Software Corrects for Inhomogeneities in Digital Radiographic Images

The corrected digital image can be used to enable computer-assisted diagnosis.

When radiographic images are recorded, they are subject to inhomogeneities due to the construction of the radiation source and to the properties of the detection medium (detector or screen). As a result, erroneous results may appear when the digital images are processed to enhance diagnostic relevance. Software described here allows for the correction of these inhomogeneities, before any further processing of the digital images is undertaken.

Digital radiography offers the possibility for computer-aided diagnosis by digital image processing (e.g. contrast enhancement). For this purpose, it is a nuisance that non-uniformities in radiation exposure and in detection affect the digital image, before it is processed further.

During exposure with X-rays in planar projection radiography, the intensity non-uniformities can largely be attributed to either the X-ray source Heel effect or the storage phosphor detector’s inhomogeneous sensitivity. These non- uniformities are characterized by an additive or multiplicative smoothly varying image background signal. Although easily corrected by the human eye, they complicate the use of automatic image processing techniques because the relative brightness of an object within the image is position dependent.

The Heel effect is a consequence of the construction of X-ray sources. For reasons of heat dissipation, an X-ray source is equipped with a rotating anode, resulting in uneven path lengths travelled by X-rays. Beams originating closer to the cathode travel a shorter path length and deliver a higher intensity on the detector. This results in an uneven background in the radiographic image detected. This effect leads to background inhomogeneities.

Other sources of non-uniformities can be caused by non-uniform sensitivity of the recording member (such as a photostimulable phosphor screen, or a direct radiography detector). Non-uniformity correction is not straightforward: it is not sufficient to take one flat field exposure for subsequent subtraction. In homo geneity varies from image to image since the X-ray source has variable positions versus the detector.

The software presented here is capable of deducing the required corrections from the actual image data and not from an extra flat field exposure. The background inhomogeneities are mathematically modeled and the digital image is then corrected by using the background model.

The inherent variations in the X-ray source are thus automatically compensated for. The resulting digital image is then ready for further image processing to enhance its diagnostic value, and to enable computer-assisted diagnosis.

The software:

  • Estimates the parameters of a mathematical model representing the physical phenomenon that produces the intensity inhomogeneity, either on the basis of an additional flat field exposure or directly from the diagnostic image using a segmentation of the direct exposure region, which is the image area having received unattenuated radiation;
  • Generates a bias field on the basis of the deduced model parameters; and
  • Applies the extrapolated bias field to correct the image data so that the spatially varying intensity inhomogeneity is removed.

This technology is offered by AGFA Health Care. For more information, click here to view the yet2.com TechPak.

The U.S. Government does not endorse any commercial product, process, or activity identified on this web site.