An Objective Pain Measurement Machine Learning Model through Facial Expressions and Physiological Signals

The widely used technique for measuring patient pain in modern medicine is based on patients' self-reported pain. Since each human body experiences pain differently and has its own distinct pain threshold, it is very difficult for professionals to objectively evaluate a patient's level of pain based on their own description. In order for pain management treatments and prescribed therapy to be efficient and effective, a precise evaluation of pain is crucial. This study developed an objective pain measurement model based on machine learning algorithms and an open-source BioVid heat pain database. More specifically, the properties of skin conductance at various pain thresholds were examined. Electrocardiogram (ECG) and facial expression data were also assessed for performance. The study's findings demonstrate that both facial expression-based and ECG-based models can detect whether a patient is in pain. The physiological response of the subject can be utilized as a nonverbal test to gauge the patient's level of discomfort. The ECG-based model is more precise than the one based on facial expressions because we are able to gather more data with the former, therefore improving the model's accuracy, by increasing the period of the pain stimulation and the number of participants.