Image quality

In all imaging methods there are many factors which influence image properties and quality. X-radiography is mainly described in this chapter, but the following concepts are usable in all medical imaging methods. Fig. 12 shows factors and parameters divided into eight groups which have significant effects on image quality.

What is image quality? An image is always a two-dimensional presentation of a three-dimensional organ (possibly 3D, if slices are taken side by side or if changes in time are taken into account). The chain from target tissue via transmitted or emitted radiation to the interpreter's brain (with a more or less definitive understanding of the normal and pathological findings) is long and complicated. There are factors which influence the interpretation in different directions. The first part of the chain is rather well known; physically measurable quantities: contrast, resolution, and noise do exist. The terminal parts of the chain, i.e. the tasks and functions of the eye and brain for the detection and recognition of findings are not so well known.

Connection between the physical quality parameters and diagnostic applicability is difficult and laborious to determine objectively. To compare two imaging methods objectively it is possible to measure sensitivity and specificity for studied objects and diagnosis, and express results in the form of a so called ROC-analysis (Receiver Operating Curve). Randomized groups of images are compared by several interpreters. The functioning point of each interpreter is situated (he uses a threshold between pathological and normal findings) somewhere in the ROC-curve.

Physical parameters of the image

Contrast, resolution, noise, and signal to noise ratio as well as the position of the image and image portions on a grey scale are important parameters in all imaging. By changing the latter in a digital image (windowing) the usability of the image can be greatly improved (see chapter on Digital image processing).

Contrast (the difference of blackness or optical density of film between nearby areas in an image, Figs. 2 and 10) is caused by properties of tissues and properties of the film or other radiation detector. The difference of transmitted X-ray intensities in Fig. l0 can be normalized with the sum of intensities resulting in contrast scale from 0 to 1. The human eye can detect differences in contrast of 0.02 under good light conditions. With digital methods it is possible to go down to the 0.001 level (Fig.13). While viewing images on a lightbox it is important to mask light from any side of the image reaching the eye. A magnifying glass or a bright light should be used when necessary. These measures improve greatly the detection of low contrast areas.

Many factors which strongly affect contrast have already earlier been described. Noise in an image also frequently plays an important role particularly at low contrast levels. Noise can be seen as a locally variant, fine or rough change in optical density which occurs even in an image of an evenly exposed water or Plexiglas phantom. There are two principal causes of noise in an X-ray image:

1. The number of X-ray quanta varies both as a function of time and site (quantum noise, statistical nature of radiation).
2. The construction of the film, screen and image amplifier, as well as the electric circuits in the imaging devices causes noise.

Relatively few X-ray quanta per exposure are collected while working with fast screen-film combinations and a very grainy image may result, like spots of rain on an asphalt surface. With normal or slow screen- film combinations (as well as with mere film) much more quanta, "raindrops" (dose increases) are detected, and an image looks as if it is "calming" out, and not so noisy. The signal-to-noise ratio is a fundamental concept, with which image characteristics can quantitatively be compared, particularly in digital imaging methods.

In the image of a sharp edge the film blackness changes to another level of blackness on a short distance serving as a measure of image sharpness. Unsharpness is caused by many factors in imaging chain; focus size, motion of the object, thickness of the screen, geometrical factors etc. Sharpness is in practice defined by spatial resolution, which tells how many details or lines (line pair, lp) can be distinguished for instance in one mm (unit lp/mm). The following values are in general use:
- 20 lp/mm (film alone)
- 10 lp/mm (normal screen-film-combination)
- 5 lp/mm (fast screen-film-combination)
- 1-2 lp/mm (image intensifier-television chain, magnetic camera)
- 1 lp/mm (CT device and ultrasound device)
- 0.1 lp/mm (gamma camera)

Resolution can also be defined as the smallest distance (mm) between two objects, which can be separated from each other in an image. The concept of modulation transfer function, MTF is also useful when comparing the contrast and resolution properties of imaging methods.

Comparison of imaging methods

In Fig. 13 general imaging methods are compared in terms of resolution and contrast. Both, i.e. low contrast level and small details, can not be achieved simultaneously. For instance contrast must be high when small details are to be separated, (e. g. in the inner ear or fine structure of bone). On a thorax image "shadows" in lung parenchyma are usually looked for;

the areas of such lesions must therefore be relatively large especially if their density does not differ much from the surroundings. Low contrast at an edge, large focus size and geometrical magnification make it more difficult to see small objects in thorax imaging. Different imaging methods can consequently be used depending on the information which one is seeking.

Digital imaging methods improve the ability to register small contrast differences (Fig. 13). The aforementioned windowing method can play a central role in the image interpretation performed in modem image work stations, where digital images from different devices (CT, DSA, magnetic resonance imaging, nuclear medicine, PET, ultrasound) are compared and analyzed. Fig. 13 contains no information regarding the dynamic characteristics of imaging methods or of noise.

Aaro Kiuru