Tracking epileptic patients in digital videos for automated video-EEG monitoring

Video EEG monitoring is considered to be the most successful application for the diagnosis of the epileptic patients. The movements of the patient are recorded in the form of a digital video along with the EEG, over a significant duration of time. This enables the analysis of the behavior of the patient during the seizures, and is critical in determining the area of the brain that is responsible for the seizures. As the video recording is extensive in time, automatic tracking of the patient is a practical need. This would not only make the monitoring process error free, but would also lower the costs associated with the need for the human intervention. The aim of this thesis is to develop a system to automate the video EEG monitoring of the epileptic patients. Due to the prior information available about the physical environment of the patient, feature-based tracking method is preferred in this thesis over the motion-based techniques. The available features are identified and analyzed for developing the tracking algorithm. The skin color is used as one of the features and a new skin color detection filter, which is shown to perform reasonably well for detecting the human skin color, has been developed. The cap worn by the patient to support the electrodes on the head is developed into a second feature by drawing a pattern on the cap. A pattern recognition technique using the Hough transform for line detection is proposed to detect this feature. The two features are jointly used together to develop an algorithm for locating the patient in the room. The tracking performance of this proposed feature-based algorithm is tested extensively under varying conditions and is shown to provide reasonable performance so that it can be used for practical implementation in a tracking system.