Summary:
Researchers in the UCLA Department of Anesthesiology have developed a deep learning model to accurately represent and visualize hemodynamic waveforms, or blood flow patterns, with minimally invasive approaches.
Background:
Swan-Ganz (SG) catheters are used for precise cardiac hemodynamic evaluations. Indicated for patients with severe cardiopulmonary disease undergoing high risk surgery, the catheter is often inserted through the jugular or subclavian vein to access the pulmonary artery. However, this placement is invasive, has risk of complications, and lacks survival benefits. As a result, emerging clinical guidelines discourage the use of SG catheters for cardiopulmonary evaluation. There is a need to develop new technical approaches to accurately assess surgical patients’ hemodynamic states while using only non-invasive procedures.
Innovation:
Researchers led by Professor Maxime Cannesson at the UCLA Department of Anesthesiology have developed a deep learning-based approach to collect key hemodynamic waveforms using exclusively minimally-invasive clinical interventions. The proposed deep learning model can estimate, represent, and visualize continuous pulmonary arterial pressure (PAP) and central venous pressure (CVP), key indicators of patient status during surgical operations. These intraoperative imputations can be calculated using input data from common, minimally invasive methods like electrocardiography (EEG), photoplethysmography (PPG), or arterial line catheter-derived blood pressure (Arterial BP) measurements. This deep-learning based model provides the advantages of SG catheter technology without many of the safety risks it imposes. This breakthrough offers a safer, less invasive alternative for monitoring critical hemodynamic parameters, potentially transforming the standard of care for high-risk surgical patients.
Potential Applications:
• Non-invasive cardiac monitoring
• Heart failure management
• Surgical/post-operative care
Advantages:
• Reduced surgical costs
• Improved safety
• High accuracy measurements
• Reduced infection risk
Development-To-Date:
Researchers have developed the deep learning algorithm and validated it using real cardiac data from human subjects.
Related Papers:
Hill BL, Rakocz N, Rudas Á, et al. Imputation of the continuous arterial line blood pressure waveform from non-invasive measurements using deep learning. Sci Rep. 2021;11(1):15755. doi:10.1038/s41598-021-94913-y
Reference:
UCLA Case No. 2024-224