| A | B | C | D | E | F | G | |
|---|---|---|---|---|---|---|---|
1 | Direction | Category | Work | Venue | Tagline | 标语 | Link |
2 | AI-Med Datasets | HEEDB | Sci Data'26 | The world's largest free, public collection of 12-lead ECGs—11.6 million recordings from 2.17 million patients—built by Harvard and Emory | 全球最大的免费公开 12 导联心电图数据库——收录 217 万病人的 1160 万份记录——由哈佛和埃默里大学建立 | https://bdsp.io/content/heedb | |
3 | AI-Med Datasets | MEETI | Sci Data'26 | The first large-scale ECG dataset—~800,000 records—that synchronizes raw signals, images, beat-level measurements, and AI-written interpretations | 首个大规模心电图数据集——约 80 万份记录——同步整合原始信号、图像、逐搏参数和 AI 解读文本 | https://www.nature.com/articles/s41597-026-06796-1 | |
4 | AI-Med Datasets | MieDB-100k | A large-scale, high-quality dataset for teaching AI to edit medical images from text instructions | 一个大规模、高质量的数据集，用于训练 AI 按文字指令编辑医学影像 | https://arxiv.org/abs/2602.09587 | ||
5 | Biosignal Foundation Models | ECG | CardioLearn | TheWebConf'20 | A plug-and-play cloud AI service that lets any device detect heart disease from ECG via a simple API | 一个即插即用的云端 AI 服务，让任何设备都能通过简单的接口（API）从心电图中检测心脏疾病 | https://dl.acm.org/doi/fullHtml/10.1145/3366424.3383529 |
6 | Biosignal Foundation Models | ECG | ECGFounder | NEJM AI'25 | A general-purpose AI foundation model that diagnoses 150 heart conditions from ECGs—including single-lead recordings from wearables | 一个通用的 AI 基础模型，能从心电图中诊断 150 种心脏疾病，包括来自可穿戴设备的单导联记录 | https://ai.nejm.org/stoken/default+domain/BFNQQWBEQRXK464GCEPW/full?redirectUri=doi/full/10.1056/AIoa2401033 |
7 | Biosignal Foundation Models | ECG | AnyECG | An AI foundation model that reads a single ECG to screen for 1,000+ diseases across the whole body—and predict future risk | 一个 AI 基础模型，仅凭一份心电图就能筛查全身 1000 多种疾病，并预测未来患病风险 | https://arxiv.org/abs/2601.10748 | |
8 | Biosignal Foundation Models | PPG | AnyPPG | KDD'26 | An AI foundation model that turns a simple wearable pulse signal (PPG) into a whole-body health screen for 1,400+ diseases | 一个 AI 基础模型，把可穿戴设备上简单的脉搏信号（PPG）变成可筛查 1400 多种疾病的全身健康检测 | https://arxiv.org/abs/2511.01747 |
9 | Biosignal Foundation Models | HR | SleepFounder | An AI foundation model that delivers gold-standard sleep assessment—and screens for multi-organ disease—from effortless heart and breathing signals at home | 一个 AI 基础模型，仅凭在家就能轻松采集的心肺信号，就能实现金标准级的睡眠评估，并筛查多器官疾病 | https://www.medrxiv.org/content/10.1101/2025.09.06.25335216v1 | |
10 | Biosignal Foundation Models | EEG | SpikeNet2 | NEJM AI'25 | An expert-level AI model that detects epilepsy spikes in EEG recordings while sharply cutting false alarms | 一个专家级的 AI 模型，能从脑电图（EEG）中检测癫痫放电（spike），同时大幅减少误报 | https://ai.nejm.org/stoken/default+domain/F8GCKVTKBECHKRHVCGTT/full?redirectUri=doi/full/10.1056/AIoa2401221 |
11 | Biosignal Foundation Models | Spirogram | DeepSpiro | npj Syst Biol Appl'25 | A deep-learning model that reads a simple breathing test to predict COPD risk years before diagnosis | 一个深度学习模型，通过简单的肺功能（呼吸）测试，提前数年预测慢阻肺（COPD）的患病风险 | https://www.nature.com/articles/s41540-025-00489-y |
