SU‐F‐BRF‐06: Density Correction for Deformable Image Registration to Improve CT Number Mapping

Abstract

Purpose:To improve the accuracy of CT number mapping inside the lung using deformable image registration for cases having substantially large lung volume differences.Methods:The deep‐inspiration‐breath‐hold (DIBH) CT image and the end‐of‐exhalation (EE) phase image in 4DCT of three thoracic cancer patients were enrolled in this study. The lung volumes were manually delineated and the lung volume ratio of DIBH‐CT over EECT is 1.7, 1.8, and 2.0 respectively, demonstrating large volume differences. A demons‐based deformable registration was first applied to register the EECT to the DIBH‐CT for each patient, and the resulting deformation field deformed the EE‐CT image to the DIBH‐CT space. Based on the assumption that the mass of lung remains the same during respiration, we created a mass preserving model to correlate the lung density variations with the lung volumetric changes, which were characterized by the Jacobian derived from the deformation field. The Jacobian is used to correct the lung CT number transferred from the EE‐CT image. The deformed images with and without the density correction were compared using absolute intensity difference created by subtracting the deformed image from the DIBH‐CT image.Results:The deformable registration could register the lung shape very well, but the CT numbers inside the lung could not be registered correctly. Without density correction, the mean and standard deviation of the absolute CT number difference inside the lung were 59±48, 69±49, and 71±42 for the three cases. After density correction, these numbers changed to 18±35, 20±35, and 14±29, respectively, demonstrating an improvement in the accuracy of CT number mapping. The accumulative histogram of the intensity difference also showed that density correction improved CT number mapping.Conclusion:Density correction improved the CT number mapping inside the lung using deformable image registration for difficult cases having large lung volume differences.This work is partially supported by the CPRIT (Cancer Prevention Research Institute of Texas) grant RP110732.