Key result
A deep learning model using ECG and PPG waveforms accurately estimated continuous arterial blood pressure, achieving correlation coefficients of 0.96 for SBP, 0.90 for DBP, and 0.92 for MAP.
Why the study?
Non-invasive blood pressure techniques yield only discrete results, whereas continuous invasive techniques carry risks of bleeding, infection, and patient discomfort.
Can a deep learning model accurately estimate continuous arterial blood pressure in real-time using ECG and PPG waveforms in critically ill patients?
Observational (n=284)
No
Can a deep learning model accurately estimate continuous arterial blood pressure in real-time using ECG and PPG waveforms in critically ill patients?
Effect estimate: R=0.96 for SBP, R=0.90 for DBP, R=0.92 for MAP
A deep learning model using ECG and PPG waveforms can accurately estimate continuous arterial blood pressure in real-time, offering a potential non-invasive alternative to invasive arterial lines.
May support non-invasive continuous BP monitoring; leaves open prospective validation in critically ill patients before clinical adoption.
It is a well-known fact that it is important to accurately measure blood pressure. In particular, it is an even more important vital sign for the elderly and patients in hospitals. Non-invasive techniques for blood pressure measurement only afford discrete results. Although invasive measurement techniques overcome this issue and provide continuous results, they involve the risk of bleeding or infection and also cause discomfort to the patient. To address this issue, in this study, we developed a deep learning model that can estimate the arterial blood pressure (ABP) in real time using the waveform signals from electrocardiograms (ECGs) and photoplethysmograms (PPGs), without handcraft setting. Data pertaining to patients with various disorders admitted in the intensive care unit (necg, ppg= 1,126,870) were used. The performance of the model was evaluated (RSBP= 0.96, RMAP=0.92, and RDBP = 0.90) and verified to meet international standards. Even if the actual value changes dramatically, the estimated value shows a graph that follows the trend. Additionally, the ABP of patients with atrial fibrillation could be measured continuously and in real time.
No takes yet. Share an insight, caveat, or question.
Seo et al. (2022) conducted an observational in Intensive care unit patients (n=284). Deep learning model (Resnet + Bi-LSTM) using ECG and PPG waveforms vs. Invasive blood pressure monitoring was evaluated on Correlation between estimated and actual arterial blood pressure (R=0.96 for SBP, R=0.90 for DBP, R=0.92 for MAP). A deep learning model using ECG and PPG waveforms accurately estimated continuous arterial blood pressure, achieving correlation coefficients of 0.96 for SBP, 0.90 for DBP, and 0.92 for MAP.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: