Factors affecting Probability of Detection with Computed Radiography

Computed Radiography (CR) can be a viable alternative for conventional film-based industrial radiography. This new technology uses flexible and re-usable storage phosphor plates that can be used in a fashion similar to film. To ensure that CR provides Probability of Detection (POD) performance for flaw indications that is equivalent to that of film radiography, international standards have been developed (ASTM E2445/E2446 and CEN 14784-1/2) to govern the assessment of the image quality. These standards classify CR systems using the concept of normalized SNR (nSNR), which is based on the hypothesis that the SNR is inversely proportional to the basic spatial resolution (SR) of the system. Because of concerns that CR plates can have a lower spatial resolution than some industrial films, the above-mentioned standards specify the maximum SR depending on the penetrated thickness. However, in practice POD is primarily determined by the minimum subject contrast threshold that can be observed. The contrast threshold depends on many parameters including primary-to-scatter ratio, attenuation coefficient of the material, flaw geometry (globular, needle etc.), SNR, and the effective contrast modulation. Of these, the effective contrast modulation is the only parameter that depends on the system resolution. We present a model that estimates the relative importance of the different factors affecting the success of a radiographic technique. Only for very small globular flaws and materials with very high X-ray attenuation is the system resolution expected to be the limiting factor. We also present experimental evaluations comparing the classification according to the standards with the observed wire IQIs sensitivity on Fe and Al from a modern CR system.