You would have to make some assumptions about the particle shape and how large an area you would need to measure to average over the scale of the hetergeneous distribution. Phase contrast tomography would be a more direct and non-destructive measurement of the actual 3D distribution and shapes. I guess it depends on how precisely you want to determine the volume fraction, and if you care about the shape, size, and locational distribution.
I agree with the tomography technique. It is better to do this using STEM, since there is difference in composition between second phase particles and the matrix.