01
Physical probes
Steel BBs are physically round. Their image shadows reveal how the LINAC/detector system locally stretches, blurs, and orients signal.
MICCAI Startup Village 2026 · Top 10 Startup
Smearlets
Not AI-generated. Physics-measured.
Coral Castle Laboratories measures image formation itself: circular physical probes reveal local blur geometry, and Smearlets use that measured structure to support deterministic, explainable correction.
Measured local blur geometry
·
Deterministic correction
·
Explainable imaging physics
·
Not neural image hallucination
·
Measured local blur geometry
·
Deterministic correction
·
Explainable imaging physics
·
Not neural image hallucination
What changes
From black-box enhancement to measured image transformation.
Smearlets start from a simple physical idea: if a known round object images as a stretched shadow, that distortion is not noise to ignore. It is a local measurement of the imaging system.
02
Local geometry
Each shadow can encode scale, anisotropy, orientation, and covariance — a map of the image formation operator across the field.
03
Physics correction
The correction is not generated by a neural network. It is derived from measured geometry and applied as an explainable transformation.
Why now
In the age of AI, trust becomes the product.
Medical imaging needs methods that can be explained, audited, and measured. Smearlets are designed for exactly that boundary: computational power without black-box image invention.
Not a learned filter trained to make images look sharper.
Not hallucinated detail or generative reconstruction.
A measured mathematical description of local blur.
A path toward quantified correction for radiotherapy imaging.
The Smearlets method
A measured chain from object to operator to correction.
The website should feel like the pitch deck, but it also needs to tell the story quickly for investors, scientists, and conference visitors.
1
Build the phantom.
A non-periodic grid of BB probes creates a field of known circular objects inside the imaging geometry.
2
Image the probes.
The physically round BBs produce elliptical image shadows with location-dependent scale and orientation.
3
Measure local blur.
Smearlets model the local operator from those observed shapes, building a physics-grounded map of distortion.
4
Correct with geometry.
The resulting image transformation is deterministic, explainable, and derived from measured imaging behavior.
Early demonstration
The fork test shows clearer structure after measured correction.
This is not positioned as finished clinical validation. It is the first visual demonstration that measured blur geometry can support correction, with clearer prong and edge structure after deconvolution.
Proof of concept
Core positioning
Make the distinction impossible to miss.
AI may accelerate software development, but the actual image transformation is not AI-generated, not learned, and not produced by a neural network.
Not AI-generated.
Physics-measured.
Coral Castle Laboratories
We are building explainable image correction for radiotherapy imaging.
Selected as a Top 10 Startup for MICCAI Startup Village 2026. Smearlets is our first step toward measured, auditable image transformation.