TwinTree Insert

09-02 Definition of Contrast

here is only one step from picture ele­ments to image contrast. Contrast de­scrib­es the relative difference of intensities of two ad­jacent regions with­in an examined ob­ject on a gray or color scale: their visibility. It is a quite con­tro­­ver­sial term in medical imaging.

Several definitions have been proposed over the years.

It is difficult to give an exact defi­nition on a conventional ana­logue x-ray image. It is merely qualitative, except when using a special measuring device — or digitizing the ana­logue image.

Digitalization of images in nuclear medicine and x-ray CT opened the door to more straight­for­ward quantitative approaches. Now, picture ele­ments are available whose gray-scale inten­sity can be expressed in numbers. The nu­merical dif­ference bet­ween two intensities allows quantitative definition of contrast.

If there is no difference between two neighboring pixels, they cannot be dis­tin­­gu­ish­ed: no contrast exists. The big­ger the difference in the intensity of two pixels, the better will be the contrast (Fig­ure 09-06).

Figure 09-06:
Two examples of four neighboring volume elements.
Top: In the first case the four voxels have the same relative signal intensity (SI), and in the resulting image they cannot be dis­tin­gui­shed from each other.
Bottom: In the second case, they have different relative intensities and thus they can be distinguished from each other in the final image.

The common quantitative definition of contrast is given by the following equation which was adapted from the original expression for optical visibility as proposed by Michelson in 1927 [⇒ Michelson 1927]:

C = (Ia - Ib) / (Ia + Ib)

where C is contrast and Ia and Ib are the signal inten­sities of two adjacent pixels or voxels.

It is important to understand that signal in­tensity in magnetic resonance ima­ging is not standardized. MR imaging does not possess any correlation to Houns­field units in x-ray CT. The signal intensity of an MR image can represent a mixture between T1-, T2-, and ρ-values, flow, diffusion, perfu­sion, and other factors influencing the sig­nal emitted by structures within a volume element.

Only normalization of images, e.g., with a water-filled vial outside the pa­tient's body, allows an approximation to be made and can be used to calculate relative sig­nal in­­ten­si­ties, which then can be compared. However, these values are only semi­quan­ti­­ta­tive. They vary between different MR equipment and have no diagnostic value.