Neuropsychiatric disorders have a profound impact on the mental health of approximately 970 million individuals, resulting in healthcare costs reaching into the trillions (1). Despite the availability of therapeutic and pharmaceutical treatments, only a fraction of patients respond positively to these conventional interventions. Standard of care achieves remission in fewer than half of patients with obsessive-compulsive disorder (OCD) (2), and a mere 30% of individuals grappling with depression attain remission (3, 4). These challenges are further compounded by issues such as adverse drug effects (5), logistical complexities in administering treatments—especially with psychotherapeutic options—and suboptimal treatment adherence (6). While symptoms are remarkably heterogeneous, treatments are often delivered based on the diagnosis, rather than the presenting functional symptoms. As neuropsychiatric diseases are increasingly understood to be disorders of dysfunctional neural circuits, there arises a need to understand the patient’s unique circuit-level pathophysiology, particularly for treatment-resistant individuals (7).

Among various alternatives, neural circuitry modulation through techniques like transcranial electrical or magnetic stimulation (tES/TMS) has emerged as a promising treatment for neuropsychiatric disorders (8–12). These interventions often target neural network dysfunction associated with neuropsychiatric conditions, potentially alleviating corresponding symptoms, such as in depression and OCD (13). However, transcranial approaches encounter inherent constraints. For example, the application of transcranial techniques restricts access to deep neural structures intricately associated with neuropsychiatric disorders, including the amygdala-hippocampus complex and nucleus accumbens (14). Further technical challenges, such as the complexity of simultaneously recording electroencephalography (EEG) and administering tES/TMS, present obstacles. Although office-based administration is effective (11), this setting is not always the most conducive, given the inaccessibility for some patients, the temporal variation in symptoms, and the inadequacy of accommodating diverse environments. This notably impacts patients whose treatment necessitates addressing specific exposures and contextual factors, such as individuals suffering from OCD and addiction.

To mitigate these issues, recent research has explored two novel approaches. First, to better target relevant neuropsychiatric structures, recent studies have attempted to obtain direct electrophysiology using implanted intracranial electrodes (15–21). Intracranial EEG (iEEG) captures neural circuits in high spatiotemporal detail (22), available both on the brain’s surface with subdural electrodes or directly into brain structures using depth electrodes (23). This approach is commonly used as a part of surgical epilepsy evaluation in those with drug-resistant epilepsy and has more recently been applied to understanding the neural correlates of neuropsychiatric diseases, such as depression (24), OCD (25), and psychosis (26).

Second, to translate these in-lab findings into more flexible at-home treatments, recent research has further explored chronically implanted closed-loop neurostimulation devices such as Responsive Neurostimulation (RNS) or adaptive deep brain stimulation (DBS) to achieve closed-loop stimulation. As opposed to an open-loop system where stimulation is delivered at a predetermined interval irrespective of neurophysiological state, closed-loop stimulation involves the detection of biomarkers related to symptom severity and the subsequent delivery of time-specific stimulation in response, within milliseconds (27). Responsive neurostimulation refers to the delivery of stimulation for a fixed duration after a triggering event, while adaptive neurostimulation involves adjustment of therapeutic parameters based upon changes to the neural signals (28). These options make it possible for patients to engage in regular life activities in an at-home setting while receiving treatment benefits, as well as paralleling treatment to the natural variations in symptom severity.

Despite great potential, many unknowns persist regarding the mechanisms, efficacy, and long-term effects of these intracranial neuromodulation effects in neuropsychiatric conditions. Therefore, to systematically map the existing literature and to better understand treatment opportunities for researchers and clinicians, we conducted a review of recent clinical studies employing intracranial, closed-loop neurostimulation as a novel treatment for neuropsychiatric diseases. Our objectives include: (1) reporting and summarizing the approaches to biomarker discovery; (2) reporting and summarizing the neurostimulation protocols and preliminary treatment outcomes; and (3) providing recommendations for consideration for future studies.

Our search yielded a total of 583 articles. After review for duplicates and relevancy, 46 articles were identified, of which 5 were included in this review. The majority of the studies excluded at full review did not apply closed-loop neurostimulation. The included studies, published between 2021 and 2024, had sample sizes ranging from a single case report to 5 participants, all with treatment-resistant neuropsychiatric diseases. Many of the patients had attendant neuropsychiatric and neurological comorbidities. Two articles investigated OCD, one article focused on depression, one article studied PTSD, and another studied binge eating disorder. Each of the studies followed a similar paradigm of intracranial electrophysiological biomarker detection by recording during various symptomatic states and through different settings, including home recordings with self-report symptoms to provide ecological validity, through provocatory behavioral tasks, or therapy sessions (depicted in Figure 2 ). For instance, Provenza and colleagues obtained recordings from self-report logging during symptomatic states, during exposure and response therapy sessions, as well as during behavioral tasks aimed at provoking an obsessive state (32). Details from each study are shared in Tables 2 , 3 . One study took an additional step to contextualize the biomarker and response to stimulation within the underlying structural and functional connectivity of the subnetwork involved (33).

