Development of Real-Time Motion-Including Dose Reconstruction on a Unity MR-Linac system

Abstract

The first part of my thesis is concerned with estimating the motion of organs and other critical anatomical structures during radiotherapy treatments. I use 2D cine images acquired during MR-Linac treatments and compare different computer vision techniques to obtain motion vectors from these.Then I calculate the radiation dose received by the patient in “real time” (as they are being treated). I use a clinically validated fast Monte Carlo dose calculation algorithm, and I iteratively compute beam doses during radiation delivery and accumulate the dose cubes on a 3D representation of the patient. In each iteration I include the current state of the patient anatomy determined in the first part. Thus I am able to compute the radiation received by moving anatomy whilst the treatment is underway. Finally, I investigate the clinical relevance of this technique. There are two avenues that I explore: the first is to facilitate the use of real-time intelligent correction measures, such as gating and tumour tracking. The second is to create prediction models to determine whether the radiation dose delivered to the target structure and organs at risk will be suboptimal due to the presence of motion. If suboptimal dosimetry is detected, it could be corrected for by various strategies, thus improving patient outcomes.