Академический Документы
Профессиональный Документы
Культура Документы
Charles E. Willis, PhD, DABR; Stephen K. Thompson, MS, DABR; S. Jeff Shepard, MS, DABR
onsidering the sales hyperbole associated with digital radiography (DR), one may wonder if it is even possible to produce a nondiagnostic digital image. Certainly DR is more tolerant of inappropriate exposure factor selection than is conventional lm-screen radiology. However, classic technical errors (such as malpositioning, patient motion, incorrect patient identication, incorrect examination, and double exposure) still occur in the usual frequency.1 The unfortunate truth is that DR is subject to many of the same inaccuracies as conventional radiography, in addition to many new ones that are a direct result of the way the DR image is generated. An understanding of the causes of both new and old problems is necessary in order to avoid these inaccuracies and recover unacceptable images.
Dr. Willis is an Associate Professor in the Department of Radiology, Baylor College of Medicine, Houston, TX. Mr. Thompson is a Medical Physicist with Memorial Medical Center, Modesto, CA. Mr. Shepard is a Senior Medical Physicist in the Department of Diagnostic Physics, University of Texas M.D. Anderson Cancer Center, Houston, TX. This article is based on material originally presented in: Willis CE. Artifacts and misadventures in digital radiography. Presented at the Society of Computer Applications in Radiology Meeting. SCAR University Course 305, 20th Symposium. Boston, MA, June 710, 2003.
(DDR) systems that make digital radiographs from the photoelectric interaction of X-rays with the detector itself, indirect digital radiography (IDR) systems that are sensitive to the light produced by an intensication screen, and optically coupled direct radiography (OCDR) systems that use optical components to focus uorescence onto charge coupled detectors (CCD). In this context, DR includes any radiographic image acquired without photographic lm, thus excluding lm digitizers whose artifacts are described elsewhere.
DR systems
Digital radiographs in this study were produced with a variety of devices in clinical service at our institutions or avail-
January 2004
www.appliedradiology.com
APPLIED RADIOLOGY
11
FIGURE 1. (A and B) Underexposed computed radiography (CR) images demonstrate increase in quantum mottle and loss of contrast in dense features. These constitute approximately 9% of all rejected images. These images were acquired on an Agfa CR System (Agfa Medical Systems, Ridgeeld Park, NJ).
FIGURE 2. (A) Underexposed direct digital radiography (DR) at 125 kVp and 4.4 mAs caused an exposure data recognition failure. (B) Repeated exposure at 7 mAs. These images were acquired on a GE DR system (GE Medical Systems, Milwaukee, WI).
able to us, including Agfa CR (Agfa Medical Systems, Ridgeeld Park, NJ), Fuji CR (Fujilm Medical Systems, Stamford, CT), GE DR (GE Medical Systems, Milwaukee, WI), and Canon DR (Canon USA, Inc., Lake Success, NY). Although not shown, we have similar experiences with Kodak CR (Eastman Kodak, Rochester, NY), Lumisys CR (now owned by Eastman Kodak), Konica CR (Konica Medical Imaging, Inc.,
12
APPLIED RADIOLOGY
www.appliedradiology.com
January 2004
FIGURE 3. (A) Overexposed computed radiography (CR) images demonstrate loss of contrast in skin and dense features. These constitute approximately 5% of all rejects. (B) Repeated exposure. These images were acquired on an Agfa CR System (Agfa Medical Systems, Ridgeeld Park, NJ).
FIGURE 4. (A) Improper calibration of a computed radiography (CR) system causes loss of contrast in dense features. (B) Exposure repeated after calibration for nonuniformity and sensitivity. These images were acquired on a Fuji CR System (Fujilm Medical Systems, Stamford, CT).
January 2004
www.appliedradiology.com
APPLIED RADIOLOGY
13
FIGURE 5. Improper calibration of digital radiography (DR). Note interfaces of four detectors visible at arrows. This image was acquired on a Canon DR system (Canon USA, Inc., Lake Success, NY).
