The first one would be if you took a bunch of points, the second if you map the waveform. The first one would not be too accurate, or only as accurate as the amount of points you decide to measure I should say. The second one would be accurate but I really don't know how you would collect that data. Remember you are looking for the average of how much the radius of your cluster deviates from the radius of a circle of the same area. Sounds like mega-calculus to me, if you find an answer I'd be interested to see what it is.