Epilepsy monitoring is vital in the accurate diagnosis of uncontrolled seizures and in preparation for epilepsy surgery. The epilepsy monitoring unit (EMU) is where continuous video EEGs are performed. The video records the patient’s physical activity, including seizures and what happens in the moments before
and after seizure. Simultaneously with the video recording, the EEG records brain activity. Both video and EEG information allow physicians to pinpoint the type of seizure that is occurring and precisely locate the source. On the one hand it is necessary to keep the patient as close as possible in the focus of the video to see also, for example, small muscle twitches; on the other hand the patient should also have the freedom to move around in the patient room
For this purpose today’s video EEG systems use pan-tilt-zoom cameras that are manually controlled by medical staff, who are usually in charge of continuously controlling the camera to keep the patient in focus while at the same time observing EEG activity for diagnosis. However, manual camera control and video monitoring over a longer period of time is exhausting and lowers operator’s attention. To overcome this problem, an innovative automated patient tracking system has been designed and developed together with NIHON KOHDEN, which consists of a two-camera setup connected to the Fraunhofer IOSB AutoTrack® patient tracking software.
The AutoTrack® software is responsible for real-time processing of video streams from all cameras available for monitoring as well as for automatic camera control.
The typical camera setup consists of two cameras: a static overview camera and a pan-tilt-zoom camera. The static wide-angle camera provides an overview of the entire monitoring room, while the pan-tilt-zoom camera is used for active high-resolution patient observation. Both camera streams are processed independently by dedicated video processing modules, which are able to detect and localize the observed patient.
Detection is performed based on color appearance features and by the use of visual tags (IOSB MXT badges), which are fi xed to the monitored patient’s clothes. MXT badges (Fig. 1) are defined visual patterns which can be identifi ed and localized in image data reliably and very quickly, even under adverse lighting conditions (e.g. low light, low contrast, or low resolution).
One unique feature of the AutoTrack® system is its ability to track the patient in the monitoring room in 3D space, instead of tracking in video (pixel coordinates) only. This approach allows a higher scalability of both the camera network and camera mounting positions. Since the images from each camera are processed by independent video processing modules, and position information is exchanged between modules based on a common 3D coordinate system, camera control can be performed by any available camera in the network.
Furthermore, MXT badges can be identifi ed by an integrated code (ID number). These IDs are used to distinguish between patients in the same monitoring room and to avoid mix-ups during automated tracking.
In 2013 the AutoTrack® system prototype has been evaluated by medical technicians at Heidelberg University Hospital (Universitätsklinik Heidelberg). After proof of concept and user-oriented software optimization the AutoTrack® software has been presented to the public and end-users at the MEDICA Düsseldorf Trade Fair, and at the AES (Annual Meeting of the American Epilepsy Society) in Washington D.C. with big success and very positive market feedback.