The complexity of radiation therapy delivery has increased over the years due to advancements in computing and technical innovation. A system of dose delivery verification has the potential to catch treatment errors and therefore improve patient safety. The goal of this thesis was to create a portal image-based in vivo dose reconstruction model for volumetric modulated arc therapy (VMAT) deliveries, specifically for stereotactic body radiation therapy (SBRT). This model-based approach should be robust and feasible within a clinical setting. VMAT involves the modulation of dose rate, gantry speed, and aperture shaping while the treatment gantry (i.e., x-ray beam) rotates about the patient. In this work, portal images were acquired using an amorphous silicon electronic portal imaging device (a-Si EPID). A geometrical characterization of the linear accelerator (linac) during VMAT delivery was performed. An angle adjustment method was determined which improves each EPID’s angular accuracy to within ±1° of the true physical angle. SBRT delivers large doses over fewer fractions than conventional radiotherapy, therefore, any error during an SBRT delivery will have a greater impact on the patient. In this work, a robust, model-based SBRT-VMAT dose reconstruction verification system using EPID images was developed. The model was determined to be clinically feasible. The accuracy of a 3D in vivo dose reconstruction, using all the EPID images acquired during treatment, is sensitive to the chosen frame averaging per EPID image: the greater the frame averaging, the larger the reconstruction error. Optimization of the EPID frame averaging number as a function of average linac gantry speed and dose per fraction were determined. The EPID-based in vivo dose reconstruction model for SBRT-VMAT developed here was determined to be robust, accurate, and clinically feasible as long as adjustments were made in order to correct for EPID image geometrical errors and frame-averaging errors. / May 2016