Applications of cone beam computed tomography in radiotherapy treatment planning.

Abstract

In recent years Image-Guided Radiotherapy (IGRT) has experienced many technical advances. One of the most significant has been the widespread implementation of kilovoltage imagers attached to the gantry of linear accelerators (LINACs); these units are capable of 2D planar imaging, fluoroscopy and 3D Cone Beam Computed Tomography (CBCT) imaging. With CBCT imaging, the treatment plan can be modified based on patient’s anatomy just before the treatment session. This method of Adaptive Radiotherapy (ART) helps in managing a patient’s treatment by compensating for the effect of daily setup variation and changes to the tumour during the course of radiotherapy. Currently the image quality of CBCT is sufficient for patient set-up verification; however the use of CBCT for dose calculations requires reproducible CT numbers in order to be used effectively during ART. The aim of this project was to investigate methods to improve the image quality of CBCT datasets in order to facilitate their use in dosimetric calculations. The project was divided into two major parts. In the first part, the conventional Feldkamp-Davis-Kress (FDK) cone-beam reconstruction algorithm was implemented in Matlab. The algorithm was then modified using weighting factors for data redundancy and for nonequal cone angles. A 2D adaptive filter was used to remove noise and to compensate for the loss of resolution. A modified in-house reconstruction algorithm was developed and the image quality obtained was comparable to reconstructed images obtained using the Varian OBI system software. The images are free of crescent artifacts and showed a maximum spatial resolution of 7 line pairs/cm. The effect of different reconstruction filters on CBCT image quality was also studied and guidelines were produced for different anatomical sites to assist in choosing appropriate filters to achieve optimal reconstructed image quality. In the next part of the research, a comparative study between Varian and in-house reconstructed images was performed using Planning CT (PCT) images as a reference dataset. The feasibility of using the Varian and in-house reconstructed images for treatment planning was investigated by acquiring CBCT images of the Rando anthropomorphic phantom. An Intensity-Modulated Radiotherapy (IMRT) treatment plan was generated using both sets of reconstructed images using the Pinnacle3 treatment planning system. Planar dose distributions were extracted from both the datasets in order to evaluate dose distributions quantitatively based on 3%/3mm Gamma analysis criteria. These distributions were then compared against the reference PCT image and it was found that in-house reconstructed images showed good agreement with the PCT images with a gamma passing rate of 99.8%. Although several pre-processing steps performed on the Varian images were not included during in-house reconstruction, the results demonstrated the potential for use of in-house reconstructed CBCT image for treatment planning. As an alternative to FDK reconstruction, iterative reconstruction using Maximum Likelihood solutions was also investigated. Since the Ordered Subsets Expectation Maximisation (OSEM) package used in this study is intended for fan-beam geometry, only the slices from the central plane of cone-beam were chosen. The projections were corrected for distance-dependent resolution and centre of rotation offset. When the number of iterations was increased to 16, the algorithm converges well and showed more uniform images. However, the images were not comparable to FDK-based images due to the intrinsic difference in data handling. The OSEM program was developed initially for emission-based measurements and did not model the scatter component effectively for transmission-based measurements. Including the scatter component more effectively may make it more realistic for CBCT geometry.