Will there be a role for neuroimaging in clinical psychiatry?

A consideration of the role of neuroimaging in clinical practice falls in the realm of discussions of personalized medicine. In reference to clinical psychiatry, personalized medicine can be simply conceptualized as falling into 3 domains: the study of genetic variation (including pharmacogenetics), the measurement of various molecular or biochemical indices of disease states (possibly including metabolomics or proteomics) and neuroimaging methods. Each of these approaches are being explored for their potential to improve the accuracy of diagnosis, but they may have a more immediate and prominent role in predicting outcomes or in matching patients with most appropriate treatment strategies. In fact, in a 2009 strategic plan for the National Institutes of Mental Health (NIMH), Insel included personalized care based on individual responses as a priority area for research, identifying a need for basic science research to enable the development of effective care. For any of these approaches to be incorporated into clinical practice, however, there must be advances in science, clinical practice and policy. The science of using neuroimaging techniques to diagnose psychiatric conditions is in a nascent stage. There are promising data from Fu and colleagues that functional magnetic resonance imaging (MRI) methods combined with a support vector machine (SVM) pattern classification method can correctly sort depressed patients and controls into their appropriate categories with a sensitivity of 84% and a specificity of 89%. More recently the same group used a general probabilistic classification method to produce measures of confidence for MRI data. Another group also used SVM applied to grey matter (structural) scans of patients with autism spectrum disorder and correctly classified affected participants with a specificity of 86.0% and a sensitivity of 88.0%. There was a relation between symptom severity and the extent to which a participant differed from the test margin. Although these are compelling results, differentiating a depressed patient from a nondepressed patient is not usually as challenging as being able to ascertain whether a first depression represents the first episode of a major depressive disorder or a bipolar disorder, or whether psychotic symptoms represent the onset of schizophrenia or a drug-induced psychosis in a young substance-abusing patient. To date, there are a limited number of studies that have specifically used SVM to differentiate between patient groups. One group was able to show that SVM was superior to radiologists in both separating patients with sporadic Alzheimer disease from normal aging and in separating patients with sporadic Alzheimer disease from patients with frontotemporal lobar degeneration. There is a need for large studies that include a range of patient populations to establish the specificity and sensitivity of these measures in distinguishing various illnesses not just from healthy brains but also from other illness states. Relative to imaging studies focusing on the accurate diagnosis of psychiatric syndromes, there are more studies examining the utility of various imaging modalities for predicting treatment responses and clinical outcomes. Structural MRI studies have reported that small hippocampal volumes are associated with poor shortand long-term clinical outcomes in patients with major depressive disorder. Reports of small hippocampal volumes being associated with poor clinical outcome are, so far, mostly confined to studies of patients with major depression, despite the fact that the hippocampus is known to be small in a variety of neuropsychiatric conditions. Functional MRI studies and other imaging modalities have shown that activity in the anterior cingulate cortex is predictive of clinical response to antidepressant medication and to cognitive behaviour therapy for depression and anxiety. Amygdala activation to emotional facial expressions among depressed patients also predicts symptom resolution. Neuroimaging methods are also being used to monitor and assess the effects of treatment. For example, cognitive enhancement therapy was recently compared against enriched supportive therapy in patients with schizophrenia. The main outcome measure was MRI-determined changes in grey matter over the course of 2 years. The potential and pitfalls of