Abstract
Background
The number of rejected images is an indicator of image quality and unnecessary imaging at a radiology department. Image reject analysis was frequent in the film era, but comparably few and small studies have been published after converting to digital radiography. One reason may be a belief that rejects have been eliminated with digitalization.
Purpose
To measure the extension of deleted images in direct digital radiography (DR), in order to assess the rates of rejects and unnecessary imaging and to analyze reasons for deletions, in order to improve the radiological services.
Material and Methods
All exposed images at two direct digital laboratories at a hospital in Norway were reviewed in January 2014. Type of examination, number of exposed images, and number of deleted images were registered. Each deleted image was analyzed separately and the reason for deleting the image was recorded.
Results
Out of 5417 exposed images, 596 were deleted, giving a deletion rate of 11%. A total of 51.3% were deleted due to positioning errors and 31.0% due to error in centering. The examinations with the highest percentage of deleted images were the knee, hip, and ankle, 20.6%, 18.5%, and 13.8% respectively.
Conclusion
The reject rate is at least as high as the deletion rate and is comparable with previous film-based imaging systems. The reasons for rejection are quite different in digital systems. This falsifies the hypothesis that digitalization would eliminates rejects. A deleted image does not contribute to diagnostics, and therefore is an unnecessary image. Hence, the high rates of deleted images have implications for management, training, education, as well as for quality.
Introduction
Rejects, deletions, and subsequent retakes of diagnostic X-ray images impose professional and ethical challenges within radiological imaging (1); it occupies unnecessary processing and personnel resources (2–5), indicates suboptimal quality management (6–8), and exposes patients to unnecessary ionizing radiation and added inconveniences (9). Traditionally reject/deletion/retake rates for film-based departments have been documented to be in the range of 10–15% (8,10–17), and their main cause has been attributed incorrect exposures due to limited dynamic range of screen/film systems. Accordingly, the digitalization of medical imaging induced expectations that the problem of image rejects, deletions, and retakes would disappear (5–7,17,18). A series of research papers have reported reject/deletion/retake rates in digital departments at around 5% (6–8,15,17,19,20), and some even at the same rate as with film systems (4). This poses the question whether the reject rates really are as high as with film systems and why the problem did not vanish with the digital revolution, as presumed.
Image digitalization significantly changed the causes of rejects. While rejects for the screen/film systems mostly were exposure-related, they are now mainly reported to relate to patient positioning errors in the digital systems. Although there are some studies with reject rates for computed radiography (CR) systems (6,8,17,19,21–23), there are few studies for direct digital radiographic (DR) systems (4). Although, one would expect the reject rates of DR systems to be below film reject rates, initial studies indicate that this is not so (4,22,24,25). In order to assess whether the high reject rate with DR only are incidental findings or represent a real challenge in digital imaging more studies are needed. The reject rate in this study was defined as images deleted on modality specific work stations or in the PACS. Accordingly, the research questions of this article are: How high is the deletion rate for DR systems, and what are the reasons for deletions?
Material and Methods
This study was registered and conducted as a Quality Assurance Project of the hospital, and is as such not subject to informed consent from patients according to the Norwegian Patient Rights Act. Employees at the Radiology Department were informed about the study in advance. Access to images and systems was supervised by the Radiology Department. Confidentiality statement was valid for the data collection.
Data were collected at two laboratories for general X-ray examinations at the radiological department of a local public hospital in the central southern part of Norway. The department makes about 25,000 general X-ray examinations per year. The two DR laboratories are part of the same department and the department’s radiographers are shared between the two laboratories. Data included all exposed images during January 2014. A registration form was developed on basis of existing literature (5,10,24–26). Some adjustments resulted from a pilot study. The registered categories are given as follows:
Positioning error (other than centering errors) Incorrect collimation Centering error Wrong exposure Artifacts Other reasons
Centering errors were differentiated from other positioning errors in order to be able to tailor education and improvement strategies. A “centering error” occurs when the object of interest is not in the center of the image, while other errors of position, such as rotation errors, are categorized as “positioning error”. Images can be deleted either at the workstation of the modality or in PACS. Images deleted on the workstations can be counted directly, as these are tagged. However, in order to collect data on additional deletions in PACS, the number of images on PACS and workstations were compared for each examination. In this study, image rejects are defined as images that do not contribute with diagnostic information with regards to the relevant clinical indication due to poor image quality (5) and they are measured as deleted images, as a deleted image has no diagnostic value as it per se is not used for diagnostic purposes. Accordingly, a deleted image is defined as an image that is deleted from the data registry either at the workstation of the modality or from the PACS (after being transferred from the workstation). A more detailed description of the relationship between deleted images, image rejects, image retakes, and unnecessary imaging can be found in the Appendix.
