L’imagerie cérébrale, un outil au service de la psychopathologie ?

Abstract

Psychopathology is worthy of consideration to inform the application of functional neuroimaging to the study of mental disorders. By promoting an approach based on mental processes underlying disorders rather than an approached based on diagnostic categories, it offers the opportunity to identify biomarkers that could be used as nosological tools. Such biomarkers may eventually inform treatment strategies. However, the question of whether functional brain imaging can be used to learn something about psychopathology, that is understanding the nature and the relationships of mental processes that underlie psychiatric disorders, remains controversial. A potential advantage of functional brain imaging could be the examination of non-conscious mental processes that could be both non reportable by the subject and difficult to capture with behavioral measures. As an example, several studies found depression and anxiety to be associated with amygdala aberrant reactivity to masked emotional stimuli. This might be interpreted as signaling automatic cognitive biases such as those proposed by cognitive theory. But when trying to deduce the presence of a mental process M from a specific pattern of brain activity A, frequently referred to as reverse inference, one has to consider several issues, regardless of the nature of the mental process. These issues can be best formalized with a Bayesian approach where the probability of the presence of M according to A, P(M|A) depends on the a priori probability of M to be present, P(M), as well as on conditional probabilities P(R|M) and P(R|M¯). Obviously, P(M) depends on the specificity of the experimental paradigm used. However, reverse inference usually takes place when A was unexpected (i.e. not supposed to be engaged by the experimental task). Conditional probabilities regarding the probability of A according to the presence or the absence of M, P(A|M) and P(A|M¯) can be estimated using methods of data mining based on growing databases. The less A is specific to M, that is to say the higher P(A|M¯) is, the less valid the reverse inference will be. This is even worse when the definition of R is loose. Despite these limitations, reverse inference is nevertheless a powerful heuristic tool when it comes to generate testable hypotheses about the nature of mental processes and their relationships. Combined with carefully designed experimental paradigms, functional brain imaging is thus likely to bring new knowledge to psychopathology.