Register for https://www.smir.ch/BRATS/Start2015 to download some brain tumor datasets in which they give files that also contain the expert segmentation. With the expert segmentation of a dataset, you can create a mask, and then you can use the techniques at that link to calculate the area covered by that mask.
Then you can do something simple like nnz() [number of non-zeros] to count the pixels set in the mask. If the number does not match the other technique then you need to figure out why not, which might get into questions of how exactly to define "area" of a region. For example if a local boundary looks like
*
**
then is the area 3 (because three pixels are set)? Or is the area 1/2 (right triangle, height 1, base 1, formula is 1/2 * base * height)? Or should we count the 3 but also deduce that about half of a pixel should be considered set on the diagonal because that is "really" a straight line that happened to get pixelized?
To put the overall matter another way: you have two questions
- did your code find the tumor boundaries properly; and
- given a set of boundaries, does your code calculate the area properly
The BRATS dataset will give you images and expert boundaries, allowing you to concentrate on question #2 for the moment, namely whether you calculate area properly assuming the boundaries are proper. Which is all you asked about in your Question, verifying the area calculation.
But you could then apply your detection algorithm to some of those BRATS images and see how well it is able to do at matching the boundaries that the experts find.
Note: if you want to get energetic on proof that your area calculation works, take a picture of an object with known area, together with a known distance measure like a ruler to allow you to calculate the ratio of pixels to cm. Binarize the object part of the image, calculate the area, multiply by the square of the real-world ratio to get the area, and compare that to the known area. If you want to work with objects of irregular shape where directly measuring the area might be tricky, then cut the object out of something stiff and weigh the object and compare to the weight of a square of the material of known size.
