IMAGE QUALITY · LESSON 22

Signal-to-noise and contrast-to-noise.

Learn why a clean image is not always a useful image, and how MRI balances signal, contrast, resolution, and time.

About 14 minutesSNR tradeoff visual3 knowledge checks
SIGNALNOISE FLOORSNR
01EXPLAIN

TWO DIFFERENT QUESTIONS

SNR asks whether signal rises above noise. CNR asks whether two tissues look different.

Signal-to-noise ratio compares useful MR signal with random variation that does not represent anatomy. A higher SNR generally produces a cleaner, less grainy image.

Contrast-to-noise ratio compares the signal difference between two tissues with the noise. High CNR makes those tissues easier to distinguish from one another.

SNRtissue signal ÷ noise
CNR| tissue A − tissue B | ÷ noise
IN PLAIN LANGUAGESNR asks, “Can I see the signal?” CNR asks, “Can I tell these two tissues apart?”
02VISUALIZE

CLEAN DOES NOT ALWAYS MEAN DISTINCT

An image can have high SNR but low CNR.

HIGH SNR · LOW CNR
Clean but similar

Both tissues are measured clearly, yet their signals are too close to separate easily.

USEFUL CNR
Different enough to see

The tissue signals are separated by more than the surrounding noise variation.

CNR depends on SNR, but it also depends on contrast weighting. Changing TR, TE, TI, flip angle, preparation pulses, or contrast agents can change tissue separation even when the overall noise level stays similar.

03CONNECT

THE SNR CONTROLS

Every way of gaining signal costs something.

VOXELLarger voxel → more SNR

More spins contribute signal, but spatial resolution decreases and partial-volume averaging increases.

NEX / NSAMore averages → more SNR

SNR rises with the square root of averages, while scan time increases directly.

BANDWIDTHLower bandwidth → more SNR

Less noise is admitted, but chemical shift, distortion, minimum TE, and other tradeoffs may change.

COILClose, appropriate coil → more SNR

Coil geometry, channel combination, loading, and distance from the anatomy affect received signal.

ACCELERATIONFaster scan → possible SNR cost

Parallel imaging samples fewer phase-encoding data and may add a spatially varying geometry-factor penalty.

SEQUENCETiming changes available signal

TR, TE, flip angle, echo train, preparation pulses, and tissue relaxation all influence signal.

1 AVERAGE1.0 × SNR2 AVERAGES1.41 × SNR4 AVERAGES2.0 × SNR

Doubling the averages does not double SNR. Because SNR follows the square root of NEX or NSA, four times the averages are needed to approximately double SNR, and the averaging portion of scan time also becomes four times longer.

OPTIMIZATION MEANS BALANCE

The goal is not the highest possible SNR. It is enough SNR and CNR to answer the clinical question while preserving needed resolution, coverage, artifact control, and a tolerable scan time.

04REMEMBER

THE CONVERSATION ANALOGY

SNR is hearing a voice. CNR is telling two voices apart.

In a noisy room, one loud voice may be easy to hear, that is good SNR. If two people sound almost identical, telling them apart may still be difficult, that is low CNR.

SSignal

The information produced by the anatomy.

NNoise

Random variation that does not represent anatomy.

CContrast

The difference between the tissues of interest.

Remember: Clean signal is SNR. Visible tissue difference is CNR.

05APPLY

CHECK YOUR UNDERSTANDING

Choose the tradeoff, not just the number.

What is the difference between SNR and CNR?

SNR compares one tissue's signal with noise. CNR compares the signal difference between two tissues with noise.

If averages increase from 1 to 4, what happens to SNR?

SNR increases by approximately the square root of 4, so it doubles. The averaging portion of scan time increases fourfold.

Why not always use larger voxels to gain SNR?

Larger voxels collect more signal but reduce spatial resolution and increase partial-volume averaging, which may hide small structures or mix tissues.

LESSON 22 COMPLETE

You understand how SNR and CNR guide image-quality decisions.

Educational references

The relationships shown here are teaching approximations with other parameters held constant. Modern coils, parallel imaging, denoising, reconstruction, filters, and magnitude-image statistics make formal SNR measurement more complex.