Effect of quantum mottle on radiographic image quality.
Radiographic image quality characterizes the ability of an imaging system to accurately depict structures in a radiographed object. Six components affect image quality -- spatial resolution (or sharpness), contrast, density, radiographic mottle (or noise), distortion and artifacts. Of these six parameters, radiographic mottle is the most difficult to understand and relate to the imaging process. Radiographic mottle creates a grainy, blotchy, textured or snowy appearance in a radiographic image. (See Fig. 1.)
If an image were obtained of a phantom of uniform composition and thickness, the resulting film would have an irregular appearance. Optical density is not consistent throughout the image of the phantom, but rather small differences are present among nearby regions. One reason for this variation is the manifestation of radiographic mottle. Consequently, the imaging process does not render a completely faithful reproduction of the object. High levels of noise produce more variation in film optical density and inhibit the depiction of low-contrast structures.
Film graininess, nonuniformity of screen phosphor and quantum mottle contribute to overall radiographic mottle. Of these three factors, quantum mottle is the dominant component and is the most important consideration regarding noise in the image. Quantum mottle is defined as the statistical fluctuation in the number of photons per unit area that contribute to image formation.
Subject contrast -- in the form of varying x-ray intensity exiting the patient -- must be detected and then converted to an observable form by the imaging system. X-ray photons absorbed in the intensifying screen become the initial information carriers for image formation. However, the detected photos exhibit statistical variations because x-ray emission from the target and interactions with matter, including the patient and the screen, are random events. Quantum mottle becomes more pronounced when the number of x-ray photos absorbed in the screen is decreased, such as by reducing the radiation dose.
The following analogy illustrates the concept of quantum mottle. Imagine that two flat, metal objects are placed on the sand at the beach. They could be of any shape or size, but suppose they are a simple circle and triangle. (See Fig. 2.) These two items are confined to a specific region with delineated boundaries -- a sandbox. A passing radiography student is blindfolded and then asked to place his or her finger at various points within the sandbox. The finger acts as a detector, sensing the nature of the items it touches. In this example, the metal objects feel hard to the student's touch and the sand feels soft. Metal or sand could be present at any location within the sandbox, and the exact locations and shapes of the metal objects are unknown to the student until he or she can sample all regions throughout the sandbox.
Suppose a second radiography student is asked to record, on a sheet of paper, the location and attributes (hard or soft) of every point that the first student touches in the sandbox. After the first student has sampled a few points, the blindfold is removed and he or she is asked to describe the shapes of the objects by studying the recorded data alone. (See Fig. 3.)
The student fails to recognize the shapes of the metal objects because the number of points he or she has sampled is insufficient to visualize their pattern. The total area probed by the student is very small compared with the total area of the sandbox.
As more points are sampled, the metal design begins to become discernible; additional sampling clarifies the structure even further. (See Fig. 4.)
Radiographic imaging is similar in many ways to the discrete sampling process employed by the two radiography students exploring the sandbox. In radiography, x-rays detected at the screen denote relative transmission through objects along the photons' paths. X-rays readily pass through radiolucent objects and are absorbed in the screen (represented in the analogy by the sand). Alternatively, x-rays completely removed by the object are not incident on the screen (represented in the analogy by the metal). The final image on film is a composite of these sampling events. Increased photon absorption by the screen (a higher number of information carriers in the imaging chain) reduces quantum mottle and improves visualization of the structure.
Medical imaging is a multistep process in which the measured parameter -- x-ray intensity -- is transferred to different types of information carriers along the imaging chain. The pathway for radiographic imaging consists of x-ray photons absorbed in the screen, visual light emitted by the screen, visual light absorbed in the photographic emulsion and converted silver halide grains by film development. The component in the imaging chain with the lowest number of information carriers establishes the level of quantum mottle in the final image. This component is called the quantum sink.
In the sandbox analogy, image formation was a two-step process: detection of hard or soft attributes and the recording of the results of sampling. Suppose the blindfolded radiography student conducts the sampling at such a rapid rate (see Fig. 5A) that some detected points fail to be recorded (see Fig. 5B). The information content of the recorded events is less than that originally detected.
Once information carriers are lost in one component of the imaging chain, no subsequent manipulation can recover this missing information. Amplification, or the conversion of x-rays to visible light in the screen, multiplies the number of information carriers but does not improve the information content. (See Fig. 5C.) The spatial distribution of interactions measured by the detector is not restored.
Because the recording step limits the information content of the final image, this step becomes the quantum sink. The quantum sink denotes the operation in which quantum mottle has the major influence upon image resolution. The initial process of detection establishes the maximum information content that cannot be enhanced by -- and often is not preserved by -- other components in the imaging chain.
Limitations in the ability to depict radiographic detail are imposed by properties of the imaging chain. The analogy can be extended to consider the effect of finger size on image sharpness when probing metal objects in the sand. The physical dimensions of the student's finger represents the area over which each measurement is obtained. Assessment of attributes is not possible for spatial dimensions smaller than the size of the finger. This dictates the minimum size of recorded events used to form the composite picture. The extent of the sampled area probed by a single finger placement constitutes the smallest detectable object and influences spatial resolution.
