For people affected by psychosis and other complex mental health concerns, early intervention can make a dramatic difference in both the substance and quality of their lives, and a new application of artificial intelligence (AI) technology may make consistent early detection, prediction, and intervention for mental health disorders – including psychosis – a reality.
In addition to easing current distress, early access to treatment can reduce the likelihood of co-occurring disorders, irreversible functional impairments, and other types of long-term harm.
Unfortunately, many people live with disruptive symptoms for years before receiving an accurate diagnosis and finding appropriate treatment for psychosis and psychotic disorders.
Mental health experts understand the value of early intervention, before they experience significant harm, but often lack the ability to accurately identify people at greatest risk.
Recent research suggests that this may change – soon.
Studies involving artificial intelligence (AI), machine learning, and other technological advances indicate the ability to predict mental health symptoms related to psychosis and psychotic disorders in people with no current symptoms may be closer than we realize.
What is Psychosis?
Before we delve into the topic of predicting psychosis risk, let’s review what, exactly, psychosis is and isn’t.
First, psychosis is not a mental health disorder. It is a set of symptoms characterized by impairments in how a person perceives their environment and interacts with others. These symptoms are often – but not always – associated with a mental health condition.
Mental illnesses that may include psychotic symptoms:
- Schizophrenia
- Schizoaffective disorder
- Schizophreniform disorder
- Delusional disorder
- Brief psychotic disorder
- Depression
- Bipolar disorder
People can also develop psychosis as a result of various medical conditions and external factors, such as:
- Substance misuse
- Prescription medication
- Alzheimer’s disease
- Parkinson’s disease
- Trauma
- Sleep deprivation
- Traumatic brain injury
- Lead poisoning
- Mercury poisoning
Psychotic symptoms may include:
Hallucinations:
These can involve seeing, hearing, or otherwise sensing phenomena that do not exist. Auditory and visual hallucinations are the most common, but some people also experience tactile (feeling/touch), gustatory (taste), and olfactory (smell) hallucinations.
Delusions:
These are rigidly held beliefs that have no basis in reality that can be easily disproved. Examples include believing that you are in a relationship with someone you’ve never met, being convinced that you have magical powers, or thinking that you’re being sent coded messages through mass media.
Disorganized speech:
This psychotic symptom refers to an inability to effectively convey information via the spoken word. Examples include responding to questions with lengthy non-sequiturs, using words because of their sound or rhyming pattern instead of their meanings, or repeating certain words or phrases.
Grossly disorganized or catatonic behavior:
This category includes actions that fall well outside of cultural norms and expectations. Examples may include dressing in a bizarre fashion, adults or non-children acting in a child-like manner, suddenly becoming agitated for no apparent reason, or staying in odd physical positions for extended periods.
Negative symptoms: Examples of this symptom category include little to no facial expressiveness, limited or nonexistent use of gestures, flat or monotone voice, and a lack of apparent interest in interacting with others.
Using Machine Learning and Brain Scans to Predict Psychosis
As we mention above, the longer a person experiences they symptoms of psychosis without receiving treatment, the greater the risk for lasting harm. This underscores the value of early detection, which experts hope AI and other advanced technologies will improve. The use of AI in the detection of mental health disorders could make a significant difference.
This topic received renewed public attention in February 2024, when the journal Molecular Psychiatry published a study that used AI machine learning and structural magnetic resonance imaging (sMRI) scans to detect elevated risk of mental health symptoms and psychosis.
This study involved brain scans from 2,194 participants. Among these, 1,165 had a clinical high risk (CHR) for psychosis. The research team divided the participants into four categories:
- CHR participants who eventually developed psychotic symptoms
- CHR participants who did not develop psychotic symptoms
- CHR participants whose psychosis status was unknown
- Non-CHR participants (the healthy control group)
Shinsuke Koike, the study’s corresponding author, said in a University of Tokyo press release about the research:
“At most only 30% of clinical high-risk individuals later have overt psychotic symptoms, while the remaining 70% do not. Therefore, clinicians need help to identify those who will go on to have psychotic symptoms using not only subclinical signs, such as changes in thinking, behavior and emotions, but also some biological markers.”
Highlights of the study include:
- The research team trained a machine learning algorithm to identify brain anatomy patterns in the participant sMRI scans.
