Neuroimaging and Psychotherapy: Tales From the Messy Middle
by Andrew J. Gerber, MD, PhD
This article is republished with permission from Psychiatric News, the newspaper of the American Psychiatric Association. It was originally published online on November 17, 2016.
Sixteen years ago I sat in the office of the director of residency at a prominent academic department of psychiatry while he asked me the question, “Can you name a neuroimaging study that has changed the clinical practice of psychiatry in a significant way?” After some anxious hemming and hawing, I ended up with the answer that he (somewhat smugly) was looking for: no. Other than the relatively rare situation in which a brain MRI or CT scan reveals a cancerous growth or a syndrome with macroscopic evidence (for example, tuberous sclerosis) that accounts for psychiatric symptoms, neuroimaging does not contribute practically to the diagnosis, understanding, or treatment of the disorders that we as psychiatrists see every day.
Today the answer to his question would be no different. While it is standard of care to obtain a neuroimaging study on a patient with new-onset psychosis, and increasingly a brain MRI is ordered for patients with newly diagnosed autism spectrum disorder, the clinical yield of such scans is low. New imaging technologies such as functional MRI (indicating patterns of activity as measured via the flow of oxygenated blood), diffusion tensor imaging (showing white matter tracts), and magnetic resonance spectroscopy (reflecting concentration of particular neurochemicals) have played a major role in research and offer exciting possible leads for researchers to follow, yet still are not useful clinically. Serious experts disagree sharply on when that will change—estimates range anywhere from a few years to several decades. Private clinics or practitioners who advertise the use of neuroimaging in diagnosis do so without scientific evidence and are widely and appropriately condemned by mainstream psychiatry as opportunistic and misleading.
The theoretical and methodological obstacles that prevent neuroimaging from being more useful clinically are varied. Though large-scale examples of localization in the brain are well known (for example, Broca’s and Wernicke’s areas for aphasia, the hippocampus for memory), consistent and identifiable localization of the more subtle deficits and compensatory mechanisms in psychiatric conditions has been elusive. Meta-analyses of imaging studies looking at the neural correlates of psychiatric conditions and even basic processes, such as emotion, implicate a wide range of areas (from subcortical structures to various portions of the prefrontal cortex), and there is little consistency even on whether hypo- vs. hypertrophy (with an anatomical lens) or hypo- vs. hyperactivation (with a functional activation lens) of particular regions is more indicative of pathology. Meanwhile, psychiatry continues to struggle between categorical divisions (as in DSM) and dimensional descriptions (as promoted by the National Institute of Mental Health in the Research Domain Criteria, or RDoC). The bottom line seems clear—despite significant progress in many areas of neuroscience, we continue to struggle describing the basic physiology and pathophysiology of brain systems.
A natural human tendency is to address difficult problems with generalities. “It’s all just 21st century phrenology! Neuroimaging will never be useful!” some say, while others are keen to suggest that the major breakthrough in clinical use of neuroimaging in psychiatry is just one paper away, or even already here and just underappreciated.
The truth is likely much more in the messy middle. As neuroimaging technology continues to improve, we as a field will need to do the painstaking work of collecting quantifiable and reliable measures of brain structure and function and then map these onto detailed clinical characterization in ever increasing numbers of patients to refine our hypotheses and ultimately reach predictions that are valuable to clinicians in assessing individual cases.
Developments in the neuroimaging field suggest that such careful work is already under way and building momentum. To this day, it is commonplace for MRI researchers to use a wide range of data collection and analytic strategies that make combining results from different laboratories or different scanners difficult to impossible. The Human Connectome Project (HCP), a large scale multisite effort funded by the National Institutes of Health and conducted at key neuroimaging centers nationwide (led by Washington University in St. Louis and the University of Minnesota), has developed, published, and validated a cutting-edge approach to collection and analysis of three key MRI modalities: anatomical, functional, and diffusion tensor imaging. This is coupled with a powerful open-source database tool for organizing MRI data (XNAT) and a set of processing pipelines that facilitate standardized use of analysis packages Freesurfer and FSL. While there will continue to be, and always should be, a range of tools to study various MRI data, the move toward standardization will help investigators pool results and move toward meaningful clinical correlations. Last but not least, these tools have been used to collect and analyze almost 1,200 subjects, and the data are being made freely available for others to use as normative standards or for other in depth analyses.
A unifying and centrally motivating pursuit in the field of psychotherapy research is to identify the mechanisms of action for change across different approaches and schools of psychotherapy. Overwhelming evidence suggests that many forms of psychotherapy, even those with superficially contrasting theories of change, are helpful to individuals with a range of problems. To date, the only consistent and powerful predictor of psychotherapeutic change with an evidence base is the alliance between therapist and patient. Perhaps no other question is as central to the daily thinking of clinicians, researchers, and even patients themselves as how to determine which approaches and techniques are most helpful to which patients. Through the type of dimensional and neurobiologically linked research outlined here, we could potentially learn more than has previously been known about psychotherapeutic mechanisms of action and the pros and cons of various techniques. The work of the HCP and the philosophy of RDoC, backed up by the data from neuroimaging studies of psychotherapeutic change, will help us to elucidate psychiatric illnesses as a fluid constellation of features and enable us to study which of those features respond to what psychotherapeutic techniques and how those features change over the course of treatment.
It is too soon to know how significant the impact of developments like the HCP and RDoC will be on clinical psychiatry or when they will be realized, but it’s still possible to fantasize what they might be. A future for psychiatry may lie in using noninvasive imaging technologies to assist in initial characterization of illness and matching those features to optimal techniques and prediction of prognosis. The effects of psychotherapy and/or psychopharmacology could then be tracked with follow-up imaging and used, along with clinical assessments, to determine progress and help make decisions as to when the approach should be changed or the initial assessment and prognosis re-evaluated. The potential impact on our field is enormous.