Haemodynamics, fluids & transfusion
Prediction of intraoperative hypotension using deep learning models based on non-invasive monitoring devices
Authors' conclusion
A deep learning model utilizing multi-channel non-invasive monitors could predict intraoperative hypotension with high accuracy.
Study context
- Population
- Non-cardiac surgery patients with non-invasive monitoring (non-invasive BP, ECG, photoplethysmography, capnography, BIS); hypotension = SBP <90 mmHg
- Participants
- 4,754 vs 421
- Comparison
- Original data vs External validation data
- Original datan = 4,754
- External validation datan = 421
- Outcome
- Performance of the deep learning model (AUROC) for predicting intraoperative hypotension 5 min before occurrence
Methodological detailsdeep learning model development, external validation
- Deep learning model development
- External validation
Performance
As reported in the abstract
AUROC for intraoperative hypotension prediction
| Test or model | AUC |
|---|---|
| Original data | 0.917 (95% CI 0.915-0.918) |
| External validation data | 0.833 (95% CI 0.830-0.836) |
AUC
Original data0.917 (95% CI 0.915-0.918)
External validation data0.833 (95% CI 0.830-0.836)
Further findings
Contribution of each monitor (attention map)
- Weights of each monitor for determining hypotension
- 8 to 22%
Original publication
Prediction of intraoperative hypotension using deep learning models based on non-invasive monitoring devices. J Clin Monit Comput. Published 19 August 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →