Likely, this data was converted from DICOM to JPG. Though I understand the reason for this (identifiability), **very** important fields in the header are removed from this process. Likely most importantly is the pixel dimensions. As we've seen with the zooming from these images, they are not the same zoom level, thus a "pixel" means different things in real world units. Thus, conversion from DICOM to JPG was not lossless in that information. This information can be crucial to define things such as erosion, which 1 pixel erosion can mean 0.5mm or 2mm depending on the zooming.
Thus, if the data is converted, shouldn't it be resampled to the same real-world dimensions?
Created by John Muschelli muschellij2 I am not sure whether these images were ever in DICOM format or whether they were directly scanned in for the purposes of this challenge. I've cced @dongmeisun and @LouBridges who can better answer this question.