12 | Biosignal Foundation Models | FHR | FHRFounder | The first AI foundation model that mines routine fetal heart-rate monitoring to predict pregnancy risks for both baby and mother | 首个 AI 基础模型，从常规的胎心监护数据中挖掘隐藏信息，预测胎儿和母亲的多种妊娠风险 | ||
13 | Biosignal MLLMs and Agents | MLLMs | HeartLang | ICLR'25 | An AI framework that reads ECGs like language—treating heartbeats as words and rhythms as sentences—to learn from unlabeled data | 一个 AI 框架，把心电图当作语言来"阅读"——将心跳视为单词、心律视为句子——从而无需标注就能学习 | https://openreview.net/forum?id=sdczcOS3n9 |
14 | Biosignal MLLMs and Agents | MLLMs | GEM | NeurIPS'25 | The first multimodal AI that reads ECG signals, images, and text together—and explains each diagnosis by pointing to the exact waveform evidence | 首个多模态 AI，能同时读懂心电图的信号、图像和文字，并通过指出具体的波形证据来解释每一个诊断 | https://openreview.net/forum?id=idtZwmjakN |
15 | Biosignal MLLMs and Agents | MLLMs | ECG-R1 | ICML'26 | The first reasoning AI for ECG interpretation—grounding each diagnosis in measurable evidence to curb the hallucinations that plague current models | 首个具备推理能力的心电图解读 AI——让每个诊断都基于可测量的证据，从而抑制现有模型普遍存在的"幻觉" | https://www.arxiv.org/abs/2602.04279 |
16 | Biosignal MLLMs and Agents | MLLMs | UniECG | An interactive AI tutor for ECG education that both explains real ECGs with evidence and generates example ECGs from a learning goal | 一个用于心电图教学的交互式 AI 助教——既能给真实心电图提供有据可循的解释，也能根据学习目标生成示例心电图 | https://arxiv.org/abs/2509.18588 | |
17 | Biosignal MLLMs and Agents | MLLMs | SpiroLLM | PLOS DH'25 | The first large language model that understands breathing-test (spirogram) curves—diagnosing COPD and explaining its reasoning in a full report | 首个能读懂肺功能（呼吸）曲线的大语言模型——既能诊断慢阻肺（COPD），还能生成完整报告解释其推理过程 | https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001300 |
18 | Biosignal MLLMs and Agents | Agents & Platforms | KidneyTalk | A no-code desktop app that runs a private, on-device AI for kidney-disease consultations—keeping patient data fully local and secure | 一个无需写代码的桌面应用，在本地设备上运行私有 AI 提供肾病咨询——病人数据全程留在本地，安全保密 | https://kidneytalk.bjmu.edu.cn/ | |
19 | Biosignal MLLMs and Agents | Agents & Platforms | ZhunXin | An AI agent that reads a phone photo of an ECG to triage the three deadliest causes of chest pain in under 3 seconds—bringing top-hospital expertise to frontline clinics | 一个只需手机拍照上传心电图，就能在 3 秒内对"胸痛致命三联征"分级预警的 AI 智能体，把三甲专家的判读能力带到基层 | https://www.zhunxinagent.com/ | |
20 | Biosignal MLLMs and Agents | Agents & Platforms | HeartOS | A "NotebookLM for the digital heart"—a unified platform that integrates the lab's data, foundation models, LLMs, and agents into one interactive entry point for cardiac-AI research and teaching | 一个对标 NotebookLM 的数字心脏智能体平台，把实验室的数据、基础模型、大模型与智能体集成为统一入口，服务心脏 AI 的科研与教学 | http://ai.heartvoice.com.cn/heartOS/ | |
21 | Digital Biomarker Discovery | Aging | PPGage | Commun Med'25 | An AI that estimates your "biological age" from a wearable pulse signal (PPG)—a non-invasive biomarker flagging higher cardiovascular risk | 一个 AI，仅凭可穿戴设备的脉搏信号（PPG）就能估算你的"生理年龄"——一种非侵入式的生物标志物，可提示更高的心血管风险 | https://www.nature.com/articles/s43856-025-01188-9 |
