Fig. 15

The mean, minimum and maximum values (over 25 noise realizations) of the normal and lesioned regions, computed on the corresponding segments of the polar maps, are compared. On the two graphs from the top, the values computed from the max-count polar maps are illustrated. On the two bottom graphs, the values computed from the mean-count polar maps are presented. For each type of polar map, the top parts (first and third graph) compare the intensity of the normal region to the intensity of the non-transmural lesion (L1), using the different reconstruction algorithms, whereas the bottom parts (second and fourth graph) show the same information for the transmural lesion (L2). The solid lines represent the mean values computed in the ground truth in the normal (red) and lesioned (yellow) tissues. The light-blue stripe highlights the reconstruction algorithm (MR-perf) that performs best in terms of similarity of the mean to the ground truth. To be noted that the mean-count polar maps of the OSEM3D+RR, despite being a good approximation of the ground truth on average, have very large error bars indicating that it can have very poor values for individual noise realisations