On the difficulty to communicate with fMRI-based protocols used to identify covert awareness

Assessment of awareness in patients with disorders of consciousness such as patients in a vegetative state (unresponsive wakefulness syndrome, UWS) and patients in a minimally conscious state (MCS) remains difficult, with a high rate of misdiagnosis (around 40%). While patients with UWS have no awareness, patients with MCS have partial preservation of conscious awareness. To improve the assessment of awareness in these patients, recent functional neuroimaging protocols have been developed. However, does the complexity of realizing and interpreting these functional magnetic resonance imaging (fMRI) investigation protocols, which are currently carried out by only a few specialist teams, permit generalizable use in clinical routine? In this study, 32 healthy volunteers, by definition perfectly conscious and able to efficiently communicate, performed the protocol proposed by Monti et al. in 2010. Four methods (comprising the method proposed by Monti et al., a mean squared error-based method, a correlation-based method, and a support vector machine-based method) were tested for correctly and accurately interpreting the communication task. Firstly, the different instructions for the localizer and the communication tasks had no effect on activations. Secondly, 25% of participants (8/32) did not provide the expected patterns of activations during fMRI tasks (four for each imagery task). However, this did not necessarily prevent the classification methods from correctly guessing the answers during the communication task. Conversely, these classification methods may fail to detect the correct answers even though participants activated the expected brain areas. None of the four methods produced 100% correct detection during the communication phases. The correlation-based method obtained the best results with an error rate of 4.2%. The results of this study demonstrate that fMRI-based communication paradigms may not be robust enough to reliably detect awareness in all aware patients. There is still a need to develop new statistical and analytical methods before considering their generalization in clinical routine.