• AI
  • Molecular Imaging
  • CT
  • X-Ray
  • Ultrasound
  • MRI
  • Facility Management
  • Mammography

Tapping power of digital x-ray boosts workflow


Systems that capture digital x-ray images possess the flexibility needed to tailor processing for optimal appearance. Sophisticated new techniques can exploit this flexibility to great advantage, while also leading to a more efficient departmental workflow. Because computed radiography and flat-panel digital radiography are still relatively new to many radiology practitioners, however, the image processing power of these systems sometimes remains untapped or not fully used.

A primer on the image processing power of CR and DR systems could help radiology practitioners get the most out of this equipment. These systems can capture a wide range of x-ray exposures in a single image, covering the spectrum of x-ray energies used in most examination types. A key benefit of this wide exposure latitude is the reduction or elimination of retakes due to under- and overexposure. While reduced retakes decrease patient dosage and improve workflow efficiency, images produced by CR and DR equipment still require processing before interpretation by a radiologist.

The fundamental purpose of image processing is to automate the transformation of raw digital data into a display-ready form. Image processing consists of two main components: automatic recognition of the diagnostically relevant regions, and automatic rendering of the region of interest. Image processing performance, as gauged by the ability to consistently produce images with optimal diagnostic quality, directly affects radiologists' confidence in making diagnoses. Because the diagnostic quality of the rendered image provides the basis for a technologist to judge whether to reprocess an image, the ability to obtain sufficient diagnostic quality automatically also improves workflow.


Sophisticated pattern recognition and segmentation techniques automatically classify x-ray images into regions that are diagnostically relevant or irrelevant. Diagnostically relevant regions correspond to anatomy, while the diagnostically irrelevant regions correspond to collimated and direct exposure areas. Direct exposure regions have higher exposure than anatomic regions, and collimated regions have lower exposure. It is therefore critical for image processing algorithms to reliably identify and then exclude these irrelevant regions to make effective use of the dynamic range of display devices. Attempting to display the full range of exposures without first excluding the diagnostically irrelevant regions almost always results in a flat-looking image with inadequate contrast for diagnosis. Robust image processing also automatically detects and excludes features such as prosthetic implants, pacemakers, and left-right markers.

Rendering the region of interest for display generally consists of gray-scale remapping, signal equalization, and sharpness restoration. All major CR and DR providers implement some form of these fundamental rendering functions, although the details of each method differ, as does the appearance of the images each renders.

Gray-scale remapping is the digital equivalent of the H and D curve in screen-film systems, and it is generally implemented as a simple lookup table that maps raw pixel values to display values. In the absence of other processing steps, gray-scale remapping establishes overall contrast, latitude, and average brightness for the image.

Even the use of this most basic function alone provides tremendous flexibility in optimizing diagnostic quality by constructing different lookup table curve shapes.

Gray-scale remapping is usually a database selectable function that is indexed by examination type. In the same way that contrast and latitude are inversely linked for screen-film systems as defined by the H and D curve, contrast and latitude are related equivalently for digital radiography via gray-scale remapping. One significant difference between the H and D curve and gray-scale remapping is that the average brightness of a digitally captured image is controlled by the accuracy of the automatic estimation of the region of interest, as opposed to exposure level. This is important because reprocessing is necessary when the automated processing fails to properly estimate the exposure, and this reprocessing can decrease technologist productivity.


Using signal equalization can lead to a sharp reduction in retake rates. Usually implemented as some form of spatial frequency decomposition and reconstruction, equalization processing has the net effect of lightening dark regions and darkening light regions while preserving detail contrast. Equalization extends the range of exposures that are rendered with full detail contrast, thereby reducing diagnostic quality sensitivity to exposure recognition errors. This relaxes the trade-off between contrast and latitude, and, consequently, fewer images will require reprocessing. The effect on image appearance is also extremely powerful.

The degree of equalization is usually controlled by a group of database parameters and can be set based on radiologists' viewing preferences. Clinical studies have demonstrated that radiologists prefer equalized images because they lead to greater diagnostic confidence. In one study, more than 90% of images processed using equalization were rated by radiologists as either satisfactory or optimal for diagnosis, compared with just 55% of images processed without equalization.1 Equalized images help improve radiologist reading efficiency as much as 20% by mitigating the need to adjust window/level settings at the diagnostic workstation.2


Sharpness restoration enhances image details by compensating for blurring effects that are introduced during radiographic image capture and display. Spatial processing methods that make use of multifrequency decomposition offer a means to boost contrast selectively for features of different sizes in the image. The visual effect of multifrequency contrast enhancement is to sharpen the image.

Sharpness restoration for very small features must be constrained in order to balance sharpness appearance with noise boost. Sharpness restoration algorithms that allow the amount of boost to vary as exposure level and background brightness change generally provide a much better overall noise and sharpness balance in the rendered image.

Collimation masking further improves the ability of radiologists to perceive contrast in the region of interest by mitigating viewing flare coming from these otherwise very bright regions. Some commercially available image processing packages also provide a feature for automatic application of a black or gray mask to collimated regions. Alternatively, some systems allow the mask to be applied manually through the user interface.


