Psychiatrists have always lacked what their colleagues in cardiology and oncology have relied on for decades: a biological test to help confirm diagnoses. Clinicians have to diagnose conditions such as schizophrenia (SCZ) and bipolar disorder (BPD) – for the most part – through patient interviews and observed behavior. Even then, treatments aren’t foolproof. And clinicians must move forward without any objective biomarkers to guide their way.
But new research reveals that all of that might be about to change.
Researchers from Johns Hopkins and Harvard universities – among others – have developed a machine-learning pipeline that identifies distinct electrical patterns in patient-derived brain cell models. By analyzing neural activity in induced pluripotent stem cell (iPSC)–derived cultures, the team achieved classification accuracies as high as 96% for schizophrenia and more than 91% when distinguishing schizophrenia and bipolar disorder from healthy controls.
The work, appearing in APL Bioengineering, represents a huge step toward objective, physiology-based diagnostics for neuropsychiatric conditions.
Why It Matters
“Schizophrenia and bipolar disorder are very hard to diagnose because no particular part of the brain goes off. No specific enzymes are going off like in Parkinson’s,” research lead and Johns Hopkins biomedical engineer Annie Kathuria explained. “Our hope is that in the future we can not only confirm a patient is schizophrenic or bipolar from brain organoids, but that we can also start testing drugs on the organoids to find out what drug concentrations might help them get to a healthy state.”
Postmortem studies have revealed some consistent hallmarks, such as deficits in GABAergic interneurons in schizophrenia, glial cell reductions in bipolar disorder, and widespread synaptic abnormalities. But revelations like these are static and don’t offer enough insight into how neural circuits misfire in living brains.
That’s where iPSC models come in. By reprogramming skin cells from patients into stem cells and coaxing them into forming cerebral organoids, or cortical neuron cultures, scientists can watch disease-related neural dynamics unfold in real time. That way, they can use the patient’s own genetic background as the blueprint.
A Digital Window Into Neural Networks
The Johns Hopkins team recorded activity from two kinds of iPSC-derived systems:
- Three-dimensional cerebral organoids (COs), which mimic early brain architecture, and
- Two-dimensional cultures of cortical interneurons (2DNs).
The researchers grew both from cells donated by patients with schizophrenia, patients with bipolar disorder, and a healthy control group.
They then captured the neural activity with the use of multi-electrode arrays (MEAs), which measure the tiny voltage fluctuations of firing neurons across 16 channels. To make sense of the overwhelming amount of raw data, the researchers built what they call a digital analysis pipeline (DAP).
This pipeline borrowed techniques from electroencephalography (EEG) while taking it a step further. Using a stimulus–response dynamic network model, the team measured how signals flowed through the networks, paying special attention to “sink” nodes – neurons that receive more input than they send. Those sink dynamics became the features fed into a support vector machine (SVM), a machine-learning classifier adept at parsing high-dimensional data.
The Power of Stimulation
The role that electrical stimulation played surprised the scientists. When the neurons were at rest, scientists detected some, albeit modest, differences between the healthy and diseased one.
But when they applied brief electrical pulses, the distinctions sharpened dramatically.
In two-dimensional cultures, schizophrenia samples were separated from controls with 95.8% accuracy under stimulation. In organoids, classification improved from 83% at baseline to 91.6% with stimulation, enabling robust discrimination among control, schizophrenia, and bipolar cohorts.
Outperforming Human Judgment
To put the results in perspective, the researchers compared their algorithm’s accuracy to that of proven clinical interviews. Structured diagnostic interviews typically achieve around 80% agreement among psychiatrists for schizophrenia and less than 60% when distinguishing schizophrenia from bipolar disorder.
The machine-learning system outperformed both of those benchmarks.
“At least molecularly, we can check what goes wrong when we are making these brains in a dish and distinguish between organoids from a healthy person, a schizophrenia patient, or a bipolar patient based on these electrophysiology signatures,” Kathuria added. “We track the electrical signals produced by neurons during development, comparing them to organoids from patients without these mental health disorders.”
The study’s authors concede that future work must leverage larger and more diverse cohorts, refined protocols to cut down on variability, and comparisons with patient imaging and behavioral data. If validated, such biomarkers could not only aid diagnosis but also help monitor treatment response and accelerate drug development.
A Broader Shift in Psychiatry
The study reflects a growing movement to ground psychiatry in objective measures of brain function. Already, EEG and fMRI studies have identified potential endophenotypes – such as reduced auditory mismatch negativity in schizophrenia – but their clinical utility remains limited.
By bridging stem-cell biology, electrophysiology, and machine learning, the Johns Hopkins–Harvard collaboration offers a fresh approach. It also aligns with the network-based view of psychiatric illness. Disorders can’t be traced back to single chemical imbalances, but from garbled communications across neural circuits.
The implications are huge. Reliable, objective biomarkers could accelerate diagnoses, curb misdiagnoses, and better inform clinicians in tailoring treatments. They also could serve as endpoints in future clinical trials, speeding the development of novel therapies.
But maybe importantly, they promise a shift toward precision psychiatry, where treatment is tailored not just to a diagnosis but to an individual’s unique neural profile.
Further Reading
Semaglutide Injections Boost Metabolic Health in Schizophrenia
Could Hepatitis C Be Hiding A Schizophrenia Connection?
Global Experts Unveil New Schizophrenia Treatment Guidelines
