Cone Beam Computed Tomography Data Acquisition during VMAT Delivery with Subsequent Respiratory Phase Sorting Based on Projection Image Cross-correlation

Localization and verification of a tumor position was conventionally performed prior to radiotherapy. The present authors, in the past a few years, proposed in-treatment verification of the tumor position by simultaneously acquiring cone beam computed tomography (CBCT) data using an on-board kilovolt imaging system. Dose delivery was either rotational conformal 1) or volumetric modulated arc therapy (VMAT). 2,3) The resulting three-dimensional (3D) volume image was appropriate for still organs such as prostate. For respiratory moving cases such as lung tumors, image blurring may be observed thereby reducing the accuracy of the tumor localization. In the meantime, Sonke et al proposed four-dimensional (4D) CBCT for respiratory moving lung tumors 4) assuming that the motion of the diaphragm correlates well with that of a lung tumor. The respiratory phase was calculated from frame-by-frame changes of projection images after enhancing diaphragm-like features on the cranio-caudal axis. Another group also published similar results later. 5) Kavanagh et al proposed an alternative method for extracting a respiratory pattern from a set of projection images without relying on edge detection of the diaphragm position, where simple pixel value summation was employed followed by high pass filtering. 6) Just recently, in ESTRO annual meeting , Sonke presented 4D CBCT during VMAT delivery by extending their previous 4D CBCT research for a lung tumor. 7) In this letter, we present our preliminary clinical 4D CBCT imaging during VMAT delivery using Elekta Synergy (Crawly, UK). As was described in our previous report, 1) the current Synergy system does not allow simultaneous delivery of kV CBCT beams and MV rotational beams. Having known that kV portal imaging is available during MV rota-tional beam delivery, we developed in-house software for CBCT reconstruction by collecting all the kV portal images during gantry rotation. We have developed a new respiratory phase sorting algorithm based on image cross-correlation between adjacent two projection images. Image cross-correlation was previously used for registering a cone beam CT image with the planning CT image inside a specified region of interest while the former image was repeatedly shifted, and the patient treatment couch was displaced according to the shift vector for the highest correlation. 8) Similarly cross-correlation between the adjacent two projection images was calculated while one of the images was shifted stepwise, and the shift vector giving the highest correlation was considered as the tumor displacement caused by breathing. In this algorithm edge detection is not employed and …