Seizure Detection Mechanism in Children

In this paper, a wearable device is presented which is designed to monitor the vital signs of young children and detect the possibility of a seizure before it occurs. Epilepsy and seizure disorders affect millions of people worldwide, and young children are particularly vulnerable to these conditions. However, it can be challenging for young children to communicate or recognize the symptoms of a seizure, which can lead to under-diagnosis and delayed treatment. The Seizure Tracker device includes an array of sensors - a biometric sensor hub for heart rate, blood pressure, and oxygen saturation, a temperature sensor, and an IMU sensor for jerk detection. The device uses a machine learning algorithm based on support vector machines to analyze the sequential data of a person both under normal and seizure conditions and classify the possibility of a seizure with an accuracy of 84%. The macro F1 score of the model on the test set was 0.8191. The misclassification rate of the model was 0.16.