Incorporation of respiratory motion estimation within a map super-resolution algorithm for image enhancement in 4D PET

Respiratory synchronized acquisitions lead to noisy images. Super-resolution (SR) aims at generating a high-resolution image or image sequence from several slightly different low-resolution images. A maximum a posteriori (MAP) SR algorithm has been implemented and applied for simultaneous motion estimation and image quality enhancement of respiratory synchronized gated frames. The algorithm was tested on GATE simulated images reconstructed using the one-pass list-mode EM (OPL-EM) algorithm. SR was performed on the gated images with the proposed MAP algorithm. The function yielded by the MAP method was optimized with respect to both the high-resolution image and the motion parameters through a steepest descent algorithm. Two motion estimation models were compared: one based on a sinusoidal model of fixed phase with the motion amplitude optimized for each voxel, the other considering motion as a linear combination of cubic b-splines. Lesion positions were recovered using profiles and image enhancement was assessed by signal-to-noise ratio (SNR) and contrast measures.