Shin Miyakawa1, Hidenobu Tachibana2, Shunsuke Moriya3, Tomoyuki Kurosawa1, Teiji Nishio1. 1. Doctoral Program in Medical Physics, Graduate School of Medicine, Tokyo Women's Medical University, Tokyo 1628666, Japan. 2. Particle Therapy Division, Exploratory Oncology Research and Clinical Trial Center, National Cancer Center, Chiba 2778577, Japan. Electronic address: htachiba@east.ncc.go.jp. 3. Doctoral Program in Biomedical Sciences, Graduate School of Comprehensive Human Sciences, University of Tsukuba, Chiba 3058577, Japan.
Abstract
PURPOSE: This study aimed to evaluate different deformable image registration (DIR) parameters for the open-source NiftyReg package in its application to DIR-based ventilation imaging. METHODS: Two three-dimensional (3D)-computed tomography (CT) scans of a non-rigid air-ventilating phantom were acquired at peak exhalation and peak inhalation, with xenon (Xe) gas being used as an air-based contrast agent. We compared four different sets of DIR parameters, including one set with two-step deformation and three sets with four-step deformation. For spatial accuracy, the target registration error (TRE) was calculated for 16 landmarks. For ventilation imaging accuracy, DIR-based ventilation images were generated using Jacobian determinant (JD) metrics, and changes in Hounsfield unit (HU) values between the two exhalation and inhalation CT images were subsequently measured. The correlation coefficients between the JD metrics and changes in HU values were calculated. RESULTS: The mean TRE values were 4.5 ± 4.7 mm (maximum, 12.3 mm), 1.47 ± 0.71 mm (maximum, 2.6 mm), 1.56 ± 0.70 mm (maximum, 2.8 mm), and 1.53 ± 0.66 mm (maximum, 2.5 mm) for the two-step deformation and three four-step deformations, respectively. The four-step deformations (R = - 0.71, -0.65, and -0.61) showed stronger correlation coefficients than the two-step deformation (R = -0.40). CONCLUSIONS: The accuracy of DIR-based ventilation imaging may vary with different DIR parameter settings, even though spatial accuracy may be tolerable and within guidelines. We found adequate parameter settings for four-step NiftyReg DIR for visualization of simulated pulmonary ventilation function.
PURPOSE: This study aimed to evaluate different deformable image registration (DIR) parameters for the open-source NiftyReg package in its application to DIR-based ventilation imaging. METHODS: Two three-dimensional (3D)-computed tomography (CT) scans of a non-rigid air-ventilating phantom were acquired at peak exhalation and peak inhalation, with xenon (Xe) gas being used as an air-based contrast agent. We compared four different sets of DIR parameters, including one set with two-step deformation and three sets with four-step deformation. For spatial accuracy, the target registration error (TRE) was calculated for 16 landmarks. For ventilation imaging accuracy, DIR-based ventilation images were generated using Jacobian determinant (JD) metrics, and changes in Hounsfield unit (HU) values between the two exhalation and inhalation CT images were subsequently measured. The correlation coefficients between the JD metrics and changes in HU values were calculated. RESULTS: The mean TRE values were 4.5 ± 4.7 mm (maximum, 12.3 mm), 1.47 ± 0.71 mm (maximum, 2.6 mm), 1.56 ± 0.70 mm (maximum, 2.8 mm), and 1.53 ± 0.66 mm (maximum, 2.5 mm) for the two-step deformation and three four-step deformations, respectively. The four-step deformations (R = - 0.71, -0.65, and -0.61) showed stronger correlation coefficients than the two-step deformation (R = -0.40). CONCLUSIONS: The accuracy of DIR-based ventilation imaging may vary with different DIR parameter settings, even though spatial accuracy may be tolerable and within guidelines. We found adequate parameter settings for four-step NiftyReg DIR for visualization of simulated pulmonary ventilation function.