Please use this identifier to cite or link to this item:
http://hdl.handle.net/11375/24047
Title: | Towards an automated measure of levels of cognitive function in unresponsive patients |
Authors: | Mah, Richard L |
Advisor: | Connolly, John F |
Department: | Cognitive Science of Language |
Publication Date: | 2018 |
Abstract: | This thesis aims to evaluate the clinical utility of several auditory paradigms designed to elicit various event-related scalp potentials (ERPs). Through four papers, this thesis (i) determines which paradigms best elicit the desired components in healthy controls, (ii) evaluates methods of confirming the presence of the MMN in clinical ERP data, and (iii) examines the use of spectral entropy, and specifically the use of wavelet signal decomposition to determine the periodicity of spectral entropy in order to target the use of these paradigms for diagnostic use. Chapter 2 first sets out a framework for extended monitoring of patients in coma by selecting paradigms that performed well in healthy control populations. From an initial group of six paradigms designed to elicit the MMN, P300, and N400, two are selected that were able to elicit the desired ERPs from the healthy controls. This study is the first to examine these various paradigms within the same participants, as well as across two different age groups (younger and older adults). Chapters 3 and 4 provide evidence that the MMN–a component previously thought to be stable over time–appears to fluctuate in detectability in patients in coma. In addition to the traditional visual inspection method of MMN detection, four other methods of verifying the presence of the MMN were evaluated: the topographic consistency test, a serial t-test, a spatiotemporal cluster analysis, and Bayesian t-tests. In all four patients presented, the MMN appears to change in detectability over a period of approximately 24 hours. The spatiotemporal cluster analysis and Bayesian t-tests both proved to be suitable for use in confirming visual inspection judgements of the presence of the MMN in this set of patient data, and were able to overcome problems from external noise in the signal. These results suggest that patients should be tested multiple times to increase the likelihood of capturing a period where the MMN is detectable and reducing the chance of a false negative. Finally, Chapter 5 examines the application of a spectral entropy signal analysis to the same patient data. Period of higher spectral entropy are indicative of a more complex EEG signal, which in turn has been thought to index conscious experience. This analysis was used to determine both if the patients had periods of higher spectral entropy, and if they did, what the periodicity of that fluctuation of spectral entropy would be. Previous work has shown that patients in a minimally conscious state (MCS) can show periods of around 70 minutes, which is similar to healthy, conscious individuals. Of the three patients whose data was appropriate to use in this analysis, one showed a periodicity of around 70 minutes, one did not show a signal with a strong main periodicity, and one had two main periodicities which was indicative of being contaminated by external noise. Even though the analysis method is extremely sensitive to external noise, it does show promise as a means of targeting the cognitive assessments, as these should be given when the patient is likely to be more conscious. This is especially important considering the evidence presented that the MMN fluctuates in comatose patients, so targeting the delivery of these tests can further reduce the false negative rate. Overall, we have established which ERP paradigms have the best chance of eliciting the components of interest, which in turn can be used for coma prognostication. We have presented evidence suggesting that the MMN fluctuates in its detectability, which provides a caution to clinicians using this methodology to perform repeat testing to better capture the MMN. As well, we have suggested methods to further confirm the presence of the MMN in noisy patient data. Finally, we provide a method of using spectral entropy for determining periods during which these tests should be performed to maximize the likelihood of capturing the components of interest. Taken together, this work brings us closer to an automated measure of the levels of cognitive function in unresponsive patients. |
URI: | http://hdl.handle.net/11375/24047 |
Appears in Collections: | Open Access Dissertations and Theses |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
Mah_R_L_2018September_PhD.pdf | 32.38 MB | Adobe PDF | View/Open |
Items in MacSphere are protected by copyright, with all rights reserved, unless otherwise indicated.