In-treatment cone-beam CT (CBCT) imaging is widely used in radiotherapy (RT) to provide 3D anatomical data for patient positioning and tumour targeting. However, because of X-ray scatter, non-ideal detectors and large regions of interest, CBCT cannot be used for essential re-calculation of delivered X-ray or proton dose when anatomical changes are observed following pre-treatment planning and during daily delivery. The proposers have previously developed and patented a software post-processing method to restore accurate tissue attenuation values in CBCT images by generation of an adaptive filter based on prior imaging. The method has proved successful for pelvic and head and neck images, which show marked improvements in visual image quality and numerical integrity for photon dose calculation. However this has not yet been achieved for thorax images, where it is particularly important to correctly account for differences in tissue attenuation due to the presence of highly variable tissue densities and of course physiological motion, primarily respiratory effects. First we will consolidate these advances, specifically by developing streamlined software tools that make the method suitable for implementation in the clinical environment and subsequent commercialization. Then we will perform Monte Carlo modelling of patient and imaging processes to advance our methodology, particularly for challenging lung imaging. Finally, we will pilot the use of corrected in-room images to address the pressing but unmet need for in-treatment proton therapy monitoring and re-planning. The resulting tools for correction of CBCT images will significantly reduce the time required to investigate the clinical impact of anatomical changes observed with CBCT (currently consuming a large amount of RT physics staff resource), and will pave the way for CBCT based adaptive photon and proton RT, where treatments are individually optimized mid-course based on imaging acquired on-treatment.