15:31 Sep 2, 2011 

but I did see after I'd posted the question that horseshoe shapes are common. I don't know if that's the case here. If so, I imagine that it would be as you suggest, the distance from either "pointe" (tip of the horseshoe?) through the centroid to the point where this line intersects the diametrically opposite curve.

09:37 Sep 2, 2011 

Do you know the tunnel shape? It might be horseshoe-shaped, i.e. with a rounded top, inward-sloping sides, and a flat bottom (like a girl I used to know). I don't know how the measure the "diameter" of such tunnels, but they are generally given as having a "diameter". Whether this is the diameter of a round tunnel of equivalent cross-sectional area or a "diameter" from one of the bottom corners (pointes?), say, to a point "diagonally" opposite, passing through the centroid, presumably, I don't know.