Abstracts

The analysis of seizure semiology is well-established in the evaluation of patients considered for epilepsy surgery. The clinical significance of several lateralizing ictal phenomena, such as head deviation, is still controversial. Visual inspection of ictal movements is usually performed by review of videocassettes in a qualitative way, which is prone to bias. Our objective was to evaluate the reliability of 2D quantitative video analysis using standard video equipment.
Videos were acquired using the routine setting of an epilepsy monitoring unit. They were recorded in analog S-VHS videocassettes using full PAL resolution. Selected video sequences were digitized to a PC computer with 720x525 pixels resolution, compressed into MPEG4 format leading to a pixelsize of 5.4 x 3.8 mm. A reference system of reflective markers was designed to provide multiple reference segments distributed in the video image. Another set of reflective markers is placed in several patient[apos]s anatomical positions. Based on these markers and the geometry model of the scene, a video analysis software is used to perform several measurements of distances, linear velocity, angles, angular velocity and movement frequency during epileptic seizures captured on digital video.
We dedicated particular attention to the upper part of the body where most of the [ldquo]movements of interest[rdquo] are located. The presented approach allows quantitative measurements like angle, angular velocity and head extension from standard 2D videos with an error below 8%. The first examples analysed allowed the quantified extraction of typical movement patterns for different types of epileptic seizures. Some comparative studies could also be performed, such as the degree of limbs jerking synchrony during a seizure.
We concluded that digital processing of standard 2D video material can extract quantified information from epileptic seizure semiology. This method introduces very little changes to the usual setup used in most of the video-EEG epilepsy monitoring units around the world. We could determine that, once chosen an appropriate reference, the error of the displacement measurements obtained from the video analysis of our setup could be kept under 8%. We consider this error sufficiently low to allow precise movement analysis. The few use of this method in real epileptic seizures we could perform until now, shows its usefulness to study seizure semiology, namely movement components. Based on this technical contribution, a study to investigate its clinical utility is now under development.
[Supported by: Work partly supported by FCT (Portuguese R[amp]D Agency) under POSI program, grant BSAB303 and POSI/33802/CPS/2000]