These tools will no longer be maintained as of December 31, 2024. Archived website can be found here. PubMed4Hh GitHub repository can be found here. Contact NLM Customer Service if you have questions.


Pubmed for Handhelds

PUBMED FOR HANDHELDS

Search MEDLINE/PubMed

  • Title: Digital portal image registration by sequential anatomical matchpoint and image correlations for real-time continuous field alignment verification.
    Author: McParland BJ, Kumaradas JC.
    Journal: Med Phys; 1995 Jul; 22(7):1063-75. PubMed ID: 7565381.
    Abstract:
    Detection of radiotherapy field misalignments with electronic portal imaging devices requires the precise initial registration of the digital portal image with a reference image indicating the prescribed field alignment. Moreover, for real-time continuous detection this registration must be performed rapidly--arguably within 250 ms. The quality of this registration is sensitive to the ability of the user to accurately identify corresponding anatomical landmarks in the image pair. To improve the accuracy of the registration and, ultimately, that of the field misalignment measurement, we have developed a sequential digital portal image registration method using both user-identified anatomical matchpoints and image information. A first pass generates registration parameters from user-provided matchpoint coordinates and explicitly accounts for the uncertainty in matchpoint identification. The second pass uses both the initial registration parameters and image information to further improve the registration quality by maximizing cross correlations between segments of the image pair. As this registration method does not use massive matrix/vector computations common to other algorithms, it is inherently faster and well-suited for real-time field placement error detection. On a platform representative of those controlling many commercial electronic portal imaging devices (486 CPU), this algorithm registers portal images in times of less than 6 ms per matchpoint with errors of less than 2% in magnification, 0.5 degree in in-plane rotation, and less than 1 pixel dimension in in-plane translation. As the algorithm assumes a rigid-body geometry, it is sensitive to out-of-plane rotations. A quantitative analysis of this algorithm is presented, indicates its accuracy, and describes its sensitivity to out-of-plane rotations.
    [Abstract] [Full Text] [Related] [New Search]