Abstract
Background
Reject analysis can be used as a quality indicator, and is an important tool in localizing areas where optimization is required. Reducing number of rejects is important yielding reduced patient exposure and increased cost-effectiveness.
Purpose
To determine rejection rates and causes in direct digital radiography.
Material and Methods
Data were collected during a three-month period in spring 2010 at two direct digital laboratories in Norway. All X-ray examinations, types, numbers, and reasons for rejections were obtained using automatic reject analysis software. Thirteen causes for rejection could be selected.
Results
Out of the 27,284 acquired images, 3206 were rejected, yielding an overall rejection rate of 12%. Highest rejection rates were found for examination of knees, shoulders, and wrist. In all, 77% of the rejected images arose from positioning errors.
Conclusion
An overall rejection rate of 12% indicates a need for optimizing radiographic practice in the department.
Reject analysis is described as ‘the critical evaluation of radiographs which are used as part of the imaging service but do not play a useful part in the diagnostic process’ (1). Rejects often leads to retakes which imposes unnecessary patient dose and increased processing and occupational costs. As such, analysis of rejected images gives an overview of the sources of radiographic errors and forms the basis for educating individual technologist (2–10).
In film-based radiography, the overall rejection rates have been reported to range between 6–15% (2–6, 11, 12). The dominating reason for retakes using film-based systems has been attributed to suboptimal exposure due to limited dynamic range of the receptor. Following digitalization of radiology departments, the rejection rate of X-ray images was anticipated to decrease since the latitude of direct detectors is wider than X-ray films. Thus, rejection due to inaccurate imaging parameters would be eliminated (1–3, 8). Replacement of film-based systems by computed radiography (CR) systems has shown to reduce rejection rates to about 5% (1–4, 8–10). For CR systems the majority of rejects arise from positioning errors and the reject rate due to exposure errors is low (1, 2, 4, 7–10, 13).
To the authors' knowledge, only one reject/retake study has been conducted for direct digital radiography (DR) systems (7). Thus, this study was undertaken in order to determine rejection rates and causes in direct digital radiography.
Material and Methods
X-ray equipment
Data for reject analysis were prospectively collected during a three-month period in spring 2010 from two direct digital laboratories at a local hospital in Norway. The digital images were obtained using Acroma acro ceil system with Canon CXD1-50G and Canon CDX1-40EG amorphous silicon TFT flat panel detectors with gadolinium oxysulfide scintillators (Canon, Tokyo, Japan). Laboratory 1 was mainly used for outpatient examinations during the week and for acute and hospitalized patients at weekends whereas laboratory 2 was used to examine acute and hospitalized patients during the week, and rarely used during weekends.
Personnel and workflow
The two DR laboratories are part of the same department and the radiographers are shared between the two laboratories. The radiographers worked together in groups of two or three. Immediately after exposure one of the radiographers had to decide whether to keep or reject the image. When rejecting an image, the radiographer had to select a reason for rejection before proceeding. Numbers of accepted and rejected images for different X-ray examinations were automatically registered by the system together with the reason for rejection. Accepted images were sent to the institutional Picture and Archiving Communication System (PACS; SECTRA IDS5 version 11.4.1, SECTRA AB, Linköping, Sweden) whereas rejected images were, for a very limited time, stored on the hard disk of the DR system. Except for a limited number of cases involving acute trauma patients, radiologists were not involved in the rejection decision.
When the rejection software was installed in our institution two plenary lessons were held explaining the use of the software and the purpose of reject analysis. A two-month training period was conducted prior to data collection. In order to ensure reporting consistency causes for rejections were reviewed during the training period.
Data collection
The Canon CXDI-1 software (Canon, Tokyo, Japan) was used for the data acquisition. Data were prospectively collected on a monthly basis. After three months all data were combined, and the final analysis was performed. In the two laboratories a large number of examinations were performed, varying both in character and frequency. By deciding to focus on high-frequency examinations (more than 150 examinations during the first month), examinations of the abdomen, pelvis, elbow, ankle, cervical spine, lumbar-sacral spine, foot, hip, hand, wrist, knee, shoulder, and thorax were included in this study. As all exposures in the two DR laboratories were automatically registered the study includes patients of all ages. When rejecting an image, the cause of rejection had to be selected among 13 pre-defined causes (Table 1) before proceeding with a new exposure. Rejected images were registered but not stored in PACS.
Reasons for rejecting X-ray images
The daily practice was observed in both laboratories to establish how the rejection analysis tools were utilized. Furthermore, a questionnaire was distributed to 15 radiographers to assess current practice with special emphasize on the selection of reject reason for different scenarios (Table 2).
Question and scenarios given to 15 radiographers to assess variation in the selection of reason for rejects
The answers do not always add up to 15 as some of the radiographers were unable to select a single reason for rejection
Data analysis
The data were analyzed using Microsoft Excel. Reject rates for individual X-ray examinations were found by dividing the number of rejected images by the total number of acquired images. Rejection rate due to positioning errors was obtained by dividing the number of rejected images due to any types of positioning errors by the total number of images acquired (Table 3). The χ2 test was used to evaluate if rejection rates for individual examinations were significantly different from the overall rejection rate (significance level 5%).
