Development and application of efficient portal imaging solutions

Abstract

This thesis describes the theoretical derivation and clinical application of methods to measure and improve patient setup in radiotherapy by means of electronic portal imaging devices (EPIDs). The focus is on methods that (1) are simple to implement and (2) add minimal workload. First, the relation between setup errors and treatment planning margins is quantified in a population‐statistics approach. A major result is that systematic errors (recurring each treatment fraction) require about three times larger margins than random errors (fluctuating from fraction to fraction). Therefore, the emphasis is on reduction of systematic setup errors using off‐line correction protocols. The new no action level (NAL) protocol, aimed at significant reduction of systematic errors using a small number of imaged fractions, is proposed and investigated in detail. It is demonstrated that the NAL protocol provides final distributions of residue systematic errors at least as good as the most widely applied comparable protocol, the shrinking action level (SAL) protocol, but uses only 3 imaged fractions per patient instead of the 8–10 required by SAL. The efficacy of NAL is demonstrated retrospectively on a database of measured setup errors involving 600 patients with weekly set‐up measurements and prospectively in a group of 30 patients. The general properties of NAL are investigated using both analytical and Monte Carlo calculations. As an add‐on to NAL, a correction verification (COVER) protocol has been developed using computer simulations combined with a risk analysis. With COVER, a single additional imaged fraction per patient is sufficient to reduce the detrimental effect of possible systematic mistakes in the execution of setup corrections to negligible levels. The high accuracy achieved with off‐line setup corrections (yielding SDs of systematic errors is demonstrated in clinical studies involving 60 lung cancer patients and 31 head‐and‐neck patients. Furthermore, potential problems arising from nonrigid body anatomy are quantified in large cervix treatment fields. Finally, the suitability of our in‐house developed (and currently commercially available) CCD‐camera‐based EPID for portal dosimetry is described. An algorithm is developed to transform EPID images into accurate dose images (SD of dose based on fundamental properties of the EPID.