Recent research has shown how technologies powered by Artificial Intelligence (AI) have made significant progress in the diagnosis of psychological disorders.
Various AI models have been successfully applied in the diagnosis of psychological disorders and mental illnesses such as schizophrenia, bipolar disorder, post-traumatic stress disorder, autism spectrum disorder, among others. These types of conditions affect more than a billion people worldwide.
The diagnosis of mental illnesses is often complicated due to their heterogeneity in clinical presentation and similar symptoms. Likewise, the practices used for the diagnosis of these diseases are based on the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) and the International Classification of Diseases manual (ICD-11), these tools seem to have been surpassed, since who propose that the diagnosis be based on patient reports and observations and interpretations made by doctors.
Likewise, its diagnosis requires the use of various resources and diagnostic tools such as interviews, as well as the preparation of clinical histories, which takes more time than expected. For this reason, Digital Health and its tools offer a series of new tools and opportunities to support the diagnosis and intervention in mental illness and mental health.
For example, AI is capable of training models that learn rules and offer conclusions about certain information. In mental health, AI has been used, for example, in social bots that support care for dementia and other disorders.
A study published in npj Digital Medicine journal, entitled: “The performance of artificial intelligence-driven technologies in the diagnosis of mental disorders: an overview”, shows the various utilities of AI in the diagnosis of these mental conditions.
“AI has great potential to reshape our understanding of mental disorders and how to diagnose them. Harnessing AI to study and make sense of complex patterns and interactions between genes, the brain, behaviors, and experiences presents an unprecedented opportunity to improve early detection of mental illness and personalize treatment options."
For the review, the authors searched 11 electronic databases, thus identifying 852 citations, of which 344 duplicates were removed and 466 citations were excluded. Finally, for this study, 15 systematic reviews were taken into account, after passing through various exclusion filters as irrelevant data in the participating population, in interventions or in type of publication.
The reviews were published between 2017 and 2020. The reviews focused on the diagnosis of ten different mental illnesses: 7 in Alzheimer's, 6 in mild cognitive impairment, 3 in schizophrenia, 2 in bipolar disorder, 1 in autism spectrum disorder, 1 in obsessive-compulsive disorder, 1 in post-traumatic stress disorder and 1 in psychotic disorders.
The performance in the reviews showed a range from 21 to 100% in the diagnosis of these mental conditions. Seven of the reviews were based on any AI approach, another seven on supervised machine learning, and one on deep learning.
The number of studies retrieved in these reviews ranged from 52 to 7991 and included studies ranged from 12 to 114.
You can check the full review and its results at the following link: https://www.nature.com/articles/s41746-022-00631-8
