Skip to content
Imaging Evidence Path

Why MRI intensity is fundamentally unlike CT HU

Tonight · ~15 min · read · energy: low · setup: none

If you take one sentence from this chapter, take this one: MRI intensity is not a physical scale. Everything hard about MRI radiomics, harmonisation and synthesis falls out of it. This short lesson establishes the central contrast with CT before the later lessons build the physics.

The anchor: CT has a scale; MRI has a contrast

In CT, a voxel value maps to the Hounsfield scale — a defined scale where water is 0 HU, air is about −1000, and the value is approximately comparable across scanners (Ch. 1). You can threshold bone at +400, measure a tumour’s mean attenuation, and compare a number across patients, because the scale is (approximately) physical.

MRI does not give you that. An MRI voxel intensity is relative — its meaning depends on the pulse sequence, the scanner, the field strength, the coil, and a dozen timing parameters. A voxel value of “420” in one MRI sequence means nothing on its own; it only means something relative to other tissues in the same image, under the same sequence. There is no “water = 0” anchor you can carry across scans.

Why this single fact causes so much trouble

Because intensity is not a physical scale, three things follow that you will meet repeatedly:

  • You cannot directly compare MRI intensities across scans. Two MRIs of the same patient on different scanners (or even the same scanner with a different protocol) produce numerically different images of the same anatomy. Pooling them as if they were on one scale is a reproducibility error.
  • MRI radiomics is especially fragile. Texture features computed on intensities that are not on a comparable scale inherit that non-comparability (Ch. 3). Intensity normalisation, bias-field correction, and harmonisation are first-class variables, not afterthoughts.
  • “Weighting” is not measurement. A routine clinical MRI does not measure T1 or T2 as numbers; it makes a picture whose contrast is dominated by T1 or T2. The next lessons make this precise.

ESTABLISHED EVIDENCE (textbook MRI physics) underlies the concepts here; no claim is novel. The point is to install the contrast — CT ≈ a scale, MRI ≈ a relative contrast — before the physics, so the physics has somewhere to land.

Stop and think — then reveal

A colleague says “this tumour is brighter on the MRI than last month’s, so it has grown”. What assumption are they making, and why is it unsafe?

They are assuming the two MRI intensities are on the same scale — that “brighter” means a real change in the tissue rather than a change in acquisition. But MRI intensity depends on sequence, scanner, field strength, coil and parameters; unless the two scans used the identical protocol on the same scanner (and even then, carefully), “brighter” can simply reflect a different acquisition. The honest reading is “the tumour looks different under a possibly-different image formation” — to claim growth you need the same protocol, geometry (Ch. 2), and ideally a quantitative measure (volume, or a quantitative map like ADC, lesson 5), not raw intensity.

What to retain

  1. MRI intensity is relative and sequence-dependent, not a physical scale. There is no “water = 0” anchor that carries across scans.
  2. Therefore you cannot directly compare MRI intensities across scans; MRI radiomics is especially fragile; and “weighting” is not measurement.
  3. CT ≈ a scale (HU); MRI ≈ a relative contrast. Hold that contrast and the rest of the chapter has somewhere to land.

Next: build the physics behind that contrast — what happens to hydrogen after the RF pulse.