Expert-level neural engineering covering neural interfaces, brain-computer interfaces, neural signal processing, deep brain stimulation, and neuroprosthetics.
Utah array: 100 electrode silicon array, acute and chronic recording in cortex. Michigan probe: silicon shank with multiple recording sites along length. EEG: scalp electrodes, non-invasive, low spatial resolution, clinical standard. ECoG: electrocorticography, subdural electrodes, better resolution than EEG. Flexible electronics: conformal arrays reduce mechanical mismatch with brain tissue.
Spike sorting: classify action potentials by waveform shape to identify neurons. LFP: local field potential, population activity, lower frequency than spikes. Common average reference: subtract mean across channels to reduce common noise. Coherence: measure of synchrony between two neural signals in frequency domain. Decoding: machine learning to predict movement or intent from neural signals.
Motor BCI: decode motor cortex signals to control prosthetic or cursor. P300 speller: EEG-based communication for locked-in patients. SSVEP: steady-state visual evoked potential, gaze-based BCI. Closed-loop: detect neural state and deliver feedback or stimulation in real time. Decoder stability: neural tuning drifts over days, adaptive decoders required.
DBS: deep brain stimulation for Parkinson, depression, OCD, high frequency pulses. Charge balance: biphasic pulses prevent electrode and tissue damage. Charge density limit: below 30 microcoulombs per cm squared to avoid damage. TMS: transcranial magnetic stimulation, non-invasive cortical stimulation.
| Pitfall | Fix |
|---|---|
| Single session decoder assuming stability | Neural tuning drifts, use adaptive or recalibrate regularly |
| Charge imbalance causing tissue damage | Always use charge-balanced stimulation waveforms |
| Ignoring electrode impedance changes | High impedance indicates poor contact or tissue reaction |
| Overinterpreting EEG spatial resolution | EEG integrates over large cortical areas |