replace SDM-PSI with peaks2images

[jdkent]

Had an inkling of an idea about transforming peaks to images and then trying to run a meta-analysis on the results: https://chatgpt.com/share/69449067-5e9c-800f-8111-02aeb6c78e1a

it would be a cool project, still have a lot to think about for calculating the variance for each image generated.

Is a solution for the problem when a researcher has some images, but not their inputs are images.

Another more straightforward idea from Nichols is to binarize the images and make them more like the coordinate map inputs.