Ex vivo human liver perfusion

Patient tissue,
alongside the trial.

Real liver tissue. Native architecture. Days of viability.
We keep intact human liver tissue alive and functioning outside the body — perfused through its own vasculature, at body temperature — so a therapy can be read out in real human tissue alongside the clinical trial.

Get in touch See the platform →
Human Native tissue
Under flow Blood-like perfusion
Sustained Working window
The platform

Intact human liver, kept alive outside the body.

We have built a miniaturized normothermic perfusion system that keeps a piece of intact human liver alive and functioning outside the body, under its own blood-like flow, at body temperature. Unlike the cell suspensions, 2D lines, and organoids the industry relies on, the tissue keeps its native three-dimensional wiring: polarized hepatocytes, the immune cells that patrol the liver, the specialised blood-vessel lining, and the microscopic sinusoidal channels blood flows through.

Because the architecture and every major cell type stay intact and connected, the tissue behaves far more like a real human liver — and it stays that way long enough to observe slow biology short-lived models simply cannot reach.

The Corperial SYSTEM 7100 — first-iteration ex vivo human liver perfusion device on the bench.
The first iteration of the perfusion device, run end-to-end with human liver tissue.
Why this is different

Every other liver model concedes a column.

Each existing human-liver model trades something away. The two things nobody else holds together are human tissue kept under blood-like flow for a sustained window.

Model Native 3D
architecture		 Patient
tissue		 Resident
immune cells		 Blood-like
perfusion		 All major
cell types		 Working
window		 Core limitation
Dissociated hepatocytes ✗polarity lost fast ✓ ✗ ✗ ✗ Hours–days Stops behaving like liver almost immediately
2D cell lines ✗monolayer ✓transformed ✗ ✗ ✗ Indefinite, non-physiological Cancer-derived; not real liver behaviour
Liver organoids Partial ✓ ✗ ✗ Partial Weeks–months Simplified and incomplete cell mix
Organ-on-a-chip ✗engineered ✓cultured cells ✗not native Partialmedium, not blood ✗few, selected Days–weeks Engineered from cultured cells — mimics tissue but isn't native
Precision-cut slices ✓ ✓when human ✓ ✗static ✓ Short (days) No flow; tissue dies quickly
Animal models ✓whole organ ✗ ✓animal ✓in vivo ✓ Long Not human — misleads on drug handling & immunity
Corperial platform ✓ ✓ ✓ ✓under flow ✓ Sustained Low-throughput

Everything else concedes at least one column. Corperial is built to be the row with no concession — high architectural fidelity and a sustained working window, in human tissue under flow.

How it works

From arrived to confirmed, even flow.

A small, sterile, incubator-resident perfusion rig. Fresh human liver tissue is received, its plumbing is verified, it is sectioned into chamber-sized pieces, and each is perfused with an oxygenated, blood-based medium under a controlled pump.

01
Prepare
Receive fresh human liver, cannulate, dye-check for even perfusion with no dead zones, then rapidly section.
02
Load & baseline
Load the multi-chamber platform, confirm flow reaches the tissue, and mark t = 0.
03
Perfuse & sample
Blood-like flow at body temperature; sample effluent and tissue over time as the run unfolds.
04
Read out
Turn samples into readouts and score them against what a model predicted the tissue would do.

The engineering lives in the details that keep a delicate vascular bed intact — careful anticoagulation and inline filtration so the channels don't clog, gentle membrane oxygenation so blood cells aren't sheared, and a perfusate tuned for low viscosity and clean optics. All adapted from clinical whole-organ perfusion and miniaturized to the bench.

What we measure

What goes in, minus what comes out.

A run produces three streams of data. Much of the signal is the difference between what enters the tissue and what leaves it — oxygen used, ammonia cleared, proteins secreted, therapeutic taken up.

Influent
The fresh medium going in — the baseline everything else is read against.
Effluent
The fluid coming out of the tissue, sampled over the course of the run.
Tissue
Sampled at set points and preserved for imaging and molecular work.

Together they cover six readout domains:

Viability & injury Liver function Vascular-lining health Perfusion physiology Immune / delivery response Molecular & imaging
Vascular-lining health
The sinusoidal lining is the first thing to fail in a struggling perfusion. We track markers specific to it, so we can tell "the whole tissue is dying" apart from "the tissue is fine but a therapy is hitting the endothelium."
Target engagement, not just proxies
The platform reports what actually happens inside human tissue — how much of a therapeutic was taken up by hepatocytes versus swept up by immune cells — rather than a downstream blood surrogate measured in an animal.

Each run is framed as a comparison against a prediction: what a model expected, lined up against what the tissue actually did. The platform is designed to be the ground-truth check on those predictions.

Live imaging under flow

The intravital view — in human tissue.

Because the tissue stays intact, perfused, and optically clear, it can be imaged live, at sinusoid resolution, while the run is happening — not only fixed and sectioned at the end. This is the view normally reserved for intravital microscopy in a living animal; here it is human tissue, without the animal's limits on optical access or how long you can watch.

That opens up dynamic biology as it unfolds: resident immune cells patrolling the sinusoids and engulfing their targets, a labelled therapeutic being taken up by hepatocytes versus swept up by immune cells, and blood-like flow moving through the microscopic channels in real time. Because the same piece of tissue lives for weeks, a field can be revisited across the run rather than captured as a single snapshot.

Current state

Built, and run end-to-end.

The first iteration of the device is built, and we have taken human liver tissue through the full workflow. Specifically, we have demonstrated that:

The last model you need.

Get in touch Explore tools →