Optimizing image quality of x-ray fluoroscopy images obtained with flat panel detectors

X-ray fluoroscopy is x-ray imaging at high frame rate (typically 30 frames/s) with low x-ray exposure per frame. Fluoroscopy is used in many image-guided diagnostic and interventional procedures. Some demanding procedures can require long x-ray fluoroscopy times leading to large cumulative patient x-ray dose. Since x-ray images are quantum limited, simply reducing dose leads to unacceptable degradation in image quality. The goal of this research is to optimize fluoroscopy image acquisition and processing so as to lower exposure while maintaining, or improving image quality.; An important component of the fluoroscopy imaging chain is the clinician viewing the images, and any image quality evaluation should include the human observer. Therefore, in these studies processing and acquisition techniques were evaluated using objective, quantitative, visual task-based measures of image quality.; Digital filtering of x-ray fluoroscopy sequences can be used to improve the quality of images acquired at low x-ray exposure. Two filters were evaluated: a spatio-temporal, non-linear filter and a temporal filter. Filtering significantly improved detection and discrimination of targets, and clinical backgrounds affected filter performance when quantum noise was significantly reduced.; Using a physics-based model of x-ray detectors, image magnification and pixel-binning in flat-panel (FP) detectors were optimized. First, image intensifier (II) analog magnification was compared with FP digital magnification in a clinically relevant task of detecting a partially deployed stent. Results show that FP digital magnification can be useful in improving detection and is dose efficient when compared to IT analog magnification.; Second, pixel-binning was examined in FP detectors. Conventional data-line (D) and gate-line (G) binning was evaluated along with a third novel method in which alternate frames in an image sequence used D and G-binning. The results suggest an exposure-dependent detector binning that switches between D-binning and alternate binning at low and high exposures, respectively.; Computational observer models were developed that incorporate features of human visual processing. Models were mostly consistent with observer data and were used to predict similar processing. A combination of such modeling and observer experiments can be used as a paradigm to develop optimal image processing and acquisition in fluoroscopy.