Heart failure & cardiomyopathies
Development of deep-learning models for real-time anaerobic threshold and peak VO2 prediction during cardiopulmonary exercise testing
Authors' conclusion
Deep-learning models for real-time CPET analysis can accurately identify AT and predict peak VO2.
Study context
- Population
- Cardiopulmonary exercise testing (CPET) records, Kyushu University Hospital, participants aged 18–90 years1,472 CPET records
- Participants
- 1,053 participants
- Comparison
- Real-time anaerobic threshold (AT) model vs Peak VO2 prediction model
- Outcome
- Real-time AT identification and peak VO2 prediction (Corr, MAE)
Methodological detailsdeep learning, single-centre, 20% held-out evaluation set
- Deep learning
- Single-centre
- 20% held-out evaluation set
Main finding
As reported in the abstract
| Group | AT Corr | AT mean absolute error (MAE) | AT time Corr | AT time MAE | Corr | MAE |
|---|---|---|---|---|---|---|
| Real-time AT model | 0.82 | 1.20 mL/kg/min | 0.86 | 0.66 min | ||
| Peak VO2 prediction model | 0.87 | 2.25 mL/kg/min |
AT Corr
Real-time AT model0.82
Peak VO2 prediction model— not stated in the abstract
AT mean absolute error (MAE)
Real-time AT model1.20 mL/kg/min
Peak VO2 prediction model— not stated in the abstract
AT time Corr
Real-time AT model0.86
Peak VO2 prediction model— not stated in the abstract
AT time MAE
Real-time AT model0.66 min
Peak VO2 prediction model— not stated in the abstract
Corr
Real-time AT model— not stated in the abstract
Peak VO2 prediction model0.87
MAE
Real-time AT model— not stated in the abstract
Peak VO2 prediction model2.25 mL/kg/min
Original publication
Development of deep-learning models for real-time anaerobic threshold and peak VO2 prediction during cardiopulmonary exercise testing. Eur J Prev Cardiol. Published 11 December 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →