Authors: Willem B. Bruin; Luke Taylor; Rajat M. Thomas; Jonathan P. Shock; Paul Zhutovsky; Yoshinari Abe; Pino Alonso; Stephanie H. Ameis; Alan Anticevic; Paul D. Arnold; Francesca Assogna; Francesco Benedetti; Jan C. Beucke; Premika S. W. Boedhoe; Irene Bollettini; Anushree Bose; Silvia Brem; Brian P. Brennan; Jan K. Buitelaar; Rosa Calvo; Yuqi Cheng; Kang Ik K. Cho; Sara Dallaspezia; Damiaan Denys; Benjamin A. Ely; Jamie D. Feusner; Kate D. Fitzgerald; Jean-Paul Fouche; Egill A. Fridgeirsson; Patricia Gruner; Deniz A. Gürsel; Tobias U. Hauser; Yoshiyuki Hirano; Marcelo Q. Hoexter; Hao Hu; Chaim Huyser; Iliyan Ivanov; Anthony James; Fern Jaspers-Fayer; Norbert Kathmann; Christian Kaufmann; Kathrin Koch; Masaru Kuno; Gerd Kvale; Jun Soo Kwon; Yanni Liu; Christine Lochner; Luisa Lázaro; Paulo Marques; Rachel Marsh; Ignacio Martínez-Zalacaín; David Mataix-Cols; José M. Menchón; Luciano Minuzzi; Pedro S. Moreira; Astrid Morer; Pedro Morgado; Akiko Nakagawa; Takashi Nakamae; Tomohiro Nakao; Janardhanan C. Narayanaswamy; Erika L. Nurmi; Joseph O'Neill; Jose C. Pariente; Chris Perriello; John Piacentini; Fabrizio Piras; Federica Piras; Y. C. Janardhan Reddy; Oana G. Rus-Oswald; Yuki Sakai; João R. Sato; Lianne Schmaal; Eiji Shimizu; H. Blair Simpson; Noam Soreni; Carles Soriano-Mas; Gianfranco Spalletta; Emily R. Stern; Michael C. Stevens; S. Evelyn Stewart; Philip R. Szeszko; David F. Tolin; Ganesan Venkatasubramanian; Zhen Wang; Je-Yeon Yun; Daan van Rooij; ENIGMA-OCD Working Group; Paul M. Thompson; Odile A. van den Heuvel; Dan J. Stein; Guido A. van Wingen · Research

Can Brain Scans Diagnose OCD? New Study Examines Potential of MRI

Large international study finds brain scans alone cannot reliably diagnose OCD, but may detect effects of medication

Source: Bruin, W. B., Taylor, L., Thomas, R. M., Shock, J. P., Zhutovsky, P., Abe, Y., Alonso, P., Ameis, S. H., Anticevic, A., Arnold, P. D., Assogna, F., Benedetti, F., Beucke, J. C., Boedhoe, P. S. W., Bollettini, I., Bose, A., Brem, S., Brennan, B. P., Buitelaar, J. K., ... van Wingen, G. A. (2020). Structural neuroimaging biomarkers for obsessive-compulsive disorder in the ENIGMA-OCD consortium: medication matters. Translational Psychiatry, 10(1), 342. https://doi.org/10.1038/s41398-020-01013-y

What you need to know

  • Brain scans alone cannot reliably diagnose obsessive-compulsive disorder (OCD), according to a large international study
  • Machine learning analysis of brain scans could not accurately distinguish between people with OCD and healthy controls
  • Brain scans were better able to detect differences between medicated and unmedicated OCD patients
  • The study highlights how medication can affect brain structure in OCD
  • More research is needed to develop reliable biomarkers for diagnosing OCD

Can brain scans diagnose OCD?

Obsessive-compulsive disorder (OCD) is a mental health condition that affects 2-3% of the population worldwide. People with OCD experience recurring, intrusive thoughts (obsessions) and repetitive behaviors or mental acts (compulsions). Currently, OCD is diagnosed based on symptoms and behaviors observed by mental health professionals. There are no biological tests or brain scans that can definitively diagnose OCD.

Researchers have been exploring whether brain imaging techniques, like MRI scans, could potentially be used to diagnose OCD or other mental health conditions. Some smaller studies have suggested that machine learning analysis of brain scans may be able to distinguish between people with OCD and those without the condition. However, these studies typically used small sample sizes from single research sites.

A large international research consortium called ENIGMA-OCD set out to test whether brain scans could reliably diagnose OCD using a much larger dataset collected from multiple sites around the world. Their study, published in the journal Translational Psychiatry, provides important insights into the potential and limitations of using brain scans to diagnose OCD.

