AI Tool Matches Psychiatrist-Level Accuracy in Mental Health Evaluations

AI Tool Achieves Breakthrough in Mental Health Evaluations

Researchers at UTHealth Houston, in collaboration with Yale University, have developed an AI tool capable of conducting mental health evaluations with psychiatrist-level accuracy. Published in npj Mental Health Research, the system demonstrates how artificial intelligence can supplement clinical psychiatric work by analyzing patient behavior, speech, and tone with remarkable precision1

1

2

2

.

The AI tool combines pretrained neural networks on Qwen3-Omni with custom software to analyze video recordings of patient interactions. "The AI tool we built can perform a diagnostic evaluation on a patient that is almost as good as a team of psychiatrists. It's very impressive," said co-first author Dr. Hammza Hamoudi, postdoctoral research fellow at McGovern Medical School1

1

. The system generates written explanations for its mental status assessments by bringing together observations across multiple visits.

Testing Against Psychiatric Teams Using Standardized Patient Cases

The evaluation process involved standardized patients portraying schizophrenia, obsessive-compulsive disorder, and bipolar disorder at varying severity levels. Both the AI model and psychiatric teams from UTHealth Houston and Yale classified patients across 10 criteria: mood, appearance, behavior and cooperation, perceptions, speech, suicidality, presence of delusions, obsessions or compulsions, and coherence and speed of thought processes2

2

.

While the AI tool made mistakes on individual criteria like patient appearance or fine motor movements, its overall diagnostic accuracy remained consistently strong. This performance gap highlights an important distinction in how the system processes information compared to human clinicians.

Designed as Supplement for Clinicians, Not Replacement

Senior author Dr. Cesar Soutullo emphasized that the goal is augmentation rather than replacement. "The point isn't, 'Is this AI as good as a psychiatrist with 30 years of experience?' The point is that both of them are doing different things, and they have strengths and weaknesses on both sides. Why pick one? You can use both," said Soutullo, who holds a John S. Dunn Professorship at McGovern Medical School1

1

.

The team envisions the system serving as a supplement for clinicians in rural areas where psychiatric expertise may be limited. "Just imagine you send this to a rural pediatrician who is seeing patients by themselves, and they can get this as an enhancement of their training," Soutullo explained. "That could be really good, because we could potentially detect symptoms that otherwise could be missed by clinicians who are not fully trained"2

2

.

Educational Settings Benefit From AI's Clinical Reasoning Gaps

Co-first author Dr. Benson Mwangi Irungu noted that the AI tool excelled at observations but struggled with domains requiring clinical reasoning that psychiatrists perform naturally. Rather than viewing this as a limitation, the research team sees educational potential in these gaps. "Even the AI's mistakes could become teaching tools. Students could compare its assessments with those of experienced clinicians, examine where the reasoning went wrong, and learn how to avoid similar errors in their own clinical assessments," Irungu said1

1

.

The next phase involves training the AI model to improve accuracy at classifying individual criteria. The research received funding from the Texas Child Mental Health Care Consortium's New and Emerging Children's Mental Health Researchers Initiative and the National Institutes of Health's Artificial Intelligence/Machine Learning Consortium to Advance Health Equity2

2

. This positions the technology for potential deployment in both educational settings and clinical practice, where it could address critical gaps in mental health care access while serving as teaching tools for the next generation of clinicians.