Simultaneous Reconstruction and Denoising of Spectral CT Images

Spectral CT based on photon-counting detectors offers the ability of obtaining CT images for different energy bins with a single X-ray excitation. The signal-to-noise ratio (SNR) of the reconstructed CT images at an energy bin is however drastically decreased. We present a study of different reconstruction methods that integrate explicit denoising in the reconstruction. The spectral CT reconstruction is then formulated as an optimization problem, in which four popular denoising techniques, including guided filter, wavelet transform, nonlocal means, and block-matching and 3D filtering, are investigated and the split Bregman algorithm is utilized to accelerate convergence speed. The so-obtained simultaneous reconstruction and denoising methods are evaluated in terms of root mean square error (RMSE), SNR and material decomposition performance. The results show that with the mechanism of simultaneous reconstruction and denoising, the images reconstructed from spectral CT have smaller RMSEs and higher SNRs while giving better performance in material decomposition.