The ENIGMA-OCD study

The study analyzed MRI brain scans and clinical data from 2,304 OCD patients and 2,068 healthy control participants. This data came from 36 research institutes across the world. The large sample size and diverse data sources allowed the researchers to test how well machine learning algorithms could distinguish between OCD patients and controls when applied to brain imaging data from multiple sites.

The researchers used advanced machine learning techniques to analyze measurements of brain structure taken from the MRI scans. They looked at the thickness and surface area of different regions of the brain’s cortex, as well as the volume of subcortical structures deeper in the brain.

Overall findings: Brain scans alone cannot reliably diagnose OCD

When the machine learning algorithms were applied to the full dataset to try to distinguish between OCD patients and healthy controls, the performance was relatively poor. The algorithms could only correctly classify people as having OCD or not about 60% of the time. This is only slightly better than chance (50%) and not accurate enough to be clinically useful for diagnosis.

The researchers tried various machine learning approaches and ways of analyzing the data, but none produced results accurate enough for a reliable diagnostic test. When they tested how well models trained on data from some research sites could classify people from other sites that weren’t used in the training data, the performance was even worse - barely better than chance.

Why can’t brain scans diagnose OCD?

There are a few potential reasons why analyzing brain structure with MRI scans was not able to reliably distinguish people with OCD:

  1. Heterogeneity: OCD is a complex disorder that can manifest in different ways. Brain differences associated with OCD may be subtle and variable across individuals.

  2. Confounding factors: Things like age, sex, medication use, and even differences in MRI scanners between sites can affect brain structure measurements. These factors may obscure OCD-related differences.

  3. Lack of specificity: Any brain differences associated with OCD may not be unique to the disorder, and could overlap with other mental health conditions.

  4. Limited brain measures: The study only looked at certain structural measures from MRI. More complex analyses or different types of brain scans may be needed.

Medication effects detected in brain scans

While the brain scans could not reliably diagnose OCD overall, the researchers did find that machine learning analysis could distinguish between medicated and unmedicated OCD patients with fairly good accuracy (around 80%).

Specifically, OCD patients taking medications like selective serotonin reuptake inhibitors (SSRIs) showed detectable differences in brain structure compared to unmedicated patients. The brain regions showing the biggest medication-related differences included areas of the frontal cortex, temporal cortex, and some deeper brain structures like the thalamus.

This finding highlights how psychiatric medications can have substantial effects on brain structure. It suggests that medication use needs to be carefully accounted for in brain imaging studies of mental health conditions. The effects of medication on the brain may sometimes obscure or complicate efforts to identify disorder-specific brain differences.

Implications and future directions

This large-scale study indicates that analyzing brain structure from MRI scans alone is not currently accurate enough to diagnose OCD. However, the research does provide some important insights:

  1. Need for large, diverse samples: The poor performance when applying models across different research sites highlights the importance of large, diverse datasets in developing reliable biomarkers.

  2. Accounting for medication: The notable effects of medication on brain structure emphasize the need to carefully consider medication use in neuroimaging studies of mental health conditions.

  3. Clinical heterogeneity matters: The difficulty in finding consistent brain differences may reflect the diverse ways OCD can manifest. Focusing on specific symptom dimensions or subtypes of OCD may be more fruitful.

  4. Beyond structure: Future studies may need to look at brain function, connectivity between regions, or more advanced imaging techniques to find reliable neural signatures of OCD.

  5. Longitudinal studies needed: To better understand how medications affect brain structure over time, studies that scan people before and after starting treatment are crucial.

While this study shows brain scans are not yet able to diagnose OCD, neuroimaging research still has an important role to play. Brain imaging studies can provide insights into the neural mechanisms underlying OCD symptoms and how treatments affect the brain. This knowledge could eventually lead to better targeted therapies.

As methods for analyzing brain scans continue to advance, it’s possible that neuroimaging may still prove useful as part of a broader approach to diagnosing and characterizing OCD. However, this study suggests we are not there yet, and more work is needed to develop reliable biomarkers for the disorder.

Conclusions

  • Current MRI brain scan analysis methods cannot reliably distinguish between people with OCD and those without the disorder
  • Machine learning algorithms performed poorly when trying to classify OCD patients vs. healthy controls using brain structure measurements
  • Brain scans could detect differences between medicated and unmedicated OCD patients, highlighting how psychiatric medications can affect brain structure
  • Large, diverse datasets are crucial for developing and testing potential biomarkers for mental health conditions
  • More research is needed to identify reliable neural signatures of OCD, potentially using different brain imaging techniques or analytical approaches

This study represents an important step in critically evaluating the potential of brain scans for diagnosing OCD. While the results suggest structural MRI alone is not sufficient for diagnosis, they provide valuable insights to guide future research efforts aimed at better understanding and potentially diagnosing OCD using neuroimaging.

Back to Blog

Related Articles

View All Articles »