Deep Learning-Based Pain Intensity Estimation from Facial Expressions

Pain estimation is of great importance for the medical doctor since it helps them in diagnosing patients and offering the adequate care and treatment. However, there is still no mechanism that accurately measures the pain intensity level. In addition, many patients are not able to express the extent of their pain verbally such as babies and people with Alzheimer disease. Indeed, pain intensity estimation will not only help in diagnosing a patient but also in measuring his response to a specific treatment or reeducation activity. For this end, building an automatic accurate pain intensity estimation system is a crucial need for medical professionals. Recently, several pain intensity estimation systems have been proposed. The most efficient of them have been based on facial expressions using deep learning models which have proven their high performance in extracting valuable facial features and modeling them. In this paper, a review of the best-performing deep learning-based pain intensity estimation systems from the face image is provided. In addition, the most commonly used datasets and performance evaluation criteria will be presented. Moreover, an experimental method analysis along with future improvement ideas will be discussed. Finally, the main findings and recommendations for further study will be spread out.