内容来源：https://research.google/blog/bringing-foundational-models-to-the-front-lines-of-tb-screening/

内容总结：

谷歌研究团队近日分享了其利用人工智能推动结核病筛查的最新进展。结核病每年夺走约123万人的生命，是全球最致命的传染病之一。由于高负担地区放射科医生严重匮乏、影像设备难以覆盖偏远社区，大量活动性病例未能被及时发现，加剧了疾病传播。

该团队自2019年起在《放射学》期刊发表基准研究，证明深度学习算法检测活动性肺结核的准确率可媲美经验丰富的放射科医生。2021年，《柳叶刀数字健康》刊发的独立评估将其模型列为全球表现最佳的结核筛查算法之一；同年，世界卫生组织推荐将计算机辅助检测用于结核筛查与分诊。2024年，《新英格兰医学杂志·AI》发表的前瞻性多中心验证进一步显示，AI辅助筛查在多元患者群体中保持高灵敏度，并能可靠识别其他偶发的肺部及心脏异常。

为推动技术落地，谷歌与南非比勒陀利亚医疗科技公司Nexus AI合作，将胸部X射线算法转化为通过认证的临床诊断工具。目前，该工具已在南非、塞拉利昂、科特迪瓦等六国的约40家医疗机构投入使用，90余名临床医生已评估超过2.5万张胸片，将诊断周转时间从数周缩短至45秒以内。双方还承诺向资源匮乏的公立诊所提供逾3000次免费筛查，目标为每年赞助1万次。在南非姆波嫩金矿——全球最深的矿井之一，Nexus AI部署了超便携数字X射线设备与离线AI系统，矿工可在井下4公里处现场完成筛查，疑似病例随即转诊进行分子确诊，将原本耗时数周的流程压缩至单次就诊。

为扩大全球影响，谷歌转向构建开放、可适配的基础模型。通过“健康AI开发者基础”计划，团队开源了胸片嵌入模型及轻量级视觉语言模型MedSigLIP，使本地团队仅凭少量区域数据即可构建专用模型。此后发布的多模态模型MedGemma支持本地离线运行，全球下载量已超过1000万次。印度尼西亚卫生部正基于本国胸片对上述模型进行微调，以推进每年筛查5000万人的目标，同时确保患者数据留在国境内。此外，谷歌研究还开发了基于3亿余条音频片段训练的声学模型HeAR，可通过普通智能手机录制的咳嗽声进行无创筛查，目前正在赞比亚、加纳等国开展合作研究。

从实验室算法到在六国诊所使用的CE认证工具，这一历程表明，将研究转化为现实影响需要严谨的科学、开放的开发者平台和专注的本地合作。谷歌表示，其角色是提供开放模型、基础研究与持续支持，帮助合作伙伴因地制宜地扩展结核病筛查与诊疗。

中文翻译：

2026年10月23日
Preeti Singh，技术项目经理负责人；Shravya Shetty，杰出软件工程师，Google Research

结核病仍然是世界上最致命的传染病之一，每年估计夺去123万人的生命。尽管全球投入了数十年的努力，但在负担最重的社区中，仍有太多病例未被发现。过去五年间，我们一直致力于改变这一轨迹，从早期研究基准走向可在诊疗现场调优和部署的开放基础模型。

能否及时获得医疗服务，往往取决于地理位置的偶然性：你住在哪里、当地诊所拥有什么资源、以及能否就近找到专科医生。

这种差距在抗击结核病（TB）的斗争中体现得尤为尖锐。尽管结核病可预防、可治愈，但每年仍夺去超过123万人的生命。一个主要瓶颈在于放射学：对于高负担地区的人群——从偏远农村地区的居民到在深井作业的矿工——放射科医生极为稀缺，影像基础设施往往需要数天才能到达。结果是，数百万活动性病例未被发现，导致传播持续蔓延。

过去几年间，我们在Google Research和Google for Health的团队一直致力于改变这一轨迹：从基于我们在医学影像领域现有工作构建的AI工具——用于识别疑似结核病患者以进行后续检测——推进到前瞻性临床试验、与当地合作伙伴共同获得监管许可，最终推出像MedGemma这样的开放基础模型，赋能全球公共卫生系统。本周在华盛顿特区举行的IDWeek 2026上，我们将分享如何将健康AI研究变为现实，以及基础模型如何在全球范围内实现结核病筛查的民主化。

