Comparison of Noninvasive Continuous Blood Pressure Measurement Using Artificial Intelligence-Assisted Pulse Oximetry and Electrocardiography Signal Combination With Invasive Arterial Blood Pressure
Comparison of Noninvasive Continuous Blood Pressure Measurement Using Artificial Intelligence-Assisted Pulse Oximetry and Electrocardiography Signal Combination With Invasive Arterial Blood Pressure
The goal of this clinical trial is to evaluate a new, noninvasive method for continuously measuring blood pressure during surgery. This method combines two common types of signals - pulse oximetry and electrocardiography (ECG) - and uses artificial intelligence (AI) to help improve accuracy. Researchers will compare this method to the standard invasive arterial blood pressure (IABP) monitoring to see if it can safely and effectively replace invasive measurements in certain situations.
In surgeries, keeping blood pressure stable is extremely important. Sudden drops (hypotension) or spikes (hypertension) in blood pressure can cause serious organ problems after surgery. That's why continuous blood pressure monitoring is necessary, especially during key moments like anesthesia induction, intubation, and surgical positioning. In high-risk patients, clinicians often use invasive monitoring by inserting a catheter into an artery. While accurate, this method can cause complications, including bleeding, infection, blood clots, and in rare cases, damage to the limb. Studies show that blood clots can occur in nearly 20% of radial arterial lines, and up to 15% of cases experience bleeding or hematoma. Even life-threatening problems such as limb ischemia and embolism still occur in about 1% of patients.
Because of these risks, noninvasive continuous blood pressure measurement (NCBPM) devices are needed. Pulse oximeters, which are already used in all surgeries, measure oxygen in the blood and produce a waveform called photoplethysmography (PPG). This waveform changes with each heartbeat and reflects blood volume and flow. It can provide useful cardiovascular information, and in recent years, researchers have explored using PPG to estimate blood pressure. However, PPG alone may not give accurate results in all patients or conditions.
One reason for this is that important heart-related signals - like the time between heartbeats (R-R interval), the heart's pumping strength, and electrical conduction - are not captured by PPG alone. These signals are available through ECG. By combining ECG with PPG, researchers hope to get a more complete understanding of each heartbeat, and use this to estimate blood pressure more accurately.
In this study, the new combined method is called CPES (combined pulse oximeter and ECG signals). It uses signals from both devices, analyzed by artificial intelligence, to generate continuous blood pressure readings. The study will compare these readings with direct invasive arterial blood pressure values.
The main question the study aims to answer is:
Can CPES accurately measure systolic and diastolic blood pressure within 5 mmHg of the invasive method? This threshold is based on international standards (AAMI) for acceptable accuracy in blood pressure monitoring during surgery.
Secondary questions include:
Can CPES track rapid blood pressure changes during critical moments, such as after anesthesia drugs are given, during laryngoscopy, intubation, and surgical maintenance? Does the accuracy of KPES stay the same when the surgical table is tilted 15 degrees up (reverse Trendelenburg) or down (Trendelenburg)? These positions are commonly used in surgery and may affect blood pressure measurements.
Participants in this study will be adult patients undergoing general anesthesia. All participants will:
Have standard invasive arterial monitoring during surgery Wear ECG electrodes and a pulse oximeter Be monitored in real time using both standard and experimental systems Be placed in different surgical positions (as tolerated) for short periods to evaluate effects on blood pressure measurement Data will be collected at various stages of surgery and in different positions. The study does not change routine care but adds additional, noninvasive sensors for research purposes.
This study may help develop safer and more comfortable ways to measure blood pressure continuously during surgery. If successful, this method could reduce the need for invasive lines in many patients, lowering the risk of complications and making surgery monitoring easier and more cost-effective. It may also be useful in settings where invasive monitoring is not possible, such as emergency departments or outpatient procedures.
Intraoperative blood pressure management plays a critical role in preventing postoperative complications, including organ dysfunction. While oscillometric intermittent noninvasive techniques are the standard for blood pressure monitoring in most surgical patients, they may fail to capture rapid hemodynamic fluctuations, especially during high-risk periods such as induction, intubation, and sudden positioning changes. In these cases, continuous invasive arterial blood pressure (IABP) monitoring is often used for real-time hemodynamic surveillance. However, IABP monitoring carries risks, including bleeding, thrombosis, hematoma, arterial occlusion, and bloodstream infections. Reported radial artery thrombosis rates reach as high as 19.7%, and catheter-related complications such as hematoma or bleeding occur in up to 15% of cases.
There is an increasing demand for noninvasive continuous blood pressure monitoring (NCBPM) technologies to avoid such complications. Photoplethysmography (PPG), derived from pulse oximeters, has emerged as a promising noninvasive technique. PPG waveforms reflect peripheral perfusion changes and offer insights into cardiac cycle dynamics. However, standalone PPG-based estimation is limited by its dependency on peripheral vascular tone, motion artifacts, and lack of electrical cardiac data.
To overcome these limitations, this study proposes a hybrid approach using synchronized photoplethysmography (PPG) and electrocardiography (ECG) signals, enhanced with artificial intelligence (AI)-based waveform analysis. This combination aims to improve the accuracy and temporal resolution of noninvasive continuous blood pressure monitoring. The proposed method, termed CPES (Combined Pulse Oximeter and ECG Signals), is based on the hypothesis that integrating electrical (ECG) and optical (PPG) signals will allow better modeling of arterial pulse wave velocity (PWV), pulse transit time (PTT), and surrogate hemodynamic parameters.
The study design includes intraoperative real-time recording of PPG and ECG signals in patients undergoing elective surgery under general anesthesia with concurrent IABP monitoring via radial artery catheterization. PPG waveforms will be acquired using a standard red/infrared dual-wavelength pulse oximeter (660 nm and 940-960 nm), and ECG will be collected using 3-lead continuous monitoring. Synchronization will be ensured using a time-stamping algorithm. Data acquisition will occur during distinct intraoperative phases: baseline (pre-induction), induction, intubation, maintenance, and positional interventions (±15° Trendelenburg and reverse Trendelenburg).
The primary outcome is the agreement between CPES-derived systolic and diastolic blood pressure values and simultaneously recorded invasive blood pressure values, assessed via Bland-Altman analysis. Clinical accuracy will be determined by comparing mean absolute error (MAE) and root mean square error (RMSE) to the AAMI/ANSI/ISO 81060-2 standard, which requires a mean difference of ≤5 mmHg and standard deviation of ≤8 mmHg.
Secondary outcomes include:
Ability of the CPES system to track rapid blood pressure changes (ΔSBP, ΔDBP) during induction and airway manipulation.
Evaluation of positional effects on blood pressure estimation accuracy using paired positional comparisons.
Exploratory modeling of derived hemodynamic parameters (e.g., PTT, HRV, PWV) and their correlations with arterial wave morphology and invasive BP metrics.
Signal processing will involve artifact removal, beat detection, normalization, and segmentation. Feature extraction will include time-domain, frequency-domain, and morphological features from both PPG and ECG. Machine learning models (e.g., gradient boosting, LSTM, or convolutional neural networks) may be trained on the dataset to enhance predictive performance.
This study aims to demonstrate that combining electrical and optical signals can yield clinically acceptable blood pressure measurements noninvasively, improving patient safety and reducing the need for invasive lines. If validated, this approach could be integrated into existing operating room monitors, expanding access to continuous BP monitoring in routine and resource-limited settings.
Inclusion Criteria:
Exclusion Criteria: