Neural network-based method for measuring the impacts of epileptic brain activities on cardiac cycles.

Although electroencephalogram (EEG) is widely used to monitor brain activity in epilepsy, limitations related to the accessibility and reproducibility of measurements may restrict its everyday use. Conversely, wearable methods, easily accessible, such as electrocardiogram (ECG), represent an alternative for indirectly monitoring brain activity through cardiac cycles. A computational model was developed based on statistical cycles and neural networks to measure changes in the morphology of ECG waves. The advantage of this approach over heart rate variability analysis is the detection of brain activity before changes in heart rate occur. In addition, using variance, skewness, and kurtosis centered on the median allowed us to achieve 100% sensitivity, specificity, and accuracy in our analyses, even using less complex algorithms, due to selecting these optimal characteristics. These findings indicate that ECG is a viable, affordable, and effective alternative for estimating epileptic brain activity. This approach's application of machine learning highlights its potential for non-invasive epilepsy monitoring, providing a cost-effective and accessible solution, especially for vulnerable populations.