JOVANA
Explore Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

Why Psychiatry Needs a Closed Loop

Depression and OCD are circuit diseases with no reliable read-out — and the first generation of open-loop brain stimulation showed exactly why sensing the brain's state may be the missing piece.

The unmet need that justifies opening a skull

You reach for an implant only when everything gentler has failed. In major depression, roughly a third of patients remain ill after adequate trials of several drugs and psychotherapy — the state clinicians call treatment-resistant depression. A similar tail of severe, disabling obsessive-compulsive disorder resists every front-line therapy. For that tail, deep brain stimulation is offered not as an enhancement but as a therapy of last resort, and OCD DBS already carries a humanitarian-device authorisation. This is the population psychiatric BCI serves: not the average patient, but the one for whom nothing else worked.

The first generation, and its honest setbacks

Open-label pilot studies of DBS for depression were genuinely exciting. Stimulating the subcallosal cingulate (Area 25) — a hub over-active in depression — or the ventral capsule / ventral striatum produced striking recoveries in some treatment-resistant patients. It looked like psychiatry had found its pacemaker.

Then the blinded, randomised pivotal trials fell short. The two large industry-sponsored studies — one targeting the subcallosal cingulate, one targeting the ventral capsule/ventral striatum — did not beat sham on their primary endpoints and were stopped early for futility. The lesson was not that DBS does nothing; it was that a fixed, open-loop dose delivered to an anatomically-fixed target, evaluated against the huge expectation effects of brain surgery, is not reliably distinguishable from placebo. That single sentence motivates almost everything in this track.

Why reading a mood is uniquely hard

Motor BCI has an enormous advantage psychiatry lacks: a fast, objective, second-by-second ground truth — the cursor either hit the target or it did not. Mood has none of this. A neural signature of mood must be learned against labels that are sparse (a few self-reports a day), subjective, slow (mood evolves over hours to days), and non-stationary (today's biomarker may drift by next month). On top of that, the read/write asymmetry bites hard: even if you could read the state, the stimulation you write back changes it, so cause and effect are tangled inside the loop.

\text{open loop: } u_t = u_0 \quad\Longrightarrow\quad \text{closed loop: } u_t = \pi\!\left(b_t\right), \;\; b_t = f_\theta\!\left(y_t\right)

The whole shift in one line. Open-loop DBS delivers a constant dose u₀. Closed-loop DBS extracts a biomarker bₜ from the neural signal yₜ via a decoder f_θ, then a policy π sets the stimulation uₜ from that state. Everything downstream is engineering this map honestly.

The shape of the frontier

The closed-loop psychiatric loop this whole track builds toward: sense a mood/network biomarker, decide with a policy, and stimulate a network node — then let the effect feed back into the next reading.

The frontier therefore has two coupled halves, and the remaining guides follow them: where to intervene (networks, not just nuclei), and when and how much to intervene (adaptive, biomarker-driven stimulation). Wrapping both is a move from population-average protocols to per-patient personalisation — because in a disease with no shared read-out, the biomarker and the target may both be individual. That is the honest promise of closed-loop DBS for depression: not magic, but a way to make the intervention answerable to the brain it is treating.