An efficient approach of epilepsy seizure alert system using IoT and machine learning

• Published in: Journal of reliable intelligent environments 2024, Vol.10 (4), p.449-461
• Main Authors: Basavaiah, Jagadeesh, Anthony, Audre Arlene, Mahadevaswamy, S, Naveen Kumar, H. N
• Format: Article
• Language: English
• Subjects: Accuracy; Algorithms; Artificial Intelligence; Computer Science; Convulsions & seizures; Epilepsy; Health care; Health Informatics; Health services; Machine learning; Machinery condition monitoring; Neurological diseases; Original Article; Performance and Reliability; Portable equipment; Remote monitoring; Seizures; Sensors; Simulation and Modeling; Software Engineering/Programming and Operating Systems; Telemedicine; User Interfaces and Human Computer Interaction; Vibration analysis; Vibration monitoring
• ISSN: 2199-4668; 2199-4676
• DOI: 10.1007/s40860-024-00228-w

Epilepsy is a neurological disorder that affects millions of people worldwide, and it is characterized by recurrent seizures that can vary in frequency and severity. The management of epilepsy requires timely and accurate diagnosis, as well as effective monitoring of symptoms and treatment. Continuous monitoring of a patient’s condition can result in better outcomes and timely intervention. Remote monitoring can reduce hospital visits and ease the burden on healthcare systems. In this work, we have developed a system that uses various sensors and equipment, such as GPS and GSM module, ECG sensor, tilt, and vibration sensors to detect and predict epilepsy. The system is based on IoT technology and includes devices like the Arduino Nano and ESP32. The monitoring devices can be worn by patients all day as they are portable and easy to use, with the system constantly recording and analyzing their data which is then sent to a secure cloud server, where healthcare professionals can access it. The system can also generate alerts when the patient is experiencing a seizure, enabling timely intervention and treatment. An essential aspect of our system involves leveraging machine learning algorithms to meticulously examine the sensor-derived data for precise diagnostics. Within our system, we’ve integrated three distinct algorithms, each yielding impressive accuracy rates. Specifically, the CNN algorithm boasts an accuracy of 95.26%, Random Forest exhibits 92.93% accuracy, and Logistic Regression achieves a commendable accuracy rate of 93.68%.