Background: A seizure is a sudden and uncontrolled electrical activity in the brain that can cause a variety of symptoms, depending on the location and severity of the abnormal activity. It can be a symptom of an underlying neurological disorder or can occur without an apparent cause. Epilepsy is one of the most common causes of seizures. Overactive electrical discharges disrupt normal brain electrical activity and interfere with nerve cell communication. Aim and Objective: A comprehensive analysis of the literature revealed that several CAD system designs have shown to be useful to radiologists in routine medical practice as second-opinion aids for epileptic seizure detection in circumstances where a clear differentiation cannot be formed subjectively. CAD systems are made to help radiologists by automating the examination of medical data and images, improving the efficiency and accuracy of diagnosis. These systems examine patterns in medical imaging using machine learning approaches, which can be quite helpful in spotting small abnormalities that the human eye can miss. Moreover, the objective of this study was to design a smart healthcare system using a combination of DWT, Hjorth, and statistical parameters for seizure detection. Method: In this research article, the authors proposed the framework of the Internet of Healthcare Things (IoHT) for performing seizure detection. The authors used different pre-processing techniques and extracted different features like Hjorth, wavelets, and statistics, which were classified using different machine-learning techniques. This novel methodology combines a number of technologies and techniques to improve seizure detection's precision and dependability. Results: DWT + Hjorth + Statistical parameters with bior 1.5 as the pre-processing technique yielding the best outcomes. 86% accuracy was obtained with kNN for k = 5, 93% accuracy was obtained with a linear kernel for an SVM classifier, and 95.5% accuracy was obtained using a decision tree and logistic regression. The authors also considered another dataset for validation and received 96.83% accuracy with decision tree and logistic regression classifiers considering the bior1.5 wavelet filter as a preprocessing technique. Conclusion: The IoHT framework offers a multi-modal, adaptive method of seizure detection that enables the dynamic modification of detection parameters and the incorporation of extra sensor signals. This improves seizure detection's precision and dependability, which has important implications for patient care and monitoring. This work shows how IoHT and machine learning can be combined to build a reliable, real-time seizure detection system. These developments, which make it possible for prompt interventions and individualized treatment plans, can significantly improve the quality of care for individuals with epilepsy.
Read full abstract