Milestone of Pain Intensity Evaluation from Facial Action Units

A self-report pain scale is very subjective and lacks precise measure to determine the important levels of pain. Identifying milestone of pain intensity could, not only help healthcare professionals in assessing the intensity of pain at a particular time but is also significant for pain management evaluation on whether the onset of pain is sudden, gradual or part of an ongoing problem. This study has shown a novel approach on evaluating the timing information of pain by classifying section of frames where the pain was triggered, reached its climax and started to diminish. Convolutional neural network was used to extract features from UNBC McMaster Shoulder Pain Archive Database and intensities were derived from the Prkachin and Solomon Pain Intensity (PSPI) metric scale. Low pass filter was applied in smoothing discrete frame series data of pain. Relevant to the development of automatic pain intensity evaluation system, the study estimated sequence and frame-level of data based on the changes of intensity since onset and has shown 60% accuracy in evaluating pain carried out by the clinical experts.