A clinical system for acquisition of averaged and single trial evoked potentials

The aim of the study was to implement an Evoked Potentials Clinical System(EPCS) for the acquisition of averaged and single-trial evoked potentials(EP) and evaluate different signal processing techniques of EP. The EPCS has been built as a Supervisor/Server systcmfl] for an existing DISAT 2000 EP/EMG system [2]. Different signal processing techniques like recursive SNR estimation and confidence intervals determination method [3], [4], quality assessment of EP, median/mean combined estimator method have been examined, developed for EPCS. Applications with simulated and real evoked potentials have been performed. The real data used in the applications were recorded with the EPCS. We have defined a recursive estimation method for the SNR of the AEP. It is used to estimate the number of sweeps needed to obtain a good estimate of the embedded signal in the EP. The confidence intervals for the SNR estimator are also given. The recursive estimation method does not only shorten the recording session time but also gives less pain to the patient and saves the time of the physician. For the cases that the signal power is low and the noise power high the number of sweeps needed is increased and a better estimate of the AEP is obtained. We also defined median/mean estimators. Two types estimators is used for on-line artifact rejection without a decrease in the SNR of the AEP compared to the mean averaging. The experiment results of block median/mean estimator with simulated and real EP data proves that artifact rejection property of median averaging and noise suppression of mean averaging is working coordinately in one tool. By the use of the block median/mean estimator without any need of an extra artifact rejection module, a good estimate of the AEP is obtained in an approximately equal time that requires for the conventional mean ayeraging. Also a running median filter/mean estimator is implemented. The running median filter has a capability of suppressing impulse like noises and high frequency components. The running median filter/mean estimator uses this facility of running median filter and the noise suppression of the mean average. Different quality assessment methods for EP were examined and coded. The intra-class correlation coefficient quality assessment method is implemented and used to asses the quality of the AEP and to shorten EP recording session. We found that intra-class correlation coefficient quality assessment method can perform well under both real and simulated EP. We also concluded that the SNR of the EP is directly related with the intra-class correlation coefficient.vııı As the SNR of the EP is increased the intra-class correlation coefficient grows rapidly and reaches the threshold value in smaller number of sweeps. We developed the split half latency difference quality assessment method for EP. This method calculates the latency differences of maximums and the minimums of the even and odd numbered sweeps' averages. The tolerance level for the differences İs given and the application with simulated EP was done. By changing the tolerance level the number of sweeps needed to process is increased or decreased. Using this method the data gathering process can be halted if the quality of the AEP achieves the predetermined quality level by the means of the tolerance level. The EPCS is being used for clinical and research purposes now in Neurology Clinic, Balcah Hospital, Çukurova University, Adana. EPCS with the enhanced signal processing modules would give the opportunities of post-processing of the recorded data, quality assessment of the recorded data and on-line artifact rejection.