Optimization of digital x-ray fluoroscopy using quantitative image quality techniques

X-ray fluoroscopy is used in a multitude of diagnostic and interventional procedures. Although the x-ray dose per acquisition is low, images are acquired at high frame rates over extended periods of time leading to significant x-ray dose to patients and staff. The adverse effects of over-exposure are well documented and there is a significant public health interest in minimizing x-ray dose. Since x-ray fluoroscopy is quantum limited, simply reducing x-ray exposure reduces the image signal-to noise ratio and leads to unacceptable image quality.; The physical factors that affect object contrast and image noise in x-ray fluoroscopic imaging is well understood, while it is less so with the perceptual characteristics. The goal of this research is to use new insights from perception studies to quantitatively optimize the acquisition and visualization of interventional devices in fluoroscopic images in order to reduce x-ray dose while maintaining, or improving image quality.; We examined observer perceptual characteristics like the role of a priori spatio-temporal phase information and ocular motor smooth pursuit on detection of moving objects and, the effects of observer uncertainties on detection. We found that oculo-motor smooth pursuit enhances the detection of thin, low contrast moving objects in noise while it degrades the detection of thick moving objects. Observer inherent uncertainty was identified as a significant reason for the sub-optimal detection of targets by human observers.; We created physics-based models of x-ray detectors and determined the effects of pixel size for image acquisition and visualization. We found that for low exposure fluoroscopy, there is a trade off between spatial resolution and image noise. For the detection of guide-wires that are commonly used in cardiac applications, a pixel size of ≈200 mum was found to be optimal. Smaller pixel sizes were found to be better suited for high exposure applications.; We developed human observer models incorporating the Human Visual System (HVS) characteristics like spatio-temporal contrast sensitivity and spatio-temporal channels. The observer model predicted many of the experimental results and will contribute to the optimization of image acquisition parameters and digital processing methods for fluoroscopy.