Production of Reversible Changes in the Central Nervous System by Ultrasound

TL;DR: In 1958 the Fry brothers and Hosea Ades showed that focused ultrasound could reversibly switch off activity in deep brain structures — suppress a response, then watch it recover — a clean dissociation between modulation and destruction. They had to remove the skull to do it. Sixty-eight years later, phased arrays and CT-based aberration correction focus ultrasound through the intact skull to a millimeter, and low-intensity transcranial focused ultrasound (tFUS) is the only tool that reaches deep human circuits reversibly without surgery. This paper is the origin point, and the dose-response and mechanism questions it raised are still open.

The claim, and why it was radical

The dominant tools for interrogating the brain in the 1950s were the electrode and the scalpel. You could record, you could stimulate with injected current, or you could ablate. All three demanded physical access: a wire in the tissue, a lesion carved by knife or heat. Reversibility — turning a region off and then on again — was essentially unavailable except pharmacologically, and drugs go everywhere.

William Fry's group at the University of Illinois had spent the early 1950s building focused ultrasound into a surgical instrument. They could converge acoustic beams from multiple quartz transducers onto a focal volume of roughly a millimeter and burn a lesion deep in an animal brain without touching the overlying tissue. That alone was remarkable: energy deposited at depth, sparing everything the beam passed through.

The 1958 Science paper is the more interesting move. By lowering the exposure, they produced changes that recovered. Neural responses in deep structures (their broader program centered on the visual pathway and lateral geniculate in cats) could be depressed by a sonication and then return over minutes. That is the whole conceptual foundation of neuromodulation as distinct from neurosurgery: a knob, not a knife. And it was spatially targeted in a way electrical stimulation — which spreads along the paths of least resistance — never is.

The framing was decades early. It said, in effect: there exists a physical modality that can transiently and focally alter deep brain activity from outside the tissue. Everything since is engineering to remove the phrase "outside the tissue" of its asterisk — the craniotomy.

Why it could not go noninvasive then

The asterisk is the skull, and the physics is unforgiving.

Ultrasound focusing works by phase: energy from many points on a transducer arrives at the focal spot in step and adds. The skull wrecks this. Sound travels at roughly 1540\ \text{m/s} in soft tissue but near 2500\text{–}2900\ \text{m/s} in cortical bone, and skull thickness varies from a few millimeters to over a centimeter across a single head. A wavefront crossing that irregular, fast layer arrives at the target smeared in phase. The focus blurs, shifts, or dissolves. At around 1\ \text{MHz} the bone also absorbs a large fraction of the energy and heats up.

In 1958 there was no way to measure a given patient's skull geometry, no way to compute the phase error, and — crucially — no way to correct it even if you knew it. A single-element or few-element transducer emits one fixed wavefront; you cannot reshape it per skull. So the Frys did the only thing available: they opened the skull and coupled the transducer to the exposed brain, and they positioned everything by hand on a stereotaxic frame, guided by anatomy and physiological recordings rather than an image of the live target.

Three missing technologies, concretely:

• Beam steering. No phased arrays. You aimed by physically moving the transducer, not by electronically re-timing hundreds of elements.
• Aberration correction. No CT to map skull density and thickness, and no computer to invert the resulting delays.
• Guidance and dosimetry. No MRI, no MR thermometry, no acoustic simulation. You could not see where the focus actually landed or how hot it got, so the safe reversible window versus the ablative one had to be found the hard way.

What changed

The enabling inventions arrived in a rough sequence:

• Phased-array skull correction. In the early 2000s Hynynen and Clement showed you could take a CT scan, convert bone density and thickness into per-element phase delays, and refocus a large multi-element array through an intact human skull. Modern clinical helmets carry on the order of 1000 elements.
• MR guidance and thermometry. MRI localizes the target to the individual's anatomy and reads out focal temperature rise in near real time, so ablation can be titrated. This produced FDA-cleared MR-guided FUS thalamotomy for essential tremor (2016) and Parkinsonian tremor.
• The low-intensity regime — the true heir to 1958. Around 2010–2014, work by Tyler, and Legon and colleagues, demonstrated reversible neuromodulation in humans at intensities well below the ablative threshold: tFUS to primary somatosensory and motor cortex changed evoked responses and behavior without lesioning. That is the Frys' reversible effect, now through the skull.

The scale of the shift is easiest to see in the number of independently timed acoustic sources you can command.

What a serious 2026 revival looks like

The paper's living project is closed-loop, noninvasive, deep-target neuromodulation, and the pieces exist to build it now.

Reuse from Fry: the core insight that a sub-ablative dose produces a reversible functional change, and the discipline of pairing sonication with a direct physiological readout to know whether you hit the target and what you did to it.

Replace: hand positioning with a CT-corrected phased array and MR/neuronavigation; open-loop single shots with a decoder-in-the-loop controller. Concretely — a tFUS helmet targeting a deep node (subgenual cingulate or amygdala for depression, an insular/thalamic node for pain), an EEG or fMRI/fNIRS decoder estimating circuit state, and a policy that adjusts pulse timing, duty cycle, and intensity to drive the state toward a target. Because tFUS is reversible and steerable, the same system doubles as a causal circuit-mapping tool: a transient, movable "virtual lesion" in the human deep brain, something no other noninvasive method offers.

Two hard problems must be respected, not waved away:

• Mechanism is genuinely unsettled. Candidate pathways include mechanical gating of stretch-sensitive channels (Piezo, TRP), membrane capacitance changes, mild heating, and cavitation. Which dominates in which regime governs whether you excite or suppress, and whether effects are cell-type specific. Sonogenetics — engineering ultrasound-sensitive channels — is one way to make the coupling deliberate rather than incidental.
• The auditory confound. Guo et al. and Sato et al. (2018) showed that a chunk of what looked like direct cortical tFUS effects in rodents was indirect activation of the auditory pathway by the pulse itself. Any credible closed-loop claim needs sham and smoothed-envelope controls that rule this out. This is the modern equivalent of the dosimetry the Frys had to establish by hand — the field's central rigor problem.

Status: vindicated in part, wide open in the rest

The ablative branch is clinical reality. The reversible branch — the actual content of this paper — is where the 2010s explosion of human tFUS studies lives, and it is still pre-consensus: effect sizes vary, mechanisms are debated, and closed-loop deep-target therapy is early-stage rather than proven. That is exactly why the paper is ripe. It stated the goal cleanly in 1958 and the field has only recently acquired the tools to test it fairly.

Where to read it, and alongside what

The paper: F. J. Fry, H. W. Ades, W. J. Fry, "Production of Reversible Changes in the Central Nervous System by Ultrasound," Science 127:3289 (1958), doi:10.1126/science.127.3289.83. Bibliographic details are as verified in OpenAlex; I have not re-checked the exact acoustic parameters in the original text, so treat frequency and focal-size figures here as approximate.

Read alongside Clement & Hynynen on transcranial phased-array skull correction (early 2000s) for the enabling physics; Legon et al., Nature Neuroscience 2014, for the first careful human reversible tFUS; Guo et al. and Sato et al., Neuron 2018, for the auditory-confound reckoning; and the sonogenetics work from Chalasani's group for where deliberate acoustic-neural coupling is heading.