The minimally-invasive bargain
Between the scalp and a craniotomy lies a productive middle ground. A minimally-invasive interface accepts a small, bounded surgical risk in exchange for a large jump in signal — reaching the brain through blood vessels, or dropping wireless motes that need no incision at all. The strategic idea is to move rightward on the invasiveness axis without paying the full cost of open-brain surgery.
Endovascular arrays: the Stentrode
The endovascular Stentrode is the clearest embodiment of this bargain. Electrodes are mounted on a self-expanding stent, delivered by catheter through the jugular vein up into a cortical vein overlying motor cortex (the superior sagittal sinus). The stent endothelialises — the vessel wall grows over it — anchoring it and recording LFP-scale activity from just outside the neurons, with no craniotomy. It has reached first-in-human trials for communication in people with paralysis.
Untethered motes: ultrasonic neural dust
Neural dust takes a different route to un-tethering: sub-millimetre motes with no wires and no battery. An external ultrasound transducer both powers a mote (a piezoelectric crystal converts incoming acoustic energy to electric charge) and reads it out — the locally recorded neural voltage modulates the amplitude of the ultrasound the mote reflects back. Detect the modulated echo and you have recovered the signal without any percutaneous connection.
I(z) = I_{0}\,e^{-2\alpha f z}, \qquad \alpha \approx 0.5\ \mathrm{dB\,cm^{-1}\,MHz^{-1}}Ultrasound attenuation in soft tissue grows with frequency f and depth z. This mild loss — far gentler than radio-frequency absorption in tissue — is exactly why acoustic power/telemetry beats RF for deep, tiny implants.
Ultrasound weakens as it goes deeper, and the higher its frequency the faster it fades. But the loss is gentle — much gentler than radio waves lose energy in tissue — which is exactly why a tiny deep implant is better powered and read out by sound than by radio. This is the case for neural dust.
- I(z)
- The ultrasound intensity remaining at depth z.
- I_0
- The intensity you started with at the surface.
- z
- Depth into the tissue.
- f
- Frequency — a higher frequency fades faster.
- \alpha
- The attenuation coefficient, roughly 0.5 dB per cm per MHz in soft tissue.
At 1 MHz, tissue costs about 0.5 dB per cm, so a mote 4 cm deep loses only a couple of dB — a small toll that leaves plenty of acoustic power to run it.
The same idea generalises: ultrasonic neural dust and its descendants aim to scatter dozens of motes through cortex, each addressed acoustically. The engineering frontier is shrinking the piezo and front-end while keeping enough backscatter modulation to resolve microvolt neural signals against tissue clutter.
The transducer frontier: acousto-opto-magnetic
Zoom out and the deeper theme is transduction. Emerging modalities turn neural activity into ultrasound, light, or magnetic contrast — magnetoelectric microstimulators, ultrasound-switchable genetically-encoded reporters, and more. Each swaps the physical channel out of the head and so re-opens the whole invasiveness–signal tradeoff on new terms.
Ultrasound also opens a write channel. Tightly focused beams can non-invasively neuromodulate deep targets, and — paired with microbubbles — transiently and reversibly open the blood–brain barrier to deliver drugs or gene-therapy vectors. See focused ultrasound neuromodulation; it is a reminder that the emerging-tech frontier is bidirectional, not read-only.