Premium
Poster — Thur Eve — 71: Evaluation of a Breathing Phantom's Tumour Motion Using Portal Images and an Optical Flow Tracking Algorithm
Author(s) -
Teo TPT,
Crow RN,
Sasaki D,
Pistorius S
Publication year - 2010
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.3476176
Subject(s) - imaging phantom , optical flow , tracking (education) , pixel , image resolution , computer vision , computer science , artificial intelligence , translation (biology) , track (disk drive) , algorithm , physics , optics , image (mathematics) , psychology , pedagogy , biochemistry , chemistry , messenger rna , gene , operating system
Purpose: The efficacy of external beam radiotherapy depends on the accuracy of radiation dose being delivered to the tumor while minimizing dose to surrounding tissues. In order to improve outcome, it is essential to have a real‐time method to track the motion of the tumor and surrounding critical organs. Methods: This paper presents the development, performance and feasibility of using an optical flow algorithm to track the motion of a tumor using Electronic Portal Images (EPI). The algorithm calculates optical flow vectors ( u,v ) associated with temporal changes of the image intensities in each pixel. Both single and multi‐resolution algorithms were developed and their performance evaluated using two methods. The first method uses a simulated image translation and the second uses an actual image sequence of a moving lung tumor in an anthropomorphic phantom (RSD RS‐1500). Results: For a simulated 7‐pixel image translation, optical flow detects a corresponding 7‐pixel shift. For the actual image sequence, a maximum velocity of 14 mm/s and 24 mm/s is detected with the single‐ and multi‐resolution algorithm respectively. This corresponds to a 61% and 33% difference compared to the measured maximum velocity. Conclusions: This study has demonstrated the potential of using an optical flow algorithm to track tumor and critical structure motion. In this novel attempt to compare the optical flow and measured velocities on EPI, it has been shown that further refinements including a modified actuator, increased EPI frame rate, and increased layers of the multi‐resolution algorithm are required to improve the tracking accuracy.