Taipei hospital touts world-first AI platform to assist schizophrenia diagnosis

Taipei, July 2 (CNA) Taipei Veterans General Hospital (TVGH) on Wednesday touted its world-first AI-powered platform designed to assist in the diagnosis of schizophrenia, saying it is over 90 percent accurate and capable of transcending the limits of traditional symptom-based assessments.

"The field of psychiatry has long hoped to identify objective biological markers that can help quantify the symptoms [of mental illness]," said Albert Yang (楊智傑), deputy director of TVGH's Medical AI Development Center.

At a press conference at the Taipei hospital, Yang said that due to the lack of objective diagnostic tools, psychiatric evaluations have long relied primarily on clinical observations and subjective interviews, making it difficult for physicians to fully understand a patient's condition and changes in brain function.

With that challenge in mind, he led the team that developed BrainProbe, the world's first AI-powered platform capable of assisting in the diagnosis of schizophrenia, a severe mental illness linked to neurotransmission abnormalities and degeneration in brain structure and function.

Based on data from over 1,500 local participants recorded since 2012 -- including both healthy individuals and those with schizophrenia -- BrainProbe is able to quantify brain abnormalities related to schizophrenia using MRI scans and deep learning algorithms, with its earliest version employed in 2019 to help diagnose patients, according to Yang.

As an example of how the platform aids in the diagnosis of schizophrenia -- which affects about 1 percent of the global population -- Yang cited the case of a 30-year-old patient who experiencing auditory hallucinations and paranoid delusions sought treatment at TVGH for suspected mental illness.

"BrainProbe was able to detect signs of degeneration in his brain function and structure -- particularly in deeper regions such as the insula and temporal lobe," Yang said, adding that abnormalities associated with schizophrenia prompted further evaluation, and the patient was later confirmed to have the disease.

In addition, Yang said that the "most important capability" of BrainProbe -- which now has an accuracy rate of 91.7 percent for diagnosing schizophrenia -- is its ability to track changes in the brain as it ages.

The platform has established a brain aging prediction index and a mechanism for monitoring pathological changes in brain structure and function, he added.

Regarding its use in Taiwan, Yang said BrainProbe -- which won gold in the Advancements in Neurological Treatments category at the 2025 Edison Awards -- is still under review by the Taiwan Food and Drug Administration, but patients can currently access the platform at TVGH on a self-pay basis through a clinical trial program.

Asked by CNA about the future development of BrainProbe, Yang said his team is currently collaborating with medical institutions abroad to incorporate data from populations beyond Taiwan.

"We hope this platform can be applied across different ethnic groups to enable more accurate research," Yang said.