Advances in artificial intelligence for early warning and monitoring of acute mountain sickness

LI Jia; Li Xia; Qi Hai-lan; Wang Chen-jing; SU Fang-ju

doi:10.3969/j.issn.1672-5301.2026.01.015

Abstract：Acute mountain sickness(AMS)remains a major threat for people who rapidly ascend to high altitude.This narrative review summarizes early-warning and monitoring studies published between 2021 and 2025,covering wearable continuous monitoring[nocturnal peripheral oxygen saturation(SpO2)and heart rate variability(HRV)],bedside lung ultrasound(LUS)with B-line scoring,multimodal magnetic resonance imaging(MRI),pre-ascent hematologic indices and multi-omics biomarkers,as well as regression-and machine-learning-based multimodal prediction models.Across modalities,a declining nocturnal SpO2 during ascent,increasing B-lines and hematologic panels combining hemoglobin,inflammatory,and metabolic indices can identify individuals at high risk of AMS before symptom onset.Multi-omics analyses of the transcriptome,proteome and metabolome reveal convergent signatures of immune activation,oxidative stress and energy-metabolism reprogramming,and have yielded compact biomarker sets composed of a few key molecules.However,most prediction models are derived from single-center,small samples of young volunteers,travelers or military personnel,use heterogeneous AMS definitions and lack rigorous external validation across different altitudes,ascent profiles and ethnic groups.Wearables,imaging,multi-omics and artificial intelligence therefore offer promising tools for proactive AMS surveillance,but future work should follow the Transparent Reporting of a multivariable prediction model for individual prognosis or diagnosis(TRIPOD)guideline and the TRIPOD+AI statement,develop and externally validate models in large multicenter cohorts,and implement them within a two-stage pathway of low-altitude baseline screening and dynamic monitoring during high-altitude exposure in high-risk populations.