Brain scans can predict who will respond to treatment for depression

A diagnosis of depression may seem like a clear and cohesive label, but people actually experience depression in different ways. A clinical diagnosis requires having 5 out of 9 different depression symptoms, so two patients with the same depression diagnosis could have just one symptom in common. This variability makes the disease difficult to understand and treat. Many depression patients waste months trying different medications and treatments in the hunt for something that works for them. A new study has found that a five-minute brain scan can be used to classify patients into different neurobiological subtypes and predict which patients will respond to brain stimulation-based treatment with up to 90 percent accuracy.

Repetitive transcranial magnetic stimulation, or rTMS, is a technique that uses a magnetic current to non-invasively stimulate brain activity to treat depression. For 30 percent of patients, rTMS is remarkably successful and fully alleviates their symptoms. Unfortunately, we previously did not know how to identify which patients fell into that 30 percent. Algorithms based on symptom severity were only 60 percent accurate in determining whether a patient would respond to rTMS – not much better than a coin-flip guess. Thus, the previous best option for patients was to undergo weeks of near-daily rTMS sessions and hope for improvement.

Researchers at Weill Cornell Medical College, Toronto Western Hospital, Emory University, and Stanford University can now better predict which depression patients will respond to rTMS by analyzing patients’ brain data. The researchers used M.R.I. to measure the brain activity of depression patients laying quietly at rest, without any instructions to think about or do anything in particular. Even in this so-called “resting-state”, the brain is constantly active and exhibits spontaneous yet systematic fluctuations in activity. With just five minutes of this resting-state brain activity, researchers were able to generate a statistical learning algorithm that predicted which patients would respond to rTMS with nearly 80 percent accuracy.

Screen Shot 2017-05-31 at 4.01.22 PM

From Figure 1 in the paper, even brains at rest exhibit spontaneous fluctuations of activity.

Though the algorithm based on resting-state brain activity was already superior to algorithms based on symptoms alone, researchers were able to make their brain-based algorithm even better by applying additional statistical classification techniques. First, they found two distinct sets of brain regions with resting-state activity that predicted different depression symptoms. The degree to which brain activity was synchronized between one set of regions, mostly from the front of the brain, predicted symptoms of decreased pleasure and slowed thought and movement. The degree of synchronization in a different set of regions, mostly from deeper structures that process emotion and reward, predicted symptoms of anxiety and insomnia.

Next, researchers applied a classification algorithm to separate patients into four different neurobiological subtypes, or biotypes, based on their brain synchronization within these two sets of regions. Though all biotypes had similar overall depression severity, each biotype had different combinations of severity for specific depressive symptoms. For example, patients in one biotype were distinguished only by their higher fatigue severity relative to the average patient, while patients in another biotype had not only higher fatigue but also higher anxiety and insomnia. Intriguingly, patients in different biotypes responded to rTMS at different rates. Only a quarter of patients in the anxiety-only biotype were rTMS responders. In contrast, more than three-quarters of patients in the high fatigue, anxiety, and insomnia biotype benefitted from rTMS.

Screen Shot 2017-05-31 at 3.58.50 PM

From Figure 3 in the paper, regions of the brain with distinct patterns of synchronization for each of the 4 biotypes.

The researchers realized that adding biotype assignment to their resting-state brain algorithms boosted accuracy up to 90 percent. Now, instead of taking a try-and-see approach to treating depression with rTMS, we only need five minutes of brain data to determine if a patient is a good candidate for rTMS. Classification algorithms and biotypes save time and money while improving the patient experience. People with depression who are identified as likely rTMS responders can start stimulation immediately and find relief as soon as possible, while those from non-responsive biotypes can be directed to seek other forms of treatment.

With this new knowledge that different depression biotypes exist, researchers can systematically study how different biotypes are affected by different forms of psychotherapy or different medications. Just as we know to treat sore throats differently depending on whether they are caused by strep bacteria or a cold virus, we may someday treat depression differently depending on whether it is caused by abnormal neural synchronization between one set of brain regions or another.

