From reading to writing
Volume I framed the BCI as a decoder: neural signals in, intention out. Neuromodulation is the inverse operation. Instead of estimating a hidden state from voltages, we inject energy to bias neural activity toward a goal. Deep brain stimulation quiets a parkinsonian tremor; intracortical microstimulation evokes an artificial touch. Both are writing to the brain — the same physical act, aimed at different ends.
The neuromodulation toolkit
It helps to organize the field by the physics of the actuator. Electrical: DBS, cortical microstimulation, vagus nerve stimulation, and transcranial electrical stimulation. Magnetic: TMS, where a changing field induces currents. Acoustic/hybrid: focused ultrasound and temporal interference. Optical/biological: optogenetics and chemogenetics, which make neurons responsive to light or a designer drug.
Every modality trades along the same axes: spatial focality, depth reach, cell-type specificity, invasiveness, and temporal precision. No technique wins on all five. Intracortical electrodes are precise but invasive; TMS is non-invasive but coarse and shallow; optogenetics is exquisitely cell-specific but needs gene delivery. The rest of this track is, in large part, the story of how each method sits on these tradeoffs — and how closing a feedback loop changes what any of them can do.
Open loop versus closed loop
Conventional DBS is open loop: a clinician sets an amplitude, frequency and pulse width, and the device delivers them around the clock regardless of what the brain is doing. A closed-loop system instead senses a neural biomarker, decides whether and how much to stimulate, and only then acts. Two clinical families already do this: responsive neurostimulation for epilepsy and adaptive DBS for movement disorders.
Read that loop as the skeleton of everything ahead. Guide 2 explains the stimulate arrow (what a pulse does to tissue). Guide 3 fleshes out the sense and decide arrows for adaptive DBS. Guide 4 replaces the electrical actuator with magnetic, acoustic and biological ones. Guide 5 asks how to learn the decision policy — and who is allowed to.
Target engagement keeps neuromodulation honest
The single most disciplining concept in the field is target engagement: did the stimulation actually change the circuit you intended to change? It separates dose delivered (milliamps into tissue) from dose absorbed (a measurable neural effect). Without an engagement readout — an evoked response, a biomarker shift, an imaging change — you cannot dose rationally, and you cannot tell a null result from a missed target.
R(d) = R_{\max}\,\dfrac{d^{\,n}}{d^{\,n} + d_{50}^{\,n}}A minimal engagement model: neural response R rises with dose d along a saturating Hill curve, with half-maximal dose d_50 and steepness n. The clinical goal is to sit on the responsive part of this curve, below the side-effect threshold.
This is a dose-response curve shaped like a stretched "S": as you turn up the stimulation dose d, the neural response R climbs, then flattens out near a ceiling R_{\max}. The clinical sweet spot is the steep middle stretch — enough dose to get a real response, but not so much that you cross into side effects.
- d
- The stimulation dose you dial in, e.g. current amplitude or pulse energy.
- R_{\max}
- The ceiling — the largest response the tissue can give no matter how hard you push.
- d_{50}
- The dose that gets you halfway to the ceiling; a natural marker of potency.
- n
- Steepness — how abruptly the curve switches on around d_{50}.
If d_{50} is a 2 mA pulse and n is large, then at 1 mA you get almost nothing, at 2 mA you reach half the maximal effect, and by 3 mA you are nearly saturated — a narrow window where small changes matter a lot.
Why writing is hard
Reading is passive; writing fights physics. You do not get to choose which neural elements a pulse recruits — axons of passage often fire before the cell bodies you were aiming at. The stimulus artifact is orders of magnitude larger than the microvolt biomarker you want to sense, so writing tends to blind reading. Biomarkers drift over hours and days. And charge you can safely inject is capped by electrochemistry and tissue damage. The next four guides are, essentially, four responses to these four obstacles.