Hi, fantastic work!. I want to see if I could apply it to my project, but I noticed that the normalisation of the PET images is done over non-standardised uptake values (PET_MAX = 32767). Since I have my PET data in SUVs (1-20 aprox), and I am actually interested in generating the ratio between two specific sessions that will give me the image view of particular pathology, by using that maximum nothing will be over the static threshold STATIC_THRESHOLD = 100), and so, no mask will be generated. I will want to know more about how did you decide this threshold, and how it affect on the attention maps generated by the u-net?
Hi, fantastic work!. I want to see if I could apply it to my project, but I noticed that the normalisation of the PET images is done over non-standardised uptake values (PET_MAX = 32767). Since I have my PET data in SUVs (1-20 aprox), and I am actually interested in generating the ratio between two specific sessions that will give me the image view of particular pathology, by using that maximum nothing will be over the static threshold STATIC_THRESHOLD = 100), and so, no mask will be generated. I will want to know more about how did you decide this threshold, and how it affect on the attention maps generated by the u-net?