Quantitative lung perfusion blood volume maps using effective Z images learned from single-kV pulmonary CT angiography data

Towards providing a “one-stop-shop” solution to anatomical and parenchymal perfusion imaging for pulmonary embolism (PE) evaluation, this work developed a method that uses a deep neural network to estimate effective atomic number (Zeff) information embedded in single-kV pulmonary CT angiography projection data. Based on the estimated Zeff map and the definition of perfusion blood volume (PBV), quantitatively accurate PBV maps can be generated. A multi-center human subject study demonstrates that the proposed single-kV CT and Zeff based PBV method provides a more sensitive and specific biomarker to quantify pulmonary perfusion defects compared with the iodine material image-based perfusion estimation method.