Technical note: Towards more realistic 4DCT(MRI) numerical lung phantoms

Abstract

AbstractBackgroundNumerical 4D phantoms, together with associated ground truth motion, offer a flexible and comprehensive data set for realistic simulations in radiotherapy and radiology in target sites affected by respiratory motion.PurposeWe present an openly available upgrade to previously reported methods for generating realistic 4DCT lung numerical phantoms, which now incorporate respiratory ribcage motion and improved lung density representation throughout the breathing cycle.MethodsDensity information of reference CTs, toget her with motion from multiple breathing cycle 4DMRIs have been combined to generate synthetic 4DCTs (4DCT(MRI)s). Inter‐subject correspondence between the CT and MRI anatomy was first established via deformable image registration (DIR) of binary masks of the lungs and ribcage. Ribcage and lung motions were extracted independently from the 4DMRIs using DIR and applied to the corresponding locations in the CT after post‐processing to preserve sliding organ motion. In addition, based on the Jacobian determinant of the resulting deformation vector fields, lung densities were scaled on a voxel‐wise basis to more accurately represent changes in local lung density. For validating this process, synthetic 4DCTs, referred to as 4DCT(CT)s, were compared to the originating 4DCTs using motion extracted from the latter, and the dosimetric impact of the new features of ribcage motion and density correction were analyzed using pencil beam scanned proton 4D dose calculations.ResultsLung density scaling led to a reduction of maximum mean lung Hounsfield units (HU) differences from 45 to 12 HU when comparing simulated 4DCT(CT)s to their originating 4DCTs. Comparing 4D dose distributions calculated on the enhanced 4DCT(CT)s to those on the original 4DCTs yielded 2%/2 mm gamma pass rates above 97% with an average improvement of 1.4% compared to previously reported phantoms.ConclusionsA previously reported 4DCT(MRI) workflow has been successfully improved and the resulting numerical phantoms exhibit more accurate lung density representations and realistic ribcage motion.