Motivation

Bidirectional brain-machine interfaces (bd-BMIs) hold the potential to restore upper-limb sensory and motor functions in individuals with paralysis by recording neural activity and delivering targeted intracortical microstimulation. However, developing the next generation of bd-BMIs requires a precise understanding of how the brain generates movement and processes sensory feedback.

Stereo-electroencephalography (sEEG) monitoring in patients with epilepsy provides a unique opportunity to investigate these mechanisms. These patients typically retain intact sensorimotor function while providing local field potential (LFP) and single unit recordings of neural activity from cortical and subcortical structures, offering a combination of spatial and temporal resolution that noninvasive methods such as EEG and fMRI cannot achieve. Because the electrodes remain implanted for several days, researchers can repeatedly examine complex sensorimotor behaviors, including individual finger movements, sensory error detection, conflict resolution, and short-term motor learning.

Ultimately, studying these intact neural systems drives two major translational goals. First, the fundamental principles of sensory perception and movement generation discovered here can be directly translated into high-impact neuroprostheses for individuals with paralysis. Second, mapping these dynamics with high-resolution sEEG pinpoints the optimal cortical and subcortical regions involved, directly informing surgical target selection and electrode implantation strategies for future bd-BMIs, while also enhancing the field of neural engineering knowledge on sensorimotor function and processing.

To investigate this, we first need a behavioral task that measures motor performance and responses to errors while synchronizing task events with neural recordings. The task must operate reliably within a clinical setting to test our core hypotheses: that sensorimotor dynamics underlying movement onset, offset, and identity can be decoded from sites like the pre-motor cortex, supplementary motor area, and thalamus; that error-related responses are detectable in regions like the anterior cingulate cortex, insula, and hippocampus for real-time prediction; and that modulating visual and tactile feedback will produce measurable changes in learning dynamics.

Our Solution

I developed Typing-Blind, a finger-based sequential typing task for studying how visual and tactile cues affect movement and performance. Participants respond to target letters with designated fingers and withhold responses to non-target letters. Across blocks, the task varies the available cues while recording touch attempts, reaction times, accuracy, and feedback outcomes. Task events are synchronized with sEEG recordings through USB-TTL pulses, allowing neural activity to be examined around movement onset and outcome feedback events. The modular setup is designed to run reliably alongside clinical monitoring.

My Contributions

I designed and implemented the Typing-Blind task within the Dareplane framework, including the experimental conditions, trial sequence, response pad, finger-to-target mapping, and scoring of correct responses and errors. I built the session-planning and operator workflow so blocks can be run and documented consistently in the clinical setting.

I also integrated task markers with Lab Streaming Layer, hardware triggers, and per-block recordings to align behavior with sEEG data. To preserve attempts that do not become scored responses, I added chronological touch-attempt logging with response-pad coordinates alongside the task event log. Finally, I developed behavioral and neural analysis workflows and figures to examine reaction time, accuracy, movement-related activity, and error-related activity across conditions.

In terms of analysis, for this project I applied preprocessing, cleaning, statistical and plotting steps through the MNE-Python or FieldTrip MATLAB toolbox, depending on neural acquisition hardware type. Primary analyses focused on comparing neural activity across conditions of tactile vs visual conditions, correct vs incorrect feedback outcome, and individual fingers and hands.