Mammographic x-ray spectra measured by Compton spectroscopy using a high resolution Schottky CdTe detector.

Analysis of x-ray spectra is important for quality assurance (QA) and quality control (QC) of radiographic systems. This is especially true for mammographic x-ray imaging systems which require low-contrast detectability. Under clinical conditions, the measurement of diagnostic x-ray spectra is difficult because of pulse pile-up due to the high fluence rate of incident x-ray photons. However, it is difficult to set a long source-to-detector distance to reduce the number of photons detected per unit time for mammographic x-ray units. Compton spectroscopy is a suitable tool to deal with this problem. Detection of 90-degrees scattered photons only, energy correction and reconstruction are easy using the Klein-Nishina formula. However, x-ray spectra measured by this method have lower energy resolution, because of the geometrical irradiation angle or electron movement in the scatterer. Moreover, spectra measured with a compound semiconductor detector, such as a high resolution Schottky CdTe detector like we used, are distorted by the detector response, which is based on detecting x-ray photon interactions and charge carrier trapping in the semiconductor crystal. While the distortion of spectra caused by the response can be easily corrected by applying a stripping procedure, it is very difficult to reconstruct the broad spectra measured by Compton spectroscopy as sharp spectra such as obtained when directly measured. Some complicated reconstruction algorithms have been reported to fit the shape of spectra obtained by the Compton spectroscopy to sharp standard spectra. However, for QA / QC of the radiographic system, it is not necessary to correct the spectra sharply if the spectral broadening is at a tolerable level and the properties of the broad spectra acquired by the Compton spectroscopy agree with those of the sharp spectra measured directly; i.e. evaluations are necessary only for estimation of spectral shape. In this paper, we compared attenuation curves calculated using Hubbell's attenuation coefficients to estimate the coincidence or difference of spectra measured by the Compton spectroscopy and directly measured in the primary beam. Our results showed that the attenuation curves acquired from the reconstructed spectra measured by the Compton spectroscopy agreed with that acquired from corrected spectra that were directly measured. Moreover, the attenuation curves acquired from the spectra actually measured by adding aluminum attenuators agreed with theoretically calculated curves.