Digital radiography image quality evaluation using various phantoms and software

Abstract

To investigate the effect of the exposure parameters on image quality (IQ) metrics of phantom images, obtained automatically using software or from visual evaluation. Methods: Three commercial phantoms and a homemade phantom constructed according to the instructions given in the IAEA Human Health Series No. 39 publication were used, along with the respective software that estimate automatically various IQ metrics. Images with various exposure parameters were acquired in a digital radiography (DR) unit. For the commercial phantoms, visual evaluations were also performed. The IQ scores obtained were analyzed to investigate the effects of increasing incident air kerma (IAK), tube potential (kVp), additional filtration, and acquisition protocol on IQ. Results: The effects of the exposure parameters on the IQ metrics, determined with the commercial and the IAEA phantoms, were not the same. For example, clear trends of improvement of IQ scores with increased IAK and reduction of most IQ scores with increased kVp were observed mostly with the IAEA phantom, but not with the commercial phantoms (for both automatic and visual scoring methods). For all phantoms, the maximum variations in IQ scores observed for repeated identical exposures were almost always below 10% with automatic evaluation whereas, for visual evaluation, reached 17%. Conclusions: Failure to detect some expected trends with the complex commercial phantoms may be attributed to the fact that IQ in DR is more strongly affected by the post-processing procedures, which may mask the effect of other parameters on IQ, something that was not observed with the simple IAEA phantom.

Automatic image quality evaluation in digital radiography using for‐processing and for‐presentation images

“…Increasing the additional filtration had no marked effect on IQ-scores of either raw or clinical images, which is good news considering the significant reduction of patient exposure obtained with increasing additional filtrations and it is in accordance with the results of the previous study regarding clinical images. 5 Overall, MTF 50%, SDNR and SNR of clinical images were larger than the respective values of raw images in most of the experiments, while the opposite was valid for d'. The observation that d' was larger for raw images than for clinical images (with a single exception shown in Figure 3e), is one of the most interesting findings of this study, since at first it seems contradictory.…”

“…1 Thus, it is not a surprise that in clinical practice when different examination protocols are used to image the same anatomical structures or different post-processing protocols are applied onto the same raw image, different image quality (IQ) levels are derived. [2][3][4][5] This is true both for patient radiographs and for radiographic images of physical phantoms used for IQ evaluation. The physical phantoms used for IQ evaluation typically contain various structures to assess spatial resolution, low contrast, and dynamic contrast.…”

“…Post‐processing allows the original image to be processed in various ways—depending on the anatomical area imaged and the respective diagnostic task—to adjust the latitude (dynamic contrast), contrast and spatial resolution, so that the visibility of various anatomical structures and consequently the diagnostic confidence of the interpreting radiologist are enhanced 1 . Thus, it is not a surprise that in clinical practice when different examination protocols are used to image the same anatomical structures or different post‐processing protocols are applied onto the same raw image, different image quality (IQ) levels are derived 2–5 . This is true both for patient radiographs and for radiographic images of physical phantoms used for IQ evaluation.…”

“…To answer this question, different exposure parameters, like tube potential (kVp), incident air‐kerma (IAK) levels, tube additional filtration and post‐processing algorithms have been used to simulate such possible changes. The differences observed between phantoms regarding the impact that the above parameters had on IQ metrics’ values (IQ‐scores), were in some cases unexpected and incomprehensible 5 . It was thought that failure to detect the expected effect that changes in IAK and kVp have on IQ with some commercial phantoms, and failure to detect any effect of additional filtration on IQ with all phantoms, may be due to post‐processing.…”

“…To investigate the effect that post‐processing algorithms have on IQ assessment using the IAEA phantom, a rationale and experimental set‐up similar to that of a previously published study was used 5 . However, the basic difference of this study from the previous study is that the DR system used in the present study produces and stores automatically two images for each acquisition: one for‐processing (henceforth referred to as raw ) and one for‐presentation (henceforth referred to as clinical ), both of which are readily available to the user.…”