Calculate ROI pixel overlap

[iwilliams91]

Hello ImageJ community,

I wrote a macro to segment blood vessels in images acquired using confocal fluorescence microscopy. I want to test the accuracy of my automated algorithm against a "ground truth" manual segmentation. For my automated algorithm, I have a mask of the segmented blood vessels and for the manual segmentation I have manually-drawn ROIs. Is there a way in ImageJ to calculate the pixel overlap between the masks and the ROIs? Ideally, I'd like to be able to calculate the % of pixels in the ROI also contained within the mask (true positive) and the % of pixels in the mask that are not in the ROI (false positive). Are there any other metrics / calculations you would recommend to assess segmentation performance?

Thanks,
Ian