Currently Available Commercial Image Processing Methods

Currently available image-processing methods produce six different effects on image appearance. They define the ''window of clinically useful exposure data,'' they affect the degree of image blackness and image contrast, they equalize image blackness in different parts of the image, and they provide edge enhancement or provide blurring of noise in parts of the image. The steps described below have been written in a logical progression, but may or may not be actually carried out in this way or in this sequence. More detail on these methods is available [17,18,19].

Defining the "Window of Clinically Useful Exposure Data"

Digital acquisition systems are designed to record data from a wide range of X-ray exposures in a relatively linear relationship of exposure to pixel value. This wide range of recorded exposure is desirable because it can correct for X-ray exposure errors and for patients with different body builds. If one used this ''raw'' data to create the images, they would be of very low contrast. By design, the width of the window of pixel values set for accepting this information is set very wide.

The raw data that has been received by the imaging device contains both useful and nonuseful exposure data. Nonuseful data, for example, include the exposure that occurs from X-ray photons that passed through the area outside of the patient and areas outside of the collimated field. Because the imaging device received unimportant exposure data, the digital data will contain unimportant exposure data. To cope with this, the first step in image processing is to define what data is likely to represent the data encoded as the X-ray photons pass through the body.

At least two different methods are involved in this analysis. One defines the edge of the collimated field and excludes the data outside of the collimated field from further analysis. Different companies use different proprietary methods for this purpose and these methods are moderately effective.

The second method uses information from the exposure histogram. The exposure histogram is a map of the different intensity values of the picture elements (pixels) that form the image. The algorithm selects those pixel values from the exposure histogram that are likely to contain clinically relevant information. For this step, the computer uses a search algorithm that is seeking the region of the remaining data that looks like chest radiograph data; one can think of this as a search for the shape within the exposure histogram that matches the shape of a chest radiograph histogram. The data for different parts of the body are different. The algorithm has been set to look for the data patterns seen in different body parts and the different common variations that are seen for each body part. Once the algorithm identifies the appropriate shape, it then uses this to defines the upper and lower boundaries of exposure data to be included in the formation of the image. This subset of the original data is then used to direct the next steps in image processing.

Poorly processed digital chest images usually result from problems occurring at these stages of the process. The most common causes are a failure to instruct the digital system as to the type of image to be processed or excessive scatter outside the edge of the collimated field. The digital system has to be told that the data set it is looking for is a chest radiograph. If it is told that a different part of the body is being imaged, it will look for data that looks like that body part and process the image data incorrectly. If there is a lot of scatter in the area outside the collimated field, it may not correctly define the edge of that field and may incorporate the scattered exposure into its defined exposure boundary. This is one of the reasons why it is important to avoid overexposure. On an overexposed image, the algorithm is more likely to include data that is due to scatter and is therefore not in the clinically relevant region of exposure.

Adjusting the Data for Differences in Exposure: The Fuji "S" Number, the Initial Adjustment for Display Optical Density

The method used by the system to define the ''clinically relevant data'' does not depend on the exposure used to obtain the original data set. In order to display the ''clinically relevant data'' as an image with proper display optical density or luminance, the system must correct for exposure differences. It does this by comparing the pixel values for the ''clinically relevant data'' to a stored expected range of pixel values. It then adjusts the ''clinically relevant data'' to lie within the range for proper display. The degree of adjustment is related to a number called ''S.'' If the extracted ''clinically relevant data'' is underexposed compared to the range of pixel values for display, the ''clinically relevant data'' will be adjusted so that it falls within the display range by amplifying each pixel value. If it is overexposed, then the computer program will decrease each pixel value. The Fuji system was designed so that an ''S'' value of 200 corresponds to the correct level for display of the image. The ''S'' value is not the speed of the system. The ''speed'' of an analog system is defined as the exposure level required for a specific degree of film optical density. In the digital systems, the optical density of the display is not determined by the exposure of the digital receptor. There is no way to equate the speed of a screen-film system to the proper exposure for a digital system. An ''S'' of 200 may have originally been based on the exposure required for a 200-speed screen-film system, but as the Fuji systems evolved, the imaging plates have become more sensitive to X-ray photons while the ''S'' number system has remained the same.

