AI artefact cleaning

Provoke the artefact, then let the network clean it.

The System tab carries an artefact simulation block — motion, chemical shift and a deep artefact model — beside an ML correction toggle. Learners degrade a real acquisition on purpose, run the correction on the same series, and compare. The lesson is both halves: how each artefact arises from a scanning decision, and how much of it a trained model can recover afterwards.

Try it free→

ScanFlo MRI exam console System tab showing the Artifact Simulation (Training) block with Motion, Chemical shift, Deep Artefact (ML) and Artifact Correction (ML) toggles beside a T2 axial brain image

How it works

Three steps, on a real console workflow.

Provoke the artefact

Switch on motion or chemical shift in the training block, or the deep artefact model, and rerun the sequence. The degradation lands on the real series, not a stock picture.

Run the correction

Enable ML artefact correction on the same acquisition and compare the two reconstructions side by side in the viewer.

Decide: rescan or accept

Some artefacts clean up convincingly; others leave residual blur or hide the finding. Making that call is the skill the department needs.

Before / after

The same acquisition, both ways.

The same T2 axial head slice shown twice: degraded by the artefact simulation, and after ML artefact correction has been applied to the same acquisition — One acquisition, two reconstructions. Drag to see what the correction recovers on the axial slice — and where residual blur means the honest answer is still a rescan.

One acquisition, two reconstructions. Drag to see what the correction recovers on the axial slice — and where residual blur means the honest answer is still a rescan.

What the training block exposes

Toggle	What it does	What it teaches
Motion	Simulates patient movement during acquisition	Why immobilisation and scan time matter
Chemical shift	Fat/water misregistration along the read direction	Bandwidth and field strength as protocol choices
Deep Artefact (ML)	Model-generated degradation for training	Recognising an artefact before naming its cause
Artifact Correction (ML)	Runs correction on the same series	What a network can recover — and what it cannot

In the classroom

How programs use it

Assign one artefact per learner: cause it, name it, correct it, and explain why correction did or did not work.
Show a corrected image where the finding is smoothed away — the safety conversation every technologist needs.
Use the rescan-or-accept decision as an OSCE station with the comparison on screen.

Learn the concepts

Short guides that pair with this feature:

Put your learners at the console today.

Free student tier. No installs. Runs in any browser.

Start scanning free→Book a demo