Processing for Medical Imaging using the OsiriX DICOM Viewer

Medical imaging is a fertile area for computer graphics, image processing and real time visualization. In this paper we present a method for reducing noise in CT (Computed Tomography) and MR (Magnetic Resonance) images that (in addition to other noise sources) is characteristic of the numerical procedures required to construct the images, namely, the (inverse) Radon Transform. In both cases, MR imaging in particular, an additional noise source is due to the process of diusion thereby predicating use of the Anisotropic Diusion method for noise suppression. This method is based on a diusion model for noise generation where the Diusivity is taken to be non-isotropic (inhomogeneous) or anisotropic and is, in the absence of a priori information, computed through application of an edge detection algorithm. In this paper we extend the approach to include the eect of fractional diusion (when the underlying statistical model associated with the diusion process is non-Gaussian) and derive a corresponding Finite Impulse Response lter. The algorithms developed are implemented using the OsiriX DICOM (Digital Imaging and Communications in Medicine), a high performance open source image data visualization system for the development of processing and visualization tools.