This work compares three methods of tracking lung tumor motion using an optical flow algorithm to analyze portal images. An earlier approach used sequential image pairs (CU) to track the motion. Errors in the position of the tumor were found to accumulate when patient traces were used and two new approaches were introduced. The first method re-calibrates the position of the tumor at the end-ofexhale (EoE) of each breathing cycle. This was done by computing the position directly between the new image and the previous reference image. The new image at the EoE is then assigned as the new reference frame, a process called reference shifting (RFSF). Between two EoE, sequential tracking was employed. The second approach derives the position of the tumor in each frame by applying the optical flow computation directly between each image and a reference image. However, since tracking is limited to a finite range (i.e. threshold), new reference images were required when the tumor exceeds a set distance from the reference position, a process called threshold shifting (THSF). Direct comparison of the position in subsequent images was made with respect to the new reference image. A 3D tumor prototype was fabricated using 3D printing techniques and seven patient traces were evaluated. Average position errors of -0.01 ± 0.65 mm, -0.10 ± 0.42 mm and -0.14 ± 0.25 were obtained for the CU, RFSF and THSF methods respectively.
CITATION STYLE
Troy Teo, P., Guo, K., Alayoubi, N., Kehler, K., & Pistorius, S. (2015). Drift correction techniques in the tracking of lung tumor motion. In IFMBE Proceedings (Vol. 51, pp. 575–578). Springer Verlag. https://doi.org/10.1007/978-3-319-19387-8_141
Mendeley helps you to discover research relevant for your work.