It is critical that default parameter settings be established based on radiologists' viewing preferences to maximize the diagnostic quality of automatic image processing.3 The ability of the image processing algorithms to consistently deliver the preferred rendering has direct impact on workflow efficiency. Most commercially available CR and DR systems provide a mechanism to index a particular rendering aim based on examination type. In practice, technologists specify the body part and projection before the image is rendered.

The consistency with which the image processing algorithms produce the desired appearance directly affects the diagnostic confidence of radiologists and workflow efficiency for staff. Establishing goals for radiographic image rendering can be challenging for several reasons: the number of parameters available for adjustment, interdependence of the parameters and their effect on image appearance, the number of body part and projection combinations that need to be optimized, and inadequate tools. Technical specialists can work with radiology departments to facilitate this process.

Technological advances offer opportunities to smooth workflow while retaining or even improving diagnostic quality. As computer speeds continue to increase and memory costs decline, image processing algorithms can perform more computationally intensive analyses and still meet high-throughput performance requirements for clinical systems. Although new development is primarily directed toward improving diagnostic quality, emphasis on improving workflow is increasing.

One key opportunity area is in the technologist's interaction with image processing techniques, which can be simplified or eliminated. Image processing software packages available today perform many functions automatically, but some repetitive user interaction is still required. The body part and projection for each image, for example, must be specified for each cassette at the cassette reader before images are scanned using CR systems. The quality-control step that technologists perform for each image is another time-consuming element in the workflow. The QC process occurs prior to image dissemination to the PACS and consists of visual assessments of positioning, patient motion, and image quality.


Emerging image processing methods address some of these repetitive functions and at the same time have the potential to improve diagnostic quality. Software is emerging that can automatically recognize radiation fields in x-ray images independently of body part and projection.4 Such an algorithm would eliminate the need for technologists to provide specific examination type information.

Algorithms that automatically segment individual radiation fields from a multiple-exposure CR image5 have the potential to increase image throughput for busy orthopedic departments. Segmentation of the radiation fields in a multiple-exposure image also logically improves diagnostic quality because the anatomy in each exposure is individually optimized. Automatic identification of image features associated with body part and projection can also be used to assess positioning in the QC process and may be exploited further to automatically reorient images to the preferred hanging protocol for the radiologist or clinician, eliminating the need for technologists to rotate and flip images.

Reengineering the algorithms used to render digital radiographs makes a dramatic simplification of the interface to image processing functionality possible. Earlier this year, scientists reported a new approach for adjusting "look preferences" for digital radiographs that is based on familiar and intuitive analog screen-film concepts.6 The parameters directly control five fundamental attributes of image quality: brightness, latitude, detail contrast, sharpness, and noise appearance.

The interdependencies among the parameters have been removed. Latitude, for example, can be increased or decreased for constant contrast. Similarly, contrast can be increased or decreased for constant latitude. The rendering method enables interactive control of these image quality attributes to facilitate the determination of rendering aims for different examination types. This new rendering algorithm and associated workstation tool will lead to improved default processing parameters and greater diagnostic quality. They could greatly reduce the need for technical specialists to help with onsite parameter optimization.

New image processing features offer the potential to achieve higher levels of diagnostic image quality while simultaneously enhancing workplace efficiencies. These features will soon be commercially available from manufacturers of CR and DR imaging systems.

Mr. Foos is director of the Health Imaging Group's Research Laboratory at Eastman Kodak Company in Rochester, NY.


1. Van Metter R, Foos D. Enhanced latitude for digital projection radiography. Proc SPIE 1999;3658:468-483.

2. Krupinski E, Radvany M, Levy A, et al. Enhanced visualization processing: effect on workflow. Acad Radiol 2001;8:1127-1133.

3. Flynn M, Couwenhoven M, Eyler W, et al. Optimal display processing for digital radiography. Proc SPIE 2001;4319:298-305.

4. Wang X, Luo H., Automatic and exam-type independent algorithm for the segmentation and extraction of foreground, background, and anatomy regions in digital radiographic images. Proc SPIE 2004;5370:1427-1434.

5. Wang X, Luo J., Senn R., Foos D. Method for recognizing multiple radiation fields in computed radiography. Proc SPIE 1999;3661:1625-1636.

6. Couwenhoven M, Senn R, Foos D. Enhancement method that provides direct and independent control of fundamental attributes of image quality for radiographic imagery. Proc SPIE 2004;5367:474-481.

Related Videos
Does Initial CCTA Provide the Best Assessment of Stable Chest Pain?
Making the Case for Intravascular Ultrasound Use in Peripheral Vascular Interventions
Can Diffusion Microstructural Imaging Provide Insights into Long Covid Beyond Conventional MRI?
Assessing the Impact of Radiology Workforce Shortages in Rural Communities
Emerging MRI and PET Research Reveals Link Between Visceral Abdominal Fat and Early Signs of Alzheimer’s Disease
Reimbursement Challenges in Radiology: An Interview with Richard Heller, MD
Nina Kottler, MD, MS
The Executive Order on AI: Promising Development for Radiology or ‘HIPAA for AI’?
Related Content
© 2024 MJH Life Sciences

All rights reserved.