Final Adjustment for Image Mean Optical Density

The ''S'' number adjustment will usually place the image data set close to the range for proper display. There is, however, some variability in laser printers and in soft-copy display devices, so the system makes it possible to adjust the final optical density or luminance and then store this change so it can be used for subsequent radiographs. In the Fuji system, this factor is called the GS factor.

Adjustment for Display Contrast

In the process of selecting the ''clinically relevant data,'' the computer program selected the highest and lowest pixel values of this data. This constitutes an initial estimate of the contrast range for the final image. It does not define the shape of the contrast curve or the steepness of its midportion. The final adjustment is made by use of a conversion method called a look-up table. In this method, there is either a stored table or a stored formula that says for a data pixel value of ''x,'' use a value of "y" for the display. The companies recommend specific look-up tables for specific parts of the body. The radiologist can accept the companies' recommendations or choose different look-up tables. Once chosen, this selection can be stored in the computer memory so that the same look up table can be consistently chosen. In the Fuji system, this factor is called the GT factor.

Once this overall look-up table has been selected, one can accept the company's default setting or one can adjust the slope of the midportion of the characteristic curve. This factor is called the GA factor in the Fuji system. The higher the GA, the steeper the slope of the midportion of the characteristic curve, and the less steep the bottom and top of the characteristic curve.

High-Spatial-Frequency Enhancement: Edge Enhancement

Mathematically, all images can be transformed (using the Fourier transform) into groups of sine and cosine waves. The waves are of different frequencies and it is therefore possible to describe images as having different spatial frequencies. All images are composed of parts with different spatial frequencies. High spatial frequencies are seen in where there are sharp edges; lower spatial frequencies are seen where there are blurred or unsharp edges. Filtering methods exist to enhance or diminish the intensity of specific groups of spatial frequencies in an image. By enhancing high spatial frequencies in an image, sharp edges are enhanced and appear sharper, but since image noise is also of high spatial frequency, it is also enhanced. If high spatial frequencies are Altered from the image, then the edges of sharply defined objects and visible noise are decreased.

Edge enhancement is an important part of the image processing of digital chest radiographs. The edges of all objects in standard chest radiographs are somewhat blurred because of the penumbra caused by the X-ray tube focal spot size and often by slight patient motion and cardiac pulsation. Slight edge enhancement restores these edges toward the appearance they would have in a specimen radiograph with a very small focal spot X-ray tube. It enhances the conspicuity of blood vessels, interstitial markings, surgical staples, and pneumothoraces (Fig. 8). This method also enhances the edges of catheters, making them easier to see. It is normally included as part of the recommendation for image-processing settings for film display. Some soft-copy display systems, however, do not provide the capabilities for edge enhancement and diagnostic accuracy in these systems may be less (20).

High-Spatial-Frequency Blurring: Noise Blurring

High-spatial-frequency blurring can be used to decrease the visibility of noise in digital images. Because noise from quantum mottle makes some digital images appear grainy, image processing methods were provided so that the images would look smoother, more like screen-film images. If such processing were applied to the entire image, then fine interstitial lines would also

Figure 8 Improved visibility of pneumothorax using edge enhancement in a patient status post surgery for recurrent spontaneous pneumothoraces. The surgical clips and the small residual pneumothorax are more conspicuous in (a) with mild edge enhancement than in (b) with no edge enhancement.

blur and be more difficult to detect. For this reason, some programs apply the image blurring only to areas of relatively low pixel values (more transparent regions of the original image. The use of these noise concealing programs unfortunately also blurs the edges of high-frequency structures, such as catheters in the mediastinum (see Figs. 7d and 7f).

Optical Density Equalization: Methods to Bring the Entire Image to an Optical Density Where High-Contrast Display Is Possible

The range of optical densities on a chest radiograph is quite wide. This is the result of the differing proportions of water, calcium, and air in the path of the X-ray photons in different parts of the image. Because of this wide range of densities, all chest images represent a compromise: To have the lungs at a proper display density, the mediastinum is too light. If one wants to see tubes in the esophagus behind the heart, then one can increase the exposure, but the lungs become too dark. Optical density equalization image processing is used to partially correct for these density differences. Current methods are moderately good, but in the future, newer methods will probably be better.

