Abstracts

Rationale: Epileptic seizures often go unnoticed, even in expert care settings. Consequences range from underestimation of seizure frequency to (lethal) damage to the patient. To improve patient care and cure a wearable cardiac monitor for real-time detection of major epileptic seizures is developed and validated.Methods: This observational study is a non-randomized, open, single-site, clinical test in 13 subjects (30 to 50 seizures) previously diagnosed with frequent (>1/week) major epileptic seizures (tonic-clonic, generalized tonic or clonic, hypermotor) with heart rate changes. Patients are admitted to the short-stay facility at Kempenhaeghe for overnight stays, for a duration of 1 to 4 weeks. A miniaturized wearable sensor (fig. 1) is used to record seizures based on heart rate changes [Mass F et al. Wireless Health 2010, ACM, New York, USA, 111-117]. The sensor analyzes the ECG signal in real-time and detects epileptic seizures based on characteristic heart rate patterns [van Elmpt WJ et al. Seizure 2006;15(6):366-75]. Events are wirelessly transmitted to a base-station within 30 ft range and subsequently to the video monitoring system. The sensor is worn on the upper arm, and from there two electrodes are connected to the chest with lead-wires. Data from the first three subjects of the study is analyzed manually to define a generic set of parameters for the detection of seizures based on heart rate patterns. The clinical study for the 10 remaining patients is divided in two phases. First the generic set of parameters is used to detect seizures in real-time and secondly the parameter set is tuned to each patient. Using this set the recorded data is processed offline by the same algorithm to re-evaluate the performance. Objectives are the sensitivity (SEN), positive predictive value (PPV) and technical feasibility. Seizures detected by the miniaturized wearable sensor are verified by visual analysis of recorded video, and comparison to previously analyzed EEG-video data.Results: Results are reported for the first 3 patients. A total of 19 major seizures (15 generalized tonic, 3 hypermotoric and 1 tonic-clonic) occurred during 45 nights. All seizures were correctly detected in Patient01 (4) and Patient02 (3), leading to a SEN and PPV of 100%. Only 3 seizures out of 12 were detected for Patient03, whereas a number of false detections (24) and missed seizures (9) are observed. Table I reflects the combined results. For calculation of the PPV all epileptic seizures that are not defined as major are regarded false positives (FP* in table I).Conclusions: Heart rate based detection of major seizures by the proposed wearable sensor is feasible, but results show a large degree of intersubject variability. Tuning the parameter settings to each patient is expected to lead to significantly better results than generic parameters in terms of SEN and PPV. Results of the remaining 10 subjects will be presented upon completion of the data analysis. For increased performance adding accelerometer and/or EMG analysis is studied. No inconvenience caused by the system has been reported.