Sensory systems as prosthesis targets
Cochlea to auditory nerve, retina to optic nerve, skin to somatosensory cortex, the vestibular organs. Where each sensory prosthesis connects, why hearing succeeded first, and what touch feedback for a prosthetic hand requires.
You are skimming: the title, the first figure, and the short version. Switch to Read in the header for the full page, or Deep to open every deep dive.
You heard through a vocoder and saw through a phosphene grid. Now the anatomy: where along each sensory pathway an electrode can be placed, what it stimulates there, and why the results differ so much between senses. The cochlear implantCochlear implantThe most successful neural prosthesis: an electrode array in the inner ear that stimulates the auditory nerve in up to 22 bands. Glossary entry works because the cochlea hands the engineer a map that is already sorted by frequency along a single line. Nothing else in the nervous system is so convenient.
Hearing: the cochlea is a spectrum analyzer
Sound sets the basilar membrane vibrating, high frequencies near the base, low near the apex: a frequency axis laid out along a 35-millimetre coil. Hair cells along it convert vibration to nerve firing; the auditory nerve fibres beneath them inherit the same tonotopic order. In most deafness the hair cells are gone but the nerve remains. An electrode array threaded along the cochlea stimulates the nerve directly, and its position along the coil chooses the pitch. That is why the implant works with so few channels: place already encodes frequency, and the electronics only have to deliver the envelope. Beyond the cochlea, auditory brainstem implants exist for people without a nerve and do far worse, because the brainstem’s map is not a line.
Vision: the retina does the first computation
Light falls on photoreceptors, which drive bipolar and then ganglion cells whose axons form the optic nerve. In retinitis pigmentosa and macular degeneration the photoreceptors die but many ganglion cells survive. A retinal prosthesisRetinal prosthesisAn electrode array on or under the retina that produces spots of light, phosphenes, in people blinded by retinal disease. Glossary entry stimulates them, on the surface (epiretinal) or beneath (subretinal). Each electrode produces a phosphenePhospheneA spot of light perceived when the visual system is stimulated electrically rather than by light. Glossary entry. The problems: the retina has already done edge detection and motion processing in circuits that stimulation bypasses; ganglion cells of opposite sign (on and off) are stimulated together; and current spreads across the retina’s axon layer, so phosphenes are streaks. Sixty to a few hundred electrodes have given navigation and letter recognition, not reading. Cortical visual prostheses, in the primary visual cortex’s retinotopic map, are the current push, with the same current-spread and layout limits and more surgery.
Why did the cochlear implant reach a million users while the retinal prosthesis reached a few hundred?
The map. The cochlea’s frequency axis is one-dimensional and already in nerve order; electrodes along it inherit the code, and speech survives with a handful of channels. Vision is two-dimensional, pre-processed by retinal circuits the implant bypasses, and useful vision needs thousands of independent points. Current spread caps independent channels in both, and vision needs far more of them.
Touch: closing the loop on a prosthetic hand
Skin mechanoreceptors send signals up peripheral nerves to the dorsal column, the thalamus, and primary somatosensory cortex, laid out as a body map behind the central sulcus, mirroring the motor map in front. Three places to restore touch to someone with an amputation or paralysis.
Peripheral nerve. Cuff electrodes or the Utah Slanted Electrode Array in the residual arm’s nerves; stimulating a fascicle produces a sensation the person feels on the missing hand, in a location that depends on which fibres you reach. This is the University of Utah’s work, and it has given amputees graded pressure sensation that lets them grasp eggs.
Somatosensory cortex. Intracortical microstimulation in the hand area of a person with spinal cord injury produces sensations on the (paralyzed) hand. Paired with a motor BCI driving a robotic arm, it halves the time to complete grasping tasks, because touch tells you when you have made contact and how hard.
The skin elsewhere. Vibrotactile or electrotactile feedback on the upper arm or chest, hapticHaptic feedbackInformation delivered through touch, such as vibration on the skin that tells a prosthesis user how hard they are gripping. Glossary entry substitution: cheap, non-invasive, and the brain learns to interpret it. This is the Phase 5 project.
The engineering problem in all three: a sensor on the prosthesis measures force, an encoder turns force into a stimulation pattern (which parameters: amplitude, frequency, pulse width), and the pattern has to feel natural enough, and be fast enough, that the user relies on it without thinking. Biomimetic encoding, patterns that imitate how real receptors fire on contact and release, works better than simple proportional encoding.
Balance: the vestibular implant
The inner ear’s semicircular canals report head rotation through three nerve branches. In bilateral vestibular loss, people cannot stabilize their gaze or stand in the dark. A vestibular implant with a gyroscope and three electrodes, one per canal, has restored gaze stabilization in a few dozen people. A small field with a clean mechanism, and a good example of a prosthesis that works because the code (rate proportional to rotation velocity) is simple.
The general lesson
Prostheses succeed in proportion to how simple and how peripheral the code is at the point of stimulation. One dimension beats two; rate beats pattern; a nerve with a known map beats cortex with a folded one. And every one of them depends on the brain’s plasticity to learn a coarse new input, which takes months and which the engineering cannot shortcut.
Deep dive Current spread and the channel ceiling 3 min
Current from a stimulating electrode spreads through tissue and excites every excitable fibre within reach; the spread depends on the electrode’s size, its distance from the neurons, and the tissue’s geometry. In the cochlea, electrodes sit in fluid a millimetre or more from the nerve, and each excites a broad swath of fibres, so 22 electrodes give roughly 8 independent channels. Bringing electrodes closer (penetrating arrays, or regrowing neurons toward them) and steering current between electrode pairs both narrow the spread. The same physics limits retinal, cortical, and peripheral stimulation. Whoever solves it doubles the resolution of every prosthesis at once.
Explain what this page was about to your roommate in three sentences. No jargon they would not know.