User introduced uncertainties in the measurement of field misalignment by a computer aided technique

While images acquired using an electronic portal imager may be visually compared to other portal or simulator images to assess patient alignment, a more quantitative comparison is desirable. Computer aided alignment tools are available which are based on user placed landmarks. The purpose of this study was to estimate uncertainties introduced by the user to the final measurement of field misalignment using such a tool. Both intra- and inter-user reproducibility were assessed. To complete this task, a number of image pairs including both phantom and patient images were compared by multiple observers. Results of the comparisons (x- and y- translation and rotation) were tabulated for each image pair and their reproducibility assessed by calculating a mean and standard deviation. User introduced uncertainty was found to be independent of the magnitude of rotation or x- or y- translation. In all cases, there was no difference between intra-observer and inter-observer uncertainty. For clinical cases, there is a significant difference between uncertainty in x- and y-translation due to both image quality and patient anatomy. In addition, the magnitude of uncertainty tracks qualitatively with image quality and number of available anatomical landmarks. The decision to make a correction in field alignment must be made considering these uncertainty estimates. Image comparison must be fully automated to eliminate uncertainties introduced by the user.