Optimization of Imaging Performance and Conspicuity in Dual-Energy X-ray Radiography

Dual-energy (DE) x-ray imaging of the chest decomposes two radiographs acquired at low- and high x-ray energies into 'soft-tissue' and 'bone' images, reducing the influence of background anatomical noise and providing increased conspicuity of subtle underlying structures compared to conventional radiography. This thesis derives a quantitative theoretical model of imaging performance in DE x-ray imaging and employs the resulting framework to system optimization in thoracic imaging. Fourier domain metrics of signal and noise performance - including the noise-power spectrum (NPS), modulation transfer function (MTF), detective quantum efficiency (DQE), and noise-equivalent quanta (NEQ) - were computed using cascaded systems analysis extended to DE imaging and combined with a quantitative model of imaging task to yield estimates of detectability across a broad range of DE image acquisition and decomposition techniques. Specifically, the detectability index provided an objective function for optimizing the selection of kVp pair, added filtration, allocation of dose between low- and high- energy views, and choice of decomposition algorithm and parameters therein. Theoretical calculations were validated in comparison to measurements of NPS, MTF, DQE, and NEQ performed on an experimental DE imaging system and through human observer studies for a variety of imaging tasks. Overall, the detectability index was found to provide a reliable predictor of human observer performance. Results identified optimal DE image acquisition and decomposition techniques that boost detectability beyond that achieved by conventional radiography or other DE imaging approaches, in many cases boosting conspicuity of subtle lesions from barely visible to highly conspicuous at fixed dose to the patient. The results are particularly encouraging, as such performance was achieved with the DE imaging dose equivalent to that of a single chest radiograph. The theoretical framework provided a valuable guide to optimization of a clinical prototype for high-performance DE chest imaging and may be extended to other DE imaging approaches, such as DE mammography and DE computed tomography.Ph