Relative intensity and quantitative limits

What the intensity value means

A CT reader can often attach a physical interpretation to a reconstructed value because the image can be expressed on the Hounsfield scale. Water is defined as 0 HU and air is conventionally near -1000 HU, although CT values still depend on energy, reconstruction, noise, and other protocol conditions (Chapter 1). The scale provides a reference against which attenuation measurements can be discussed; it does not make every CT value independent of how the image was formed.

An ordinary MR magnitude image has a different status. Its stored intensities are the result of a signal-generation and reconstruction process in which proton density, relaxation, sequence timing, flip angle, field strength, receive-coil sensitivity, receiver scaling, and reconstruction all matter. The resulting number is usually a relative or arbitrary intensity. A value such as 420 has no useful standalone meaning: it becomes interpretable only in relation to other tissues in that image and to the acquisition that produced it. There is no universal water-zero anchor for carrying that number across MRI examinations.

Relative does not mean meaningless. Within one image, differences in intensity are the very contrast that allows anatomy and pathology to be seen. The mistake is to promote a within-image relationship into a cross-scan calibration without checking the image- formation conditions. A lesion that is brighter in a later examination may have changed, but it may also have been measured with a different sequence, coil, receiver setting, geometry, or reconstruction. Even when a protocol is intended to be repeated, the comparison should record what was actually acquired rather than relying on a display impression.

Quantitative limits are not a dead end

The absence of a universal raw-intensity scale changes how quantitative work is designed. For a question about size, a geometry-based volume is a different and usually more direct measurement than the mean brightness of a lesion. For a question about tissue signal properties, a quantitative map can be estimated from several acquisitions and a stated model. An ADC map, developed later in this chapter, illustrates the compromise: its values have physical units, but they remain dependent on the b-values, fit, noise, distortion, and other acquisition conditions.

Intensity normalisation, bias-field correction, and harmonisation can make a dataset more suitable for a comparison, but they do not turn the original intensity into a universal physical measurement. Each transformation defines another processing step, with assumptions and parameters that belong in provenance and must be kept separate from the tissue property being inferred. A radiomic feature computed after that transformation is a feature of the transformed image, not of an abstract scanner-independent intensity.

This is the practical mental model: CT and MRI both produce useful images, but their values enter quantitative analysis through different contracts. MRI contrast is shaped by how magnetisation is prepared and allowed to evolve. The next section opens that process at its physical starting point: excitation and relaxation.

Sources

• Abi Berger, “Magnetic resonance imaging” — magnetic/RF signal formation, pulse sequences, and the dependence of image contrast on tissue relaxation.
• Wiltgen T, Voon C, Van Leemput K, Wiestler B, Muehlau M. Intensity scaling of conventional brain magnetic resonance images avoiding cerebral reference regions: A systematic review. PLoS One. 2024;19:e0298642. doi:10.1371/journal.pone.0298642. (Full text)
• DenOtter TD, Schubert J. Hounsfield Unit. StatPearls. Updated 2023 Mar 6. PMID 31613501. (NCBI Bookshelf)
• Boss MA, Malyarenko D, Partridge S, et al. The QIBA Profile for Diffusion-Weighted MRI: Apparent Diffusion Coefficient as a Quantitative Imaging Biomarker. Radiology. 2024;313:e233055. doi:10.1148/radiol.233055. (Full text)