Comparison of two- and three-dimensional filtering methods to improve image quality in multiplanar reconstruction of cone-beam computed tomography

Two- and three-dimensional (2D and 3D, respectively) filtering methods were examined to improve the accuracy of bone morphology depicted in dental conebeam computed tomography (CBCT) images. An attempt to improve multiplanar reconstruction (MPR) image quality was carried out by reducing the image noise. CBCT examinations were performed with the following principal exposure parameters: I-mode, field of view (FOV) 10 cm in diameter, 120 kV, 15 mA, 0.2 mm slice thickness, and exposure time of 10 s. 2D and 3D filtering procedures for averaging, median smoothing, and Gaussian smoothing were applied for improvement of MPR images. For comparison, 2D and 3D Laplacian sharpening for images preprocessed by Gaussian sharpening was also tested. MPR images at the midsagittal plane on the maxilla are presented. Three smoothing filters yielded improvements in slightly different ways. The Gaussian filter clearly showed moderate changes. Small but obvious differences were observed between 2D and 3D filtering. When we focused on the depiction of bone contours, the effects of these noise reduction filters seemed to be minimal in morphological bone depiction. The Laplacian filter was useful for sharpening and emphasized noise in the resulting images, in contrast to those processed by smoothing filters. Various smoothing filtering methods reduced the noise on MPR images of CBCT and also functioned differently between 2D and 3D filtering matrices. (author)