Essentials: Compulsive Behaviors & Deep Brain Stimulation | Dr. Casey Halpern

Halpern describes this protocol as a direct translation of epilepsy monitoring, where seizures are provoked and recorded. Here, mood provocation is used because emotional dysregulation is a trigger for eating episodes. A psychiatrist tailors the provocation to each patient’s history, inducing guilt, stress, or sadness that typically leads to a binge. The patient is placed in a “food monitoring unit” with a one‑way mirror; video, audio, and an eye‑tracker synchronize with the brain recording from the implanted lead. This allows the researchers to see exactly what the patient fixates on and the neural activity milliseconds before they take a bite. Halpern emphasizes that even under these conditions, patients still binge, demonstrating the automaticity of the compulsion. The ultimate goal is to train a machine learning classifier on this signal so the DBS device can deliver responsive stimulation to interrupt the cycle.

Halpern explains that in Parkinson’s surgery, neurosurgeons routinely listen for ‘tremor cells’ whose electrical signal, when converted to audio, sounds like the hand shaking. Confirming these cells guarantees they are in the subthalamic nucleus and that stimulation will abolish tremor. He extended this method to craving: in an OCD patient they recorded cells while provoking obsessions and found neurons whose activity correlated with the urge. This not only validates the target but also indicates which subregion to stimulate. The exploration is done carefully, one microelectrode poke at a time, with the patient awake to report symptom changes. The ability to immediately reverse side effects by turning off the electrode makes this mapping safer than a permanent lesion. The approach sets the stage for personalized DBS where the electrode is placed based on a patient’s own disease‑specific electrophysiology.

Halpern describes focused ultrasound as an FDA‑approved, non‑invasive alternative to DBS. He personally performs these procedures and finds them ‘fabulously effective,’ often a ‘miracle’ because there is no incision. Using MRI guidance, the focused ultrasound precisely targets the ventral intermediate nucleus of the thalamus, heating the tissue until it is ablated. The procedure is immediate; tremor disappears during the treatment. However, the main challenge for psychiatric disease is that we don’t yet know the optimal ablation target. Halpern envisions that invasive mapping studies will eventually define targets for OCD or obesity that could then be treated non‑invasively. He also mentions that research is exploring sub‑ablative ultrasound to open the blood‑brain barrier or to modulate neural activity, but those applications are still experimental.

Halpern notes that TMS is already FDA‑approved for depression, OCD, and nicotine addiction. His team is working to turn TMS into a mapping tool: if a patient’s OCD symptoms improve transiently with TMS over a specific prefrontal region, that suggests the underlying circuit is modifiable and an implanted DBS electrode in the connected subcortical target (like the ventral striatum) might produce sustained relief. This mirrors how epilepsy monitoring maps seizure foci before resection. He sees this as a way to increase the responder rate of DBS, ensuring that only patients with a proven responsive circuit undergo surgery. While still under development, he believes neurosurgeons should help refine this non‑invasive mapping to make invasive procedures more precise and justified.

Halpern explains that DBS for psychiatric indications targets the ventral striatum/nucleus accumbens region. The surgery itself is only the first step; the real therapy is the finely tuned electrical stimulation delivered over months of programming. He acknowledges that outcomes are suboptimal, motivating his research into closed‑loop systems and better target identification. He discusses the alternative ablation (capsulotomy) that some centers prefer, noting it can be effective but is irreversible. He favors modulation because it is reversible, but has seen patients benefit from both. The major limitation is access: with only ~200,000 DBS surgeries performed globally, it cannot meet the needs of the estimated 50 million Americans with severe psychiatric conditions, underscoring the need for scalable non‑invasive methods.

Halpern describes the operating room experience: advancing a microelectrode step‑by‑step, converting electrical activity to audio, and listening for disease‑specific signatures. In a single case, they recorded neural activity while provoking OCD‑related distress and found cells whose firing mirrored the obsession. This proof‑of‑concept validates that the brain encodes pathological urges in distinct neuronal populations. The vision is to build a closed‑loop DBS device that detects the craving signal and stimulates only when needed, reducing side effects and improving efficacy. He acknowledges the gap between single‑cell recording and chronic population‑level detection, but argues machine learning can bridge it.

Halpern explains the nucleus accumbens is interconnected with prefrontal and orbitofrontal cortices—often hyperactive in OCD—and the dorsal striatum. In health, the circuit prioritizes adaptive rewards; in pathology, the balance tips toward compulsion. The common thread is ‘urge despite the risk’: drug seeking despite overdose, binge eating despite metabolic harm, checking a lock thirty times despite exhaustion. This is why targeting the same deep brain region can help multiple disorders. He notes that DBS for Parkinson’s sometimes unexpectedly improves comorbid gambling or mood symptoms, offering further evidence. The challenge is that the nucleus accumbens has subregions mediating different aspects of compulsion, so precise targeting is critical.

Unlike epilepsy, where patients are monitored for days to capture spontaneous seizures, psychiatric mapping requires provoking symptoms—mood induction or exposure to OCD triggers—to record neural correlates of the pathological state. By studying patients with comorbid depression and epilepsy, researchers can already validate targets. The long‑term vision is that robust targets identified from invasive studies will inform non‑invasive treatments like focused ultrasound or TMS, making therapy scalable. He emphasizes that the invasive step is surprisingly well tolerated and, because electrodes can be removed or turned off, it is reversible. Aggregating data from many patients could reveal a consistent ultrasound target, reducing the need for lifelong implants.

After identifying craving cells, Halpern’s lab brought patients with implanted DBS electrodes to the lab, provoked binges via mood induction, and recorded brain activity. They trained algorithms to detect the craving signal. The device would listen for this pattern and stimulate only when the risk of losing control is high, effectively reinforcing the brain’s ability to inhibit the behavior. He notes that even under video surveillance, patients aware of being observed still binge, demonstrating the involuntary nature of the urge. This closed‑loop approach could eventually be embedded in a non‑invasive wearable if the signal can be detected externally.

Halpern acknowledges the value of CBT and ERP—he calls them “fabulously helpful”—but points out that when patients with severe OCD stop therapy, many revert to their old behaviors. This is not a criticism of the therapy itself but a recognition of its limitation in the most resistant cases. The patients he sees as a surgeon have already tried medications and behavior management and remain fully aware of their condition, yet still lose control—even in a lab setting designed to make them self‑conscious. Therefore, he argues that a durable, direct neuromodulation is needed to rebuild the brain’s inhibitory capacity so that behavioral strategies can take lasting effect.