Machine-Learning System Monitors Patient Pain During Surgery

Researchers developed a contactless machine-learning system that monitors patient pain during surgery by analyzing facial expressions and heart rate data via remote photoplethysmogram (rPPG). The system achieved 45% accuracy when tested on realistic surgical footage, offering a non-invasive alternative to traditional pain monitoring methods that require wired sensors.

• →The system uses camera-based analysis of facial expressions and heart rate variability to estimate patient pain levels without physical sensors.
• →Researchers trained the model on realistic surgical footage lasting 30 minutes to 3 hours, including challenging conditions like poor lighting and patient movement.
• →The model achieved 45% accuracy despite using suboptimal video conditions, which is notable given the complexity of real-world surgical environments.
• →The contactless approach could benefit patients who cannot communicate pain levels, such as infants or those with dementia.
• →Researchers suggest that more complex neural network approaches could significantly improve the system's performance beyond the current statistical model.