Artificial intelligence is going to become a great diagnostics tool in medicine. It can notice details in images that go unnoticed for a human eye. However, up until now it was almost impossible to imagine AI being used to diagnose mental conditions.
Now scientists at the University of Birmingham have developed a new AI-based diagnostics method, which can identify patients with many psychotic and depressive symptoms more accurately.
A lot of people experience symptoms of both depression and psychosis. Image credit: Сања Малохоџиќ via Wikimedia (CC BY-SA 4.0)
There are many people who are suffering from symptoms of both psychosis and depression. This makes accurate diagnosis very difficult. In many cases doctors would identify one ‘primary’ and one ‘secondary’ illness, which sometimes may lead to insufficient treatments.
Depression is usually viewed as that secondary illness, which means that treatment focuses on symptoms of psychosis (such as hallucinations or delusions). Scientists have figured that this is a very inefficient approach and treating both conditions systematically would be better, but diagnosis would be a huge challenge.
Now scientists explored the possibility of using machine learning to develop accurate models of pure diseases to see how they correlate. This could help clinicians diagnose people more accurately, because sometimes symptoms don’t mean that there is that ‘secondary’ disease, but instead there is just a different manifestation of the primary condition.
This new study involved 300 patients. Scientists used machine learning to develop models of ‘pure’ depression, and ‘pure’ psychosis. Scientists found that diagnoses of depression as the primary illness were more likely to be accurate than the ones where psychosis was considered the main condition.
Paris Alexandros Lalousis, lead author of the study, said: “There is a pressing need for better treatments for psychosis and depression, conditions which constitute a major mental health challenge worldwide. Our study highlights the need for clinicians to understand better the complex neurobiology of these conditions, and the role of ‘co-morbid’ symptoms; in particular considering carefully the role that depression is playing in the illness”.
AI-based models could help create better, more accurate diagnosis. And the same process that was used for these models could be introduced for other conditions as well. The result would be people getting treated for conditions they actually have, which would lead to better outcomes.
Individual experiences and neurobiology are very important for accurate diagnosis, but they are also very difficult to include in a systematic method. AI should help solving this problem. In fact, it is likely that AI-based approaches are going to be very common in medical diagnostics in the very near future.
Source: University of Birmingham