We have different datasets with different image sizes because they were taken with different scanners (1.5T and 3T)? I was thinking of using the nearest neighbor interpolation method to reduce the size of the largest images to match the small ones and then, perform the statistical analysis. But I have not found any of this in literature.