The context and intelligence layer for human anatomy.
Toralis Labs turns patient imaging and surgical expertise into knowledge that surgery, research and medical-device teams can act on.
We're starting with abdominal aortic aneurysms. Scroll to follow one scan.
It starts with a scan the hospital already takes.
A CT angiogram is a stack of hundreds of cross-sections through the body. Contrast dye makes flowing blood bright.
This one shows an abdominal aortic aneurysm: a weak section of the body's main artery, ballooned to several times its normal width.
Our models read every slice.
They label each part of the anatomy that matters for a repair, across the whole stack, without anyone tracing it by hand.
Lumen. The channel where blood is still flowing. It looks like the whole aneurysm, but it isn't.
Thrombus. Old clot lining the inside of the sac. It takes up no dye, so a view of the bright blood alone makes the aneurysm look far smaller than it is.
Calcification. Hard plaque in the vessel wall. A stent has to seal against that wall, so where the calcium sits changes where a device can land.
Now the patient has a digital twin.
Those labels become a 3D model of this patient's aorta, with every branch where it really is.
Pick it up.
Drag to turn it. Switch layers to see what's inside.
And what a scan can't show.
History, labs, earlier imaging, operative notes and surgeon annotations join the same record, so every finding keeps its clinical context.
- Earlier imaging
- Sac grew 5 mm in 2 years
- History
- 72, ex-smoker, CKD stage 3
- Labs
- eGFR 54 mL/min
- Operative notes
- Open hernia repair, 2019
- Surgeon annotation
- "Review the proximal seal"
This record is invented for illustration. The scan is real and public, and nothing is known about the patient.
Then we work out how blood moves through it.
Full simulations of blood flow are slow. We train graph neural networks on simulated flow, so velocity and pressure can be predicted on each patient's own geometry.
Flow shown here is illustrative, not a result for this patient.
Wall shear stress shows where blood drags hardest on the wall.
It's fast through the narrow neck and slow across the wide sac. Walls and devices behave differently under each, which is worth knowing before anyone goes in.
Colours are illustrative. We show no values on purpose.
One twin. Two ways to use it.
Every measurement an IFU asks for.
A device's instructions for use (IFU) set the anatomy it can treat. Each measurement is taken along the vessel's own centre line, so a bent neck reads at its true length and angle.
Model estimates from a public research scan. Not reviewed by a clinician.
Test each device against its instructions for use.
| Requirement | This scan | Device limit | Result |
|---|
Devices and limits are illustrative. In the product, each limit links to the IFU page it came from.
Every scan becomes a comparable twin.
Run a cohort through the same pipeline, and patients measured the same way become one dataset. Plot it against today's devices, and every orange point is a patient they miss. This patient is one of them.
Test a device before you build it.
Drag the handle on the chart, or the sliders under it, to set a new device's limits. The cohort answers straight away, and green points are the patients it brings in.
Hand R&D a brief backed by anatomy.
The twin knows the exact shape of each aorta, so a graft can be drawn to match it: here, the seal laid along this patient's bent neck, each limb sized to its own iliac.
Concept drawing, not a manufactured device.
Anatomy, turned into knowledge you can act on.
We're starting with abdominal aortic and iliac aneurysms. The same approach extends across the vascular system.
- Surgery
- Plan each case on the patient's own anatomy.
- Research
- Compare anatomy across cohorts, measured one way.
- Medical devices
- Design for the patients current devices miss.