Dissecting the coma spectrum using Bayesian classification

Abstract

A patient who does not regain full consciousness after coma is typically classified as being in a vegetative state or a minimally conscious state. While the key determinants in this differential diagnosis are inferred uniquely from the observed behaviour of the patient, nothing can, in principle, be known about the patient's awareness of the external world. Given the subjective nature of current diagnostic practice, the quest for neurophysiological markers that could complement the nosology of the coma spectrum is becoming more and more acute. We here present a method for the classification of patients based on electrophysiological responses using Bayesian model selection. We validate the method in a sample of fourteen patients with a clinical disorder of consciousness (DoC) and a control group of fifteen healthy adults. By formally comparing a set of alternative hypotheses about the nosology of DoC patients, the results of our validation study show that we can disambiguate between alternative models of how patients are classified. Although limited to this small sample of patients, this allowed us to assert that there is no evidence of subgroups when looking at the MMN response in this sample of patients. We believe that the methods presented in this article are an important contribution to testing alternative hypotheses about how patients are grouped at both the group and single-patient level and propose that electrophysiological responses, recorded invasively or non-invasively, may be informative for the nosology of the coma spectrum on a par with behavioural diagnosis.