Towards automatic detection of movement during sleep in pediatric patients with epilepsy by means of video recordings and the optical flow algorithm

The detection and analysis of epileptic seizures is typically done by video-electroencephalogram monitoring. Although it is considered as the Golden Standard, it has disadvantages: the electroencephalogram electrodes are uncomfortable to wear for a longer period of time and hospitalization is often required. The aim of our work is to investigate whether the optical flow algorithm applied to video recordings can be used to detect movement during sleep in pediatric patients with epilepsy. The optical flow algorithm allocates intensities to pixels proportional to their involvement in movement of an object in a scene. The average of the 0.06% highest of these intensities was plotted as a function of time R(t). As dataset we used simulated video recordings (640×480 30fps) consisting of normal sleep movement and seizure-like movement. We investigated R(t) as a function of the acquisition parameters (such as spatial resolution, frame rate and compression) and environment parameters (such as illumination and camera viewpoint). We found that we still could make a distinction between movement and non-movement, when we spatially downsampled to 320×240. Furthermore we found that we could reduce the frame rate to 10 frames per second). The reduction in spatial and temporal resolution eases the computational load of the algorithm which puts future real-time applications in reach. MPEG-compression and reduced illumination increased R(t) for non-movement frames. Camera positions where in-plain movement is larger than out-of-plain movement yield a larger R(t) than vice versa. From our findings above we extracted the optimal acquisition parameters which guaranty clear distinction between movement (high R(t)) and non-movement (low R(t)). © 2009 Springer Berlin Heidelberg.status: publishe