Noise reduction by temporal estimation in perfusion computed tomography

Perfusion CT protocols consist in an injection of contrast agent, a continuously acquisition of projection data during typically 40 seconds, and a subsequent reconstruction of a temporal sequence of images. The time evolution of the attenuation value (time attenuation curve) is used to compute functional parameters of the patient. Due to the long irradiation time, the dose per time frame has to be strongly reduced. In consequence, typically the contrast enhancement is of the same amplitude as the noise. The aim of this contribution is to better exploit the information in the measured data either to further reduce dose or to increase the signal to noise ratio (SNR) of the reconstructed images. A priori information about the bandwidth of the spectrum of time attenuation curves is considered in the framework of sampling theory. Two approaches are presented. In the first approach, temporal smoothing is used to increase the SNR. In the second one, the time between two frames is increased while keeping the total applied dose. Efficient temporal estimation schemes are applied to account for the sparse temporal sampling. Results on clinical data show an increase of the SNR without any significant loss in temporal or spatial resolution