The goal of this proposal is to develop and evaluate new CT image reconstruction algorithms for the purpose of detecting and characterizing malignant tumors in the lung. New, accurate and efficient algorithms are needed to take advantage of the large cone-angle acceptance of novel detectors that are expected to be employed in future medical CT systems. The developed and existing algorithms will be tested against physical factors such as partial volume averaging, polychromatic x-ray sources, signal noise and x-ray scatter. These physical factors are present in real CT systems and are known to degrade image quality. Algorithms for taking advantage of possible, x-ray energy resolving capability of future detectors will also be explored. Such a capability may alleviate beam-hardening artifacts, and provide additional diagnostic information. One clinical area that the new helical, cone-beam CT technology might be applied to is lung cancer screening. The rapid volume scanning will reduce motion artifacts, and detection of x-rays in different energy windows may assist radiologists in distinguishing malignant and benign nodules in the lung. The developed and existing image reconstruction algorithms will be assessed initially by performance of specific detection and estimation tasks on simple computer-generated phantoms. In order to provide a better means for assessing image reconstruction algorithms for the purpose of lung cancer screening, computer simulated, anthropomorphic chest phantoms, modeling the physical properties of normal and pathological structure of the lung, will be developed. Such phantoms will provide a unique opportunity to assess image reconstruction algorithms, because the data sets will have similar features to actual patient data, the number of generated data sets can be arbitrarily large, and the pathology of each data set is known. Computer-aided diagnosis programs for lung nodule detection may be employed as an observer to assess various image reconstruction and processing algorithms.
