Respiratory Signal Extraction From Thoracic Cone Beam CT Projections
H Yan1*, X Wang1, W Yin2, T Pan3, M Ahmad3, X Mou4, L Cervino1, X Jia1, S Jiang1, (1) University of California San Diego, La Jolla, CA, (2) Rice University, Houston, TX,(3) UT MD Anderson Cancer Center, Houston, TX, (4) Xi'an Jiaotong University, Xi'an, SHAANXI, CHINATU-C-213CD-12 Tuesday 10:30:00 AM - 12:30:00 PM Room: 213CD
Purpose: To obtain accurate and robust respiratory signals from thoracic cone beam CT (CBCT) projections, and to facilitate projection phase binning for 4D CBCT reconstruction.
Methods: During CBCT imaging, the gantry rotates around the patient while the patient is breathing. The CBCT projections corresponds to a 2-torus in a high-dimensional ambient space due to the periodicities of respiration and gantry rotation. Local principle component analysis (LPCA) is utilized to extract the respiration signals from this 2-torus. Two other methods, Amsterdam shroud (AS) and intensity analysis (IA), were compared with the LPCA method. All methods were evaluated on six patient data sets, by visually referring to the internal anatomy movement and the external RPM signals.
Results: The AS method works well for the three cases with the diaphragm in the field of view. While both IA and LPCA methods generally yield consistent respiration signals with the internal anatomy movement and RPM signal for all six cases, the LPCA method provides more accurate respiration signals for some extremely difficult cases.
Conclusions: 1) The proposed LPCA method is promising for detecting the breathing phase signals from the CBCT projections, especially for the cases with very weak signals such as 1 minute half-fan scan and the cases with no diaphragm in the field of view. 2) The AS method relies on a strong feature like the diaphragm, so it is suitable for cases with inclusion of the diaphragm. 3) The IA method utilizes the intensity variation in the projection images, working well for most cases. 4) Based on the internal anatomy information contained in projections, accurate respiratory signals can be extracted from the internal anatomy information in the projection data. To improve the robustness of the LPCA, we are investigating to incorporate the AS, IA in the LPCA methods.