Estimation of patient setup deviation with the lines algorithm using images in the AP view. An overlay of the portal image and the DRR in its original pose for weeks 1–8 is shown in the first and third columns. Images of the fused portal image and the DRR computed at the estimated pose for weeks 1–8 are shown in the second and fourth columns. Linear features from the portal image are green in color and features from the DRR are red. At alignment, linear features in both images should overlap minimizing the amount of red and green structures in the fused registered images and maximizing the amount of yellow. Note that the bony anatomy was registered correctly for all . The images displayed here do not include the correction for the treatment field offset.
An overlay of the portal image and the DRR in its original pose for weeks 6 and 7 is shown in the first column. Images of the fused portal image and the DRR computed at the estimated pose for these using the AP and RL image separately are shown in the second and third columns, respectively. Linear features from the portal image are green in color and features from the DRR are red. The AP-only registration algorithm was able to recover the pelvic tilt present in week 6 accurately, while the RL-only registration algorithm incorrectly aligned the femurs instead of the pubic symphysis and sciatic notch. The registration with the lateral image was wrong by approximately 2° and in its estimation of the rotation around the right-lateral axis and the translation along the posterior–anterior axis, respectively. No femur motion or pelvic tilt was observed in week 7. In this case, estimates of setup deviations obtained with the AP image and lateral image were comparable.
Table of displacements and registration errors for poses 1 through 15. The true pose, estimated pose, and registration error in estimating the true pose are listed for each of the six transformation parameters. The correlation coefficient, , obtained at the estimated pose, the total distance moved, , and the total registration error, , are also shown.
Mean, standard deviation, minimum, maximum, and median of the absolute value of the registration errors for each of the six transformation parameters and also for the total registration error for poses 1 through 15.
Article metrics loading...
Full text loading...