A Systematic Approach to the Design and Development of Wearable Sensor for Detection of Arthritis using Gait Pattern

Abstract

"Arthritis is recognized as one of the most widespread articular diseases affecting a large population in the world and is the main cause of articular disability. Arthritis is a chronic disorder that can create problems in any bone .ioint. It arises frequently in the neck, thumb base, finger joints. hip. lower back, knee, and at the base of the big toes. In every bone joint. the carlilage covers the ends of each bone. Arthritis also results in lower limb traurna. The normal gait of a human being is affected when the person is sub.iected to any injury or pain in the lower limbs. To reveal the underlying conditions of the subject and to aid in the diagnosis of the patient, the gait patterns of the sub.iects are captured and subsequently analysed. Gait analysis is one of the important tools used to dedr-rce subtle and evident gait imegulalities exhibited by the subject. The knee-pressure and foot-pressure distribution signals are collected from the individuals to identify the ones affected ivitl-r knee pain. The irregularities in gait often manif-est as knee pain or physical discomfort experienced by the patient. Existing vision and floor sensor-based systems have the limitations of operational complexity and high cost that make them uncomfortable for individual use. Wearable sensors can be used to address this ditficulty. A Thermoplastic Polyurethane (TPU) - Carbon Nano Fiber (CNF) based nanocomposite piezoresistive pressure sensing film is developed by solution casting method. It is characterized by surface resistance, bulk resistance, impedance, and fiequency dependence. The aim was to utilise the developed sensor to capture the pressure distribution around the knees during gait. Prepure sensing mechanisms are arranged in an appropriate knee cap utilising a TPU-CNF based nanocomposite piezoresistive (20% w/w) sensing film with a bulk resistivity of 192f!-cm in this research. The reliability of this arrangement for detecting the change in resistance owing to pressure fluctuation during knee movement is also examined. The voltage changes in various compressive stacking lab experiments indicated that this developed piezoresistive material has a wide range of applications. As a result, it can be efficiently used for signal acquisition applications in arthritis patient knee movement monitoring. TPU-CNF20 has been proven to be suitable fbr monitoring the gait of arthritis patients. It has a lower bulk resistivity (192 + 6.5 Q-cm) at the same time it has a higher linear fit of R2:0.99306. The developed flexible piezoresistive sensing films are sirnply sandrviched between two copper foil electrodes fbr improved conductivity in the experiments. In real-tirne, signals fbr fbur types of knee joint movements are collected: Knee extension, bend, half squat, and knee raise. Using statistical time-domain f-eatures, the main characteristics of these lour movements are identified. The data was collected fiorn the TPU-CNF sensor, however, the subjects exhibited restricted movements. These limitations led to research interest in the design and development of insole embedded with 102 piezoelectric sensors to capture foot-pressure distribution images and to detect lower limb disorders. From the morpholo-qical analysis and the resistance response curve, a foot pressure sensor was developed at a lamination temperature of 120oC. The developed piezoresistive-based sensorized insole with excellent sensing applications was successfully used for acquiring the hi gh-resolution pressure distribution prolile. The quality of these heat images is enhanced by a Hybrid Filter (RMSE:2.748 and PSNR:39.35) and a feature extraction technique is utilised on the enhanced foot pressure images for classification. The k nearest neighbours (KIIIN) learner model yields an accuracy of 99.4% in the detection of knee pain. Gray level co-occurrence matrix (GLCM) based and pre-trained DNN based approaches are used for feature extraction. GLCM based feature extraction, the GLCM-Full & GLCM-HR approaches KNN leamer model yieldedanaccuracyofgg.ll%.GLCM-VRapproachKNNlearnermodel yielded an accuracY of 99.44%' In the pre-trained deep neural network (DNN) based feature extraction, InceptionV3, VGG16, and VGG19 DNNs are involved. Inception V3 and vGG19 based approaches, the KNN learner model yietded an accuracy of gg,sg%.In the VGGl6 based approach, the KNN learner model yielded an accuracy of 99.83%. TheresultsshowthatthedevelopedTPU-CNFbasedpiezoresistive sensors and velostat based sensorized insole are the most suitable for analysing the gait Pattern."