Performance comparison of water phantom based flat field correction methods for photon-counting spectral CT images: experimental results

Photon-counting detector (PCD) is one of the leading candidates for the next generation X-ray detector. The PCD enables effective multiple material decomposition and low dose CT imaging, but several correction techniques to compensate non-ideal effects of the PCD should be preceded to maximize the imaging performance. In this study, we present experimental results of water phantom based flat field correction methods to reduce energy distortion for photon-counting spectral CT images. We compared the performance of the flat field correction using single and multiple water phantoms, and examined the robustness of the correction method for various imaging tasks. The flat field correction using single water phantom was conducted by normalizing measured projection data of an object with those of the centered large water phantom. Then, ideal water phantom image was added to the reconstructed image of the water normalized projection data, which is equivalent to the conducting first order fitting of detector pixel gain. The dynamic range of detector pixel gain was increased by using multiple water phantoms. Each detector pixel gain was modeled using 2nd order polynomial, and the coefficients were estimated by comparing measured projection data of water phantoms with those of ideal water phantoms. The estimated gain of each detector pixel was tested for various imaging tasks. The results show that the single water phantom based method reduces the ring artifacts for a centered object effectively, but residual ring artifacts are introduced for the off-centered multiple water phantoms. In contrast, the multiple water phantoms based method reduces the ring artifacts effectively regardless of the size and location of the object. In addition, the multiple water phantoms based method provides improved SNR by 38.3 to 41.9% and CNR by 52.8 to 56.5%, compared to the single water phantom based method.