PURPOSE: To establish a practical and accurate motion tracking method for the development of rigid motion correction methods in helical x-ray computed tomography (CT). METHODS: A commercially available optical motion tracking system provided 6 degrees of freedom pose measurements at 60 Hz. A 4 × 4 calibration matrix was determined to convert raw pose data acquired in tracker coordinates to a fixed CT coordinate system with origin at the isocenter of the scanner. Two calibration methods, absolute orientation (AO), and a new method based on image registration (IR), were compared by means of landmark analysis and correlation coefficient in phantom images coregistered using the derived motion transformations. RESULTS: Transformations calculated using the IR-derived calibration matrix were found to be more accurate, with positional errors less than 0.5 mm (mean RMS), and highly correlated image voxel intensities. The AO-derived calibration matrix yielded larger mean RMS positional errors (≈ 1.0 mm), and poorer correlation coefficients. CONCLUSIONS: The authors have demonstrated the feasibility of accurate motion tracking for retrospective motion correction in helical CT. Their new IR-based calibration method based on image registration and function minimization was simpler to perform and delivered more accurate calibration matrices. This technique is a useful tool for future work on rigid motion correction in helical CT and potentially also other imaging modalities.
PURPOSE: To establish a practical and accurate motion tracking method for the development of rigid motion correction methods in helical x-ray computed tomography (CT). METHODS: A commercially available optical motion tracking system provided 6 degrees of freedom pose measurements at 60 Hz. A 4 × 4 calibration matrix was determined to convert raw pose data acquired in tracker coordinates to a fixed CT coordinate system with origin at the isocenter of the scanner. Two calibration methods, absolute orientation (AO), and a new method based on image registration (IR), were compared by means of landmark analysis and correlation coefficient in phantom images coregistered using the derived motion transformations. RESULTS: Transformations calculated using the IR-derived calibration matrix were found to be more accurate, with positional errors less than 0.5 mm (mean RMS), and highly correlated image voxel intensities. The AO-derived calibration matrix yielded larger mean RMS positional errors (≈ 1.0 mm), and poorer correlation coefficients. CONCLUSIONS: The authors have demonstrated the feasibility of accurate motion tracking for retrospective motion correction in helical CT. Their new IR-based calibration method based on image registration and function minimization was simpler to perform and delivered more accurate calibration matrices. This technique is a useful tool for future work on rigid motion correction in helical CT and potentially also other imaging modalities.
Authors: Jang-Hwan Choi; Rebecca Fahrig; Andreas Keil; Thor F Besier; Saikat Pal; Emily J McWalter; Gary S Beaupré; Andreas Maier Journal: Med Phys Date: 2013-09 Impact factor: 4.071