A Hybrid Approach for Detection of Epileptic Seizures in Real-time using LSTM and IoT

In recent years epileptic seizures are considered as a major neurological condition among 50 million people worldwide due to their abnormal brain activity patterns. The unpredictable patterns of seizures interrupt daily operations thus requiring prompt detection to improve care methods. The research develops a deep learning detection paradigm which merges IoT hardware components and LSTM neural networks to perform real-time EEG signal analysis. The system uses two steps of preprocessing including band-pass filtering alongside notch filtering before extracting frequency bands from the signal for subsequent temporal segmentation. LSTM receives previously processed EEG segments to establish training that classifies EEG states into interictal and preictal and ictal states. Experimental trials verify that the proposed LSTM neural network delivers higher performance metrics than traditional and deep learning methods that include Support Vector Machine (SVM), Random Forest, Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU). The suggested model surpasses CNN and GRU approaches with $96 \%$ accuracy and $94 \%$ recall and $96 \%$ F1 score. The framework operates in real-time with wearable EEG headbands to detect seizures automatically and delivers immediate alert notifications through mobile applications to caregivers. The combination of LSTM network temporal learning strengths with IoT device continuous data logging boosts seizure prediction precision while improving overall reliability. The proposed method helps develop better clinical treatment methods and provides direction for tailored healthcare delivery to epilepsy patients.