How about Root (node with children, but no parent), Node (node with children and parent) and Leaf (node with no children and parent)?
You can then distinguish by name and position within the tree structure (E.g. DepartmentRoot, DepartmentNode, DepartmentLeaf) if need be..
Update Following Comment from OP
Looking at your question, you said that "some" are special, and in your diagram, you have different nodes looking differently at different levels. The nodes may be different in their design, you can build a tree structure many ways. For example, a single abstract class that can have child nodes, if no children, its a leaf, if no parent, its a root but this can change in its lifetime. Or, a fixed class structure in which leafs are a specific class type that cannot have children added to them in any way.
IF your design does not need you to distinguish nodes differently depending on their position (relative to the root) it suggests that you have an abstract class used for them all.
In which case, it raises the question, how is it different?
If it is simply the same as the standard node everywhere else, but with a bit of styling, how about StyledNode? Do you even need it to be seperate (no style == no big deal, it doesn't render).
Since I don't know the mechanics of how the tree is architected, there could possibly be several factors to consider when naming.