Abstract:
"
Differentiating artefactual signals from iEEG signals has become a time/labour-consuming task
to be handled by humans which makes it impractical for a large amount of data. By identifying
the limitations of the existing solutions and introducing a new method that is accurate and
efficient in identifying artefacts will make the noise cancellation of the iEEG signal more
successful.
This project intends to substantially streamline the process, allowing for more effective
research, by analysing and developing methods for smoothing misrepresented, noisy data, as
well as algorithms for primitive pattern recognition. Numerous signal processing techniques
were analysed, and novel iEEG signal processing algorithms were developed to address the
issues. The basic premise of this research was that there was minimal chance of properly
detecting each noise spike, and the only logical alternative was to emphasize all probable
spikes.
The approach has a baseline precision of 74.6 per cent and increases in accuracy as the
occurrence of spikes increases, reaching its highest precision of 87.5 per cent. The result is a
solution that accepts a path of raw Intracranial Electroencephalogram input and graphically
depicts all potential spikes for researchers by simultaneously analysing and comparing several
data streams."
