AI Shows Promise for Early Schizophrenia Diagnosis Through Speech Analysis
Research is underway to utilize artificial intelligence for improving the diagnosis and monitoring of schizophrenia, a condition that impacts approximately 23 million individuals worldwide, typically identified in late adolescence to early adulthood. Traditional diagnostic methods depend on subjective evaluations, often requiring about 1.5 years after symptoms first appear. In the Netherlands, researchers developed an AI system that analyzed 88 vocal characteristics from diagnosed patients, achieving an accuracy rate of 86.2% in differentiating schizophrenia from healthy individuals. At the Feinstein Institutes, psychiatrist Sunny Tang's machine learning approach assessed speech content, attaining 87% accuracy, compared to 68% for human evaluators. Thomas Insel emphasizes the potential for precise remote monitoring by 2030, though challenges like data diversity and ethical issues remain. Experts such as Vijay Mittal caution against viewing AI as a comprehensive solution.
Key facts
- Schizophrenia affects about 23 million people worldwide and is usually diagnosed between the late teens and early 30s.
- Americans with psychotic disorders receive a diagnosis an average of one and a half years after first symptoms appear.
- Dutch researchers used AI to analyze 88 vocal features and achieved 86.2 percent accuracy in distinguishing schizophrenia patients from healthy controls.
- Sunny Tang's AI model analyzed speech content and reached 87 percent accuracy, while clinical raters without AI were only 68 percent accurate.
- Thomas Insel led the U.S. National Institute of Mental Health for 13 years and founded several mental health startups.
- A survey of AI psychiatric diagnosis studies found many use small, unrepresentative samples, according to Jeffrey Girard of the University of Kansas.
- Sandra Just of UiT the Arctic University of Norway notes that speech can slow due to age, second language, stress, medication, or physical illness.
- Tang hopes to have a tool ready for clinical trials in 2030.
Entities
Artists
- Thomas Insel
- Alban Voppel
- Sunny Tang
- Jeffrey Girard
- Sandra Just
- John Torous
- Brita Elvevåg
- Vijay Mittal
- Bryan Charnley
Institutions
- Knowable Magazine
- U.S. National Institute of Mental Health
- McGill University
- Feinstein Institutes for Medical Research
- University of Kansas
- UiT the Arctic University of Norway
- Harvard's Beth Israel Deaconess Medical Center
- National Alliance on Mental Illness
- Northwestern University
- Annual Review of Clinical Psychology
- Wellcome Collection
- Smithsonian Magazine
Locations
- Netherlands
- Montreal
- Canada
- New York City
- Ontario
- Norway
- United States