. Hence, we must also democratise access to AI capabilities.
Consider MerMed-FM, a foundational model that SingHealth and A*STAR are co-developing. The model aims to make AI development more efficient by requiring lesser training data than traditional methods. If successful, MerMED-FM could enable smaller healthcare facilities to build AI applications without starting from scratch.
Responsible scaling
But even as we solve pain points, we need to do it responsibly.
After all, while Generative AI brings powerful new capabilities, it also raises new risks about privacy, safety and trust which we must address.
Enigma, a healthcare AI platform by Enigma Health, automates hospital administrative work. At KK Women's and Children's Hospital and PRISM 1 , a pilot with Enigma cut genetic reporting time from 30 minutes per report to just seconds, or 1,400 reports in an hour, instead of weeks. Importantly, data privacy is preserved.
ELVF-FM, a research collaboration between SingHealth and A*STAR, seeks to help doctors interpret medical images, and verify clinical results. If it succeeds, the tool will highlight the abnormalities that it detects and explains its reasoning. It could overcome the ‘black box’ nature of many AI systems today, enabling safer AI applications.
Beyond individual applications, we need to build secure infrastructure that enables innovations. For instance, health research can help us understand more about health conditions, develop new medical treatments, plan health programmes and improve public health policy, but it must be done in a manner that upholds public trust and preserves individuals’ privacy. TRUST 2 , Singapore’s national health-related data exchange platform, supports this by putting in place secure infrastructure and a robust framework to enable anonymised at-scale analytics. Research is done with controlled access and vetted outputs, but without involving personal identifiers.
Enigma, ELVF-FM and TRUST offer a glimpse of how we can harness the power of AI while remaining trusted.
Good governance is just as crucial as technological advances in advancing the adoption of AI in healthcare. Without clear rules, companies hesitate to invest, and doctors hesitate to adopt new technologies.
This is why Singapore's Health Sciences Authority (HSA) provided clear regulatory pathways when it introduced AI medical device guidelines in 2019. In 2021, the Ministry of Health, the then-Integrated Health information Systems (IHiS), now Synapxe, and the HSA built on this foundation by publishing the Artificial Intelligence in Healthcare Guidelines.
Collaborate widely
Yet even the most sophisticated technology and robust governance frameworks cannot succeed in isolation. Healthcare transformation requires collective effort and shared expertise.
No single institution can tackle the complexity of healthcare AI alone. We need to collaborate across sectors and stakeholders, between the public and private sectors.
Today, I am glad to have witnessed the two MOU signings between Enigma Health and Roche as well as Enigma Health and ST Engineering.
The two MOUs exemplify our collaborative approach to healthcare innovation.
Through their partnership, Enigma Health and Roche will combine AI capabilities with pharmaceutical expertise to strengthen clinical trials and market access. This collaboration will help bring innovative treatments to patients more efficiently.
The partnership between Enigma Health and ST Engineering will make healthcare innovation more accessible to those who understand patient needs best.
Collectively, these efforts to solve real problems, scale responsibly, and collaborate widely enables Singapore to leverage on AI to augment human expertise in healthcare, thereby delivering better outcomes for Singapore and Singaporeans.
Thank you for your contributions to this symposium, and for being part of Singapore’s journey to build better healthcare for generations to come.
I truly believe that together we can build a healthier future for all. Have a good day ahead.
1 SingHealth Duke-NUS Institute of Precision Medicine
2 Trusted Research and Real World Data Utilisation and Sharing Tech