Epileptic Seizure Detection

Motivation

Epilepsy is a neurological disorder in which involuntary seizures occur. Around 1% of people worldwide suffer from various forms of this disease. In 30-40% of cases, the epileptic seizures that people suffer cannot be stopped by any form of medication. A seizure, if it happens at the wrong time and in the wrong place, can have devastating consequences: A harmless trip on a bicycle to the nearest supermarket can thus become a danger for the sufferer and those around them. A study by Mahler et al. analyzed accidents involving people with epilepsy. Another study by Schulze-Bonhage et al. surveyed 141 people with epilepsy from Germany and Portugal and found that over 90% would like a seizure detection device.

Research Objective

The diagnosis of epilepsy usually requires a thorough medical examination, including electroencephalographic (EEG) monitoring. Recently, however, it has been shown that machine learning techniques can be an effective method for detecting epilepsy based on electrocardiographic (ECG) data. The Epileptic Seizure Detection project addresses the question of how machine learning techniques can be used to detect epilepsy based on ECG data and presents a solution approach.

The aim is to develop a system suitable for everyday use that measures the patient's heartbeat and analyzes it for signs of an epileptic seizure. If an epileptic seizure is detected by the system, an alarm signal is to be emitted and the seizure recorded in order to inform the person affected and relevant contact persons and to enable subsequent analysis by a specialist.

This requires a smart device (such as a smartwatch) that measures the patient's heartbeat and analyzes it for signs of an epileptic seizure. Seizure detection is based on heart rate variability, which determines the variance in the time between two heartbeats. To measure the time, the duration of an RR interval is determined, which measures the time between two R peaks in ECG signals. An R-peak is the highest point of a so-called QRS complex wave, which represents a part of the ECG signal and reflects the electrical activity of the heart during a heartbeat.

Practical Relevance

Seizure detection by analyzing EEG data is considered the gold standard in practice. However, this method has a number of disadvantages. Thus, a seizure detection system based on EEG data outside a medical facility proves to be very impractical. A system that makes predictions based on heart rate could be implemented via an app solution on a smartphone and an ECG-enabled smartwatch, for example, and would therefore significantly improve the quality of life of those affected. At the same time, everyday life is made safer and those affected are less affected.

The study by Schulze-Bonhage et al. also shows that such a system, which works independently of EEG data, is needed by those affected: around 80% of those surveyed stated that they did not want to wear EEG electrodes on their scalp in the long term for seizure detection.