Sign in Subscribe
Cardiac patients

Deep Learning Model Predicts Peak Oxygen Uptake in CHD Patients

Strategic Insights

— 2 min read

Researchers at Peking University Third Hospital have developed and validated a deep learning model to predict peak oxygen uptake (VO peak) in patients with coronary heart disease (CHD) using submaximal cardiopulmonary exercise testing (CPET) indicators. The study, published in BMJ Open, aimed to develop and validate prediction models for VO peak in CHD patients using submaximal CPET indicators and deep learning methods. The research team analyzed data from 10,538 patients with CHD who underwent CPET between January 2014 and December 2019, and developed multiple machine learning and deep learning models to predict VO peak.

Key Takeaways:

  • The study aimed to develop and validate prediction models for VO peak in CHD patients using submaximal CPET indicators and deep learning methods.
  • The research team analyzed data from 10,538 patients with CHD who underwent CPET between January 2014 and December 2019.
  • The best-performing deep learning model achieved an R² of 0.82, a mean absolute error (MAE) of 1.55 mL/kg/min, and a bias of 0.08.
  • The XGBoost algorithm was the best-performing traditional machine learning model, achieving an R² of 0.74.
  • SHAP analysis identified eight top-ranked features, including VO @AT, OUES, weight, VE/VCO slope, VE/VCO @AT, age, gender, and HR@AT.
  • The study concluded that the CPET deep learning model shows potential for predicting VO peak in CHD patients, but requires further external validation and prospective studies before clinical application.
  • The research was supported by the Beijing Haidian District Health Development Research And Cultivation Plan, Peking University Third Hospital Eagle Project, Youth Incubation Fund of The Peking University Third Hospital of Peking University, and Clinical Key Project of The Peking University Third Hospital of Peking University.

Statistics:

  • 10,538 patients with CHD were included in the study, who underwent CPET between January 2014 and December 2019.
  • The best-performing deep learning model achieved an R² of 0.82.
  • The XGBoost algorithm was the best-performing traditional machine learning model, achieving an R² of 0.74.
  • SHAP analysis identified eight top-ranked features: VO @AT, OUES, weight, VE/VCO slope, VE/VCO @AT, age, gender, and HR@AT.
  • The study concluded that further external validation and prospective studies are required before clinical application of the CPET deep learning model.

Sources:

  • NewsRx. Researchers at Peking University Third Hospital Publish New Data on Heart Disease (Deep learning prediction of peak oxygen uptake in patients with coronary heart disease: a retrospective study). Clinical Trials Week. October 20, 2025; p 140.
  • BMJ Open. Deep learning prediction of peak oxygen uptake in patients with coronary heart disease: a retrospective study. 2025,15(10). (BMJ Open - http://bmjopen.bmj.com/).
  • https://doi-org.sdpl.idm.oclc.org/10.1136/bmjopen-2025-098878.

Read more