Design and Development of Computed Tomography Image Synthesiser using Supervised Deep Generative Models

Abstract

Radiotherapy is a medical treatment that uses radiation to destroy cancer cells in newlinethe human body. Image-Guided Radiation Therapy (IGRT) is a type of radiotherapy newlinethat utilises medical imaging techniques, such as Computed Tomography (CT) and newlineMagnetic Resonance Imaging (MRI), to deliver precise doses of radiation to target newlinecancer cells. MRI is particularly suitable in IGRT planning as it has good soft-tissue newlinecontrast and can accurately outline the planning target volume and organs-at-risk. newlineFan Beam Computed Tomography (FBCT) scans are often used in IGRT to obtain newlineelectron density information for radiation dose calculation. The use of real-time newlineMRI-guided Radiation Therapy (MRIgRT ) with an MR-LINAC (Linear Accelerator) newlineis a recent advancement in cancer treatment. Still, it requires a synthesiser to generate newlineFBCT data from MRI images obtained through the MR-LINAC. In the IGRT process, newlinea planning phase using FBCT is usually followed by a radiation delivery phase newlineguided by Cone Beam Computed Tomography (CBCT) using CBCT-LINAC. The newlineIGRT treatment is often fractionated, taking several weeks and consisting of several newlinetreatment fractions. Low-dose CBCT images are used for intra-fractional imaging for newlineprecise beam positioning, but they are unsuitable for dose calculations. Therefore, if newlinethere is tumour shrinkage after a fraction, it is necessary to retake FBCT for treatment newlinereplanning. In these cases, medical image synthesis, such as synthesising FBCT newlineimages from MRI images and synthesising FBCT images from CBCT images, can newlinebe helpful to avoid repeated FBCT cycles. This research aims to develop synthesisers newlineusing supervised deep generative models to create high-quality FBCT images from newlineMRI and CBCT images. newline