An "optical flow" method based on pressure sensors data for quantification of Parkinson's disease characteristics. (March 2023)

• Record Type: Journal Article
• Authors: Dong, Chenhui; Chen, Ying; Huan, Zhan; Li, Zhixin; Gao, Ge; Zhou, Bangwen
• Is Part Of: Biomedical signal processing and control. Volume 81(2023)
• Publication Date: 2023-03
• Subjects: Parkinson's disease -- Pressure sensor -- Optical flow method -- Multi-sensor -- Gait analysis; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.104377
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: An "optical flow" method applied to the VGRF time series data is proposed. A gait symmetry evaluation index is proposed to extract more symmetry features in dynamics. The results show that the proposed model has better performance than several other state-of-the-art methods. Abstract: Most patients with Parkinson's disease (PD) have different degrees of movement disorders, and effective gait analysis is beneficial to find the abnormal gait of patients to achieve the diagnosis of patients with Parkinson's disease. In this paper, an "optical flow" method based on Vertical Ground Reaction Force (VGRF) time series data is proposed, the algorithm takes the force point as the detection target, regards the transfer of the force point on the sole as the optical flow, and combines the optical flow of the multi-level force points to form optic flow field. To quantify the optical flow direction, the direction histogram is used to extract the direction information, and the symmetry information is further extracted according to the optical flow difference between the left and right foot, which not only realizes the fusion of multi-sensors but also extracts highly interpretable motion information. Finally, the model is trained by combining optical flow features and spatial-temporal features. The results show that the proposed model has better performance in gait detection of Parkinson's disease patients than several other state-of-the-art methods previously studied. Among them, the accuracy of Parkinson's disease diagnosis reached 93.3%, and the accuracy of severity assessment of Parkinson's disease reached 91.4%.