soft-copy devices, and displays used for quality control (QC).12 Processing algorithms used by the DR system must be designed to anticipate the use of this display function in order to properly render the image for display. Not all vendors adhere to this standard. Calibrated, high-quality QC monitors are essential on every acquisition system and QC workstation. Adjusting image processing on an uncalibrated monitor leads to unsatisfactory images. During the initial installation, it is important to make sure that the DR system is properly congured with the most up-todate version of software, hardware, and durable goods. Multiple vintages of imaging plates exist for CR, and some are not universally compatible with CR hardware. The software and settings should be consistent with the versions and settings in operation with other individual DR systems at the site, including examination-specic parameter settings (Figure 6). All DR systems have an internally calculated estimate of exposure. The DR system may need to be congured to report this value to the digital image management system, and the image management system may need to be congured to display it to the radiologist. When CR is introduced into an imaging operation, phototimers in all X-ray rooms need to be recalibrated to deliver the appropriate exposure.13 The methodology for phototimer calibration is different because DR density is adjustable. Scheduled and unscheduled service should be done in a thorough and timely manner, including reporting and documenting, cleaning, and repairs. Operator functions include cleaning, reporting service interruptions, removing the unit from clinical service, re-introducing the unit into clinical service, and documenting service events (Figure 7). Service engineer functions include
FIGURE 6. (A) Computed radiography (CR) corduroy artifact (arrows) is an interference pattern generated from the interaction of the line rate of a xed scatter reduction grid, the sampling rate of the detector, and display zoom factor and pixel dimensions. This image was taken on an Agfa CR system (Agfa Medical Systems, Ridgeeld Park, NJ). (B) These CR cassettes are marked to indicate orientation when inserted into the CR scanner. Each orientation was appropriate only for a specic scanner model. Inserting in the incorrect orientation caused jams. (C) CR image displayed according to exam-specic processing parameters in the main department. (D) Same image displayed according to settings in orthopedic department. Images B, C and D were acquired on a Fuji CR System (Fujilm Medical Systems, Stamford, CT) .
14
APPLIED RADIOLOGY
www.appliedradiology.com
January 2004
FIGURE 7. Artifacts on clinical images constitute 3% of all rejects. (A) Artifact observed (arrow) after technologist frees jam from computed radiography (CR). (B) CR plate discoloration (arrow). (C) CR plate defect in two different orientations (arrows). Images A, B, and C were acquired on an Agfa CR system (Agfa Medical Systems, Ridgeeld Park, NJ). (D) Dust on Fuji CR collection optics (arrows) (Fujilm Medical Systems, Stamford, CT).
performing scheduled maintenance that includes preventive maintenance and software and hardware upgrades, as well as unscheduled maintenance or repairs. Onsite support for DR requires a team with expertise not only in image acquisition, but in picture archiving and communications systems (PACS), radiology information systems (RIS), technologist workow, image quality, and networks, as well.
images. Technologists are often unfamiliar with DR features and functions, and may require additional training beyond vendor applications training. Technologists need to select the proper examination; in addition, they must properly associate demographic and examination information to the image, properly manipulate the detector, and review the image before releasing it to the image management system. Beyond this, technologists need to know how to recover from errors without repeating examinations and need to follow exposure factor control limits. Quality control processes
must be in place to detect and correct unsatisfactory images (Figures 8, 9, and 10). Digital radiography requires a new approach to QC and study reject analysis.14 Electronic images can disappear without a trace: counting the lms remaining in the lm bin is no longer a useful method for determining the number of repeated images. Double exposure is a classic operator error that constitutes approximately 2% of all rejected images. The consequence of double exposure can be either a single repeated examination, when an inanimate object is involved (Figure 11), or
January 2004
www.appliedradiology.com
APPLIED RADIOLOGY
15
FIGURE 9. It is difficult to distinguish debris on this computed radiography (CR) imaging plate from foreign bodies (arrows). The technologist must open the cassette and inspect the plate or re-expose the cassette. These errors constitute <1% of all rejects, probably because radiologists usually tolerated them. This Image was acquired on an Agfa CR system (Agfa Medical Systems, Ridgeeld Park, NJ).
two repeated examinations when two patients are involved (Figure 12). In DR, double exposures can also be caused by power interruptions and communications errors, as well as by inadequate erasure secondary to overexposure or erasure mechanism failure.
Image processing
FIGURE 8. (A) Artifact on computed radiography (CR) image (arrows) released by technologist and reported by the radiologist. (B) Same artifact (arrows) reported later on another image from the same machine by the same radiologist. Images were acquired on an Agfa CR system (Agfa Medical Systems, Ridgeeld Park, NJ).
Appropriate digital image processing is key to producing good DR images. All DR systems have extremely wide latitude, which means that connected to a display system with a relatively narrow dynamic range, DR images have extremely low contrast. The primary purpose of image processing is to maximize the contrast of the part of the image that contains relevant clinical details.15,16 To do this, the DR system locates either the boundary of collimation or the border of the projected anatomy, and disregards details outside this
16
APPLIED RADIOLOGY
www.appliedradiology.com
January 2004
FIGURE 10. (A and B) Missing lines or pixels in computed radiography (CR) can indicate memory problems, digitization problems, or communication errors. Images were acquired on an Agfa CR system (Agfa Medical Systems, Ridgeeld Park, NJ).
FIGURE 11. Double exposures with computed radiography (CR) requiring a single repeat study. (A) Backboard or gurney side rails. (B) Cassette left in buckey tray during uoroscopy. Images were acquired on an Agfa CR system (Agfa Medical Systems, Ridgeeld Park, NJ).
January 2004
www.appliedradiology.com
APPLIED RADIOLOGY
17
FIGURE 12. (A and B) Double exposures with computed radiography (CR) requiring two repeats. Images were acquired on an Agfa CR system (Agfa Medical Systems, Ridgeeld Park, NJ).
