Investigating the potential of routine image-guided radiotherapy data to improve clinical outcomes in lung cancer

Abstract

Image-guidance is now commonly utilised for patient positioning and as such, there are vast archives of Image-Guided Radiotherapy (IGRT) images, with data captured at many time points throughout patients’ treatment schedules. In this thesis image-guidance data has been incorporated into data-mining techniques to account for changes that occur over the course of treatment. That is, the aim of this thesis was to study how clinical outcomes of lung cancer radiotherapy patients can be improved based upon learning from large cohorts of routinely treated patients including their IGRT data. First, a method to synthesise 3D-planned gross tumour volumes (GTV) from 4D-planned internal target volumes (ITV) was developed to allow older lung cancer datasets to be utilized. Estimates of the GTV volume were successfully obtained, and models built using a combination of the true and synthetic data performed significantly better than all others. The potential value of utilising IGRT data was then demonstrated through a study on the effect of residual setup errors following image-guidance on survival. This study highlighted the significance of small setup errors which move the heart closer or further away from the high dose region, and thus suggest the importance of strict imaging protocols and heart dose constraints. This work was validated in an oesophageal cancer cohort. Building on this analysis, the effect of residual setup errors on the delivered dose, and the difference from planned dose, was assessed to investigate the cause behind the observed survival difference. Image-based data-mining revealed a significant region in the base of the heart associated with survival. This result provides further evidence that the base of the heart is a radiation-sensitive organ at risk. Finally, the effect of residual setup errors was analysed in two additional lung cohorts. The first looked at a cohort treated with a stricter imaging protocol, and it was found that the survival effect appeared to be removed, showing the power of routine data to update and evaluate clinical practice iteratively. The second was a cohort treated with a hypofractionated regime, and baseline shifts of the tumour were studied. Once again, a significant survival effect was observed in patients with residual setup errors in the direction of the heart, suggesting not only that the dose to the heart should be considered in the treatment plan, but that a margin to account for the effect of baseline shifts is also required. This finding is very exciting because baseline shifts are unavoidable. This means that any change in practice could be tested in routine clinical data for lung cancer, using IGRT data. Our results show, for the first time, the direct clinical benefit of IGRT for lung cancer patients whilst adding to the weight of evidence showing that cardiac toxicity in lung cancer patients is more important than previously thought, in both advanced stage and inoperable early stage disease settings.

Date of Award	31 Dec 2018
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Gareth Price (Supervisor) & Marcel Van Herk (Supervisor)