One of the mainstay technologies of the medical digital imaging digital transformation over the past 15 years has been computed radiography (CR). This technology has, for the most part, taken the place of wet chemistry-based film processors in the projection radiography workflow. With industry estimates of between 40% and 70% of all radiology imaging workload being performed through projection radiography, CR clearly is a serious contributor to the total volume of images being acquired, interpreted, and distributed for clinical purposes. This technology is not to be confused with direct radiography (DR), which produces digital images via a solid-state image receptor instead of using photostimulable phosphor-based imaging plates, as does CR. The purpose of this article is not to differentiate between CR and DR but to explore some of the quality issues associated with CR technology and the management of that technology in the clinical setting.

John C. Weiser, PhD (left), and John Romlein, MS

To this end, a survey of 10 topic areas pertaining to CR quality issues was developed and distributed to CR vendors. While there are many more topic areas and questions that we could have asked, these were deemed to be pertinent to the general quality of the CR devices and their ongoing quality-assurance management. The vendor responses were collected, tabulated, and analyzed with the goal of developing an overview of the state of the industry as to CR-related quality issues. The questions, analysis, and table are presented below.

1) Established QC Tools and Procedures

An ongoing quality control (QC) program is an essential part of the operation of a CR system.

Questions posed to vendors:

  1. Do you have a published QC program for your product?
  2. If so, does your QC program include a particular test phantom or phantoms, and the acceptable range of QC test values for these phantoms?
  3. Does your QC program include software that can automatically analyze the test phantom images? If so, which parameters are automatically measured?
  4. What parameters are measured using the test phantoms?
  5. Does the QC program produce an exportable set of data or a report?

Discussion:

Basic image-quality parameters, such as spatial resolution, noise, and low contrast object detectability, are included in most vendors’ QC programs. However, the methods by which these parameters are measured and interpreted vary between vendors. All vendors provide or specify a particular phantom test object to be used as part of their QC program, and two offer automated analysis and reporting software linked to a specific phantom as an extra cost option.

2) Spatial-Resolution QC Measurements

Spatial-resolution measurements are an important element of a QC program. Since the readout of the image is a 2D process, it is important to measure the spatial resolution in two directions at right angles. Sometimes, a single measurement is made at a 45° angle, but this type of measurement might mask a problem in one direction or the other.

Questions posed to vendors:

  1. In your QC program, do you measure spatial resolution in two directions?
  2. If you measure your spatial resolution at a 45° angle, how do you compare and reconcile the result to the vertical and horizontal spatial-resolution specifications of your system?

Discussion:

All but one vendor measure spatial resolution in two directions instead of using the shortcut method of measuring at 45°. In those cases where the 45° method is used, it is used as a comparison method against prior readings or a subsequent calculation is required to arrive at the horizontal and vertical values. The recommended method is to perform this measurement in both directions to arrive at a direct result instead of a derived or estimated value.

3) Bit Depth

Bit depth is an indicator of contrast resolution. Due to the wide dynamic range of CR image receptors, often there is a difference in the number of bits used for initial acquisition of the image and the number of bits used for the final processed image that is output to a printer or a PACS.

Questions posed to vendors:

  1. What is the “bit depth” at which the initial readout is obtained?
  2. What is the “bit depth” at which the image output is produced?

Discussion:

The answers to the question about the bit depth of the original readout ranged from 12 bits to 16 bits. The answers for the bit depth of the output image ranged from 10 bits to 16 bits. It is difficult to make any direct comparisons between products, because each vendor has a unique method of processing the acquired image and providing an output image that is mapped to the appropriate grayscale values for display. This is a case where “the more bits, the better” is not necessarily true. When considering the purchase of a CR system to use with a PACS, it is very important to understand the composition of the output image and how it will be displayed on the PACS workstations. Before deciding on a CR purchase, it is useful to determine if there are existing sites where the CR system is already operating with the same type of PACS as yours, and if so, to get some feedback from some of these sites. Otherwise, the interoperability of the CR with the PACS should be demonstrated prior to making a purchase commitment.

4) Repeat Analysis

The American College of Radiology (ACR of Reston, Va) requires a monthly retake analysis to be performed in ionizing radiation imaging environments. Computerized imaging systems, including CR systems, could facilitate this process if QC-reporting tools were built into the acquisition workstations of each modality.

Questions posed to vendors:

  1. What, if any, method do you provide to enable retakes to be monitored and analyzed?
  2. On what device does this retake date reside, and who operates the application?
  3. Of what QC parameters does the application allow monitoring?
  4. Does the retake-analysis software produce exportable data or a report? If so, please describe the output.

