Automatic Pain Severity Level Classification Using The Convolutional Neural Network

Abstract

Postoperative pain management of patients has become a major medical and nursing challenge. The management of postoperative pain continues to remain ambiguous and disappointing. Nowadays, hospitals have taken initiatives to measure acute pain using self-report measures like the Visual Analogue Scale and Numeric Pain Intensity Scale. But, these methods are inaccurate as it depends on patients input. Therefore, there is a requirement for an objective, quantitative method to monitor acute pain continuously. Since human facial expressions are the best indicators of pain, researchers are more focused on utilizing facial expressions for the automatic detection of the severity of pain. In this work, a shallow Convolutional Neural Network (CNN) architecture is developed to perform classification of acute pain levels by achieving good classification accuracy with less computational time. The proposed CNN architecture has been evaluated on Part-A BioVid Heat Pain database (only facial expression data are used for the study). This architecture utilizes only two Convolutional layers and achieves 94.15% of classification accuracy. The performance signifies that the proposed CNN architecture outperforms