Efficient Spike Detection with Singular Spectrum Analysis Filter

Several technological tools have been developed to aid neurologists in accurately diagnosing epilepsy, which is among the diseases commonly addressed in neurological clinics. In real-life conditions, the recording of surface or depth electroencephalogram (EEG) is always disrupted by artifacts, making it difficult to analyze critical and interictal paroxysmal events (IPE) or spikes of short durations. Artifact synchronizations are commonly observed, which can lead to medical interpretation errors. Often, data analysis is limited to visual inspection of EEG tracings, which does not always enable the identification of certain epilepsy events. Therefore, it is necessary to reduce or even eliminate noise for better data processing. Given the structure of EEG signals, we will study filters that allow us to highlight transient events. In this paper, We suggest utilizing the Singular Spectrum Analysis (SSA) filter for improved detection of spikes within EEG signals. Before choosing the SSA filter, a comparison with other filters such as high-pass, and Kalman filters was conducted. SSA filter is paired with the fractal dimension (FD) detector for automated spike detection. FD is based spike detection method using adaptive threshold.