Discussion:

The incorporation of features for retake and rejected image data collection and analysis is supported by all but one of the vendors surveyed, but the implementation methods and functionality vary greatly. In some cases, the data collection and analysis is an optional software package, and in some cases, it comes standard with the CR system. In all cases, these functions reside on the individual CR-acquisition workstations with a privileged login required.

The amount and type of data collected varied from “all image/file data” to a simple log file with a listing of “rejected image type, operator, date, and time.” Two vendors saved the rejected images for later review—one in thumbnail size, one in full resolution. One vendor can configure an internal reporting feature to support reject reporting by the technologist. One vendor points to a future release where the analysis function can be performed from a centralized QA workstation. Although there is general support of the retake and reject analysis functionality, there is no general functional concept for what data is needed to perform this task efficiently and effectively.

5) DICOM GSDF Calibration

Calibration of the display device on the CR-acquisition workstation is critical to accurate image adjustments and quality control.

Questions posed to vendors:

  1. Can the display monitor of your acquisition workstation be calibrated to the DICOM grayscale display function (GSDF)?
  2. If so, does it require an external photometer, or is there an internal photometer?
  3. Can the display device be monitored using a third-party calibration-monitoring application?

Discussion:

It is useful to keep in mind that the American Association of Physicists in Medicine (AAPM of College Park, Md) considers the acquisition workstation monitor on a CR system to have the same calibration requirements as a diagnostic display if adjustments are made to the images prior to sending them to the radiologist for interpretation. All but one vendor’s systems allow the display monitor to be calibrated to the DICOM GSDF. Vendors are beginning to offer higher-resolution displays on the acquisition workstations with internal photometers and support the ability to include them in the centralized display QA program using centralized monitoring methods. The general level of importance placed on the display quality of the CR systems is not as high as it should be, considering the importance of this component in the clinical workflow.

6) Use of a Standard Test Pattern

Displaying and transmitting a standard test pattern is important in display QC and downstream analysis. The ACR and the AAPM recommend the ability to store, display, and transmit a standard test pattern from acquisition workstations on modalities.

Questions posed to vendors:

  1. Is a Society of Motion Picture and Television Engineers (SMPTE) or similar test pattern available as a file on your acquisition workstation for monitor QC?
  2. Can third-party test patterns, such as the AAPM test patterns, be imported and used for this purpose?
  3. Can these same test patterns be transmitted to a DICOM printer or PACS for downstream analysis?

Discussion:

The support of this functionality is overwhelmingly positive, but could still be improved upon through the standardization of QA procedures and the test images used to perform them. All but one vendor supports the presence of a standardized test pattern on the CR acquisition workstation. All but two vendors allow the importing of third-party test patterns into the CR system for QA purposes. All vendors support the transmission of stored test patterns for downstream testing of PACS and laser file printers.

7) IP and Cassette Tracking

Tracking down imaging plates or cassettes that are producing artifacts is a major issue in imaging departments.

Questions posed to vendors:

  1. If a cassette or imaging-plate artifact is discovered on an image transmitted to PACS, what information is available that would allow the cassette or imaging plate to be identified?
  2. Where can that information be viewed—for example, on the acquisition workstation, from a PACS workstation, or on a printed film?

Discussion:

All vendors have unique serial numbers assigned to the imaging plates and or the associated cassettes. All but one can configure their systems to add these numbers to the DICOM header information in a manner that it can be visualized at the CR-acquisition workstation, by a PACS workstation, and laser printer film. However, this is not the default configuration for the most part, and it must be requested.

8) Spatial Resolution

Spatial resolution is a key factor in the ability of a CR system to detect and digitize small objects and rapidly changing features in anatomy.

Questions posed to vendors:

  1. What is the measured spatial resolution of each CR system? If the resolution varies with cassette size or operating mode (for example, mammography, pediatrics, extremity), please provide information for each size.
  2. Please list all imaging-plate technologies currently in production for each listed CR reader and their respective maximum spatial resolutions.

Discussion:

CR systems’ spatial resolution varied from a minimum of 2.8 lp/mm to a maximum of 10.3 lp/mm. Some vendors achieve varying resolutions by operating their CR readers in different modes and using different cassette sizes, while others have a fixed resolution no matter what size cassette is used. Some vendors did not respond with a measured spatial resolution but provided a spot size or line-resolution value that cannot be directly compared. Two vendors offer more than one scanning technology coupled with new imaging-plate technology that produces higher-resolution images. Users should be aware of the methods by which different vendors achieve higher resolutions and factor it into their CR-purchase considerations.