22 | Digital Biomarker Discovery | Aging | ECGage | An AI that reads an ECG to gauge your heart's "biological age"—a non-invasive biomarker that flags hidden accelerated aging and cardiovascular risk | 一个 AI，通过解读心电图来评估心脏的"生理年龄"——一种非侵入式生物标志物，可揭示隐匿的心脏加速衰老和心血管风险 | https://www.medrxiv.org/content/10.64898/2026.03.24.26349186v2 | |
23 | Digital Biomarker Discovery | Aging | CTGage | An AI that reads routine fetal heart-rate monitoring (CTG) to estimate a "biological age"—a non-invasive biomarker predicting future pregnancy risks | 一个 AI，通过解读常规胎心监护（CTG）来估算"生理年龄"——一种非侵入式生物标志物，可预测未来的妊娠风险 | https://arxiv.org/abs/2509.14242 | |
24 | Digital Biomarker Discovery | Indices | Sleep Depth Index | npj Digit Med'25 | A deep-learning method that turns coarse sleep stages into a continuous "sleep depth index"—revealing finer sleep structure and new health-risk biomarkers | 一个深度学习方法，把粗糙的睡眠分期转化为连续的"睡眠深度指数"——揭示更精细的睡眠结构，并产生预测健康风险的新型数字生物标志物 | https://doi.org/10.1038/s41746-025-01607-0 |
25 | Digital Biomarker Discovery | Omics | ECGomics | HDS'25 | An "omics-level" framework for ECG analysis that fuses expert-defined features with AI embeddings—delivering interpretable digital biomarkers across four dimensions | 一个"组学级"的心电图分析框架，把专家定义的特征与 AI 的深度表征相融合——从四个维度产生可解释的数字生物标志物 | https://spj.science.org/doi/10.34133/hds.0427 |
26 | Digital Biomarker Discovery | Omics | ECG-gene | A deep-learning framework that reads a routine ECG to flag inherited cancer-risk genes—turning a standard heart test into a low-cost cancer-screening tool | 一个深度学习框架，通过解读常规心电图来识别遗传性癌症易感基因——把一份标准的心脏检查变成低成本的癌症风险筛查工具 | ||
27 | AI-Enhanced Health Devices | AI-ECG Device 问心无恙®️ | Portable ECG Device | A pocket-sized, home-use medical ECG monitor that lets anyone record their heart's signal and screen for atrial fibrillation anytime, anywhere | 问心无恙 A1s——一款手掌大小、家用便携的医用心电图检测仪，让用户随时随地自测心电、筛查房颤 | https://item.jd.com/100077268268.html | |
28 | AI-Enhanced Health Devices | AI-ECG Device 问心无恙®️ | Flexible ECG Patch | A home-use wearable Holter monitor that records your heart's rhythm continuously for 24 hours—catching irregular beats that a quick check would miss | 问心无恙 E-PC01——一款家用便携的动态心电图仪，可连续 24 小时监测心率与心电，捕捉日常容易漏掉的心律异常 | https://item.jd.com/100075297613.html | |
29 | AI-Enhanced Health Devices | AI-ECG Device 问心无恙®️ | ECG Pad | A smart ECG terminal that reads your heart rhythm in 30 seconds with a touch of your hands—now deployed in 30 hospitals for rapid triage and cardiac screening | 一台只需把手放上去、30 秒即可测出心律并云端 AI 分析的智能心电仪，用于快速分诊与心脏筛查 | https://m.peopledailyhealth.com/articleDetailShare?articleId=d82aa59146874d03bf9278852dcea240&_t=1736855177391 | |
30 | AI-Enhanced Health Devices | Mobile Phone as a Medical Device | ScanECG | An automated ECG image enhancement tool that transforms low-quality ECG photos into clean, standardized, high-resolution ECG images by reconstructing grids, correcting distortion, and extracting waveforms | 一个自动化心电图图像增强工具，可通过网格重建、图像拉伸校正与波形提取，将低质量拍摄心电图转换为清晰、标准化的高质量心电图图像 | ||