Rejects during a three-month period in two direct digital laboratories
*Rejection due to positioning errors refers to the overall percentage of the rejected images caused by any type of positioning errors
Results
During the three-month period a total of 27,284 digital X-ray images were taken. The number of acquired images, the number of rejected images, and the rejection rates for the different imaging procedures after the high-frequency categorization are summarized in Table 3. The overall rejection rate was 12%. Rejection rates for pelvis, hip, hand, and thorax were not significantly different from the overall rejection rate (P > 0.05). Examination of the chest accounted for the majority of rejects. The body areas with the highest rejection rate were the knee, shoulder, and wrist. There was no correlation between the number of examinations and percentage of rejected imaging.
All 15 radiographers completed the questionnaire giving a response rate of 100%. The answers to the questions are summarized in Table 2. Only three of the radiographers had less than five years of radiography experience. Thirteen of the 15 radiographers reported to have received adequate training to reliably perform reject analysis. However, when the radiographers were presented with five scenarios of suboptimal image quality their selection of reason for rejecting the images revealed large variety. For the case where a parent's finger is projected onto the image when holding a child during the examination, the reasons for rejection were reported as positioning error (3), image artifacts (5), patient movement (3), improper patient preparation (1), and additional image (3).
Discussion
Reject analyses has been used as a quality indicator for many years. Previous reject/retake analyses have been performed mostly in film-based radiology departments (1, 4–6, 11, 12) and in departments using CR systems (1, 3, 4, 8, 9). In the present reject analysis, performed in two direct digital radiography laboratories, the overall rejection rate was 12%. The main reason for rejects was patient-positioning error. The total number of rejected images corresponds to one month of full-time work at each of the two laboratories during a year. Reducing number of rejects is important yielding reduced patient exposure and increased cost-effectiveness.
Rejection rates are known to depend on the patient population, type of examination performed, the equipment used, how the rejects are registered, and the skills of the radiographer (1–3, 6–10, 14). An increased number of rejects could be expected when both hospitalized patients and outpatients are included since higher rejection rates are expected in the hospitalized patient group (4). This was not the case in our study. The fact that the examinations included in this study were all frequently performed suggests that differences in rejection rates are not attributed to the number of procedures performed. However, the complexity of individual procedures highly influenced the rejection rate: in our study the percentage of rejected images ranged from 4% for examinations of feet and hands to 20% for knee examinations.
Albeit low rejection rates for some examinations, the overall rejection rate is about twice as high as reported in other studies (1–7, 10). Despite differences in patient population and examinations, the most frequently rejected examinations and the main causes for the rejection are supported by the literature. However, change from CR to DR was anticipated to yield reduced rejection rates due to the elimination of errors arising from the post-acquisition process and the even greater opportunity of digital manipulation.
Strength of our study is the large number of included procedures, examinations, and images. For each of the 13 high-frequency examinations the number of included images ranged from 677 to 7672. The American Association of Physicists in Medicine claims that data from at least 250 patients should be collected to ensure meaningful results (14). To our knowledge, only three studies have included a larger number of images than our study (5, 8, 9). When using the automatic data acquisition software, we ensured that all consecutive examinations during the collection period were registered, minimizing the possibility of the operator influencing the data. When rejecting an image, a reject cause had to be selected prior to continue the examination. Also, no pre-selection of single examination types or patient groups were performed. Thus, the obtained results reflect a realistic description of the daily work in the two DR laboratories.
The rejection rates found in this study may be underestimated. The radiographers could delete images directly from PACS after the examination had ended. As the rejection software records all images being sent to PACS as accepted images, deleting suboptimal images directly in PACS would yield an underestimated rejection rate. Supplementary images are images not included in the standard protocol. Supplementary images were acquired if additional pathology was suspected or if an image was accepted despite limited diagnostic value leading to a second exposure. As supplementary images were not accounted for separately but were included in the total amount of images taken, this may have led to an under-estimation of the rejection rate. It gives rise to concern that despite several reasons for possible under-estimation an overall rejection rate of 12% was found.
Implementation of digital advanced equipment gives the radiographer many options and eases the radiological procedure. A possible explanation for the high rejection rates can be that the direct digital detectors make it too easy to retake images and thus a lower threshold for rejecting suboptimal pictures just to be sure may be established (2). Also, reduced contact between radiographers and radiologists results in less direct feedback on the radiographic work (2, 3, 5). This could lead to increased focus on technical image criteria or demands and radiographers being too self-confident or too reliant on the advanced technology (2, 3, 6, 15). There is thus a possibility that radiographers reject images that in fact possess enough diagnostic information to describe the actual pathology (2, 3, 5, 6, 10, 15).
The answers to the questionnaire revealed that there was a lack of common understanding for selecting the cause for rejecting an image (Table 2, question 5). This lack of common understanding has also been reported as a major problem in earlier studies (2, 3, 8). Both technical and anatomical factors should be taken into consideration when analyzing the reasons for rejects. Image perception and observer performance should also be evaluated (2, 14). However, even though it is difficult to establish specific reasons for the rejection, it is evident that the main reason for rejection was positioning errors (77%). This finding is in accordance with results from other CR studies (1, 6–10).
In the past, reject analysis has been an important quality indicator (1–10). It is accepted that a reject analysis is an important tool in localizing areas where optimization is required, and can be used as a basis for staff training and education. However, such quantitative analyses give no information of the diagnostic quality of the rejected images. Investigations and further studies of the diagnostic value would yield important information of the rejected images. Also, performing reject analysis in other hospitals with DR technology will increase the basis for comparison.
In conclusion, the overall rejection rate for direct digital radiography in our study was 12%. Positioning errors accounted for 77% of all rejections. Highest rejection rates were found for wrist and knee examinations. The high rejection rate indicates a need for further optimization in the digital radiology department.