By combining cerebral organoids (human cells programmed to reproduce certain characteristics of the brain) with artificial intelligence algorithms, researchers are proposing a new approach that would facilitate the diagnosis of schizophrenia and bipolar disorders.
Diagnosing schizophrenia and bipolar disorder remains a major challenge in psychiatry today. In the absence of reliable biological biomarkers, clinicians rely mainly on observation of symptoms and clinical interviews, with a risk of errors and confusion between pathologies. A team of researchers from Johns Hopkins University (Baltimore, United States) is proposing an innovative approach: combining cerebral organoids (see below), sometimes simplified into “mini-brains” – cellular models grown in the laboratory that reproduce certain characteristics of human brain tissue – with artificial intelligence algorithms to identify neuronal signatures specific to each disorder. Ultimately, this approach could complement current clinical tools, providing support for diagnosis.
Schizophrenia and bipolarity, disorders difficult to diagnose
Today, no biological examination, no blood test or no imaging test makes it possible to establish with certainty a diagnostic of schizophrenia or bipolar disorder. These pathologies are based on a clinical assessment of symptoms, who may be close to others neurological or psychiatric diseases. This complexity explains frequent diagnostic errors.
Cerebral organoids: 3D cellular models of the human brain
To overcome these limitations, researchers are turning to cerebral organoids. Cultivated in the laboratory from reprogrammed human cells (e.g. from skin or blood), these three-dimensional structures reproduce certain characteristics of the brainsuch as the organization of neural networks and electrical activity. However, they constitute neither a complete organ nor a functional brain.
In this study, scientists generated organoids mimicking the cell composition of the prefrontal cortexa key brain region involved in planning, decision-making and regulation of behavior. These organoids were produced from cells from people with schizophrenia, bipolar disorderbut also people without psychiatric disorders, serving as a control group, in order to compare the results.
When artificial intelligence “listens” to neuronal activity
The organoids were recorded using microelectrodes capable of capture their electrical activity. These signals were then analyzed by machine learning algorithms (machine learning), capable of recognizing complex patterns invisible to the human eye.
Result: the researchers identified distinct electrical signatures between healthy organoids and those derived from patients suffering from schizophrenia or from bipolar disorder. These differences become even more marked when the neural networks are stimulated by electrical impulses, mimicking a strong demand on the brain.
Artificial intelligence thus achieves distinguish organoids from schizophrenia patients from those of healthy peoplebased solely on their electrophysiological signatures. Under some experimental conditions, the classification accuracy reaches more than 92%.
Towards neural signatures of psychiatric illnesses
These results suggest the existence of real “ signatures neuronales » specific to each disorder. Ultimately, this approach could complement current clinical tools, providing support for diagnosis.
Beyond diagnosis, these organoids open the way to new form of personalized medicine : it becomes possible
the of test different treatments directly on the organoids from a patient’s cells, in order to observe their effect on neuronal activity, without risk for the person. The objective is no longer just to treat symptoms, but tailor prescriptions to individual biological responses.
A major challenge for the neurosciences and psychiatry of tomorrow : personalized medicine
This research illustrates an evolution in neuroscience: the transition from psychiatry based solely on the observation of behavior to an approach integrating biological and computational data. By combining cerebral organoids, artificial intelligence and neuroscienceresearchers are developing tools capable of linking the functioning of neural networks to psychological disorders. Ultimately, these approaches could transform the management of psychiatric and neurodevelopmental illnesses, by enabling more reliable diagnostics, better targeted treatments and more personalized medicine.
Article from the Brain Research Foundation – written by Océane Delvarre, March 12, 2026
