Optimised detail detectability in computerised tomography
M. Mangard and P Hammersberg
ABSTRACT It is of interest to know the minimum discernible detail size when performing non-destructive testing with highresolution computerised tomography. But it is difficult to empirically find optimal parameter settings that maximise detectability for each individual imaging task. In this work a method to determine the optimal performance for a high-resolution computerised tomography system has been developed. It reveals the detectability limit for specific contrasting details in terms of imaged object diameter when the signal-to-noise ratio, SNRDs,cT, between the contrasting detail and its surroundings has been maximised using optimal data collection parameter settings. This work includes modelling and verification of the total unsharpness of the CT-system in terms of modulation transferfunction, MTF. Together with the limiting perception factor and maximised SNRDproj in the CT projection data, the detectability limits for any specific contrasting detail are determined as a function of imaged object material and geometry. Maximised SNRdproj was obtained by optimising the parameters: X-ray source tube potential, X-ray filter, exposure time and optical aperture. The importance of using optimal settings when performing CT-investigations, especially when investigating objects with large diameters, is demonstrated. It was found that it is possible to predict the detectability for any specific contrasting detail. It was also discovered that the optimal settings for a given object diameter and material are dependent on the detail material. In some cases an improvement of the detail signal -to-noise-ratio was obtained by using a combination of X-ray filter materials. It was also found that SNRDproj is very sensitive to the thickness of denser X-ray filter materials (higher atomic number and density). Keywords: Computerised tomography, CT, detectability, optimal performance
Publication Source: Journal of X-Ray Science and Technology, ISSN: 0895-3996. Volume 8, Number 2, 1998. Pages: 51...
Publisher: IOS Press - Amsterdam
© NDTnet
