Hybrid KNN-CNN Framework for EEG-Based Epileptic Seizure Detection

The proposed paper refers to a neurological condition known as epilepsy. In other words, it concerns the unpredictable nature of seizures which pose risks to patients' safety and quality of life. To detect seizures automatically, EEG signals can be employed since they represent brain activity which could be used as input data. Unfortunately, the process of manual evaluation of EEG signals is very labor-intensive, complicated, and associated with a risk of mistakes. Therefore, this paper attempts to introduce an efficient hybrid solution to detect seizures in patients using the KNN and CNN models. As part of the proposed approach, statistical features of preprocessed EEG signals will be recognized by the KNN algorithm, while the CNN model will analyze the spectrograms (i.e., time-frequency representations). The outputs produced by both algorithms will be subsequently merged to ensure more accurate results. In conclusion, experimental findings have shown that this hybrid solution works better than traditional ones.