CHAPTER 5 Graph convolutional networks for pain detection via telehealth Suzan Anwar 1,2 , Mariofanna Milanova 3 , Shereen Adbulla 4 and Saja Ataallah Muhammed 5 1 Computer Science Department, Philander Smith College, Little Rock, AR, United States 2 Computer Science Department, Salahaddin University, Erbil, Iraq 3 Computer Science Department, University of Arkansas, Little Rock, AR, United States 4 Computer Science Department, Ploy Tech University, Erbil, Iraq 5 Computer Science and Information Technology Department, Salahaddin University, Erbil, Iraq 5.1 Introduction In 2020, a severe crisis was created because of the highly contiguous COVID-19 and its fast spread [1]. The public’s concerns are increased due to the absence of effective treatment for this pandemic [2]. Around the world, the healthcare deliv- ery systems were significantly impacted by COVID-19. In early 2020, the Secretary of Health and Human Services declared a public health emergency and allowed beneficiaries to receive telehealth services in their home [3]. The need to detect pain during telehealth is increased to help the physician improve outcomes for outpatients and more control back to them with more appropriate medications [4]. Deep learning is a crucial machine learning tool to solve complex problems and achieve supervised learning. Many ways are using different measures to describe pain levels such as speech, physiological, facial expression, and body gesture. The long short-term memory (LSTM) and neural networks (NNs) to detect pain from speech analysis are used in Ref. [5] and consider the first research to use speech for pain detection. The vocal features are expected using Chinese corpus, and the unsupervised learning NN is employed in the first layer. To obtain the sentence-level acoustic representation as an output for each patient, the vocal features are fine-tuned using a triage dataset and NNs. Finally, a support vector machine (SVM) is used in the classes to detect pain levels. The experimental findings showed that the proposed method achieved 72.3% weight average recall in two classes. A physiological signal is one of the viral approaches for pain assessments. It depends on the physiological response of the patient’s body. Brain dynamic or vital signals such as muscle activity, heart rate, and blood pressure are examples of physiological signals. A model using NN techniques is used in Ref. [6]. The author in Ref. [7] proposed a multimodel data system that used signals from Artificial Intelligence in Healthcare and COVID-19. DOI: https://doi.org/10.1016/B978-0-323-90531-2.00006-0 © 2023 Elsevier Inc. All rights reserved. 93