spatial accuracy measure and the completeness measure can be used to quantify the correctness and the accuracy of when segmenting an image to multiple regions. Check in the following paper which uses the same measurement for evaluating a segmentation problem that has been formulated as a clustering problem.Chapter Visual Feature Clustering Using Temporal, Color and Spatial ...

Article Multilevel Thresholding for Image Segmentation using Krill H...

Accuracy related measures with statistical test of significance can be used to detect which algorithm is providing better segmentation.

To compare segmentations, you need the golden standard: what is the 'true' ROI?

Then you could use Hausdorff, Jaccard or Dice (index/coefficient/distance).

Is there any difference between golden standard and groundtruth

Leena Silvoster

I think those two terms can be interchanged. In oncology, the golden standard of tumor segmentations is provided by manual delineations by experts. Ground truth rarely is provided and comes from pathology.

The following paper can be refereed

https://www.hindawi.com/journals/jam/2013/575414

Also when the ground truth image is available, dice coefficient , jaccard index can be used.

Many websites are having original and ground truth image for validation