Quality assessment and enhancement for cone-beam computed tomography in dental imaging

Abstract

Cone-beam CT will become increasingly important in diagnostic imaging modality in the dental practice over the next decade. For dental diagnostic imaging, cone-beam computed tomography (CBCT) system based on large area flat panel imager has been designed and developed for three-dimensional volumetric image. The new CBCT system can provide a 3-D volumetric image during only one circular scanning with relatively short times (20-30 seconds) and requires less radiation dose than that of conventional CT. To reconstruct volumetric image from 2-D projection images, FDK algorithm was employed. The prototype of our CBCT system gives the promising results that can be efficiently diagnosed. This dissertation deals with assessment, enhancement, and optimization for dental cone-beam computed tomography with high performance. A new blur estimation method was proposed, namely model based estimation algorithm. Based on the empirical model of the PSF, an image restoration is applied to radiological images. The accuracy of the PSF estimation under Poisson noise and readout electronic noise is significantly better for the R-L estimator than the Wiener estimator. In the image restoration experiment, the result showed much better improvement in the low and middle range of spatial frequency. Our proposed algorithm is more simple and effective method to determine 2-D PSF of the x-ray imaging system than traditional methods. Image based scatter correction scheme to reduce the scatter effects was proposed. This algorithm corrects scatter on projection images based on convolution, scatter fraction, and angular interpolation. The scatter signal was estimated by convolving a projection image with scatter point spread function (SPSF) followed by multiplication with scatter fraction. Scatter fraction was estimated using collimator which is similar to SPECS method. This method does not require extra x-ray dose and any additional phantom. Maximum estimated error for interpolation was le...