Implementation,Validation and Application of an Automated Motion Correction Algorithm in Hepatic Perfusion CT for Image-Guided Radiotherapy

Functional imaging holds promise in guiding, evaluating and predicting efficacy of cancer treatment. Perfusion imaging provides high resolution functional information generated by applying a model to a dynamic series of contrast enhanced anatomical images. The objectives of this thesis are: 1) to develop and validate a registration algorithm to reduce breathing motion artifacts in hepatic perfusion CT, and 2) to evaluate the potential of perfusion CT to improve radiotherapy of liver cancer. An automated one-dimensional correlation-based registration algorithm was developed and validated in phantom and pig studies and against manual registration of patient scans. This algorithm was used to register dynamic contrast enhanced CT (DCE-CT) image series, which consisted of 40 CT volumes acquired sequentially over 2 minutes. Perfusion maps were generated from unregistered and registered DCE-CT scans and compared to determine impact of motion correction on tumour-normal tissue contrast. The motion corrected perfusion maps were used in a radiotherapy target volume delineation study. Three observers delineated tumours on perfusion images and standard contrast enhanced 4D-CT scans. Interobserver concordance was used as surrogate for accuracy. Knowledge of contrast kinetics from DCE-CT scans was also applied to model contrast enhancement in radiotherapy alignment cone-beam CT. Contrast enhanced cone-beam CT scans were acquired of two rabbits to determine improvements in target alignment. Reproducibility of registration and spatial accuracy was less than the slice thickness in phantom and patient experiments. Motion correction of pig scans showed increased accuracy in 5 of 6 cases. In patient scans, arterial and total blood flow were significantly elevated, and blood volume was significantly reduced in tumours compared to normal tissue. In uncorrected maps, differences between tumour and normal tissue blood flow were significantly reduced. Target delineation interobserver variability was significantly reduced with perfusion maps compared to standard clinical 4D-CT. Tumour alignment was improved by increased tumour-normal tissue contrast in contrast-enhanced cone-beam CT. In conclusion, it is possible to obtain useful hepatic perfusion maps from DCE-CT scans of free breathing patients provided motion correction is applied. DCE-CT also improves targeting and patient alignment for radiotherapy and could be a valuable addition to current clinical imaging