Preliminary study of quantitative X-ray spectral imaging with spectral deconvolution

Abstract

The degrading factors of photon counting detectors such as charge-sharing, K-escape, resulting in spectrum distortion, set the crucial limitation for its application in quantitative imaging. We present a method to eliminate the degrading factors by spectral deconvolution. Simulation study and experimental study were carried out to verify its effectiveness and robustness. In simulation study, the method can significantly reduce the bias between detected photon counts and the theoretical value, which implies the effectiveness. Poisson noise was added to test the robustness and we find the result shows no significant difference to noiseless result when up to 106 photons are detected. In experimental study, theoretical spectrum is not easily accessible, instead we verified the effectiveness by calculating the width of an aluminum plate based on Beer-Lambert Law, which is supposed to be consistent among the whole energy range. An improvement can be seen after deconvolution and the output shows a good agreement with simulation result. Furthermore, we simulated a CT reconstruction process using the proposed method and found the reconstructed μ fits the theoretical value very well, which manifests a promising potential in quantitative X-ray spectral imaging.