IMPROVE HUMAN SEIZURE DETECTION ACCURACY USING A HYBRID MODEL INVOLVING RESNET50 AND SUPPORT VECTOR MACHINES

Abstract

The term improve human seizure detection refers to the creation and application of cutting-edge, reliable, and effective techniques, technologies, or systems that improve the capacity to recognize and diagnose seizures in people with epilepsy or other neurological disorders. Traditional detection algorithms may produce false positives or false negatives as a result of this variability. To overcome these limitations, we propose a Hybrid ResNet50+Support Vector Machine (HResNet50 + SVM) that recognize and diagnoses seizures in people. Initially, we gather dataset from the EEG observation and Min-Max normalization is used to process data and extract the relevant features from pre-processed data using temporal and spectral. Hybrid Particle swarm optimization (PSO)-Whale optimization is used for feature selection. To assess the given approach performs at the level of accuracy (93.8%), error (27.44%), sensitivity (93.9%), specificity (93.1%), precision (95.9%), false rate (6.86%) and F1 score (96.8%). As a result, addressing problems in human seizure detection through sophisticated machine learning algorithms, reliable data collection techniques and cooperative efforts between medical professionals and technologists holds the promise of developing precise and usable seizure detection systems.