我们的工作始于一个根本性问题：基于AI的工具能否以专家放射科医生的准确度解读胸部X光片（CXR）用于结核病筛查？

胸部X光摄影是肺部筛查的一线影像手段，但检测结核病的细微征象——如肺尖浸润（肺上部的液体）和空洞性病变（组织损伤）——需要专业技能，而这在农村诊所中极少具备。2019年，我们在《Radiology》上发表了一项基准研究，证明深度学习算法能够以与经验丰富的放射科医生相匹配的性能水平检测活动性肺结核。

然而，实际应用需要独立验证。2021年8月，Stop TB Partnership在《柳叶刀数字健康》上对多种自动化计算机辅助检测（CAD）软件进行了跨多样患者队列的评估，我们的模型被评为全球表现最佳的结核病筛查算法之一。同年，世界卫生组织也推荐将CAD用于结核病筛查和分诊。2024年9月，我们在《NEJM AI》上详细介绍了了一项前瞻性、多站点验证，证明AI辅助筛查在多样患者群体中保持高灵敏度（最大限度减少漏诊），同时还能可靠地检测其他偶发性肺部和心脏异常。

在医学AI领域，发表论文只是开始。在算法模型和可信的临床工具之间存在着“转化鸿沟”：使软件在各种X光机和离线环境中保持可靠、将其整合到临床工作流程中、以及满足医疗器械法规要求。

考虑到这一点，我们与总部位于比勒陀利亚的医疗技术公司Nexus AI合作，他们将我们的胸部X光算法转化为经过认证的临床诊断工具。应对监管环境需要大量审计：

如今，Nexus Intelligence正在诊断需求最紧迫的地方积极部署这项技术。

在六个国家——南非、塞拉利昂和科特迪瓦，并正在刚果民主共和国、赞比亚和喀麦隆扩展——Nexus AI的工具在约40家医疗机构中运行。迄今为止，超过90名临床医生已评估了超过25,000张胸部X光片，将周转时间从数周缩短至45秒以内，加速了快速分诊，同时赋能医疗专业人员做出最终临床决策。

我们的合作致力于健康公平：Nexus AI已为资源不足的公立诊所提供了超过3,000次免费筛查，正朝着每年10,000次赞助筛查的目标推进。

也许没有什么部署比在南非Mponeng金矿——世界上最深的矿山之一——所做的工作更能体现“在患者所在之处提供服务”的理念。深井矿工面临慢性硅尘暴露和肺部疾病风险升高的问题。Nexus AI与Stop TB Partnership、Harmony Gold Mine和南非卫生部合作，在地面以下四公里处部署了超便携数字X光套件和离线AI。矿工在现场数秒内完成筛查，被标记为疑似结核病的人员可立即转诊进行确认性分子检测。筛查、分诊和确认性检测现在可以在一次就诊中完成，将通常需要数周的流程大幅压缩。

与单一合作伙伴定制的解决方案无法独自应对全球结核病负担。为扩大影响力，我们从孤立部署转向构建开放、可适配的基础平台，使开发者和卫生部能够创建适合当地需求的解决方案。

通过健康AI开发者基础（HAI-DEF），我们开源了胸部X光嵌入和MedSigLIP——一个轻量级视觉-语言模型，经过训练将医学图像（如胸部X光片）与描述它们的临床语言对齐。由于MedSigLIP同时理解图像和文本，开发者可以用自然语言标签对其进行图像分类或搜索，或用少量本地样本针对特定任务进行微调。这使本地团队能够在适度的区域数据集上构建专用模型，无需大规模计算资源即可达到强劲性能。

在此基础上，我们发布了MedGemma——一系列开放权重多模态模型（40亿和270亿参数配置）。MedGemma从基础分类进阶到多模态临床推理：它将影像学发现与临床病史综合，用自然语言回答临床医生的查询，并可在无网络连接的设备上本地运行。自发布以来，MedGemma在全球已被下载超过1000万次。

开放模型有助于公共卫生系统构建保护患者隐私的主权解决方案。在东南亚，印度尼西亚卫生部正在应对世界第二高的结核病负担，目标雄心勃勃：每年筛查5000万公民以根除该疾病。

人工审阅5000万张X光片是不可行的，而专有云软件带来数据驻留挑战。因此，该部正在利用国内胸部X光片对MedGemma和MedSigLIP进行微调。由于模型在本地基础设施上运行，患者数据保留在国境之内，使卫生领导者对临床部署拥有完全主权。

