This project proposes a novel approach for brain oxygen saturation (SO₂) mapping by integrating quantitative BOLD (qBOLD) modeling with a Physics-Informed Neural Network (PINN). Unlike conventional curve fitting, which is sensitive to noise and limited in generalization, the proposed PINN incorporates physical constraints to produce stable and physiologically plausible SO₂ maps. Both simulation and patient studies demonstrated higher accuracy, robustness, and reliability.