Neural Signal Acquisition and Sensor Modalities
Recording and Stimulation Modalities
Modalities overview
Computer-Brain AI systems use a range of modalities with different tradeoffs in spatial resolution, temporal resolution, invasiveness, portability, and signal-to-noise ratio. The choice of modality determines what can be decoded and how safely a system can be deployed.
Common modalities
| Modality | Strengths | Limitations |
|---|---|---|
| EEG | Portable, inexpensive, millisecond timing | Low spatial resolution, artifact-prone |
| MEG | High temporal precision, better source localization than EEG | Expensive, non-portable |
| fNIRS | Safer optical hemodynamics, wearable | Slow response, shallow penetration |
| ECoG | Better SNR than EEG, moderate invasiveness | Requires surgery |
| Intracortical arrays | Single-unit and multi-unit access | Highly invasive, long-term stability issues |
Tradeoff principle
Higher signal fidelity usually comes with higher invasiveness, greater cost, or lower portability.
Artifacts and preprocessing challenges
Noise sources
Artifacts may arise from eye movements, muscle activity, line noise, motion, electrode drift, hemodynamic lag, and stimulation interference. Robust preprocessing often requires artifact rejection, adaptive filtering, re-referencing, independent component analysis, and quality control metrics.
Important terms
Which modality generally provides the best access to single-neuron activity?
Intracortical microelectrodes can capture spikes and multi-unit activity near individual neurons.
Correct answer: Intracortical microelectrodes
Which issue is most characteristic of EEG in real-world use?
EEG is noninvasive but highly vulnerable to artifacts from eye and muscle activity.
Correct answer: Strong susceptibility to ocular and muscle artifacts
Give one reason fNIRS is attractive for wearable neurotechnology.
fNIRS is often used when portability and safety matter more than speed.
Correct answer: It is relatively safe and wearable while providing information about cortical hemodynamics.
Typical acquisition workflow
-
1
Step 1: Select modality based on task and risk profile.
-
2
Step 2: Place sensors or implant hardware.
-
3
Step 3: Calibrate and record baseline activity.
-
4
Step 4: Monitor signal quality and artifacts.
-
5
Step 5: Stream data to decoding and control modules.