The use of a radiotherapy portal imaging device for patient setup verification through cone beam reconstruction

English: In this study the feasibility of patient setup verification through cone-beam computed tomography (CBCT) using an Elekta Precise linear accelerator (linac) and electronic portal imaging device (EPID) was investigated. CBCT images are used for treatment verification i.e. determination of patient set-up errors prior to treatment delivery. An 8 MV photon beam was used to acquire planar images of an anthropomorphic head phantom as the system rotates in a 200° arc around the phantom. The Feldkamp-type algorithm which, through a weighting function, is an approximation of the filtered backprojection algorithm was used to perform reconstruction of 2D transversal CT images from the projection/planar images. Reconstruction was done using tools developed with the Interactive Data Language (IDL) software package. The reconstruction technique was evaluated in terms of image quality, dose imparted during image acquisition and positional accuracy. Geometric calibration of the imaging system (linac and EPID) was performed to derive a set of parameters that fully describes the geometry of the system. These parameters include piercing point (projection of isocenter on EPID), detector rotation around its normal axis, detector tilt angles and gantry angle variation. A dedicated calibration phantom was manufactured and used to determine and correct for the above-mentioned parameters during the reconstruction process. These corrections ensure accurate reconstruction and avoidance of image artefacts due to geometrical misalignments of the system. Image quality was evaluated using a standard image quality phantom (Catphan®). Parameters such as signal-to-noise ratio (SNR), spatial resolution, contrast resolution and uniformity were used to quantify image quality. Due to the high energy of the photon beam, reconstructed images yield relatively poor image quality compared to kilovoltage photons. Although soft tissue contrast were very poor, image quality was sufficient for visualization of bony landmarks, air cavities or fiducial markers. Image quality can be enhanced by increasing the number of monitor units (MUs) used but, this will lead to an increase in dose which is not desirable from a clinical point of view. Dose imparted during image acquisition was measured using an ionization chamber placed at the treatment isocenter. An overall shape of the dose distribution was obtained by film measurements and a simulated plan computed by a treatment planning system (TPS). The dose measured at the centre of the phantom was ±160 cGy which is clinically unacceptable. There exist several dose reduction techniques; in this study a reduction in number of projection images were used. The dose was reduced to ± 30 cGy when 40 projection images instead of 200 were used. Due a trade-off between dose and acceptable image quality the dose could not be reduced any further without degrading the image quality beyond clinically acceptable levels. Positional accuracy was evaluated by simulation of phantom shifts and rotation. The reconstructed images were then used to determine these shifts and rotation. Shifts and rotation could be determined within 2 mm and one degree respectively for images reconstructed with 40 projection images. Larger angular increments yield poor image quality and severe reconstruction artefacts which resulted in false positive values (absolute difference from simulated shift and rotation smaller than 2 mm and one degree). Intra- and inter observer dependency were evaluated during positional accuracy determination. Results showed that the method for determination of position and rotation is observer independent. The clinical feasibility of this method is limited by image acquisition time and dose imparted during image acquisition. Further investigations (which are beyond the scope of this study) should be conducted to explore means of reduction in image acquisition time and dose i.e. improvement of EPID detection efficiency and linac output (fractional MUs).