认识到影像设备无法到达每个偏远卫生站，Google Research还开发了健康声学表征（HeAR）。HeAR在超过3亿个音频片段上训练，能够通过标准智能手机麦克风录制的咳嗽声音进行非侵入性声学筛查。

HeAR正通过撒哈拉以南非洲的合作研究加以应用，合作方包括CIDRZ（赞比亚）、Aurum Institute（加纳，获Google.org 25万美元资助）以及HELINA的150万美元区域合作项目。声学分诊（HeAR）、移动X光筛查（HAI-DEF）和多模态临床推理（MedGemma）共同为服务不足的人群提供了整合式诊断级联工具。

从实验室算法到在六个国家诊所使用的CE认证工具的历程，向我们展示了将研究转化为现实影响所需要素：严谨的科学、开放的开发者平台和专注的本地合作。因为我们的合作伙伴最了解他们的社区，他们现在正在将这项工作推向更远。Nexus AI正在向新国家扩展，印度尼西亚正在将结核病筛查纳入其国家项目，CIDRZ、Aurum Institute和HELINA正在为基于咳嗽的分诊构建临床证据。我们的角色是帮助他们获得开放模型、基础研究和持续支持，让更多结核病患者更早得到诊断和治疗。

在IDWeek 2026上，我们期待与传染病和公共卫生界建立联系，共同扩大公平医疗服务。

我们感谢Nexus AI、Stop TB Partnership、UNOPS、CIDRZ、Aurum Institute、HELINA、Harmony Gold Mine、南非卫生部和印度尼西亚卫生部的合作伙伴，感谢他们的临床领导力和对全球健康公平的奉献。

