EEG Data reduction with Epileptic Seizure Detection based machine learning in IoMT Networks

There is a greater demand to enhance intelligent healthcare systems in order to keep track of the medical condition of patients. This demand leads to increase the interest in remote healthcare systems that make use of biosensors. These biosensors generate a significant amount of vital sensed data, which is then sent to the edge of the Internet of Medical Things (IoMT) for processing before being sent on to the data centers of the cloud. In this paper, we propose an EEG data reduction and epileptic seizures detection in IoMT networks. The suggested method operates on 2 levels within the network: edge and fog levels. At the edge gateway, The EEG data is transformed into time-frequency representations using Short Time Fourier Transformation (STFT). Then, a set of statistical texture feature based on the gray level co-occurrence matrix (GLCM) is extracted. These EEG based features are then compressed before being sent to the fog gateway using a binary-encoded representative method. In Fog Gateway, the received encoded features are decoded and passed into a classifier to determine the patient's situation. A set of classifiers are used to evaluate the proposed method using the Bonn University dataset. The proposed technique demonstrates the outperformance of the proposed method by achieving an accuracy of (100%) in multiple cases and an acceptable accuracy for other cases while considering fewer features compared to existing work.