AI tool predicts antidepressant response with 97% accuracy using just minutes of brain activity

A new breakthrough in mental health research has shown that artificial intelligence can predict whether a patient will respond to antidepressants with nearly 97% accuracy, using only short segments of resting-state EEG data.

The study, conducted on patients receiving selective serotonin reuptake inhibitors (SSRIs), suggests that AI-guided treatment planning could dramatically improve outcomes for millions living with depression.

⭐ Why this matters

Depression is one of the most complex psychiatric conditions, and up to one-third of patients do not respond to first-line SSRI medications.

Doctors usually rely on trial-and-error — a process that can take weeks or months, often worsening distress or delaying recovery.

A reliable biological predictor could help clinicians:

avoid ineffective medications

personalise treatment faster

reduce side-effects

shorten the time to symptom improvement

This research offers an early but promising path toward precision psychiatry.

⭐ How the study worked

Researchers collected resting-state EEG (electroencephalography) data — a non-invasive measurement of electrical activity in the brain — from:

27 patients with major depression

plus 5 additional patients used as a validation cohort

All patients were treated with standard SSRIs, and their progress was measured using the Hamilton Depression Rating Scale-17 (HAMD-17).

Based on the reduction in scores after treatment, patients were classified into:

drug-effective (responders)

drug-ineffective (non-responders)

A machine learning model was then trained using multidimensional EEG features.

⭐ What the AI model achieved

The AI system was able to:

analyse brief EEG recordings

detect subtle brain-signal patterns

distinguish responders from non-responders

achieve up to 97% prediction accuracy

Researchers say the results demonstrate the potential for objective neural biomarkers to guide antidepressant prescriptions.

⭐ Why EEG works for predicting treatment response

Resting-state EEG captures:

electrical rhythms

connectivity patterns between brain regions

markers of cortical excitability

network-level signals associated with mood regulation

These features change differently in patients who will respond to SSRIs versus those who will not, giving AI a measurable signal to interpret.

Traditional clinical assessments cannot detect such fine-grained neural variations.