Estimation of pain threshold from eeg signals of patients in physical therapy using long-short-term memory deep learning model

Pain is a natural stimulation to protect the whole body. Excessive reaction to this warning causes damage to the tissue. When a patient with shoulder pain goes to physical therapy, joint range of motion (EHA) measurement is routinely performed during the first examination. In this measurement, between 0 degrees and 90 degrees, the patient usually does not feel pain, but between 90 degrees and 120 degrees, the patient feels pain. If the patient deliberately or unknowingly pushes himself too hard during this measurement, it can cause tissue damage. In this study, it was aimed to determine the patient's pain status from electroencephalography (EEG) signals in the measurement of joint range of motion, which was routinely performed during the first examination of patients who came to the physical therapy unit because they had shoulder pain. During the first examination, the effect of pain due to joint range of motion on EEG signals was examined from volunteer patients who came for physical therapy. Determination of pain from EEG signals has been shown to help prevent tissue damage in the patient and to determine the angle at which the patient's pain occurs by physiotherapists. Our Data Simav Assoc. Dr. 43 volunteer patients admitted to the physical therapy and Rehabilitation Department of Ismail KARAKUYU State Hospital due to shoulder pain were taken with a 14-channel wireless Emotive brand EEG device registered with the simav Faculty of Technology Electrical and Electronic Engineering Electronic Laboratory fixtures. EEG signals were taken from the patient when there was pain in the first 10 seconds in record 1 without pain in the next 10 seconds. In record 2, the first 10 seconds were pain without pain, while the second 10 seconds were pain, and the last 10 seconds were a lot of pain, EEG signals were taken from the patient. In order to determine the pain threshold, the power spectrum of EEG signals was obtained by Welch, Periodogram and Multitaper methods. Delta, Theta, Alpha and beta power density values are included in the spectrum values obtained from 1 to 49Hz. The LSTM (Long-Short Term Memory) algorithm, one of the artificial intelligence methods, was used to classify the values of this spectrum. The LSTM deep learning model has been used extensively in signal processing studies in recent years. Because LSTM has short-and long-term memory units, the fact that it showed more successful results in signal classification studies than other methods motivated us to use the method in the study. With the Welch method, the success of record 1 is 95%. The success of record 2 is 75%. The success of registration 1 with the Periodogram method is 76%. The success of record 2 is 65%. The success of enrollment 1 with the multitasking method is 70%. Record 2's success is 61%. As a result, the LSTM classifier achieved the highest success when the Welch method property extraction was performed. There are no studies in the literature on physical therapy and pain. In this study, patients who come to the physical therapy unit for shoulder pain will be able to prevent pain-related tissue damage during EHA measurements and physiotherapists will be able to provide information about the angle at which the pain occurs. Keywords : Pain, EEG signal processing, Physiotherapy therapy, LSTM Deep Learning Model, Multitaper, Periodogram, Welch.