Evoked Potentials SNR maximization by PCA and genetic algorithms

Middle Latency Auditory Evoked Potentials are bioelectrical signals that constitute a key technique for the assessment and diagnosis of various clinical conditions, nonetheless background electrical activity and artifacts prevent clinically relevant information to be revealed. Many state of the art techniques allow physicians to remove noise and help them to reach assertive diagnosis, however those techniques are often related to a priori signal templates that induce bias on the resulting information. In this paper a technique based on the iterative evaluation and combination of the available EEG trials is presented and performance is evaluated by comparing the results obtained from processing data from normo-hearing and clinically deaf patients. Results suggest that the proposed methodology improves the SNR ratio of the records by the only usage of the patient's EEG and a few fixed parameters.