Abstracts

Rationale: A prominent goal of recent work in human neural prosthetics is to develop closed-loop feedback systems capable of delivering stimulation based on real-time decoding of ongoing brain activity. Neural structures with predictable activity profiles are important for the development and evaluation of these sense-and-stimulate systems. In the human brain, one possibility is the motor system, where voluntary movement is preceded by preparatory activity in various subregions. In particular, premotor preparatory motor activity can precede movement by hundreds of milliseconds. The interval between premotor activity onset and observable motor behavior provides a potential test bed for the evaluation of a closed-loop sense-and stimulate system. Here, we tested a sense-and stimulate method that aimed to detect pre-movement activity and then safely deliver cortical stimulation resulting in observable changes in subsequent behavior. Methods: We worked with electrodes placed in medial premotor cortex in a cohort of 6 patients undergoing seizure screening for pharmacologically intractable epilepsy. Patients were first engaged in a localizer task meant to activate motor planning circuits. This task involved a series of coordinated bimanual finger movements in response to an instruction cue. Electrocorticographic (ECoG) signals were analyzed for pre-movement gamma frequency power increases. The premotor electrode showing the strongest pre-movement increases in gamma frequency (~60-180 Hz) was then selected for online analysis in the closed-loop phase of the experiment, and two adjacent electrodes were chosen to deliver bipolar direct current stimulation. Our closed loop control system was configured to quantify gamma power in real time. When gamma power exceeded a pre-defined threshold, a transient electrical stimulus was delivered to premotor electrodes. On control trials, no stimulation occurred. For both stimulation and control trials, we measured the time when the contralateral hand first pressed a button. Results: All six patients had one or more premotor electrodes showing pre-movement increases in gamma power. Subsequent spectral analysis of the ECoG signal time locked to the delivery of closed loop stimulation showed that in five of the six patients, the closed-loop system successfully detected pre-movement gamma activity. One patient was excluded from further analysis due to excessively long RTs throughout the task (x = 3037132 ms). In four of the five remaining patients, stimulation based on the detection of pre-movement activity resulted in significant slowing of behavior in stimulated versus control conditions (mean for four subjects=19385 ms, all p < 0.05). Conclusions: Our results clearly show that, within a trial, impending movement can be detected and used to trigger direct current stimulation of premotor cortex to influence ongoing motor behavior safely and in a manner imperceptible to the subject. Funding: This work is supported by RO1 DA026452.