Though this study’s algorithms are specific to depression, the statistical learning techniques that the authors applied to resting-state brain data are generalizable. The concept of biotypes may unlock new ways of understanding and treating other highly variable and difficult to treat mental disorders such as schizophrenia. Neural biotypes are potential keys to making mental illness treatment more personalized and consequently more effective.

The abstract: Biomarkers have transformed modern medicine but remain largely elusive in psychiatry, partly because there is a weak correspondence between diagnostic labels and their neurobiological substrates. Like other neuropsychiatric disorders, depression is not a unitary disease, but rather a heterogeneous syndrome that encompasses varied, co-occurring symptoms and divergent responses to treatment. By using functional magnetic resonance imaging (fMRI) in a large multisite sample (n = 1,188), we show here that patients with depression can be subdivided into four neurophysiological subtypes (‘biotypes’) defined by distinct patterns of dysfunctional connectivity in limbic and frontostriatal networks. Clustering patients on this basis enabled the development of diagnostic classifiers (biomarkers) with high (82–93%) sensitivity and specificity for depression subtypes in multisite validation (n = 711) and out-of-sample replication (n = 477) data sets. These biotypes cannot be differentiated solely on the basis of clinical features, but they are associated with differing clinical-symptom profiles. They also predict responsiveness to transcranial magnetic stimulation therapy (n = 154). Our results define novel subtypes of depression that transcend current diagnostic boundaries and may be useful for identifying the individuals who are most likely to benefit from targeted neurostimulation therapies.

ResearchBlogging.orgDrysdale, A., Grosenick, L., Downar, J., Dunlop, K., Mansouri, F., Meng, Y., Fetcho, R., Zebley, B., Oathes, D., Etkin, A., Schatzberg, A., Sudheimer, K., Keller, J., Mayberg, H., Gunning, F., Alexopoulos, G., Fox, M., Pascual-Leone, A., Voss, H., Casey, B., Dubin, M., & Liston, C. (2016). Resting-state connectivity biomarkers define neurophysiological subtypes of depression Nature Medicine, 23 (1), 28-38 DOI: 10.1038/nm.4246


How to Spot Stellar Brain Science Infographic

In my last post, I promised that my next pop science write-up would be about my own research. That piece has been written for months, but I am working on getting it published in a legit media outlet, so stay tuned.

In the meantime, I was inspired by this infographic about spotting bad science to create my own infographic about How to Spot Stellar Brain Science. The content is based on an article by Russ Poldrack, one of my neuroscience heroes.

Click for zoomable version

Click for zoomable version

fMRI studies can generate flashy popular press coverage, and studies have shown that just adding brain images makes scientific reports seem more credible. Thus, it’s important to know how to critically evaluate the quality of neuroimaging studies.

Dr. Poldrack’s piece is a great teaching tool for undergraduates or anyone new to cognitive neuroscience. Last semester, I had my students read Dr. Poldrack’s piece and then analyze this extremely popular NY Times Op-Ed about how fMRI data purportedly show that we literally love our iPhones.

Does the Op-Ed pass the infographic test? I’ll refer you to Dr. Poldrack himself. If you want a second opinion, I’ll refer you to these dozens of other neuroscience PhDs and professors.

2015 has been a busy year

I have done a terrible job at keeping up my science writing. My folder of cool studies to write about keeps growing, and my spare time to do such writing has continued to dwindle. Here are some the science-related things I have done in 2015:

  • Published my first first-author paper about how kids do not fear the unknown. That will be my next pop science write-up!
  • Led the organization and execution of ComSciCon-Triangle 2015, a local, 2-day science communications workshop for other grad students. You can see how that went on Twitter @ComSciConTri and #ComSciCon.
  • Had my first undergraduate thesis student successfully write and defend her thesis! She gets to Graduate with Distinction next week. I am so proud of her.
  • Was published in Science twice (in January and April). Unfortunately, it was not my research getting into the pages of Science, but it’s still fun to say that I’ve been published in one of the top scientific journals. And you can download my face as a PowerPoint slide for teaching at the Science website, which I find hilarious.

Here’s what’s on deck for the rest of 2015