Automatic detection of pain intensity

Abstract

Previous efforts suggest that occurrence of pain can be detected from the face. Can intensity of pain be detected as well? The Prkachin and Solomon Pain Intensity (PSPI) metric was used to classify four levels of pain intensity (none, trace, weak, and strong) in 25 participants with previous shoulder injury (McMaster-UNBC Pain Archive). Participants were recorded while they completed a series of movements of their affected and unaffected shoulders. From the video recordings, canonical normalized appearance of the face (CAPP) was extracted using active appearance modeling. To control for variation in face size, all CAPP were rescaled to 96x96 pixels. CAPP then was passed through a set of Log-Normal filters consisting of 7 frequencies and 15 orientations to extract 9216 features. To detect pain level, 4 support vector machines (SVMs) were separately trained for the automatic measurement of pain intensity on a frame-by-frame level using both 5-folds cross-validation and leave-one-subjectout cross-validation. F1 for each level of pain intensity ranged from 91% to 96% and from 40% to 67% for 5-folds and leaveone-subject-out cross-validation, respectively. Intra-class correlation, which assesses the consistency of continuous pain intensity between manual and automatic PSPI was 0.85 and 0.55 for 5-folds and leave-one-subject-out cross-validation, respectively, which suggests moderate to high consistency. These findings show that pain intensity can be reliably measured from facial expression in participants with orthopedic injury.

Go-with-the-Flow: Tracking, Analysis and Sonification of Movement and Breathing to Build Confidence in Activity Despite Chronic Pain

“…At the same time, our results showed that external representation by sound can enhance patients' understanding of their own movements and breathing patterns (if embodied), and help with providing personalized explanations and advice, facilitating pacing and goal-setting. The supervisory support by the device could be further enhanced by using functionalities to automatically detect increased pain or more subtle cues of fear of pain from body cues (Aung et al, in press;Olugbade, Aung, Marquardt, De C. Williams, & Bianchi-Berthouze, 2014 and from facial expressions (Hammal & Cohn, 2012;Kaltwang, Rudovic, & Pantic, 2012;Meng & Bianchi-Berthouze, 2014;Romera-Paredes et al, 2013) and suggest or guide recalibration. Indeed, in a recent follow-up study we carried out on sensing wearable devices, people with CP confirmed the role of technology as a support to learning supervision skills and even to share such the supervisory role in real-life situation where the task at hand requires much attention (Felipe, Singh, Bradley, Williams, & Bianchi-Berthouze, 2015).…”

Spatiotemporal Facial Super-Pixels for Pain Detection

“…The results of the proposed system are compared against two state-of-the-art pain detection systems of [7] and [8]. Following these two works, the obtained pain index of section 3.5, is classified into three different categories of no pain (if the pain index is zero), weak pain (if the pain index is either 1 or 2), severe pain (if the pain index is larger than or equal three).…”

“…In [10], three feature sets of Facial landmarks (PTS), Discrete Cosine Transform coefficients (DCT) and Local Binary Patterns (LBP) are extracted from the facial images, and are then fed to a Relevance Vector Regression (RVR) to estimate the pain intensity. In [7] canonical appearance of the face using AAM are passed through a set of lognormal filters to get a discriminative energy-based representation of the facial expression which is then used to estimate the pain level. Inspired by this work, in [8] another energy-based system has been developed for pain estimation which uses spatiotemporal filters.…”

Machine Learning Methods for Social Signal Processing

“…Recent release of the pain-intensity coded data (Lucey et al 2011) has motivated research into automated estimation of the pain intensity levels (Hammal & Cohn 2012, Kaltwang et al 2012, Rudovic et al 2013a). For example, (Hammal & Cohn 2012) performed estimation of 4 pain intensity levels, with the levels greater than 3 on the 16-level scale being grouped together.…”

“…For example, (Hammal & Cohn 2012) performed estimation of 4 pain intensity levels, with the levels greater than 3 on the 16-level scale being grouped together. The authors applied Log-Normal filters to the normalized facial appearance to extract the image features, which were then used to train binary SVM classifiers for each pain intensity level, on a frame-by-frame basis.…”