Protocol Design Challenges in the Detection of Awareness in Aware Subjects Using EEG Signals

Abstract

Recent studies have evidenced serious difficulties in detecting covert awareness with electroencephalography-based techniques both in unresponsive patients and in healthy control subjects. This work reproduces the protocol design in two recent mental imagery studies with a larger group comprising 20 healthy volunteers. The main goal is assessing if modifications in the signal extraction techniques, training-testing/cross-validation routines, and hypotheses evoked in the statistical analysis, can provide solutions to the serious difficulties documented in the literature. The lack of robustness in the results advises for further search of alternative protocols more suitable for machine learning classification and of better performing signal treatment techniques. Specific recommendations are made using the findings in this work.

“Look at my classifier's result”: Disentangling unresponsive from (minimally) conscious patients

“…Using this approach, data from individual trials are more closely matched within a block than across multiple blocks. Furthermore, temporal dependence between blocks may have a confounding effect (Goldfine et al, 2013;Henriques et al, 2014) especially if the number of blocks is small. In general, the block structure of the data must be taken into account when statistically evaluating a classifier's performance (Lemm et al, 2011;Goldfine et al, 2013).…”

Substitute or complement? Defining the relative place of EEG and fMRI in the detection of voluntary brain reactions

“…These high discrepancies of sensitivity which depend on the statistical method used are a current concern in the interpretation of EEG data. Additionally, the test used for statistical analysis by Cruse et al (2011) involves assumptions that cannot be justified and that, as it is shown in (Henriques et al, 2014), are likely to be violated which hence may produce misleading results. Indeed, in this protocol, the cross-validation design is a particularly delicate issue that has already been pointed out (Goldfine et al, 2013;Noirhomme et al, 2014) and that shows the strong influence of the temporal dependence between the test-set blocks on the classification accuracy.…”

Brain–Computer Interfaces for Awareness Detection, Auxiliary Diagnosis, Prognosis, and Rehabilitation in Patients with Disorders of Consciousness

“…Furthermore, the paradigm design needs to consider the possibility of fluctuating arousal, rapid fatigue, and limited attention span of patients with brain injury. 47 In this regard, complexity and duration are important factors to be considered when designing a BCI paradigm.…”