AI image enhancement
Deep-learning reconstruction: less noise, more detail, same acquisition.
Vendor consoles call it different names; the principle is the same — a trained network denoises and sharpens the reconstruction so a faster or lower-dose acquisition reads like a longer one. ScanFlo lets learners switch it on and off and see exactly what it buys, and what it cannot.
ScanFlo MRI — Viewing · T2 TSE, 3-up
How it works
Three steps, on a real console workflow.
01
Scan at the trade-off
Cut the averages, shorten the scan, or lower the mAs — acquire the noisy version on purpose.
02
Reconstruct with enhancement
Apply deep-learning denoising and resolution enhancement to the same raw acquisition and compare side by side.
03
Judge what changed
SNR, apparent resolution, texture — and the limits: enhancement cannot invent anatomy the acquisition never captured.
What learners compare
| Setting | Effect | What it teaches |
|---|---|---|
| Denoising strength | Lower noise at the same scan time | Where SNR comes from and what it costs |
| Resolution enhancement | Sharper edges from the same matrix | Apparent vs. acquired resolution |
| Off vs. on | Same raw data, two reconstructions | Enhancement is a reconstruction choice, not a sequence |
| Dose / time saved | Equivalent quality at lower input | How AI reconstruction changes protocol design |
In the classroom
How programs use it
- Run the same case with enhancement on and off and have learners write what is real and what is smoothed.
- Combine with dose training: find the lowest mAs that still reads acceptably with enhancement.
- Discuss the department policy question — when is an AI-reconstructed image the record of truth?
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.