I think an expression like yours above will work in the Decision Tree.  If you start a new tree, in your first node make your decision to be 'b1 gt .2 and b1 lt .5'.  Map your index to b1.  Add a child node to the result of the first binary decision.  Add a new decision of 'b2  gt .8'.  This should split classes between not classified and classified as forest for the first decision.  Then within the forest class, class 1 or class 2 depending on whether or not it meets the second decision criteria.