PROJECT SUMMARY Neuropsychiatric disorders are a leading cause of disability worldwide, with depressive disorders being the most disabling among them. Many patients are resistant to all current treatments. Invasive electrical brain stimulation for treatment-resistant depression showed early promise in open-label studies but has had variable efficacy in controlled clinical trials. To date, stimulation in neuropsychiatric disorders has been limited to an open-loop approach that applies a fixed pattern of continuous stimulation regardless of symptom levels. One limitation is that open-loop stimulation does not track the inter- and intra-subject variabilities in neuropsychiatric symptoms, which can change rapidly in an individual. Another limitation is the lack of an input-output model that can guide stimulation by predicting how ongoing stimulation input drives large-scale neural activity and the symptoms it underlies in an individual. We will address these limitations to enable precise invasive electrical brain stimulation for neuropsychiatric disorders by developing a novel real-time model-based neural control system. We will provide proof-of-concept demonstration for acute control of neural biomarkers of mood states related to depression symptoms in epilepsy patients with implanted intracranial electroencephalography (iEEG) electrodes, in whom we will obtain repeated mood self-reports and perform stimulation simultaneously with neural recording. The system will continuously adjust the stimulation parameters, for the first time, based on 2 elements: (i) Novel personalized input-output model learned on recorded brain network response while delivering a new stochastic stimulation waveform to excite network activity. (ii) Personalized decoder trained on multi-day continuous iEEG recordings and simultaneous mood self-reports to estimate mood state variations from neural activity as feedback. Combining these, we will build a real-time model-based closed-loop system to precisely drive the neurally-decoded mood state—the neural biomarker of mood—to a target level. Our system generalizes to any stimulation site. Here, we will demonstrate the system with orbitofrontal stimulation as we have shown it to acutely improve mood and modulate large-scale mood-relevant brain activity. We will run real-time closed-loop experiments in each patient. The system will estimate the neural biomarker from iEEG and adjust the stimulation amplitude and frequency in real-time based on the input-output model to drive the estimated biomarker to a target level. We will also develop model-free closed-loop on-off stimulation that turns stimulation on-off based on the neural biomarker. We will compare model-based, on-off and open-loop stimulations. Success of this program will enable precisely regulating a desired brain state by developing the first model-based closed- loop invasive brain stimulation system and advancing neuromodulation technology. It will also directly inform elec...