Detection of Freezing of Gait in Parkinson's Disease by Squeeze-and-Excitation Convolutional Neural Network with Wearable Sensors

Abstract

It is one of the most severe motor indications of Parkinson's disease that one's stride becomes freezing of gait (FOG). Patients' quality of life is negatively impacted by FOG, which may lead to falls. Typically, questionnaires have been used to diagnose FOG; however, this method is subjective and may not correctly represent the severity of this disorder. It is possible to monitor symptoms using sensor-based devices, which can provide reliable and objective data. In this paper, the SE-DeepConvNet, a compact deep convolutional neural network including squeeze-and-excite components for fog detection, was proposed. In conducted to evaluate SE-DeepConvNet, we employed Daphnet, a publicly accessible benchmark FOG dataset. In terms of effectiveness, the SE-DeepConvNet excels most traditional deep learning models, receiving a score of 95.66% on the accuracy evaluation.