Image Guided, Motion-free Patient Body Setup using 3D Volumetric Image Registration of Classified Stable Bony Landmarks

Abstract

Purpose/Objective(s)Different motions and motion artifacts residing in planning and setup CT images show altered and distorted patient anatomies, which undermine the reliability of rigid image registration and introduce hidden uncertainties in image-guided radiation therapy (IGRT). A motion-free approach is pursued to improve the IGRT patient setup.Materials/MethodsThe 3D volumetric image registration (3DVIR) utilizes classified, motion-less bony landmarks for rigid image registration. Image classification is performed by controlling voxel transparency using an opacity lookup table based on image histogram in real-time. Both visual and quantitative measures of color homogeneity distributed on the volumetric landmarks are used as registration criteria. Mutual information (MI) based registration is used for comparison. For respiratory motion characterization, four sets of 4D-CT images with 10 phases under normal respiration are studied. For IGRT patient setup, daily Tomotherapy MVCT images are aligned with the planning kVCT image. Two patients' daily MVCT and planning kVCT images are used for IGRT setup for treatment of lesions in or near the spine.ResultsUsing the 4D-CT images, stable bony landmarks in the averaged respiratory cycle have been identified. The native alignment of the 3D bony structures in different respiratory phases shows that the spine, clavicle, scapula and upper sternum are motion-free under normal respiration, with an uncertainty of < ± 0.2 mm, while the ribs and lower sternum may have >2 mm displacement. MI-based rigid image registration, however, produces an overall 1-2 mm shift from the native alignment. Using these stable bony landmarks, registrations of the MVCT and kVCT images, free from respiratory, cardiac and digestive motions, have been achieved. The MVCT and kVCT carry distinctive motions and motion artifacts, as they are acquired in different time frames. The kVCT (scans at 1 second per rotation, or 1 SPR) represents an anatomy in an arbitrary breathing stage and contains image distortion artifacts (imaging anatomies at different respiratory phases); while the MVCT (scans at 10 SPR) has image blurring artifacts (averaged over respiratory cycles). This 3DVIR approach does not use moving anatomies for alignment, eliminating the motion/deformation uncertainties, and therefore is superior to any rigid image registration that does. This result indicates that the use of stable bony landmarks should provide improved accuracy and reproducibility for IGRT patient setup. These improvements can be critical in stereotactic body radiation therapy (SBRT).ConclusionsThis image-guided, motion-free approach eliminates the interference of moving soft tissues and ribs, providing accurate and reproducible IGRT patient setup. Purpose/Objective(s)Different motions and motion artifacts residing in planning and setup CT images show altered and distorted patient anatomies, which undermine the reliability of rigid image registration and introduce hidden uncertainties in image-guided radiation therapy (IGRT). A motion-free approach is pursued to improve the IGRT patient setup. Different motions and motion artifacts residing in planning and setup CT images show altered and distorted patient anatomies, which undermine the reliability of rigid image registration and introduce hidden uncertainties in image-guided radiation therapy (IGRT). A motion-free approach is pursued to improve the IGRT patient setup. Materials/MethodsThe 3D volumetric image registration (3DVIR) utilizes classified, motion-less bony landmarks for rigid image registration. Image classification is performed by controlling voxel transparency using an opacity lookup table based on image histogram in real-time. Both visual and quantitative measures of color homogeneity distributed on the volumetric landmarks are used as registration criteria. Mutual information (MI) based registration is used for comparison. For respiratory motion characterization, four sets of 4D-CT images with 10 phases under normal respiration are studied. For IGRT patient setup, daily Tomotherapy MVCT images are aligned with the planning kVCT image. Two patients' daily MVCT and planning kVCT images are used for IGRT setup for treatment of lesions in or near the spine. The 3D volumetric image registration (3DVIR) utilizes classified, motion-less bony landmarks for rigid image registration. Image classification is performed by controlling voxel transparency using an opacity lookup table based on image histogram in real-time. Both visual and quantitative measures of color homogeneity distributed on the volumetric landmarks are used as registration criteria. Mutual information (MI) based registration is used for comparison. For respiratory motion characterization, four sets of 4D-CT images with 10 phases under normal respiration are studied. For IGRT patient setup, daily Tomotherapy MVCT images are aligned with the planning kVCT image. Two patients' daily MVCT and planning kVCT images are used for IGRT setup for treatment of lesions in or near the spine. ResultsUsing the 4D-CT images, stable bony landmarks in the averaged respiratory cycle have been identified. The native alignment of the 3D bony structures in different respiratory phases shows that the spine, clavicle, scapula and upper sternum are motion-free under normal respiration, with an uncertainty of < ± 0.2 mm, while the ribs and lower sternum may have >2 mm displacement. MI-based rigid image registration, however, produces an overall 1-2 mm shift from the native alignment. Using these stable bony landmarks, registrations of the MVCT and kVCT images, free from respiratory, cardiac and digestive motions, have been achieved. The MVCT and kVCT carry distinctive motions and motion artifacts, as they are acquired in different time frames. The kVCT (scans at 1 second per rotation, or 1 SPR) represents an anatomy in an arbitrary breathing stage and contains image distortion artifacts (imaging anatomies at different respiratory phases); while the MVCT (scans at 10 SPR) has image blurring artifacts (averaged over respiratory cycles). This 3DVIR approach does not use moving anatomies for alignment, eliminating the motion/deformation uncertainties, and therefore is superior to any rigid image registration that does. This result indicates that the use of stable bony landmarks should provide improved accuracy and reproducibility for IGRT patient setup. These improvements can be critical in stereotactic body radiation therapy (SBRT). Using the 4D-CT images, stable bony landmarks in the averaged respiratory cycle have been identified. The native alignment of the 3D bony structures in different respiratory phases shows that the spine, clavicle, scapula and upper sternum are motion-free under normal respiration, with an uncertainty of < ± 0.2 mm, while the ribs and lower sternum may have >2 mm displacement. MI-based rigid image registration, however, produces an overall 1-2 mm shift from the native alignment. Using these stable bony landmarks, registrations of the MVCT and kVCT images, free from respiratory, cardiac and digestive motions, have been achieved. The MVCT and kVCT carry distinctive motions and motion artifacts, as they are acquired in different time frames. The kVCT (scans at 1 second per rotation, or 1 SPR) represents an anatomy in an arbitrary breathing stage and contains image distortion artifacts (imaging anatomies at different respiratory phases); while the MVCT (scans at 10 SPR) has image blurring artifacts (averaged over respiratory cycles). This 3DVIR approach does not use moving anatomies for alignment, eliminating the motion/deformation uncertainties, and therefore is superior to any rigid image registration that does. This result indicates that the use of stable bony landmarks should provide improved accuracy and reproducibility for IGRT patient setup. These improvements can be critical in stereotactic body radiation therapy (SBRT). ConclusionsThis image-guided, motion-free approach eliminates the interference of moving soft tissues and ribs, providing accurate and reproducible IGRT patient setup. This image-guided, motion-free approach eliminates the interference of moving soft tissues and ribs, providing accurate and reproducible IGRT patient setup.