Michael W. Dupin

Observer study of a noise suppression algorithm for computed radiography images

Lower x-ray exposures are commonly used in radiographic exams to reduce the patient radiation dose. An unwanted side effect is that the noise level increases as the exposure level is reduced. Image enhancement techniques increasing image contrast, such as sharpening and dynamic range compression tend to increase the appearance of noise. A Gaussian filter-based noise suppression algorithm using an adaptive soft threshold has been designed to reduce the noise appearance in low-exposure images. The advantage of this technique is that the algorithm is signal-dependent, and therefore will only impact image areas with low signal-to-noise ratio. Computed radiography images captured with lower exposure levels were collected from clinical sites, and used as controls in an observer study. The noise suppression algorithm was applied to each of the control images to generate test images. Hardcopy printed film versions of control and test images were presented side-by-side on a film alternator to six radiologists. The radiologists were asked to rate the control and test images using a 9-point diagnostic quality rating scale and a 7-point delta-preference rating scale. The results showed that the algorithm reduced noise appearance, which was preferred, while preserving the diagnostic image quality. This paper describes the noise suppression algorithm and reports on the results of the observer study.

Clinical evaluation of new workflow-efficient image processing for digital radiography

An observer study was conducted on a randomly selected sampling of 152 digital projection radiographs of varying body parts obtained from four medical institutions for the purpose of assessing a new workflow-efficient imageprocessing framework. Five rendering treatments were compared to measure the performance of a new processing algorithm against the control condition. A key feature of the new image processing is the capability of processing without specifying the exam. Randomized image pairs were presented at a softcopy workstation equipped with two diagnosticquality flat-panel monitors. Five board-certified radiologists and one radiology resident independently reviewed each image pair blinded to the specific processing used and provided a diagnostic-quality rating using a subjective rank-order scale for each image. In addition, a relative preference rating was used to indicate rendering preference. Aggregate results indicate that the new fully automated processing is preferred (sign test for median = 0 (α = 0.05): p < 0.0001 preference in favor of the control).

Statistical interpretation of ISO TC42 dynamic range: risky business

Recently, two ISO electronic imaging standards aimed at digital capture device dynamic range metrology have been issued. Both ISO 15739 (digital still camera noise) and ISO 21550 (film scanner dynamic range) adopt a signal-to-noise ratio (SNR) criterion for specifying dynamic range. To resiliently compare systems with differing mean-signal transfer, or Electro-Optical Conversion Functions (OECF), an incremental SNR (SNRi) is used. The exposure levels that correspond to threshold-SNR values are used as endpoints to determine measured dynamic range. While these thresholds were developed through committee consensus with generic device applications in mind, the methodology of these standards is flexible enough to accommodate different application requirements. This can be done by setting the SNR thresholds according to particular signal-detection requirements. We will show how dynamic range metrology, as defined in the above standards, can be interpreted in terms of statistical hypothesis testing and confidence interval methods for mean signal values. We provide an interpretation of dynamic range that can be related to particular applications based on contributing influences of variance, confidence intervals, and sample size variables. In particular, we introduce the role of the spatial-correlation statistics for both signal and noise sources, not covered in previous discussions of these ISO standards. This can be interpreted in terms of a signals spatial frequency spectrum and noise power spectrum (NPS) respectively. It is this frequency aspect to dynamic range evaluation that may well influence future standards. We maintain that this is important when comparing systems with different sampling settings, since the above noise statistics are currently computed on a per-pixel basis.