Key words: radiotherapy, dose-guided radiotherapy, treatment verification, artificial intelligence, error detection

"Quantitative methods for improved error detection in doseguided radiotherapy"

With the increasing complexity of radiotherapy treatments, it becomes increasingly important to verify that the desired radiation dose is delivered as planned (high dose in tumours, as little as possible in healthy tissues). This dissertation focuses on dose-guided radiotherapy using dose measurements with electronic X-ray cameras (EPID dosimetry) to identify treatments in which errors occur, so that these treatments can be adjusted. This dissertation shows that automatic error detection with EPID dosimetry can be significantly improved. It contributes to this improvement by providing a framework for analysing the uncertainties of dose measurements by quantifying the performance of simple error classification methods, and by applying advanced artificial intelligence algorithms for error classification. These results will ultimately lead to improved radiotherapy treatments.