31 | AI-Enhanced Health Devices | Mobile Phone as a Medical Device | TraceECG | An ECG image digitization tool that faithfully extracts waveforms from ECG images, identifies lead layouts, and reports confidence based on waveform completeness | 一个心电图图像数字化工具，可忠实提取图片中的心电波形，识别导联排布，并根据波形缺失程度给出数字化置信度 | ||
32 | AI-Enhanced Health Devices | Mobile Phone as a Medical Device | ImputeECG | An AI that reconstructs complete 12-lead ECGs from partial or corrupted recordings—restoring near-full diagnostic accuracy and rescuing incomplete ECG archives | 一个 AI，能从残缺或损坏的心电记录中重建完整的 12 导联心电图——恢复近乎完整的诊断准确度，让不完整的心电档案重新可用 | ||
33 | AI-Enhanced Health Devices | Mobile Phone as a Medical Device | PhysWave | A lightweight AI that recovers high-quality pulse waveforms from ordinary facial video—delivering contactless heart monitoring with a tenth of the usual compute | 一个轻量级 AI，仅凭普通的人脸视频就能还原高质量的脉搏波形——以十分之一的算力实现非接触式心脏监测 | ||
34 | AI-Enhanced Health Devices | Device Repurposing | Holter-to-Sleep | A framework that turns a single-lead Holter ECG into both an overnight sleep assessment and a cardiac screen—enabling low-burden, at-home cardio-sleep monitoring | 一个框架，仅凭单导联 Holter 心电图，就能同时完成整夜睡眠评估和心脏筛查——实现低负担、可居家的"心脏-睡眠"联合监测 | https://arxiv.org/abs/2603.18714 | |
35 | AI-Enhanced Health Devices | Device Repurposing | AnyECG-Echo | An AI that detects 13 types of structural heart disease from a wearable single-lead ECG—bringing echocardiography-level screening to population scale | 一个 AI，仅凭可穿戴单导联心电图就能检测 13 种结构性心脏病——把超声心动图级别的筛查推向人群规模 | https://arxiv.org/abs/2606.09332 | |
36 | AI-Enhanced Health Devices | Device Repurposing | AnyECG-CCTA | An AI that reads a standard ECG to estimate vessel-specific coronary artery blockage—offering a low-cost, radiation-free first screen for heart disease | 一个 AI，通过解读标准心电图来估算各支冠脉的狭窄程度——为冠心病提供一种低成本、无辐射的初筛工具 | https://arxiv.org/abs/2512.05136 | |
37 | AI-Enhanced Health Devices | Device Repurposing | AnyECG-Lab | An AI that reads an ECG to estimate dozens of blood-test values—pointing toward fast, needle-free lab screening | 一个 AI，通过解读心电图来估算数十项血液化验指标——为快速、无需抽血的化验筛查提供了可能 | https://arxiv.org/abs/2510.22301 | |
38 | AI-Enhanced Health Devices | Data Generation | DiffuSETS | Patterns'25 | An AI generator that creates realistic, clinically faithful ECG signals from text reports—easing the data shortage that holds back cardiac AI | 一个 AI 生成器，能根据临床文本报告生成逼真且具临床意义的心电信号——缓解制约心脏 AI 发展的数据短缺难题 | https://www.cell.com/patterns/fulltext/S2666-3899(25)00139-4 |
39 | AI-Enhanced Health Devices | Data Generation | PPGFlowECG | An AI that translates a wearable's pulse signal (PPG) into a clinically useful ECG—bringing gold-standard cardiac screening to everyday devices | 一个 AI，能把可穿戴设备的脉搏信号（PPG）转换成具临床价值的心电图——让日常设备也能实现金标准级的心脏筛查 | https://arxiv.org/abs/2509.19774 | |
