You can have a clinician(s) to manually segment the region of interests for a ground truth.

Sir do we have any standard repository from where we can use images to test the segmentation technique we have designed?

Maybe, you could use the TCGA images. Here is a random link I took as an example. There are over 1k metastatic breast cancers there:

http://www.cbioportal.org/case.do#/patient?caseId=TCGA-AR-A24Q&studyId=brca_tcga&tab=tissueImageTab

In that case your algo should distinguish tumor from necrosis if it is working properly.

Hello,

You can find the data in kaggle.com (has a lot of information about several topics) :

https://www.kaggle.com/paultimothymooney/breast-histopathology-images

and as a extra information here is a post to handle this data:

https://www.pyimagesearch.com/2019/02/18/breast-cancer-classification-with-keras-and-deep-learning/