Pain Expression Detection System for Non-Communicable Patients

This paper presents a system for detecting facial expressions of pain using a computer vision algorithm and a camera connected to a microcontroller. The prototype was designed to alert physicians to facial expressions of pain in patients during surgical procedures. Currently, the detection of pain signals through hospital technologies has been done by monitoring variables linked to heart rate variability in combination with blood pressure. New technologies have been developed using cameras and algorithms that detect patient movements and facial expressions, aiming to obtain more data for patient assessment by the medical team. Maintaining the well-being of hospitalized individuals is a challenge in cases of non-communicative or poorly communicative patients, such as infants, adults with cognitive impairments that affect communication, or some elderly patients. The algorithm demonstrated high precision (0.99), indicating strong reliability in identifying pain without generating false positives. However, its recall or sensitivity was moderate (0.57), missing a significant number of real pain events (79 false negatives). The F1-score (0.72) and accuracy (0.78) reveal a reasonable balance between precision and recall, though with limitations in sensitivity. These results suggest that while the model is effective at confirming pain when detected, it fails to identify all occurrences. Despite this, the system holds potential as a didactic tool for educational purposes.