The fundamental concept underlying methods for optical density equalization is that one can separate an image into components of different spatial frequencies. High spatial frequencies show the edges of objects, medium spatial frequencies demonstrate the shape of moderate sized structures, and low spatial frequencies tend to show broad effects across an image. It is the low-spatial-frequency information that is modified in optical density equalization. While different methods are used by different companies, in concept, if one takes the low-spatial-frequency image and changes areas that are black to white (and vice versa) and then partially adds it back into the original image, then larger areas of density differences will be decreased in their intensity and the whole image will show a lower range of optical densities. The resulting image will show a lower range of optical densities across the image allowing one to then enhance the contrast across the entire image. Enhancing the contrast results in smaller details being more conspicuous. This method is very helpful for enhancing interpretation of bedside chest radiographs. It has the disadvantage that large areas of lung consolidation or large pleural effusions will also be affected by the processing, making them less intensely white. Sometimes faint but larger areas of air space disease will be less conspicuous on these images. Also, the area over the liver sometimes looks more radiolu-cent and could appear similar to the radiolucency that can occur on supine films with free abdominal air.

Figure 9 Energy subtraction used to confirm calcification in nodules. (a) A digital chest image with standard processing shows several small lung nodules. (b) The calcium emphasis image is shown and the small nodules can be seen to contain calcium.

Figure 9 Energy subtraction used to confirm calcification in nodules. (a) A digital chest image with standard processing shows several small lung nodules. (b) The calcium emphasis image is shown and the small nodules can be seen to contain calcium.

Energy-Subtraction Imaging

Energy-subtraction imaging is based on the kilovolt dependence of absorption of calcium and water. At lower kilovolt levels, calcium absorbs or scatters moderately more X-ray photons than water. At higher kilovolt levels, the absorption levels are more similar. By obtaining a chest radiograph at two different energy spectra, one can process the images to create a bone-emphasis and a soft-tissue-emphasis image. This acquisition can be done either using one exposure where two imaging plates are exposed, but with a layer of copper

Figure 10 Energy subtraction used to enhance visibility of a central mass. A squa-mous-cell lung cancer involving the right upper hilum is seen. The energy subtraction processing has greatly decreased the visibility of the ribs and enhance the conspicuity of the mass.

Figure 11 Energy subtraction used to enhance visibility of airways disease. Digital chest radiographs with (a) standard and (b) soft-tissue-emphasis energy-subtraction processing in a patient with mucus plugs and bronchiectasis from cystic fibrosis. The soft-tissue-emphasis image largely eliminates the conspicuity of the ribs. The lung disease is more conspicuous.

Figure 11 Energy subtraction used to enhance visibility of airways disease. Digital chest radiographs with (a) standard and (b) soft-tissue-emphasis energy-subtraction processing in a patient with mucus plugs and bronchiectasis from cystic fibrosis. The soft-tissue-emphasis image largely eliminates the conspicuity of the ribs. The lung disease is more conspicuous.

or other absorber between them so that the energy spectra are different, or by taking two exposures one right after the other at different kilovolt peak settings. The images must then be closely matched so that the chest is exactly the same size and position. This is done by a process called warping, where one image is distorted (warped) to be the same size and shape as the other. Subsequently, the images are registered (matched in position) as closely as possible. They can then be processed through a complex series of steps that eventually yield the bone-emphasis and soft-tissue-emphasis images (Figs. 911). With the current commercial system, slight image blurring occurs in this process, so that the finest intersitial lines may be less conspicuous, but at the same time, areas of infiltrate and small masses become more conspicuous. It is likely that future developments both in acquisition devices and software will improve the quality of energy subtraction and increase its value.

AMBER system

The AMBER system is a special type of digital chest radiographic system that scans the chest adjusting the intensity of the X-ray beam according to how much radiation has penetrated the chest. It continually adjusts the amount of exposure so that more radiation is applied where there is more absorption (as in the mediastinum) and less over the lungs. It produces very high quality images.

Was this article helpful?

0 0
5 Easy Ways To Stop Smoking

5 Easy Ways To Stop Smoking

Your first day without cigarettes can be difficult, but having a plan will make it easier! Learn what steps to take on the day you quit smoking.

Get My Free Ebook


Post a comment