I'm looking for a citable reference for the following (perhaps folkloric?) result on topological field theories.

(There are obviously generalizations to other dimensions; I'm happy with just the 2-dimensional case described below.)

Let $Y$ be a 2-manifold, $c$ be a finite set of oriented points in $\partial Y$. A *spine* for $(Y; c)$ is a oriented 1-complex $G\subset Y$ such that
$G \cap \partial Y = c$ and $Y$ deformation retracts onto $G$.

If $C$ is a semisimple pivotal category (or something similar enough for the following to make sense), and now $c$ is a finite set of oriented points in $\partial Y$ labelled by simple objects of $C$, then a *labeling* of a spine $G$ is a labeling of the edges of $G$ by simple objects of $C$ (compatible with $c$) and labeling of the vertices of $G$ by (appropriate) morphisms of $C$.

**Spine Lemma** Suppose $(Y, c)$ has a spine (exactly when each connected component has non-empty boundary). The Turaev-Viro vector space for $(Y,c)$ has a basis the set of labellings of any fixed spine $G$, where the vertex labels should be drawn from a basis for the relevant morphism space.

Can anyone point to a place in the literature this (or a generalization) has been proved?

(Sketch: take an arbitrary element of the Turaev-Viro vector space, compress it onto the spine, using basic relations. In the other direction, labellings are linearly independent, using a nondegenerate pairing on each morphism space.)