CARMEN:基于十种生理信号预训练的心肺基础模型
A Cardiorespiratory Foundation Model Pretrained on Ten Physiological Signals for Perioperative and Critical Care
这篇来自 arXiv 的论文做了个挺有意思的模型:把 ICU 里的动脉压、颅内压、呼吸机波形这十种信号一起预训练,术中低血压预测 AUROC 到 0.894,19 个任务上没有输过。做生理信号或医疗 AI 的开发者可以仔细看看 LSCNorm 这个设计。
CARMEN 是一个针对围术期和重症监护场景的心肺波形基础模型,在 227 万 modality-hours 的数据上进行了无标签预训练,覆盖动脉压、静脉压、肺动脉压、颅内压和呼吸机波形等十种生理信号。此前生理波形基础模型几乎全部基于心电图和光电容积脉搏波构建,CARMEN 将范围扩展到手术室和 ICU 中实际指导决策的更多信号类型。
模型采用单一编码器架构,可接受从 10 秒到 5 分钟的任意组合共录信号。其关键创新是 LSCNorm 条件机制,用于恢复归一化过程中丢失的绝对信号水平——消融实验显示,去掉该机制后术中低血压预测的 AUROC 从 0.894 降至 0.780。
在来自韩国和美国队列的 19 个任务上,冻结编码器在 15 个任务中的 12 个上超过监督模型,在 19 个任务中的 14 到 17 个上超过三个基础模型(一个通用、两个生物信号专用),且没有任何任务被超越。仅用心电图或光电容积脉搏波时,它在多数对比中达到或超过信号专用基础模型,并在通道缺失时保持性能。
A Cardiorespiratory Foundation Model Pretrained on Ten Physiological Signals for Perioperative and Critical Care
Arterial, venous, pulmonary artery and intracranial pressures and ventilator waveforms guide decisions in operating rooms and intensive care units, yet foundation models for physiological waveforms have been built almost entirely on the electrocardiogram and photoplethysmogram. We present CARMEN, a single encoder pretrained without labels on ten cardiorespiratory waveforms (2.27 million modality-hours). CARMEN accepts any combination of co-recorded signals from 10 s to 5 min, and a conditioning mechanism, LSCNorm, restores the absolute signal level that normalization removes; without it, the AUROC for intraoperative hypotension fell from 0.894 to 0.780. Across 19 tasks from Korean and US cohorts, the frozen encoder outperformed supervised models on 12 of 15 and three foundation models, one general-purpose and two for biosignals, on 14 to 17 of 19, with none exceeding it on any task. Given the electrocardiogram or photoplethysmogram alone, it matched or exceeded signal-specific foundation models in most comparisons, and it retained its performance when channels were missing.
本条情报归属 AITOP「论文」核心追踪栏目。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:CARMEN、生理信号、医疗AI、基础模型,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明:本文原始线索采编自「arXiv cs.LG」(@Cheol-Hui Lee, Wookyung Kim, Hyeonhoon Lee, Garam Lee, Da-in Eun, Hyung-Chul Lee),经 AITOP 多维评分管线结构化提炼。