The main purpose of this work is to explore whether and how much multichannel signal processing strategies can be beneficial for improving the detection procedure for auditory late response (ALR) in clinical applications in comparison with single channel recording. To achieve this target, four multichannel noise reduction methods based on independent component analysis (ICA) were proposed for noise reduction for multichannel recording of ALR. The four alternative component selection strategies introduced in this work are: Magnitude Squared Coherence (MSC) [based on coherency of the ICs with an evoking stimulus], the maximum Signal to Noise Ratio (Max-SNR) of ICs over a particular interval, the kurtosis (maximum non-Gaussianity of the ICs), and minimum entropy of the ICs. The proposed methods are applied for the noise reduction of auditory late response (ALR) data captured using 63 channel EEG from 10 normal hearing participants. The performances of the proposed methods for improving signal quality were compared with each other and also with the single channel alternatives. All automated component selection approaches produced high SNR for multichannel ALR data. MSC-ICs produced significantly higher SNR than Max-Kurt-ICs or Min-Entropy-ICs. However the performance of MSC-ICs and Max-Fmp-ICs were not significantly different. Therefore, the MSC-ICs approach was selected for further work. MSC-ICs were used for three different clinical applications: Finding hearing threshold level, exploring the effect of attention and exploring inter- and intra- subject variability. The results for MSC-ICs were compared to the single channel signal processing alternative of weighted averaging. The results confirm that the multichannel signal processing can significantly improve the detection procedure for threshold measurement and for measuring the effects of attention. However, no significant enhancement was found for detecting inter- and intra- subject variability with multichannel processing over single channel alternative. The MSC-ICs method was also used in an application for removing cardiac artifact from the ALR recordings and the results was compared with an existing artifact rejection platform based on constraint ICA (cICA). The results of this comparison show that the proposed method can significantly improve the quality of cardiac artifact rejection from ALR data. Finally, the use of MSC-ICs was explored for reducing the required time for recording ALR. Time reduction was investigated in two senses: 1. reducing the number of stimulus repetitions. 2. Optimizing the position and the number of the recording electrodes in multichannel recordings (potentially saving the time required to place many electrodes on the scalp). The results show that using multichannel processing can significantly reduce the number of stimulus repetitions and consequently the time of recording in comparison with the single channel alternative. Minimum required number of stimulus repetition (average over10 subjects) for having SNR equal to single channel processing at Cz was found to be 74 for un-weighted averaging and 85 for weighted averaging. Moreover, the results of optimal electrode placement procedure confirm that, the ALR can be recorded form the vertex (with the same SNR as when ALR is recorded using 63 channels) by using fewer numbers of electrodes. For the data set of this study (10 normal hearing adults) the same SNR as with 63 channels was achieved by using 40 channels. Placing 40 electrodes (instead of 63) on the scalp decreases the required time for recording ALR considerably, i.e. 53% improved.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:665127 |
| Date | January 2015 |
| Creators | Mirahmadizoghi, Seyedsiavash |
| Contributors | Bell, Steven |
| Publisher | University of Southampton |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://eprints.soton.ac.uk/379884/ |
Page generated in 0.0018 seconds