40 | AI-Enhanced Health Devices | Data Generation | MCMA | npj Cardiovasc Health'24 | An AI that reconstructs a full 12-lead ECG from any single-lead wearable recording—closing the gap between clinical and at-home heart monitoring | 一个 AI，能从任意单导联可穿戴心电记录重建完整的 12 导联心电图——弥合临床与居家心脏监测之间的差距 | https://www.nature.com/articles/s44325-024-00036-4 |
41 | AI-Enhanced Health Devices | Data Generation | WearECG | PLOS DH'25 | An AI that reconstructs a full 12-lead ECG from just three wearable leads—preserving clinically meaningful detail for scalable, low-cost cardiac screening | 一个 AI，仅凭三个可穿戴导联就能重建完整的 12 导联心电图——保留具有临床意义的细节，实现可规模化、低成本的心脏筛查 | https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001335 |
42 | AI-Enhanced Health Devices | Data Generation | ECG360 | A generative tool that expands standard 12-lead ECGs into synchronized high-density body-surface ECGs, helping clinicians detect spatial cardiac electrical patterns beyond conventional 12-lead views | 一个基于生成模型的心电扩展工具，可将临床常规 12 导联心电图扩展为同步的高密度体表心电图，帮助医生发现传统 12 导联难以呈现的空间电活动异常 | ||
43 | AI-Enhanced Health Devices | Data Generation | ECGFlowCMR | KDD'26 | An AI that generates moving cardiac MRI videos from a cheap, routine ECG—offering a scalable, low-cost window into heart structure and motion | 一个 AI，能从廉价、常规的心电图生成动态的心脏 MRI 影像——以可规模化、低成本的方式呈现心脏的结构与运动 | https://arxiv.org/abs/2601.20904 |
44 | Clinical Validation | Screening | AF | A large real-world study showing AI can detect hidden atrial fibrillation from a portable single-lead ECG in normal rhythm—flagging risk up to 2.5 years before the first episode | 一项大规模真实世界研究证明，AI 仅凭便携单导联设备在正常心律下采集的心电图，就能识别隐匿的房颤——在患者首次发作前 2.5 年即可预警 | ||
45 | Clinical Validation | Screening | SVT | A large real-world study showing an AI-ECG phone system, paired with patient-requested cardiologist review, reliably detects supraventricular tachycardia while cutting review workload by ~99% and patient costs by ~80% | 一项大规模真实世界研究表明，AI-ECG 手机系统结合患者主动发起的医生复核，能可靠检测室上性心动过速（SVT），同时将医生复核工作量减少约 99%、患者成本降低约 80% | https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5775869 | |
46 | Clinical Validation | Screening | WPW | A large real-world study showing a portable AI-ECG system, paired with cardiologist review, can screen for rare but deadly WPW syndrome—cutting manual review by ~99.5% while concentrating risk among true cases | 一项大规模真实世界研究表明，便携 AI-ECG 系统结合医生复核，可筛查罕见却致命的预激综合征（WPW）——将人工复核量减少约 99.5%，同时把高风险精准集中在真正的患者身上 | https://arxiv.org/abs/2510.24750v2 | |
47 | Clinical Validation | Screening | VT | PLOS DH'25 | A deep-learning approach that detects life-threatening ventricular arrhythmia from an ECG and adapts to a new individual with just a handful of heartbeats | 一个深度学习方法，能从心电图中检测危及生命的室性心律失常，并仅凭少量心跳就快速适配到新的个体 | https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001037 |
48 | Clinical Validation | Screening | AVB、LQTS | npj Cardiovasc Health'26 | An on-device algorithm that measures key ECG intervals from a single-lead wearable at physician-level accuracy—enabling continuous cardiac monitoring outside the hospital | 一个在设备端运行的算法，仅凭单导联可穿戴设备就能以医生级精度测量关键心电参数——支持院外的连续心脏监测与远程筛查 | https://www.nature.com/articles/s44325-026-00145-2 |
