PURPOSE: A solution for automatic registration of 3D Rotational Angiography (XA) to CT/MR of the liver. Targeted for use in treatment planning of liver interventions. METHODS: A shape-based approach to registration is proposed that does not require specification of landmarks nor is it prone to local minima like purely intensity-based registration methods. Through the use of vessel characteristics, accurate registration is possible even in the presence of deformations induced by catheters and respiratory motion. RESULTS: Registration was performed on eight pairs of multiphase CT angiography and 3D rotational digital angiography datasets. Quantitative validation of the registration accuracy using vessel landmarks was performed on these datasets. The validation study showed that the method has a registration error of 9.41 +/- 4.13 mm. In addition, the computation time is well below 60 s making it attractive for clinical application. CONCLUSION: A new method for fully automatic 3DXA to CT/MR image registration was developed and found to be efficient and accurate using clinically realistic datasets.
PURPOSE: A solution for automatic registration of 3D Rotational Angiography (XA) to CT/MR of the liver. Targeted for use in treatment planning of liver interventions. METHODS: A shape-based approach to registration is proposed that does not require specification of landmarks nor is it prone to local minima like purely intensity-based registration methods. Through the use of vessel characteristics, accurate registration is possible even in the presence of deformations induced by catheters and respiratory motion. RESULTS: Registration was performed on eight pairs of multiphase CT angiography and 3D rotational digital angiography datasets. Quantitative validation of the registration accuracy using vessel landmarks was performed on these datasets. The validation study showed that the method has a registration error of 9.41 +/- 4.13 mm. In addition, the computation time is well below 60 s making it attractive for clinical application. CONCLUSION: A new method for fully automatic 3DXA to CT/MR image registration was developed and found to be efficient and accurate using clinically realistic datasets.