Automatic spontaneous pain recognition using supervised classification learning algorithms

Self reporting is the standard measure of pain in most medical institutes. However, pain evaluation by mere patient self reporting is not a good measure of the amount of pain a person is suffering from. Some prefer not to give a lot of information pertaining to their actual level of pain, due to fear of being diagnosed with severe diseases. Even a seasoned care giver cannot properly evaluate the amount of pain a mute patient is going through because they cannot communicate verbally. The Prkachin and Solomon pain intensity (PSPI) scale is one of the main contributions towards evaluating pain to date that complements the process of self reporting. It evaluates active facial action units along with their intensities and give a score. Manually assessing PSPI scores is time consuming and error prone. This paper seeks to minimise the problem of pain assessment using supervised classification learning algorithms, to notice subtle muscle changes on the patient's face as they go through high levels of spontaneous pain. The paper will explore algorithms such as Extreme learning machines and Support vector machines for quick and effective pain assessment based on performance, precision and reliability. ELM average accuracy rates were 95.96% and 86.44%, while the SVM scored 79.78% and 60.30% in both frame and sequence based tests, respectively.