Program Information
Extraction of Diaphragm Motion Curve From Sinogram of KV Cone Beam Projection Images Using Connected Threshold Image Filter (CTIF), a Validation Study
J Zhou*, J Liang , D Zhou , D Yan , William Beaumont Hospital, Royal Oak, MI
Presentations
TU-L-GePD-J(A)-4 (Tuesday, August 1, 2017) 1:15 PM - 1:45 PM Room: Joint Imaging-Therapy ePoster Lounge - A
Purpose: Diaphragm motion has been reported to be a reliable surrogate for lung tumor motion. We investigate extraction of diaphragm motion curve from sinogram of kV cone-beam projections using CTIF.
Methods: Eight on-board cone beam (CB) projection image sets of lung tumor were used in this study. Acquisition sampling frequency was 5.6Hz. A Gaussian gradient operator in the cranial-caudal (CC) direction was applied to each projection and the projection was projected to one line in the CC direction. All lines from all projections within one scan were aligned to form a 2D sinogram image (‘Amsterdam Shroud’ image). One or several seed points were manually selected within the diaphragm motion curve on the sinogram and a CTIF was used to extract the breathing curve. To evaluate the accuracy, reference curves were created with semi-automatically determined and manually verified diaphragm position on each projection. The mean and the Pearson correlation coefficient between the extracted and the reference motion curve were analyzed.
Results: Mean diaphragm motions (RMS value) were 0.477 ± 0.084 cm and 0.492 ± 0.074 cm for the extracted and reference curves, respectively. Mean differences between them and the Pearson correlation coefficient were 0.015 ± 0.231 cm and 0.863 (p<0.001), respectively. 75% of the differences were < 0.250 cm. The majority of higher discrepancies occurred at the max in-hale positions.
Conclusion: Diaphragm motion can be reliably and efficiently extracted from sinogram of the CB projection images. The effect of less accuracy at the max inhale positions could be insignificant due to the short period of tumor staying at those positions and the limited phase resolution of 4D-CT image sets. The extraction algorithm can be improved in the future with the help of pre-knowledge and advanced imaging process.
Contact Email: