As Sargas said, you can use an other object (with small size and low point count) to build an interior and exterior frame that detects collisions. Use the Pin behavior to make the instances of this collision object follow and rotate with your Sprite.
A high detail collision polygon however won't be inefficient if there is only one object. Even if there are many, if not on screen at the same time, you can disable them and only activate the rotation, collisions and physics when they are on screen, thus gaining performance.
Here is an example using both methods at the same time. I don't know which is more efficient, you can play around and figure it out