跳到正文
原文
Nature 健康护理主题·· 1 天前精选AI 评分68

丹麦急诊科研究:深度学习分析27,660张面部照片预测死亡率,AUC达0.80

Facial photographs reveal mortality risk beyond triage and clinical risk scores

AI 导读

丹麦哥本哈根大学团队利用27,660张智能手机面部照片训练深度神经网络,从中位年龄73岁的急诊入院患者中预测死亡率。面部得分为3天死亡率AUC 0.80、30天死亡率AUC 0.77,优于分诊级别、早期预警评分(EWS)、生命体征、血液生化及估算面部年龄。

推荐理由

原文首次用大规模真实急诊数据验证面部图像预测死亡率,为临床风险分层提供新线索与可迁移方法。

正文

Nature Communications (2026) Cite this article

We’re sharing this article early to provide faster access to peer-reviewed, accepted research. It is citable and carries a permanent DOI. This version is subject to further edits and will be replaced automatically by the final Version of Record. All legal disclaimers apply.

Abstract

Rapid risk stratification is essential in the clinic, yet vital signs, laboratory tests, and triage scores often miss important risk at presentation. Here we show that a single facial photograph contains prognostic information about mortality risk. Using 27,660 smartphone facial photographs obtained from patients admitted to Danish emergency departments (median age 73 years; 52% male), we train deep neural networks to identify mortality risk. Face-derived scores show strong discrimination for mortality, with AUCs of 0.80 for 3-day mortality and 0.77 for 30-day mortality, outperforming triage level, early warning scores (EWS), vital signs, blood chemistry, and estimated face age. Patients with high EWS scores and the highest face score quartile have a thirty-fold increase in 30-day mortality compared to low-risk patients, and among low-EWS patients, the highest face score quartile identifies substantially elevated risk (OR of 13.39). These findings indicate that unstructured facial photographs encode clinically relevant prognostic information and may complement existing triage and monitoring strategies.

Funding

MS-K was supported by the Novo Nordisk Foundation (NNF0089176 and NNF17OC0027812) and the Danish Cancer Society (R368-A21521, R302-A17379_001, and R167-A11015_001).

Author information

Author notes

  1. These authors contributed equally: Indra Heckenbach, Rasmus Bo Hasselbalch.

  2. These authors jointly supervised this work Kasper Karmark Iversen, Morten Scheibye-Knudsen

Authors and Affiliations

  1. Center for Healthy Aging, Department of Cellular and Molecular Medicine, University of Copenhagen, Copenhagen, Denmark

    Indra Heckenbach, Michael Ben Ezra & Morten Scheibye-Knudsen

  2. Department of Cardiology, Copenhagen University Hospital – Rigshospitalet, Copenhagen, Denmark

    Kasper Karmark Iversen, Henning Bundgaard & Rasmus Bo Hasselbalch

  3. Department of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark

    Kasper Karmark Iversen

  4. Copenhagen University Hospital – Steno Diabetes Center Copenhagen, Herlev, Denmark

    Mikkel Porsborg Andersen

  5. Prehospital Center, Region Zealand, Næstved, Denmark

    Mikkel Porsborg Andersen & Henning Bundgaard

  6. Department of Health Science and Technology, Aalborg University, Aalborg, Denmark

    Mikkel Porsborg Andersen

  7. Department of Cardiology, Copenhagen University Hospital – Herlev-Gentofte, Copenhagen, Denmark

    Rasmus Bo Hasselbalch

Authors

  1. Indra Heckenbach
  2. Kasper Karmark Iversen
  3. Michael Ben Ezra
  4. Mikkel Porsborg Andersen
  5. Henning Bundgaard
  6. Morten Scheibye-Knudsen
  7. Rasmus Bo Hasselbalch

Corresponding authors

Correspondence to Indra Heckenbach, Kasper Karmark Iversen, Morten Scheibye-Knudsen or Rasmus Bo Hasselbalch.

Ethics declarations

Competing interests

IH and MS-K have filed a provisional patent related to risk prediction methods. All other authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Heckenbach, I., Iversen, K.K., Ben Ezra, M. et al. Facial photographs reveal mortality risk beyond triage and clinical risk scores. Nat Commun (2026). https://doi.org/10.1038/s41467-026-79581-8

Download citation

  • Received: 13 January 2026

  • Accepted: 23 October 2026

  • Published: 19 November 2026

  • DOI: https://doi.org/10.1038/s41467-026-79581-8

来源:Nature 健康护理主题 · nature.com