Improved reslicing by voxel‐project‐method

Abstract

AbstractThe projection processing has frequently been applied in recent years, where the three‐dimensional image composed of a large number slices, such as X‐ray CT and MRI, is used as the input. In such processings, the cross section is produced with arbitrary position and inclination, or the three‐dimensional image is reprojected on an arbitrary plane. In the generation of the cross section especially, the highspeed processing is required as a pre‐processing for the image understanding. The voxel‐search method and the voxel‐projection method have been employed as the methods for projection. The voxel‐search method has been used widely because of the better image quality, but has a problem in that the speed improvement by parallel structure is difficult. The voxel‐projection method, on the other hand, can easily be implemented on a parallel system, but the deterioration of the image quality has been a problem.This paper aims at the method to generate cross section which can realize the same image quality as in the conventional voxel‐search method while maintaining the parallelism of the voxel‐projection method. The image deterioration, such as the missing of the image data and the striped pattern of intensities on the image, are remedied. The scanning place determination is proposed for the former problem, which can determine the object of projection according to the inclination of the cross section. The average distribution technique is proposed for the latter problem, and it is shown that the cross section with a satisfactory image quality can be generated. Thus, it is shown that the voxel‐projection method can be applied to the cross section generation, paving the way to the realization of the cross section generation system with the speed improved by the parallel processing.