Authors :
First Author: Anna Mira Loesch-Biffar, MD – University of Munich
Presenting Author: Jan Rémi, – Department of Neurology
Tamás Karácsony, Faculty of Engineering (FEUP) and 3Center for Biomedical Engineering Research – University of Porto, Porto, Portugal and Institute for Systems’ Engineering and Computers, Technology and Science (INESC TEC), Porto, Portugal; Leah Sattlegger, Epilepsy Center, Department of Neurology – University of Munich, Munich, Germany; Christian Vollmar, Epilepsy Center, Department of Neurology – University of Munich, Munich, Germany; Jan Rémi, Epilepsy Center, Department of Neurology – University of Munich, Munich, Germany; João Paulo Silva Cunha, Faculty of Engineering (FEUP) and 3Center for Biomedical Engineering Research – University of Porto, Porto, Portugal and Institute for Systems’ Engineering and Computers, Technology and Science (INESC TEC), Porto, Portugal; Soheyl Noachtar, Epilepsy Center, Department of Neurology – University of Munich, Munich, Germany
Rationale:
This study aims to test the capability of the novel NeuroKinect 3D-method and to introduce new movement visualization techniques to support diagnosis. Furthermore, to utilize the approach to quantitatively differentiate the movements during epileptic seizures between hyperkinetic and automotor movements in a clinical routine Epilepsy-Monitoring-Unit (EMU).
Methods:
The utilized dataset is extracted from the NeuroKinect dataset, which is a unique RGB-D-IR dataset of epileptic seizures acquired in a collaboration by INESC TEC and the EMU of the University of Munich Epilepsy Center. The dataset is recorded with Kinect v2 and consists of RGB, Infrared (IR) and depth streams. It is recorded and stored in 512x424 resolution for IR and Depth, and 640x480 for RGB data with a full size of 2.7 TB per day and labeled with epileptic seizure types and semiology annotations. We performed quantitative 3D-movement analysis of upper limb movements of 10 hyperkinetic (n=7 patients) and 10 automotor (n=10 patients) seizures. The patients in which we recorded automotor seizures were mainly temporal lobe epilepsies (7 of 10 patients, 70%), one patient had a focal epilepsy originating also from temporal and frontal lobe, the other two patients had a bilateral focal epilepsy also evolving from temporal and frontal lobe. The seven patients with hyperkinetic seizures were either frontal lobe (three patients, 43%) or bilateral focal epilepsies (two bilateral frontal lobe, two bilateral frontal and temporal lobe epilepsies).
Results:
Velocity, acceleration, jerk, covered distance, displacement, and movement extent of different region of interests (ROI: head, right hand, left hand and trunk) were captured. Hyperkinetic seizures showed a significantly higher mean velocity, mean acceleration and mean jerk (p < 0.05) in every ROI (Table 1). The movement extent of the trunk was also significantly higher in hyperkinetic than in automotor seizures (0.27 +/- 0.67 versus 0.02 +/- 0.03, p=0.05). The duration of movements (80 seconds in automotor +/- 38 versus 26 seconds in hyperkinetic +/- 14; p=0.001) as well as the total seizure duration was longer in automotor seizures (152 seconds +/- 135 versus 40 +/- 18; p=0.02). We developed new visualization techniques that make it possible to reconstruct the tracked movement via 3D viewer in space. Our study shows that quantitative 3D movement analysis applied in a clinical routine set up of a standard EMU allows the objective differentiation of hyperkinetic and automotor seizures, an information which is important for the evaluation of candidates for epilepsy surgery.