Fig. 2From: ParaPET: non-invasive deep learning method for direct parametric brain PET reconstruction using histoimagesDetail of the first three steps in Fig. 1 (i.e. a–c) showing how histomimages were deblurred and then cleaned using a U-Net for an example subject. Histoimages in a were generated using a Gaussian filter as described in the methods section, followed by iterative deconvolution via the Richardson-Lucy algorithm (b). The U-Net architecture used to denoise images is showcased in c, resulting in dynamic denoised images as shown in dBack to article page