FIGURE 13. Exposure Data Recognizer (EDR) failure, an inability of the software to determine collimation boundaries. This type of failure can be caused when the operator fails to follow collimation rules or when interferences prevent the software from detecting the collimation boundaries. It results in incorrect histogram analysis and inappropriate rescaling. Image acquired on the Fuji computed radiography system (Fujilm Medical Systems, Stamford, CT).
FIGURE 14. (A) Error in computed radiography (CR) automatic detection of the collimation eld. (B) Manually collimated images. Images were acquired on an Agfa CR system (Agfa Medical Systems, Ridgeeld Park, NJ).
FIGURE 15. (A) Inappropriate processing of pediatric digital radiography image by vendor-supplied parameters suitable for adults. (B) Image processed by parameters modied by the customer. Images acquired on GE DR system (GE Medical Systems, Milwaukee, WI).
18
APPLIED RADIOLOGY
www.appliedradiology.com
January 2004
*Fuji CR (Fujilm Medical Systems, Stamford, CT); Agfa CR (Agfa Medical Systems, Ridgeeld Park, NJ); Kodak CR (Eastman Kodak, Rochester, NY). These items are different methods, but have equivalent purposes.
boundary. Errors in collimation can cause mistakes in detection of the boundary, with a dramatic loss of image contrast (Figures 13 and 14). The secondary function of image processing is to customize contrast in the region of interest (Table 1). This type of image processing includes modifying the image to enhance the contrast and sharpness of some features while compromising the contrast and sharpness of others, as well as modifying the image to make it appear more like a conventional transilluminated lm. This secondary image processing is applied in a manner that is usually specic to the anatomic projection. Errors in the selection of the anatomic projection can cause inappropriate processing (Figure 15). An auxiliary purpose of image processing is to improve the usability of the digital image.17 This includes imprinting demographic overlays, adding annotations, applying borders and shadow masks, ipping and rotating, increasing magnication, conjoining images for special examinations like scoliosis, and modifying the sequence of views. This processing may require a separate QC workstation. Image processing is not a panacea. Misuses of image processing include compensating for inappropriate radiographic technique, compensating for poor calibration of acquisition and display devices, and surreptitious deletion of nondiagnostic images. Image processing
January 2004
to recover nondiagnostic images to prevent re-exposure should be a last resort, not a routine activity. Routine reprocessing indicates a problem with automatic image processing or technical practice. Access to image-processing software is essential to develop and maintain appropriate processing parameters. Automatic image processing involves assumptions about the radiographic technique, the composition of anatomic region imaged, and the use of collimation. A number of factors can interfere with the automatic detection of the boundaries of the radiation eld, including nonparallel collimation, use of multiple elds on a single imaging plate, poor centering, implants (especially when they overlie the boundary), and violation of collimation rules provided by the vendor. For example, placement of gonadal shields is no longer trivial, but may adversely affect image quality.
sure indicator values, instead of brightness and contrast. Without this attention, patient dose will escalate. If exposure indicator logs are available, they need to be evaluated. If they arent, this will need to be done manually. Vendors need to make such logs available in convenient digital form. New technologies should be developed for dealing with pediatric examinations and patients with prosthetic devices. New image processing strategies may be needed with these special patients, as well. Thorough training and active onsite support of the technical staff are crucial. For many, this is a completely different way of thinking about the imaging process. Technologists are generally eager to become involved and master this new technology, but they need proper training and guidance to use it effectively to produce diagnostic-quality images.
REFERENCES
1. Willis CE, Mercier J, Patel M. Modication of conventional quality assurance procedures to accommodate computed radiography. Presented at the 13th Conference on Computer Applications in Radiology. Denver, Colorado. June 7, 1996: 275-281. 2. Oestmann JW, Prokop M, Schaefer CM, Galanski M. Hardware and software artifacts in storage phosphor radiography. RadioGraphics. 1991;11:795-805. 3. Solomon SL, Jost RG, Glazer HS, et al. Artifacts in computed radiography. AJR: Am J Roentgenol. 1991;157:181-185. 4. Volpe JP, Storto ML, Andriole KP, Famsu G. Artifacts in chest radiography with a third generation computed radiography system. AJR: Am J Roentgenol. 1996;166:653-657. 5. Tucker DM, Souto M, Barnes GT. Scatter in com-
Conclusions
A multitude of factors affect DR image quality, and no device or operator is immune to unacceptable images. A strategy for addressing images that just didnt turn out right must be implemented. Responsibilities for documenting, reporting, and taking corrective action must be clearly established. Wider dynamic range means that technologists have to pay attention to expo-
www.appliedradiology.com
APPLIED RADIOLOGY
19
APPLIED RADIOLOGY
www.appliedradiology.com
January 2004