In this review, we examined 5 pioneering studies that integrate intracranial recording and closed-loop stimulation for treating neuropsychiatric conditions. Overall, biomarker identification and stimulation methodologies represented a consistent paradigm across studies. However, our findings indicate a paucity of research focusing specifically on closed-loop approaches. Despite three authors involved in paper selection - one performing the search, two selecting the articles, and one confirming the relevance - potential misses remain possible. The small number of studies limits our ability to make definitive conclusions about the broader applicability and efficacy of these interventions across diverse populations and conditions. Regardless of lingering uncertainties, the potential treatment benefits observed in these 11 participants across these 5 studies underscores the urgent need for further elucidating the mechanisms, treatment efficacy, and long-term viability of this treatment approach.

First, of primary interest, further research should seek to understand the effects of intracranial stimulation in the context of the neuropsychiatric circuitry. Despite numerous past studies utilizing functional neuroimaging, the direct link between brain structures and neuropsychiatric symptoms has remained limited (47, 48). Consequently, many neuropsychiatric symptoms have been reconceptualized as dysfunctions within specific functional networks (49). As such, research that employs preliminary mapping of biomarkers and potential therapeutic sites across broader circuits may enhance neural circuit modulation efficacy. For instance, Scangos and colleagues implanted depth iEEG electrode probes throughout various sites within the emotion circuitry of their patient. Among the electrode locations, they identified amygdala activity as the strongest predictor of depression symptoms. They then performed stimulus-response mapping and discovered that stimulation of the VC/VS led to consistent and sustained symptomatic improvement. To support these findings, they performed evoked potential mapping and found that the VC/VS and amygdala constitute a structurally and functionally connected subnetwork, with the VC/VS influencing numerous distant brain regions (33). This suggests the importance of the VC/VS in modulating the associated network at large, given its significant connections and effectiveness as the optimal site for stimulation (33). This network approach suggests that the success of neuromodulatory treatments hinges on their effects on the entire network (50). Therefore, exploring the broader network as a stimulation target based on preliminary mapping may be crucial for future successes in utilizing intracranial stimulation as a treatment for neuropsychiatric diseases.

Second, future research may also consider complementary approaches to interrogate the heterogeneous neuropsychiatric circuits for effective neuromodulation. Neuropsychiatric symptoms are frequently comorbid across various diseases and widely vary. As such, the Research Domain Criteria (RDoC) was developed as a framework for researchers to explore neuropsychiatric diseases based on broad neurobehavioral functioning domains, rather than operationally defined based on symptoms (51). One key aspect is the characterization of neural circuits that drive specific functional problems (48). In addition to identifying electrophysiological features time-locked to a state of high symptom severity, future research may benefit from probing explicit behavioral tasks that arise from a given functional circuit (1, 52, 53). For instance, dysphoria can be interrogated by electrophysiological responses to emotional imagery tasks, while response inhibition - variably impacted across neuropsychiatric disorders including depression, anxiety, and PTSD - could be examined using the Go/NoGo task (52, 54). In the context of neuromodulation, this approach offers a way to precisely characterize, target, and modulate these behavioral circuits directly. Additionally, modulating the behaviors that arise from these neural circuits may address symptoms common to multiple diseases (55). Bolstering the RDoC framework for closed-loop neuromodulation may provide additional insights into behavior across various disorders, contributing to a better understanding of trans-diagnostic behavioral circuits.

Third, evaluating the long-term viability of this intracranial approach as a treatment for neuropsychiatric conditions is crucial. While some intracranial electroencephalography (iEEG) implants remain in the patient’s brain for an extended period (e.g., RNS), the efficacy of long-term stimulation effects over years remains unknown. For instance, while Scangos and colleagues demonstrated that intracranial effects can reduce patients’ depression symptoms for up to 60 days (33), data from further longitudinal follow-ups are not yet available. It remains uncertain whether direct brain stimulation over an extended period (e.g., a few years) can lead to alterations in the underlying neural circuitry, some of which may even involve structural changes. This raises technical and ethical considerations regarding the extent of this treatment as a long-term solution without introducing potential side effects. Longitudinal follow-up of patients from clinical trials may shed more light on these issues in the upcoming years.

Furthermore, it’s important to note these additional caveats of the intracranial approach. First, the intracranial approach is inherently invasive, significantly limiting its accessibility. Currently, this technique is reserved for individuals with clear treatment resistance; neuromodulation is not intended as a replacement for standard pharmacotherapy but rather as an option for those who haven’t benefited from initial treatments. Given the involvement of pathological networks in symptom production, it may be possible to engage deeper network connections indirectly by modulating cortical projections to influence symptoms effectively. Identifying specific cortical entry points for modulating deeper structures through non-invasive stimulation techniques could represent a promising path for enhancing the applicability and accessibility of this approach in future research. Second, further research is needed to elucidate biomarkers across various patient populations (40). While precision medicine operates at an individual level, identifying common patterns may be necessary to objectively confirm biomarker significance as disease process hallmarks. This will be crucial for scalability across diagnoses and generalizability across demographic groups. Last, identifying the best stimulation targets within the network is essential, considering the possible distinction between the biomarker representing symptomatology and the area offering the best treatment response. Investigating efficient treatment parameters capable of adequately perturbing the targeted circuit is crucial. Collectively, clarifying the mechanisms, treatment efficacy, and long-term viability of the intracranial approach along with addressing these additional caveats will provide necessary insights into neuropsychiatric disorder physiology and contribute to much-needed innovations in future neuropsychiatric treatments.