Can artificial intelligence (AI) machine learning provide clinicians with predictions of patients who will develop psychosis? In a new European study released last month in Jama Psychiatry led by Dr. Nikolaos Koutsouleris at the Max-Planck Institute of Psychiatry in Munich, Germany, researchers used AI machine learning and human intelligence to predict mental illness.
“Our study showed for the first time, to our knowledge, that the augmentation of human prognostic abilities with algorithmic pattern recognition improves prognostic accuracy to margins that likely justify the clinical implementation of cybernetic decision-support tools,” wrote the researchers.
Psychosis is a treatable condition with many causes that affects how the brain processes information. It is characterized by a loss of connection with reality. Psychosis is a symptom of health issues such as schizophrenia and bipolar disorder, which are considered primary psychotic illnesses. Secondary psychosis is the term used to describe when a person experiences a disconnect from reality for triggers other than primary psychotic illnesses.
Psychosis may arise from sleep deprivation, medical conditions, trauma, stress, or substance abuse of alcohol and drugs. Psychosis may accompany anxiety, depression, insomnia, and social isolation. Symptoms of psychosis may include hallucinations, delusions, nonsensical speech, and inappropriate behaviors.
Roughly three out of 100 Americans will experience psychosis during their lives, and each year 100,000 young American adults and adolescents experience their first psychotic episode according to the National Institute of Mental Health (NIH).
The global antipsychotic drugs market will grow at a CAGR of 4.1 percent between 2020 and 2027 and is projected to reach USD 21.8 billion by 2027, according to a September 2020 report by Research and Markets. Last year, the U.S. had more than 28.8 percent market share of the global antipsychotic drug market, per the same report.
For this study, the researchers used NeuroMiner, a machine learning software available on GitHub, to develop “a sequential prognostic algorithm that identified the optimal sequence of predictive components to be combined into a stacked model.”
To find the optimal set of predict features for the risk calculators, the researchers used the Support Vector Machine (SVM) provided by the Liblinear library in NeuroMiner.
Support Vector Machines are supervised machine learning models that are used to find discernible patterns in complex datasets. It is a machine learning method that has a relatively reduced risk of overfitting high-dimensional imaging data such as neuroimaging. With its flexible method for handling classification, Support Vector Machines is an AI machine learning method particularly well-suited for neuroscience research for precision psychiatry for AI prediction of depression, Alzheimer’s disease, and schizophrenia.
“In this prognostic study of 334 patients and 334 control individuals, machine learning models sequentially combining clinical and biological data with clinicians’ estimates correctly predicted disease transitions in 85.9 percent of cases across geographically distinct patient populations,” reported the researchers.
The intent of the researchers was to provide an early intervention tool to help medical professionals in determining which patients require therapeutic interventions. Now there is a proof-of-concept for a potential new AI-assisted method for early detection of patient psychosis that may one day lead to improved mental health treatment in the future.
Copyright © 2021 Cami Rosso All rights reserved.
Credit: Google News