Operating a cardiopulmonary bypass (CPB) system requires high-level multitasking while simultaneously maintaining continuous situation awareness. Among these, venous reservoir level is particularly critical. This level is monitored using sensors fixed to arbitrary positions on the reservoir. When a sudden drop in level, the rate of decline accelerates. Conventional sensor threshold settings do not allow sufficient time for the operator to recognize the danger and intervene, so a means of detecting dangerous drops in level earlier is necessary.

The purpose of this study was to develop a system that can detect a sudden drop in liquid level early, regardless of the shape of the reservoir, based on image analysis using a webcam and mathematical models, and to verify its effectiveness.

We used a webcam to quantify the level fluctuation rate in two types of venous reservoirs (Manufacturer T and Manufacturer M) with different shapes, and predicted level fluctuations from the resulting mathematical model. We further developed a system that uses the reservoir level fluctuation rate as a threshold. This system predicts dangerous situations based on the rate of level decline caused by poor drainage or massive bleeding of unknown cause, and issues alerts and alarms.

The rate of reservoir drawdown accelerated below the intermediate reservoir level. Our proposed method triggered a danger alarm an average of 7 s earlier (Manufacturer T) and 2 seconds earlier (Manufacturer M) than the conventional method (300 mL threshold).

The efficacy of an early warning system that issues alerts in advance of the conventional volume-based threshold was demonstrated by utilizing the slope (rate of change) of the level decline as the detection threshold. Implementation of this system may improve the safety of extracorporeal circulation and reduce the burden on the operator during cardiopulmonary bypass procedures.