Study On Cross-sectional Interpolation Technology Based On Low-dose CT Images

Cross-sectional interpolation is one of the key technologies in3D reconstruction. It can reduce the distance of slice images by increasing the interpolation image and improve the resolution of volume data. As a result, the quality of3D reconstruction images can be improved greatly. In recent years, the widespread use of multi-detector CT results in a marked increase in radiation exposure by the CT-scanners. Studies have shown that the CT examination with high-dose X-ray increases the risk of cancers, leukemia or other genetic diseases. It is desirable to lower the clinical dosages to eliminate the potential danger of radiation. However, Low-dose CT scanning results in increased quantum noise and reduced spatial resolution of volume data. The quality of3D reconstruction images will be affected by noise and terrace effect seriously.The traditional gray-based cross-sectional interpolation, shape-based cross-sectional interpolation and wavelet-based cross-sectional interpolation algorithm were first introduced in this paper. Base on the comparative study of traditional interpolation method, we proposed two new cross-sectional interpolation algorithms for Low-dose CT:a wavelet-based matching interpolation method and a curvelet-based bilinear interpolation method.The wavelet-based matching interpolation method is combining wavelet transform and matching interpolation. Firstly, the slice images are decomposed using Wavelet Transform. Secondly, we first process the detail sub-images using the Donoho’s threshold method and then obtain the interpolated detail sub-images by linear interpolation. Thirdly, the accurate interpolated low-frequency sub-image is obtained by the matching interpolation. Finally, the interpolated image is reconstructed base on the interpolated sub-images by the inverse Wavelet Transform. This method not only can improve the accuracy of interpolation, but also can effectively suppress the noise.Because curvelet transform is a multi-scale geometrical transform and it can sparsely characterize the high-dimensional signals. This paper proposed a curvelet-based bilinear interpolation method. In order to suppress the Gibbs phenomenon produced by the curvelet transform, we first process slice images using circular shift. Then process the coarse scale layer coefficients after curvelet decomposition using the modified bilinear interpolation. The fine scale and detail scale layer coefficients are processed using the adaptive threshold method and then obtain the interpolated fine scale and detail scale layer coefficients by linear interpolation. Finally, the interpolated image is reconstructed base on the interpolated scale layer coefficients by the inverse curvelet transform and inverse circular shift. The experimental results show that this method has better noise immunity than the wavelet-based interpolation method and it expend less time.In order to validate the efficiency of the proposed algorithms, two new interpolation methods are applied to3D reconstruction. The original Low-dose CT images and interpolation images constitute a new sequence of images. We respectively use the Marching Cubes algorithm of surface rendering and Ray Casting algorithm of volume rendering for3D reconstruction to the new sequence images and comparative with the interpolation algorithms in the literature. The comparative experiments of3D-reconstruction show that the two proposed interpolation methods can effectively reduce the noise and eliminate the terracing effect in three dimensional reconstruction images.In the end of the article, a summary of my work are given and the future work are discussed.