An evaluation of two methods of anatomical alignment of radiotherapy portal images.
ABSTRACT Two techniques have been developed at our institution to allow anatomical registration of digitized portal images to a simulation film. Accuracy of the portal image alignment methods is tested and single intrauser and multiple interuser variation is examined using each technique.
Method one requires the identification of anatomical fiducial points on a simulation image and its corresponding portal image. The parameters required to align the corresponding points are calculated by a least squares fit algorithm. Method two uses an anatomical template generated from the simulation image and superimposing it upon a portal image. The template is then adjusted by a computer mouse to obtain the best subjective anatomical fit on the portal image. Megavoltage portal images of a skull phantom with various known shifts and eight clinical image files were aligned by each method. Each data set was aligned several times by both a single user and multiple users.
Alignment of the anatomical phantom portal images demonstrates an accuracy of less than 0.8 +/- 0.9 mm and 0.7 +/- 1.0 degrees with either method. As out of plane rotation increased from 0 to 5 degrees, simulating out of plane malpositioning, alignment orthogonal to the plane of rotation worsened to 1.5 +/- 1.1 mm with the point method and 2.4 +/- 1.6 mm with the template method. Alignment parallel to the axis of the gantry rotation was insensitive to this change and remained constant as did the rotational alignment parameters. For the clinical image files the magnitude of variation for a single user is typically less than +/- 1 mm or +/- 1 degree. The magnitude of variation of alignment increased when multiple users aligned the same image files. The variation was dependent upon anatomical site and to a lesser degree the method of alignment used. The root mean square deviation of translational shifts range from +/- 0.68 mm when using the template method in the pelvis to as high as +/- 2.94 mm with the template method to align abdominal portal images. In the thorax and pelvis translational alignments along the horizontal axis were more precise than along the vertical axis. Multiple user variability was in part due to poor image quality, user experience, non rigidity of the anatomical features, and the difficulty in locating an exact point on a continuous anatomical structure.
In well controlled phantom studies both the fiducial point and template method provide similar and adequate results. The phantom studies show that alignment error and variance increase with distortion in anatomical features secondary to out of plane rotations. In clinical situations intrauser variation is small, however, multiple interuser variation is larger. The magnitude of variation is dependent upon the anatomical site aligned.
[show abstract] [hide abstract]
ABSTRACT: Conformal radiotherapy requires accurate patient positioning with reference to the initial three-dimensional (3D) CT image. Patient setup is controlled by comparison with portal images acquired immediately before patient treatment. Several automatic methods have been proposed, generally based on segmentation procedures. However, portal images are of very low contrast, leading to segmentation inaccuracies. In this study, we propose an intensity-based (with no segmentation), fully automatic, 3D method, associating two portal images and a 3D CT scan to estimate patient setup. Images of an anthropomorphic phantom were used. A CT scan of the pelvic area was first acquired, then the phantom was installed in seven positions. The process is a 3D optimization of a similarity measure in the space of rigid transformations. To avoid time-consuming digitally reconstructed radiograph generation at each iteration, we used two-dimensional transformations and two sets of specific and pregenerated digitally reconstructed radiographs. We also propose a technique for computing intensity-based similarity measures between several couples of images. A correlation coefficient, chi-square, mutual information, and correlation ratio were used. The best results were obtained with the correlation ratio. The median root mean square error was 2.0 mm for the seven positions tested and was, respectively, 3.6, 4.4, and 5.1 for correlation coefficient, chi-square, and mutual information. Full 3D analysis of setup errors is feasible without any segmentation step. It is fast and accurate and could therefore be used before each treatment session. The method presents three main advantages for clinical implementation-it is fully automatic, applicable to all tumor sites, and requires no additional device.International Journal of Radiation OncologyBiologyPhysics 06/2003; 56(1):259-65. · 4.11 Impact Factor