Neural Network-based Estimation of Microbubbles Generated in Cardiopulmonary Bypass Circuit: A Clinical Application...

Area of Science:

• Cardiovascular Surgery
• Biomedical Engineering
• Medical Physics

Background:

• Cardiopulmonary bypass systems can generate microbubbles (MBs).
• Microbubbles pose risks, including neurocognitive dysfunction.
• Accurate estimation of MBs is needed for surgical management.

Purpose of the Study:

• To evaluate a previously developed neural network model for estimating MBs.
• To adapt and validate the model using data from actual clinical cardiac surgery cases.
• To assess the model's accuracy in a real-world surgical environment.

Main Methods:

• A neural network model was utilized, incorporating suction flow rate, venous reservoir level, blood viscosity, and perfusion flow rate.
• The model's estimation of MBs was compared against measured MBs in four clinical cases.
• Analysis included assessing the coefficient of determination (R²) and the impact of surgical interventions.

Main Results:

• The initial model achieved a coefficient of determination (R²) of 0.558 (p<0.001) across all surgical procedures.
• Surgical treatments like drug administration, fluid, and blood transfusions were found to increase measured MBs.
• Excluding these treatment periods significantly improved the model's accuracy, yielding an R² of 0.8762 (p<0.001).

Conclusions:

• The neural network model demonstrates potential for accurate MB estimation during cardiac surgery.
• Accounting for transient increases in MBs due to surgical interventions is key to improving model performance.
• The refined model shows high accuracy in estimating MBs within a clinical setting.