49 | Clinical Validation | Opportunistic Screening | ECG-CKD | An AI-ECG opportunistic screening system that turns routine 12-lead ECGs into early CKD risk alerts—automatically identifying high-risk patients, routing them to community physicians, and testing whether closed-loop outreach increases new CKD diagnosis in a pragmatic cluster-randomized trial | 一个 AI 心电机会性筛查系统，能将常规 12 导联心电图转化为慢性肾脏病早期风险预警——自动识别高危患者并推送给属地基层医生，通过闭环随访与确诊检测，在真实世界整群随机试验中评估其是否提高 CKD 早期诊断率 | ||
50 | Clinical Validation | Opportunistic Screening | PPG-AVD | An AI that screens for aortic valve disease from a wearable's pulse signal (PPG)—using physiology-guided self-supervised learning to overcome scarce labeled data, offering a low-cost alternative to echocardiography | 一个 AI，仅凭可穿戴设备的脉搏信号（PPG）就能筛查主动脉瓣疾病——借助生理知识引导的自监督学习克服标注数据稀缺，为超声心动图提供低成本替代方案 | https://arxiv.org/abs/2602.04266 | |
51 | Clinical Validation | Opportunistic Screening | Spiro-RHF | A self-supervised AI that uses a simple breathing test (spirogram) to flag right heart failure early in patients with lung disease—offering a low-cost screen for those at elevated risk | 一个自监督 AI，仅凭简单的肺功能（呼吸）测试就能在肺病患者中早期识别右心衰竭——为高风险人群提供低成本的筛查手段 | https://arxiv.org/abs/2511.13457 | |
52 | Clinical Validation | RCTs | Pocket-K | A single-lead AI-ECG system that non-invasively screens for life-threatening high potassium (hyperkalemia)—running on a handheld device for monitoring outside the hospital | 一个单导联 AI-ECG 系统，无创筛查危及生命的高钾血症——可在手持设备上运行，实现院外监测 | https://arxiv.org/abs/2603.14177 | |
53 | Clinical Validation | RCTs | PocketED-K | A single-lead AI-ECG system that prescreens for life-threatening low potassium (hypokalemia) in the ER—a low-burden, handheld tool to prioritize who needs a blood test first | 一个单导联 AI-ECG 系统，在急诊中预筛查危及生命的低钾血症——一种低负担的手持工具，用于优先确定谁需要先做血钾检测 | https://www.medrxiv.org/content/10.64898/2026.05.23.26353774.abstract | |
54 | Clinical Validation | RCTs | Pocket-NT-proBNP | A single-lead AI-ECG model that noninvasively flags elevated NT-proBNP—a key heart-failure blood marker—in under a minute on a handheld device, extending risk screening to home and community settings | 一个单导联 AI-ECG 模型，无需抽血就能在手持设备上不到一分钟内识别 NT-proBNP（关键的心衰血液标志物）升高——把风险筛查延伸到居家和社区场景 | ||
55 | Clinical Validation | RCTs | ECG-RVEF | Pulm Circ'26 | A prospective multicenter study that will track right-heart recovery after pulmonary embolism using daily portable-ECG and AI—aiming to enable earlier risk stratification and personalized care | 一项前瞻性多中心研究，将通过每日便携心电图结合 AI 追踪肺栓塞后右心功能的恢复过程——旨在实现更早的风险分层和个体化管理 | https://onlinelibrary.wiley.com/doi/abs/10.1002/pul2.70312 |
56 | Clinical Validation | Special Populations | Pregnant Women | BMC MIDM'25 | A clinical validation study showing the portable WenXinWuYang single-lead AI-ECG reliably measures heart rate and detects arrhythmias in pregnant women, closely matching standard 12-lead ECG | 一项临床验证研究表明，便携式"问心无恙"单导联 AI-ECG 在孕妇中能可靠测量心率、检测心律失常，结果与标准 12 导联心电图高度一致 | https://link.springer.com/article/10.1186/s12911-025-02952-6 |