This paper introduces a novel imaging knowledge-informed dynamic optimal transport framework incorporating imging physics into both transport cost and equation, enabling to learn from unpaired real data for prospective medical image reconstruction.