Detection freezing of gait (FOG) in Parkinson's patients using wearable sensors and deep learning

Freezing of the gait (FOG) is a complication of Parkinson's disease (PD) that leads to the patient's inability to perform motor activities. The occurrence of FOG reduces patients' independence in daily activities and generally reduces their quality of life. The use of computational methods can provide non-pharmacological support and complementary information about the disease to neurologists by carefully examining patients’ FOG status and increasing the likelihood of a more effective treatment. This paper presents a method for FOG detection based on deep learning and signal processing techniques. The data used for this paper is the Daphnet data collection, which is collected by the wearable sensors on the patient's body. The proposed methoddetects FOG by providing a deep neural network architecture based on two-way short-term memory networks (BDL-FOG). Experimental results show that the proposed method, due to its better compatibility with time-series data, has been able to improve the FOG detection process to achieve higher accuracy than the best available methods.