Processing of NMR Slices for Preparation of Multi-dimensional Model

Abstract

The paper describes the pre-processing and subsequent segmentation of NMR images of the human head in the region of temporomandibular joint in several slices. Image obtained by means of the tomograph used are of very low resolution and contrast, and their processing may prove to be difficult. A suitable algorithm was found, which consists in pre-processing the image by a smoothing filter, sharpening, and the four-phase level set segmentation. This method segments the image on the basis of the intensity of regions and is thus suitable for processing the above-mentioned NMR images, in which there are no sharp edges. The method also has some filtering capability. It is described by partial differential equations that have been transformed into corresponding difference equations, which are solved numerically. In the next stage, segmented slices of temporomandibular joint will be used to create a multidimensional model. A spatial model gives a better idea of the situation, structure and properties of tissues in the scanned part of patient’s body. Multidimensional modeling mainly represents a change in describing input NMR data. The segmented regions need to be transformed from discrete to vector description i. e. the outer geometry of objects needs to be described mathematically.