Real-time in vivo dosimetry and error detection during afterloading brachytherapy

Image guided afterloaded brachytherapy (BT) allows for conformal and patient specific radiotherapy (RT) treatments against cancer, where high dose concentrations are administered to the tumor volume and small doses to organs at risk (OARs). In afterloaded BT, ionizing radiation is delivered by means of a radionuclide attached to a source chain that is placed inside source catheters implanted in the target region. As for any RT treatment modality, BT treatments are subject to discrepancies between the delivered and planned treatments. Given the localized and high dose concentration near BT sources, even small discrepancies of the planned source position may result in largely modified dose distributions that could lead to an insufficient dose to the tumor and/or increased doses to OARs. One way to monitor the integrity of a BT treatment delivery and to detect potential treatment errors, is to perform real-time in vivo dosimetry (IVD) inside the target region during the treatment. That way, an independent and patient specific verification of the agreement between delivered and planned treatments can be performed. If a treatment error is detected, modifications of the treatment parameters or even a treatment termination could be warranted. However, widespread implementation of IVD for BT is hampered by important limitations of current IVD systems. Further developments are thus needed such that IVD can offer improved safety for more patients that are treated with afterloaded BT. The work presented in this thesis involves the development of an IVD implementation which targets current important problems for IVD. Initially, a dosimetry protocol appropriate for BT irradiation conditions was developed. Subsequently, the dosimetry protocol was exposed to various simulated clinically relevant error scenarios, in order to quantify the error detection sensitivity of the real-time point dosimetry system used by means of a statistical error detection concept that incorporated a full uncertainty analysis. The limiting effects of the dependence on the a priori reconstruction of the dosimeter position were targeted by the development of an adaptive error detection algorithm (AEDA) which does not rely on an a priori reconstruction. The performance of the AEDA was validated against a static error detection algorithm which used a static a priori reconstruction in order to judge the treatment. The contribu- tions mentioned are described in three publications or manuscripts which have been published in or submitted to international peer-review journals. Tools and methods were developed in order to facilitate IVD on a routine basis for image guided pulsed dose rate (PDR) BT of locally advanced cervical cancer at the Aarhus University Hospital. The tools and methods developed for the implementation targeted requirements for accurate IVD and the demands for a time-efficient and straightforward clinical approach. The performance of all developments was explored based on IVD results for 20 PDR BT treatments.