Data collection was performed during evening time in order not to influence the workflow or the deletion rate. Deleted images were categorized by two persons or three persons when there was doubt. Descriptive statistics was used with Microsoft Excel 2010 to calculate deletion rate and confidence intervals. A detailed description of the X-ray equipment and PACS is given in the Appendix.
Results
In total, 1911 examinations with 5417 images were registered during January 2014. Of these 596 images were deleted during this period. Accordingly, the deletion rate was 11.0% (95% CI, 10.2–11.8].
The number of images and deletions for the 10 most frequent types of examinations.
Distribution of identified reasons for deletion.
It was not possible to decide why the image was deleted.
Distribution of identified reasons for deletion on various examination types.
Discussion
Our results show a deletion rate which is quite high compared to international studies on CR systems (6–8,11,17,19,21), but very much in line with existing Norwegian studies. Leffmann et al. found a deletion rate of 13.1% for wrist images with a CR system (27), and Andersen et al. found a reject rate of 17% for wrist images with DR (4), while we found a deletion rate of 12.4% (95% CI, 9.7–15.1). In line with both Leffmann and Andersen’s findings, our study shows that the main reason for deletion of wrist images was positioning errors. Leffmann’s study does not report whether deletions in PACS are included. If they are not, as the article indicates, their real rate may be significantly higher. This also goes for Andersen’s study which does not include deletions in PACS. Therefore, the real reject rate may be higher (4).
Our overall results are also in line with the overall deletion rate of 12.5% found in 2009 at one of the labs included in our study (25), and 12% found in the study by Andersen and colleagues (4). The finding show that the deletion rate is on level with the retake rate with film systems, but the reasons for deletions are different: from incorrect exposure to positioning error. This can indicate poorer quality of work among radiographers.
There are some discrepancies in the results of the reasons for deleting images in our and Andersen’s study. For example, Andersen et al. found an overall positioning error of 77% while our results showed 82.3% (centering and other positioning errors) (4). This may of course be due to real differences between the sites, but can also be due to difference in interpretations of the categories and the mode of registration. In Andersen et al.’s study the radiographers registered the reason for deletion themselves, while we registered a retrospective interpretation of the radiographers’ reasons for deleting the images. This weakness in our study is only relevant for the interpretation of the reasons for deletions, and not for the deletion rate, where our study is more complete than comparable studies (4). Hence, there is a tradeoff between the validity of the results on reject rate and on reasons for rejects.
The categories of reasons for deletions are quite coarse in our study. Radiographers may have more subtle reasons for deleting, which cannot be identified by the study. However, the pilot study showed that a more detailed list of reasons was not feasible with the interpretative method chosen. Nevertheless, our categories correspond well with those of other studies. In addition to registering the type of examination, it is valuable to have information on the projections of the deleted images.
This study has not measured unnecessary imaging, but only how large proportion of the images that were deleted. However, a deleted image has no diagnostic value as it per se is not used for diagnostic purposes. It is therefore unnecessary. The number of deleted digital images will therefore be an underestimation of image reject, of retakes, and of unnecessary imaging, simply because many original non-used images are not deleted. Nevertheless, the number of deleted images provides a useful estimate of the lowest possible rate of unnecessary imaging. If the number of deleted images is high, the number of unnecessary images is alarming. Fig. 1 illustrates the relationship between the number of rejects, retakes, and unnecessary images. There are of course many reasons why images are not deleted: abundant storing capacity; one forgets to delete them; one believes that they may be of some value in the future; the old image may in the end showed up to be better than the new one; time pressure; or because deleting too many pictures would give the impression of poor quality work.
The relationship between unnecessary images, retakes, rejects, and deleted images.
In conclusion, we find a deletion rate of 11%. This indicates that the reject and the retake rate, as well as the rate of unnecessary images is higher than 11%. We found deletion rates comparable with reject rates of previous film based imaging systems, but that the reasons for reject rates are different. This falsifies the hypothesis that rejects and retakes would be abolished with digitalization of radiographs. For some examination types the deletion rate is over 20% and the main reasons for deletions are positioning and centering errors (together 82.3%). Monitoring unnecessary images is highly relevant to verify and improve the quality in modern radiographic imaging. It is of great importance for management, training, education, and for quality improvement.
Footnotes
Acknowledgements
The authors thank the employees at the hospital where the study was conducted for their facilitation and support. They also thank their colleague Dag Waaler who assisted with the statistical analysis.
Contributions: All authors designed the study. TBR, CJ and KHCW did data collection and primary analysis. BH supervised the study and drafted the article.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.