The lab I was affiliated with published a similar paper on gender dimorphism for pediatric hand and wrist radiographs in 2018. The experience of working on tha

The lab I was affiliated with published a similar paper on gender dimorphism for pediatric hand and wrist radiographs in 2018. The experience of working on that paper made me realize saliency/class-activation maps are only really useful if you want to know where an object is in the image, but not that helpful if you are trying to legitimately try and figure out what the AI is using to classify. I spent days looking at male and female hand radiographs + class activation maps trying to figure out what the system was using to tell the genders apart. I never figured it out. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6646498/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6646498/