Diagnosing mental health issues accurately and in a timely way has historically been a challenge for the health care industry. Unlike many physical ailments, which can be diagnosed with the aid of objective tests and tools, mental illness has long required a more subjective approach to evaluating symptoms. In recent years, magnetic resonance imaging (MRI) has allowed practitioners to make steady progress in improving their diagnostic approach.
Neuromarkers of mental health disorders
One of the foundational elements of MRI use for mental health diagnostics is neuromarkers. Neuromarkers are patterns or images on brain scans that scientists believe indicate certain activity, functionality or dysfunction within the brain. The markers show up on a variety of scans, including those captured with MRI, qEEG, ERP and PET machines, and researchers have identified markers related to a range of psychiatric disorders.1
By 2013, scientists had identified neuromarkers for a variety of mental health conditions in MRI and other brain scans of people with schizophrenia, ADHD, bipolar disorder, depressive disorder and Tourette's. Scientists then used scans from those studies to run tests to see if MRI results could lead to accurate mental health diagnoses. At the time of that study, computer-aided diagnostics based on brain scans left something to be desired.
Specifically, the system was able to make almost perfect diagnoses when presented with only two options, such as a choice between two mental health conditions or a choice between one condition and a clean bill of health. When researchers added in a third option, the system did not fare as well.2
More data means better diagnostics
But scientists didn't stop with those studies. Once neuromarkers were identified for various mental health conditions, researchers continued to gather data, make new discoveries and enhance their ability to diagnose mental illness with brain scans.
One reason more data leads to better diagnostics is the way MRI scans measure brain activity and how they can be used to compare differences in mental states. For example, fMRIs can take scans every two to three seconds, creating a time series of brain images. Those images are first run through computer programs to remove differences based on physical movement and other factors, leaving changes in brain activity over time that result only from mental status. Statistical analysis programs are run on the scans to discover what areas in the brain were functioning to a greater or lesser extent than other areas, which can help scientists identify abnormal neuromarkers.3
But those statistical changes can be very, very small, which means a computerized diagnostic system capable of high accuracy in mental health diagnoses would need results from many people, each of who are scanned while performing certain tasks multiple times.
The role of machine learning in MRI-aided mental health diagnostics
Exponential jumps in the capability of machine learning in recent years has increased the relevance of MRIs and other brain scans in mental health diagnostics. Machine learning refers to the ability of computers to accept and analyze data in such a way that they can make decisions and provide relevant feedback.
Machine learning is what drives your Netflix or Amazon recommendations. Based on your viewing, browsing and shopping habits, which the system gathers more data on each time you use it, the computer becomes increasingly adept at placing movies or products in front of you that you're likely to enjoy. In short, computers can now "behave intelligently," and in select circumstances, they can emulate the computations a human brain does to solve a problem.4
Can brain scans and computers exceed human diagnostics?
Psychiatrists and other mental health professionals have been diagnosing mental health disorders based on observation of symptoms and the person in question and comparison of those observations with known facts about specific mental health disorders. Today, researchers are hoping that computers can do the same type of diagnostic work, only eventually more accurately than humans can alone.
Specifically, computers will take brain scans with certain neuromarkers and compare it to known facts about neuromarkers for specific mental health disorders. As it stands today, MRI scans and machine-based diagnostics remain an aid to other tools (such as a clinician's observations). Research published in Front Psychiatry in 2018 notes that MRI capability, machine learning and knowledge of neuromarkers has reached the point that "it is possible to create well-validated neuroimaging biomarkers that augment existing prognostic capabilities."5
MRIs and related technology are becoming increasingly adept at diagnosing mental illness. Currently, magnetic resonance imaging can play an important role alongside the observations of physicians and other mental health care professionals. Individuals who may be facing a mental health issue can talk to their providers about the potential use of MRI in diagnosis.
1. Juri D. Kropotov. "Functional Neuromarkers for Psychiatry: Applications for Diagnosis and Treatment." Elsevier Press. 2016; Pp. xxiii-xxiv.
2. Simon Makin. Can Brain Scans Diagnose Mental Illness? Scientific American. 2013. Web. 20 October 2018. <https://www.scientificamerican.com/article/can-brain-scans-diagnose-mental-illness/>.
3. Kevin Sitek. "Can Computers Use Brain Scans to Diagnose Psychiatric Disorders?" Harvard University Graduate School of Arts and Sciences blog. 2016. Web. 20 October 2018. <http://sitn.hms.harvard.edu/flash/2016/can-computers-use-brain-scans-to-diagnose-psychiatric-disorders/>.
4. Stephen Thornquist. "Intuition in silico: How Ideas From Computational Neuroscience Help Programmers Build Smarter Computers." Harvard University Graduate School of Arts and Sciences blog. 2016. Web. 20 October 2018. <http://sitn.hms.harvard.edu/flash/2016/intuition-in-silico-how-ideas-from-computational-neuroscience-help-programmers-build-smarter-computers-2/>.
5. Lee Jollans & Robert Whelan. "Neuromarkers for Mental Disorders: Harnessing Population Neuroscience." Front Psychiatry. 2018. Web. 20 October 2018. <https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5998767/>.