I found this comment
I would also like to point out that invariance of domain and the statement that every continuous bijection from $\mathbf{R}^n$ to $\mathbf{R}^n$ is a homeomorphism are equivalent. Therefore I would not expect there to be a proof that avoids this machinery.
but couldn't prove it myself using fundamental methods, i.e. no algebraic topology. As far as I can tell, invariance of domain is equivalent to the statement restricted to open balls in $\mathbf{R}^n$. But from there, I don't see how to prove this using only the statement about continuous bijections and no algebraic topology. Clearly, the preimage of an open ball is open, but how can I conclude that it's homeomorphic to an open ball?