9) Dynamic Range

Dynamic range varies greatly across the various projection radiographs. A system with wide dynamic-range capabilities and the ability to manage the dynamic range efficiently against the range of image data is an important feature. The dynamic range also can be affected by the type of amplification that is applied to the light-measurement signal before it is digitized. For example, a linear amplification or a logarithmic amplification could be applied.

Questions posed to vendors:

  1. What is the acceptable exposure range (that is, the dynamic range) or your system?
  2. How is dynamic range verified?
  3. What type of amplification is applied to the light-measurement signal before it is digitized?

Discussion:

Some vendors take the approach that the CR readers should be able to automatically operate over a broad range of exposures to compensate for variations in x-ray technique factors, and therefore operate over a large dynamic range. Others require the operator to control the x-ray techniques factors more carefully, and require less dynamic range. Automatic compensation for over- and underexposure is not a critical need in situations where automatic exposure control is used, but may be of importance when using portable x-ray in areas such as intensive care units.

10) Procedure Mapping

Most radiology orders are placed by procedure description and procedure code, and a procedure might consist of several projections. For example, a “screening Chest” procedure might consist of a PA Chest projection and a lateral Chest projection. Radiology workflow is greatly enhanced if the CR user interface can be programmed to automatically bring up the appropriate projections, based on the procedure code received from a modality worklist. This capability often is referred to as “procedure code mapping.” Additionally, each projection will have different image-processing requirements, and the ability to adjust image processing automatically for each projection is highly desirable.

Questions posed to vendors:

  1. Is the user interface on your system configurable for procedure code mapping in a manner similar to the above description?
  2. Does your system automatically adjust the image-processing parameters for each projection?
  3. Are the processing parameters for a projection able to be modified by the users to develop locally optimized image processing?
  4. Can the procedure code mapping features listed above be downloaded and exported to all CR systems in an enterprise?

Discussion:

All but one vendor supports automatic procedure code mapping from the modality worklist to the CR-acquisition workstation with concurrent processing algorithm assignments based upon individual projections listed in each ordered study. All but one vendor report the ability to automatically apply image processing to unique projections of studies based on locally developed tables. All vendors reported the ability to export developed procedure code and processing tables to all CR devices within an enterprise. These levels of configuration are not the default service provided by CR vendors and must be requested and possibly paid for. Considerable effort will required by both the vendor and site to achieve complete procedure code mapping and processing table distribution, but the result is a much smoother imaging operation with much more consistent image quality.

Conclusion

We are pleased to report that progress has been made in the quality issues associated with CR systems. CR system-performance characteristics, workflow configurability, procedure code mapping, and QA-tool availability are generally impressive and well beyond what was available only a few years ago. CR vendors clearly have targeted design characteristics of good-quality CR systems, coupled with a variety of useful QA features and functions. Many of the quality features and functions are optional; however, some are third-party products (QC phantoms and DICOM GSDF calibration), and most require considerable configuration effort and technical understanding. So, while the products are improving in terms of quality features and performance, the level of effort to properly implement QA operations through the use of these features should not be underestimated. Additionally, there are vast differences in the form, fit, and function, and CR vendor familiarity with these features, and users should not assume that the mere listing of features and performance factors by vendors means they are equally implemented and understood by those vendors—much the same as the use of the term “DICOM compliant” by nearly all medical-imaging device vendors.

There are also great differences in the spatial resolution, dynamic range, and bit depth across the various systems. Users should keep this in mind when making purchasing plans. CR systems intended to be used with specific x-ray systems (fixed and portable) or imaging applications (extremity, body, neonatal, pediatrics, mammography, and PACS) should definitely be validated in that venue before making a purchase.

Overall, the state of CR quality is better now than ever before, and there is continuing progress on the part of standards organizations to define requirements, and on the part of vendors to meet or exceed those requirements. You, the users, are the catalysts that accelerate this developmental process by applying feedback and market pressures.

John C. Weiser, PhD, and John Romlein, MS, are partners and cofounders of Qualiteering Labs LLP (QLABS LLP of Thurmont, Md). QLABS LLP is dedicated to the betterment of medical-imaging opeerations through quality-process reengineering and education. E-mail John Weiser at ; e-mail John Romlein at .