In Fig. 6A, the circle plus triangle pattern has been modified to include two metal strips placed very close together. A student with a large finger would not be able to detect the region of sand between the two narrow strips. (See Fig. 6B.) Thus, objects separated by a distance less than the dimensions of the finger would appear as one structure in the recorded image. Sampling with a small finger reduces the area corresponding to a detected event and improves the spatial resolution, as shown in Fig. 6C.
The principle demonstrated by variation in finger size is applicable for understanding the effect of phosphor layer thickness upon image quality. Light photons produced by the absorption of an x-ray diverge toward the film. Film darkening is not confined to the point of x-ray interaction, but encompasses a larger area. Increasing the screen thickness degrades the spatial resolution by allowing the light photons to expand over a wider area before reaching the film. Many other factors, including focal spot size, screen/film contact and screen phosphor crystal size, also contribute to image blurring by effectively increasing the sampled area for each detected event.
Detection of Small Density Differences
Until now, one of two possible results has been recorded each time the radiography student places his or her finger in the sandbox: The Student either touches metal or touches sand. Changing the analogy to replace the student's finger with an x-ray beam, the x-rays either interact with the metal (and are removed from the beam) or do not interact with the sand (and are absorbed in the screen).
In this example, it's all or nothing -- either 100% of the photons pass through a structure or none do. The anatomical structures in a patient, however, do not exhibit such a large difference in the attenuation of x-rays. For example, 90% of the incident photons might pass through one anatomical structure but only 70% percent might pass through a neighboring structure. A good quality radiograph must provide visualization of these small differences in subject contrast.
The complexity of the sampling process its compounded by the random nature of attenuation. Photons incident on the structure of interest may penetrate the structure and subsequently be detected by the screen, in which case they participate in image formation. However, other similar photons may be absorbed in the object and excluded from contributing to the image. The outcome for each photon is uncertain and varies according to probability of events along a path.
Assume metal and sand are replaced with muscle and water, but the same geometric shapes are maintained. These structures now have nearly identical transmission properties, creating lower subject contrast. The detected x-ray intensity is similar for each tissue type. The differentiation between muscle and water in the radiographic image becomes more difficult. Fig. 7 denotes the information carriers in similar shades of gray.
Contrast resolution describes the ability of the imaging system to distinguish structures with similar x-ray transmission as separate entities. At a particular noise level, large structures are clearly defined; small structures are less perceptible. As the noise level is decreased by increasing either the radiation dose or the absorption coefficient of the screen, smaller structures become visible. Since clarity of structures is enhanced by using more photons in image formation, the detectability of low-contrast structures is limited by the noise level.
Speed vs Noise
The relative number of x-ray photons necessary to cause film darkening is characterized by the term "speed." A screen or film classified as "fast" requires less radiation dose to produce a specific radiographic density. The evaluation of speed vs noise for each type of examination influences radiation dose and image quality.
Changes in system speed do not always alter the noise content in the image. For example, replacing a screen with one with a thicker phosphor layer creates a faster system without increasing the noise. Even though the number of x-ray photons incident on the screen is decreased, thus producing a lower radiation dose, the fraction absorbed by the screen is higher. Consequently, the total number of detected photons, or information carriers, introduced to the imaging chain is unchanged. Increased phosphor layer thickness affects image quality by reducing sharpness. The use of a phosphor with a higher absorption coefficient also can improve speed with no change in quantum mottle.
Some increases in system speed, such as faster film or phosphors with higher x-ray to light conversion efficiency, cause an increase in the noise level. By using a more sensitive film or by producing more light in the screen, fewer detected x-rays are necessary to darken the film, and thus quantum mottle becomes more pronounced.
Quantum mottle is inherent in the radiographic imaging process and cannot be totally eliminated. An understanding of quantum mottle and the interrelationship between spatial resolution and contrast resolution is essential for any critique of image quality.
The illustrative analogy described in this article shows the effect of discrete sampling on pattern recognition to help demonstrate the concept of quantum mottle.
[Figures 1 to 7 ILLUSTRATION OMITTED]
[1.] Sprawls P Jr. Physical Principles of Medical Imaging. Rockville, Md: Aspen Publications; 1987:297. [2.] Curry TS III, Dowdey JE, Murry RC Jr. Christensen's Physics of Diagnostic Radiology. 4th ed. Philadelphia, Pa: Lea & Febiger; 1990:202. [3.] Hendee WR, Ritenour ER. Medical Imaging Physics. 3rd ed. St. Louis, Mo: Mosby Year Book; 1992:447. [4.] Macovski A. Medical Imaging Systems. Englewood Cliffs, NJ: Prentice-Hall; 1983:76-77. [5.] Thompson MA, Hattaway MP, Hall JD, Dowd SB. Principles of Imaging Science and Protection. Philadelphia, Pa: W.B. Saunders; 1994:323.
W.R. Hedrick, Ph.D., is a medical physicist in the Radiology Department at Aultman Hospital in Canton, Ohio, and a professor in the Medical Radiation Biophysics Department at Northeastern Ohio Universities College of Medicine, Rootstown, Ohio.
Reprint requests may be sent to the American Society of Radiologic Technologists, Publications Department, 15000 Central Ave. SE, Albuquerque, NM 87123-3917.
|Printer friendly Cite/link Email Feedback|
|Date:||May 1, 1996|
|Previous Article:||The resurgence of TB and its implications for radiology.|
|Next Article:||Responsible dissemination of scholarly work in radiology.|