One common complaint about psychiatry is its subjective nature: it lacks definitive tests for many diseases. So the idea of diagnosing disorders using only brain scans holds great appeal. A paper published recently in PLOS ONE describes such a system, although it was presented only as an initial proof of concept. News reports, however, trumpeted the advent of “objective” psychiatric diagnoses.
The paper used data from several earlier studies, in which researchers outlined key brain regions in MRI scans of people with bipolar disorder, ADHD, schizophrenia or Tourette’s syndrome; people with low or high risk of developing major depressive disorder; and a healthy group. The scans were also labeled with the disorder ordepression risk level of the original study participant.
In the new study, scientists divided the scans randomly into two sets, one to build the diagnostic system and the other to test it. Their software then grouped the scans in the first set by the shape of various regions. Each group was labeled with the most common diagnosis found within it.
During testing, the system analyzed the shapes of brain regions in each test scan and assigned it to the group it most resembled. The scientists checked its work by comparing the new labels on the test scans with the original clinical diagnoses. They repeated the procedure several times with different randomly generated sets. When the system chose between two disorders or one ailment and a clean bill of health, its accuracy was nearly perfect. When deciding among three alternatives, it did much worse.
□ The brain-scan system will probably struggle with patients who have more than one diagnosis. Only chronically ill patients with a single, unambiguous diagnosis were used in the study. In the real world, individuals often cope with confusing lists of symptoms and receive multiple diagnoses, some of which may change over time. Seth Gillihan, assistant professor of psychology at Haverford College, explains, “I’d expect [multiple diagnoses] would be a considerable problem, given the difficulty the system had trying to assign people to one of three [rather than two] categories.” Because these patients would presumably benefit most from an objective method of diagnosis, the clinical usefulness of the system remains questionable.
□ Brain scans may not be able to detect early stages of a disease. In theory, an objective method of psychiatric diagnosis could avoid the suffering and cost that an initial misdiagnosis and subsequent mistreatment might cause. The patients in this study, however, had been ill for an average of more than 10 years, so whether the system can detect a disorder in its early stages is unknown. Gillihan speculates that early diagnosis “is likely to be a harder problem assuming that these conditions change the brain.”
□ The system uses diagnostic categories that might not be biologically valid. Our understanding of brain disorders continues to evolve as genetics reveals how disorders overlap and relate to one another. For instance, some patients exist at the extremes of the distinction between schizophrenia and bipolar disorder, but “in the real world, a very large fraction have shared or overlapping symptoms,” and the two disorders share a majority of their genetic risk factors, according to Steven Hyman, director of the Broad Institute’s Stanley Center for Psychiatric Research in Cambridge, Mass. Hyman thinks we will eventually be able to classify disorders by their underlying biology, but our diagnostic system will likely undergo radical changes once that starts happening.
70 Percentage of children with autism who are diagnosed with another mental disorder, most commonly social anxiety disorder or attention-deficit hyperactivity disorder
50 Percentage of people with schizophrenia who also receive a diagnosis of depression
38 Percentage of women with an anxiety disorder who are also diagnosed with a mood disorder; 30 percent of men are diagnosed with both