- Based on these brain patterns, the team directed the algorithm to separate the participants into two groups – CHR participants most likely to develop psychosis and participants with risk minimal of psychosis.
- During the training sessions, the machine learning tool correctly classified 85% of the brain scans.
- In a test involving new data, the algorithm had a 73% success rate of predicting symptoms of psychosis among study participants.
Researchers report the next step is to build an algorithm that can analyze brain scans from different sites and a variety of imaging machines. Success in that area, the research team indicates, would bring them much closer to a tool that clinicians can use to assess psychosis risk. Koike adds:
“If we can do this successfully, we can create more robust classifiers for new data sets, which can then be applied to real-life and routine clinical settings.”
Using AI and Language Analysis to Assess and Predict Mental Heatlh Disorder and Psychosis Risk
Before Koike’s team released the results of their research, the open access journal Schizophrenia published a study that also used artificial intelligence to identify individuals with increased risk of developing symptoms of psychosis.
Conducted by Neguine Rezaii, Elaine Walker, and Phillip Wolff at Emory University in Atlanta, Georgia, the study applied AI analysis to speech samples rather than brain scans.
Rezaii, Walker, and Wolff trained a machine learning tool to assess semantic density, which refers to word combinations and complexity, as well as how often a person talked about voices or sounds. The semantic density training involved a process called vector unpacking, which breaks audio samples into component parts, such as words, sentence structure, and meaning.
In a release about her team’s work, Rezaii said:
“Trying to hear these subtleties in conversations with people is like trying to see microscopic germs with your eyes.”
Once they completed the training, the researchers used the machine learning tool to review audio samples from 40 participants from the North American Prodrome Longitudinal Study (NAPLS). This effort produced the following results:
- The algorithm had a 93% success rate at predicting symptoms of psychosis within a two-year follow-up window.
- Audio samples that showed low semantic density (or poverty of content) and increased focus on voices and sounds indicated an elevated risk of psychosis.
- Low semantic density was associated with negative symptoms of psychosis, while increased talk about voices or sounds indicated increased likelihood of positive symptoms, such as hallucinations, delusions, disorganized speech, and grossly disorganized or catatonic behavior.
- Semantic density analysis demonstrated that the way a person structures words into sentences may be as important as which words they use.
Here’s how the research team described their results:
“Our approach included the development of a mathematical algorithm for unpacking the meaning components of a sentence as well as a computational pipeline for identifying the kinds of thought content that are potentially diagnostic of mental illness. Finally, we showed how the linguistic indicators of mental health, semantic density and talk about voices, could predict the onset of psychosis at high levels of accuracy.”
How Will Research on AI and Mental Health Benefit Patients?
The research we discuss in this article hasn’t yielded tangible changes in how we identify and treat psychosis – yet. Therefore, predictions about their potential benefits are at best educated guesses.
With that caveat in mind, research into the use of AI, machine learning, and other advanced technologies suggests the following possible developments:
- AI brain scan analyses and language assessments may improve our ability to identify patients at greatest risk of developing mental health disorders with psychotic symptoms later in their lives. Reducing the time between symptom onset and treatment delivery can minimize risk of long-term harm.
- Language-focused research predicted psychosis onset and provided researchers with insights into which symptoms a person is more likely to experience and how those symptoms may affect them. The more a clinician understands specific needs, the better they can develop a focused treatment plan.
- Clinicians may be able to combine the findings from language learning, brain scans, and related studies to construct a comprehensive model for identifying and treating patients with elevated risk of psychosis.
Find Treatment for Psychosis Today
If you, a family member, or loved one experiences hallucinations, delusions, or other psychotic symptoms, you don’t have to wait for AI advances in diagnosis to get help.
Crownview Psychiatric Institute is a trusted provider of comprehensive, personalized care for adults with psychotic disorders and other complex mental health conditions. Our treatment center in Southern California is a safe and supportive environment where patients receive personalized care from a team of skilled and compassionate professionals.
Treatment at CPI is a dynamic experience that incorporates multiple evidence-based therapies, innovative adjunct services, life- and work-skills education, and true wraparound support, all in a welcoming community-like environment.
To learn more about us or to discuss how we can help, please visit our Contact page or call us today.
Gianna Melendez
Jodie Dahl, CpHT