Chapter Ten - Video and optoelectronics in movement disorders

Abstract

Movement disorders are routinely analyzed visually by neurologists as part of the assessment methods used. These visual analyses are subjective, depend on the expertise and condition of the expert, and can lack repeatability. During the last years, several efforts to objectively quantify motor test-assessments have been developed, including inertial measurement units (IMUs) which are becoming the reference standard. Video acquisition and analysis is a low-cost, non-invasive approach that allows remote evaluation of motion. To obtain quantitative data from the videos, cameras that provide depth information of the scene have used. Recently, artificial intelligence algorithms have been developed for accurate measurement from standard video sequences, offering great promise for automatic motion analysis in the evaluation of patients with movement disorders, including hyperkinetic disorders. There are several types of artificial intelligence techniques that are used for analyzing video data in the evaluation of patients with movement disorders. One common technique is computer vision, which is used to extract features from the video frames. Deep learning algorithms use neural networks to learn features directly from the video data. Convolutional neural networks (CNNs) are commonly used for video analysis, as they can automatically learn features from the raw video frames. There are also algorithms that combine computer vision and pose estimation techniques with machine learning or deep learning models to provide a comprehensive analysis of movement disorders. In this chapter we describe video acquisition devices, explain the main video analysis methods and review the application of these techniques, including computer vision algorithms, pose estimation algorithms, and machine learning or deep learning models, to the evaluation of patients with movement disorders.