Three-Dimensional Image Quality Evaluation and Improvement in Flat-Panel Detector Based Cone-Beam CT Imaging

Yan Zhang

Professor Ruola Ning

Goergen 109

Abstract

Due to the development of flat-panel detector (FPD) in recent years, FPD-based cone beam computed tomography (CBCT) made a breakthrough in CT imaging in terms of large volume reconstruction and isotropic resolution. This technique provides three-dimensional (3D) visualization of the imaged anatomy with one single circular scan, and slices reformatted at any arbitrary orientation with high spatial resolution. Due to all these advantages, FDP-based CBCT has become the most promising imaging modality in many medical fields including small animal imaging, breast imaging, bone imaging, etc.

As a potential diagnostic imaging tool, FDP-based CBCT faces challenges to consistently provide high-quality images for doctors to make right diagnosis. Therefore, it's important to fully evaluate its imaging performance and to correct the image artifacts caused by various sources to ensure the image quality.

This work focuses on two primary issues faced by FPD-based CBCT: 3D image quality evaluation and the image quality improvement with artifacts correction. It includes five parts. (1) New image quality evaluation methods are proposed to evaluate major image quality factors of CBCT based on its 3D volume reconstruction characteristic. (2) A noise model is developed to study the noise level in 3D reconstructed image and its quantitative relationships with various parameters of CBCT system. (3) The relationship of image lag in CBCT with different detector parameters and x-ray techniques is investigated as well as the quantitative effect of image lag on the image quality in breast imaging. (4) A correction method is developed to correct the streaking and shading artifacts caused by metal implant in FDP-based CBCT bone imaging. (5) A simple artifact correction method is proposed to correct the dark strip at the chest-wall in breast imaging due to incomplete projections.