The system
The child sits on a parent’s lap and wears a soft cap. The session is designed to last under five minutes, with no sedation. The clinician receives a 3D map of active and quieter regions — a monitoring view, not a diagnosis and not a treatment.
Physics and design follow. MATLAB, ANSYS, and k-Wave cases are simulations. The wearable has not been fabricated.
How the information is obtained
QuietScan is a pediatric transcranial ultrasound platform for children ages 0–5. It does not ask what the tissue looks like. The long-term aim is to ask what a region is doing — whether local blood volume and flow changed with activity. That measurement does not exist yet. Today there is no patient data, no phantom, and no hardware.
1. Transmit
A short pulse leaves the 64-element array. The pulse is kept short so heating stays low and the returning echo can be separated from the send.
2. Cross the skull
The wave passes coupling gel, then bone, meninges, and the fluid around the brain. Bone conducts sound at about 2,900 m/s; soft tissue at about 1,540 m/s.
3. Scatter at the region
At the intended depth, a fraction of the energy scatters off tissue and blood. The rest is reflected or absorbed, mostly in bone.
4. Record the return
Part of the scattered wave travels back and is recorded on the array. That echo is the information. Nothing is injected, and nothing is taken from the child.
5. Line the arrivals up
Skull thickness is not uniform, so each element is delayed by a different amount. A per-element time correction is meant to make those arrivals meet at one point.
6. Compare over time
Later, the question is whether the echo from that spot changed — more blood, or less. Active brain regions get more blood. That is the long-term aim. It is not something the device can do today.
Why correction is required
Bone carries sound almost twice as fast as brain tissue. A child’s skull is also thicker in some places than others. So the wave from each of the 64 sensors arrives at a slightly different time. The bright spot smears, and it lands off to the side. A map made that way would be a map of the wrong place.
The fix is a small wait on each sensor — a different wait for each one — so every wave arrives at the chosen point together.
Our first pictures were only a sketch. The computer blurred the beam, then un-blurred it. That showed the idea. It was not proof, because the same program made the blur and took it away.
September’s Stage B sent a wave through an analytic skull plate. The October measurement replaces that plate with thickness taken from pediatric CT.
One full cycle · 2 minutes
How the 64 sensors send a wave, how a child’s skull smears the spot, and how time reversal puts it back — in a computer only. Turn the sound on.
Simulation. No patient data. The wearable has not been fabricated. Time reversal needs a source at the target, which a child does not provide.
Research update · October 2026
Skull thickness is now measured from pediatric CT, not estimated. Through a measured infant skull, 10.9% of the beam energy stays on target, against 12.4% with no skull. Simulation only. No clinical claim.
Read the October updateStage B · September 2026
In September 2026, Stage B replaced an image-space sketch with k-Wave through an analytic three-layer skull plate — outer cortical bone, diploe, inner cortical bone — at pediatric-scale thicknesses of 1.0, 1.5, and 1.0 mm, with a smooth change in thickness from side to side. A 64-element array at 0.3 mm pitch transmits into the model. k-Wave solves the acoustic wave equations; it is the method most published transcranial work uses. This plate is not the measured infant geometry in the October research update.
A point source sits at the target inside the simulated brain. Arrival time is recorded at each of 64 elements, once with no skull and once through the skull. The difference is the delay the bone adds — a number the earlier image-space model could not produce. In water, arrival time must equal distance divided by the speed of sound. That control matched to within 0.17 percent. Through the skull, peak pressure fell to 54 percent of the water value. The aim shifted 4.1 mm off target. Energy landing within 3 mm of the target fell from 12.4 percent to 1.6 percent.
Timing each element to cancel its measured delay restored aim from 4.1 mm off target to 0.4 mm, and raised energy inside the 3 mm box from 1.6 percent to 6.8 percent — just over half the no-skull value of 12.4 percent. It did not narrow the focal spot. One delay per element cannot cancel a wave that bounces inside the thin bone and arrives more than once. A control with the delay sign reversed scored 2.1 percent against 6.8 percent, which confirms the sign by experiment.
Time reversal — recording the field at each element and playing it backwards — recovered 9.9 percent of energy on target, about 80 percent of the ideal, and a focal spot narrower than the water beam. It is a simulation benchmark, not a method for a child: it needs a sound source at the target, which a child does not provide. Its value is that the remaining gap is physics, not a coding error. The next stage is to close that gap without a source inside the head.
A two-dimensional, lossless simulation through an analytic skull plate. No patient data. No phantom data. No hardware. No trained model — corrections in the current code are estimated or blind, not learned. No clinical claim and no regulatory clearance.
Stage B v2 — k-Wave scorecard
Each beam is scored against an ideal no-skull case. Delay correction restores aim and about half the target energy. Time reversal shows the rest is physically recoverable. 2D, lossless, analytic three-layer skull. No patient or phantom data. No trained model.
Stage B — wave through a three-layer skull
Same physics, earlier layout: 1 MHz waves through outer bone, spongy middle, and inner bone, with pediatric-scale thickness variation. The delay curve is measured from the wave, not assumed.
Safety, by design
Energy that is not reflected is absorbed, and absorbed energy becomes heat. Two indices have to be reported. Reporting only one is a gap. These are design limits for a device that does not exist yet — not a measured exposure, and not a clearance.
TIC
For a child, this is the binding constraint. The skull is thin, at the surface, and absorbs strongly. Internal working margin: TIC no higher than 0.5. The regulatory Track 3 ceiling is 6.0. 0.5 is not a claim that 6.0 would be acceptable.
TIS / TIB
Thermal Index is not a temperature. It is a relative risk indicator against the exposure expected to raise tissue by 1 °C. Soft tissue and bone-at-focus are reported separately.
MI
Mechanical Index is peak rarefactional pressure, derated, divided by the square root of frequency in MHz. Higher frequency and lower pressure both reduce it. The pulse stays short, and the duty cycle stays low, so average heating stays low.
A functional measurement needs repeated looks, not one scan. The exposure has to be compatible with that. ALARA: lowest output and shortest time that still returns a usable echo.
Earlier MATLAB cases
Working simulation cases. Not human clinical data. Not a diagnostic output.
Cases 01–08 model the beam in image space: distortion is a broadening and a shift, and correction blends back toward the ideal. That was a concept check. It is circular if treated as proof, because the distortion is inserted and then removed by construction. Green is energy in the right place. Red is energy that missed.
Case 4 — Image-based cortical target
Earlier MATLAB work. Image-based focusing on a synthetic cortical mask. After the skull, the beam should stay in the region — not smear into neighboring tissue.
Case 5 — Estimated delays
Earlier image-space work. An estimated delay map is compared with the true delays. Correction helps and does not reach the ceiling. This is not a trained model, and it is not the k-Wave Stage B result above.