英文来源：

October 23, 2026
Preeti Singh, Lead, Technical Program Manager, and Shravya Shetty, Distinguished Software Engineer, Google Research
Tuberculosis remains one of the world's deadliest infectious diseases, claiming an estimated 1.23 million lives each year. Despite decades of global investment, too many cases go undetected in the communities that carry the highest burden. Over the past five years, we’ve worked to change that trajectory, moving from early research benchmarks toward open foundation models that could be tuned and deployed at the point of care.
Getting access to timely healthcare often comes down to an accident of geography: where you live, what resources your local clinic possesses, and whether a specialist is within reach.
Nowhere is this disparity more acute than in the fight against tuberculosis (TB). Despite being preventable and curable, TB claims more than 1.23 million lives annually. A primary bottleneck is radiology: For those in high-burden settings, from people who live in remote rural areas to miners working deep underground, radiologists are scarce, and imaging infrastructure is often days away. As a result, millions of active cases go undetected, perpetuating transmission.
Over the past several years, our team across Google Research and Google for Health has worked to change that trajectory: moving from an AI-based tool that builds on our existing work in medical imaging to identify people with presumptive TB for follow-up testing to prospective clinical trials, regulatory clearance with local partners, and ultimately open foundation models like MedGemma that empower public health systems worldwide. This week at IDWeek 2026 in Washington, D.C., we are sharing how we’re bringing health AI research to reality, and how foundational models can democratize TB screening at global scale.
Our work began with a fundamental question: Can AI-based tools interpret chest radiographs (CXRs) with the accuracy of an expert radiologist for TB screening?
Chest radiography is the frontline imaging modality for pulmonary screening, but detecting subtle signs of TB like apical infiltrates (fluid in the top portion of the lung) and cavitary lesions (tissue damage) requires specialized expertise that is rarely available in rural clinics. In 2019, we published a benchmark in Radiology, demonstrating that deep learning algorithms could detect active pulmonary TB at performance levels matching experienced radiologists.
Real-world utility, however, demands independent verification. In August 2021, the Stop TB Partnership evaluated automated computer-aided detection (CAD) software across diverse patient cohorts in The Lancet Digital Health, ranking our model among the top-performing TB screening algorithms globally. That year, WHO also recommended CAD for TB screening and triage. In September 2024, in NEJM AI, we detailed a prospective, multi-site validation demonstrating that AI-assisted screening maintains high sensitivity (minimizing missed cases) across diverse patient populations, while also reliably detecting other incidental pulmonary and cardiac abnormalities.
In medical AI, publishing a paper is only the beginning. Between an algorithmic model and a trusted bedside tool lies the "translation gap": making software reliable across varied X-ray machines and offline settings, integrating it into clinical workflows, and meeting medical device regulations.
With this in mind, we partnered with Nexus AI, a Pretoria-based medical technology company, and they translated our chest X-ray algorithms into a certified clinical diagnostic tool. Navigating the regulatory landscape required extensive audits:
Today, Nexus Intelligence is actively deploying this technology where diagnostic access is most urgently needed.
Across six countries — South Africa, Sierra Leone, and Côte d'Ivoire, with expansions underway in the Democratic Republic of the Congo, Zambia, and Cameroon — Nexus AI’s tool operates in approximately 40 healthcare facilities. To date, over 90 clinicians have evaluated more than 25,000 chest radiographs, reducing turnaround times from weeks to under 45 seconds, accelerating rapid triage while empowering healthcare professionals to make final clinical decisions.
Our collaboration commits to health equity: Nexus AI has delivered more than 3,000 free screenings to under-resourced public clinics, progressing toward a goal of 10,000 sponsored screenings annually.
Perhaps no deployment better illustrates meeting patients where they are than the work done at the Mponeng Gold Mine in South Africa, one of the deepest mines in the world. Deep-shaft miners face chronic silica dust exposure and an elevated risk of pulmonary disease. In partnership with the Stop TB Partnership, Harmony Gold Mine, and the South African Department of Health, Nexus AI deployed ultra-portable digital X-ray kits and offline AI four kilometers underground. Miners are screened on-site within seconds, and those flagged with presumptive TB can be immediately referred for confirmatory molecular testing. Screening, triage and confirmatory testing can now happen in a single visit, compressing a pathway that often takes weeks.
A custom solution with a single partner cannot address the global TB burden alone. To scale impact, we shifted from isolated deployments to building open, adaptable foundations that allow developers and health ministries to create solutions tailored to local needs.
Through the Health AI Developer Foundations (HAI-DEF), we open-sourced our chest X-ray embeddings and MedSigLIP, a lightweight vision-language model trained to align medical images, such as chest X-rays, with the clinical language used to describe them. Because MedSigLIP understands both images and text, developers can use it to classify or search images with plain-language labels, or fine-tune it for a specific task with only a small set of local examples. This lets local teams build specialized models on modest regional datasets and reach strong performance without massive compute resources.
Building on this, we released MedGemma, a family of open-weight multimodal models (4B and 27B parameter configurations). MedGemma advances from basic classification to multimodal clinical reasoning: it synthesizes radiographic findings with clinical history, answers clinician queries in natural language, and runs locally on-device without internet access. Since its launch, MedGemma has seen more than 10 million downloads worldwide.
Open models help enable public health systems to build sovereign solutions that protect patient privacy. In Southeast Asia, the Ministry of Health of Indonesia is addressing the world's second-highest TB burden with an ambitious goal: screening 50 million citizens annually to eradicate the disease.
Reviewing 50 million X-rays manually is infeasible, and proprietary cloud software poses data residency challenges. Instead, the Ministry is fine-tuning MedGemma and MedSigLIP on domestic chest radiographs. Because the models run on local infrastructure, patient data remains within national borders, giving health leaders full sovereignty over clinical deployment.
Recognizing that imaging equipment cannot reach every remote health post, Google Research also developed Health Acoustic Representations (HeAR). Trained on over 300 million audio segments, HeAR enables non-invasive acoustic screening from cough sounds recorded on a standard smartphone microphone.
HeAR is being utilized through collaborative research across sub-Saharan Africa with CIDRZ (Zambia), the Aurum Institute (Ghana, supported by a $250,000 Google.org grant), and HELINA’s $1.5 million regional collaborative. Together, acoustic triage (HeAR), mobile X-ray screening (HAI-DEF), and multimodal clinical reasoning (MedGemma) offer communities an integrated diagnostic cascade of tools for underserved populations.
The journey from a laboratory algorithm to a CE-certified tool used in clinics across six countries has shown us what it takes to turn research into real-world impact: rigorous science, open developer platforms, and dedicated local partnerships. Because our partners know their communities best, they are now taking this work further. Nexus AI is expanding to new countries, Indonesia is integrating TB screening into its national program, and CIDRZ, the Aurum Institute and HELINA are building the clinical evidence for cough-based triage. Our role is to help equip them with open models, foundational research and ongoing support, so that more people with TB are diagnosed and treated sooner.
At IDWeek 2026, we look forward to connecting with the infectious disease and public health communities as we work together to expand equitable care.
We thank our collaborators at Nexus AI, Stop TB Partnership, UNOPS, CIDRZ, Aurum Institute, HELINA, Harmony Gold Mine, the South African Department of Health, and the Indonesian Ministry of Health for their clinical leadership and dedication to global health equity.