Machine learning approaches for the prediction and detection of epilepsy

Abstract

The one of the most difficult factors associated with epilepsy is the unanticipated nature of seizures triggered by epilepsy. The techniques which can identify tremors before several minutes of its occurrence, may be helpful in reducing the harmful effects of seizure and even can reduce the sudden death rate. Thus, early detection of seizures can enhance the quality of life of an epileptic patient.EEG has been utilized for forecasting seizures by employing contemporary computer tools, artificial intelligence, and deep learning techniques. despite this, ambient noise can contaminate EEG readings, and artefacts like flickers of the eye and contractions of the muscles can cause "bumps" in the signal that result in electromagnetic interference that is difficult to see for more-duration recordings. These restrictions on computerized interictal peak and epileptic attack identification, a crucial tool for closely studying and analyzing the EEG data, are recommended. The deep learning models in this research aim to enhance epileptic seizure detection using a Field Programmable Gate Array (FPGA) implementation of the quick Fourier transform interfere with.For detection of seizures, the following steps have been taken: (1) time-frequency evaluation of EEG portions using STFT; (2) collection of spectrum bands and characteristics that are of interest; and (3) seizure identification by using a convolutional neural network, and bilateral transient-term memory with long-term retention. This brief overview also makes suggestions for how neurological doctors should actively work to achieve advancements in EEG-based ML seizure detection.

Keywords: Electroencephalography, artificial intelligence (AI), Epilepsy and machine learning (ML), seizure onset zone (SOZ).