Estimating person-specific neural correlates of mental rotation: A machine learning approach