Skip to content
Imaging Evidence Path

What diffusion and ADC actually measure

Tonight · ~20 min · read · energy: medium · setup: none

Of routine clinical MRI outputs, the ADC map is the closest thing to a quantitative number. This lesson is what diffusion-weighted imaging (DWI) actually measures, what the b-value does, and why “quantitative-ish” is the honest word for ADC.

The anchor: water molecules jitter, and that blurs the MRI signal

Water molecules in tissue are in constant microscopic motion (diffusion). DWI sensitises the MRI signal to that motion: it applies diffusion-sensitising gradients so that freely-diffusing water loses signal (its spins dephase), while restricted water (packed cells, acute cytotoxic oedema) retains signal. The result is a contrast that reflects microscopic water mobility — something CT cannot see at all.

The b-value

The strength of diffusion weighting is set by the b-value (in s/mm²). A b-value of ~0 is essentially a non-diffusion-weighted image; higher b-values weight the image more heavily toward diffusion. Critically, a single diffusion-weighted image is still a relative intensity image (lesson 1) — to get a number you need at least two.

ADC: from relative image to (almost) a number

The Apparent Diffusion Coefficient (ADC) map is computed from ≥2 b-values by fitting the signal decay:

S(b) = S0 · exp(−b · ADC)

With two or more b-values you solve for ADC per voxel, producing a map whose values are in physical units (mm²/s) — one of the more directly quantitative routine MRI-derived maps. Restricted diffusion (low ADC) flags cellularity (e.g. tumour, acute stroke); free diffusion (high ADC) flags loose fluid (e.g. cysts, vasogenic oedema).

Why “quantitative-ish”

ADC is more comparable across scans than T1/T2-weighted intensities, but it is still not a clean physical constant:

  • it depends on sequence parameters, noise, distortion (notably echo-planar/DWI geometric distortion), and motion;
  • the “≥2 b-values” choice and the fit model affect the number;
  • it is still an apparent coefficient, averaged over a voxel containing many microenvironments.

So treat ADC values with the same robustness discipline as radiomic features (Ch. 3): record the parameters, do not over-interpret small differences, and remember the map inherits acquisition variability. “Quantitative-ish” — more comparable than T1/T2 images, less comparable than CT HU — is the honest framing.

Stop and think — then reveal

An acute stroke shows bright on the diffusion-weighted image (high signal). Does that mean the ischaemic tissue has high ADC? Explain the apparent contradiction.

No — it has low ADC, and the bright DWI signal is the clue, not the contradiction. In acute stroke, cytotoxic oedema packs cells and restricts water diffusion, so diffusing water dephases less under the diffusion gradients and the DWI image is bright (high signal). Low ADC means less signal loss; the DWI brightness reflects restricted diffusion. The ADC map shows the stroke as dark (low ADC), which is the quantitative read. This is exactly why you read DWI together with ADC: the DWI image is “bright = restricted”, and ADC is the number. (Also beware “T2 shine-through”: a bright DWI can also come from high underlying T2 signal, which ADC resolves.)

What to retain

  1. DWI sensitises the signal to microscopic water diffusion; the b-value sets the diffusion weighting strength.
  2. ADC is computed from ≥2 b-values (S(b)=S0·exp(−b·ADC)), in physical units — one of the more directly quantitative routine MRI maps.
  3. Restricted diffusion (low ADC) → cellularity (tumour, acute stroke); free diffusion (high ADC) → loose fluid.
  4. ADC is “quantitative-ish”: parameter/distortion-sensitive, still an apparent voxel-averaged coefficient — treat with the radiomics robustness discipline.

Next: the part people find most abstract — what k-space is, without an MRI physics degree.