Study On Volume Rendering Algorithm Of Low-dose CT Volumedata Based On GPU

Diagnosis using 3D reconstructed images is an important method for reducing CT repeat examinations and improving precision of clinical diagnosis. In recent years, patients were exposed to large X-ray radiation by CT scanning, and the radiation is harmful to human health. Besides, the traditional 3D reconstruction algorithms cost very long time for rendering. In order to obtain the best diagnostic results with minimum radiation dose and shortest reconstruction time, this paper studies on volume rendering algorithm of low-dose CT(LDCT) volume data based on GPU.LDCT scanning is a potential way to reduce X-ray radiation dose. However, it will produce strong quantum noise which can affect the CT image quality and diagnostic accuracy seriously. Besides, traditional volume rendering algorithms are almost impossible to achieve real-time rendering because of high calculation complex and huge reconsruction data. The quality of the reconstructed image and the speed of 3D reconstruction algorithms are the main bottlenecks of limiting LDCT reconstruction technique is widely used in clinical application. Therefore, This paper will focus on the high quality preprocessing algorithm and high speed 3D reconstruction algorithm of LDCT volume data. The main work of the paper is as follows:The preprocessing algorithm is the foundation of 3D image quality and affects the diagnosis accuracy directly. LDCT images have strong quantum noise which can affect the image quality and diagnostic accuracy seriously. In this paper, we propose an effective preprocessing algorithm for improving LDCT image quality using shearlet transform before 3D reconstruction. Because the quantum noise can be simulated by Poisson process, we first transform the quantum noise by using anscombe variance stabilizing transform(VST), producing an approximately Gaussian noise with unitary variance. Second, the non-noise shearlet coefficients are obtained by adaptive hard-threshold processing in shearlet domain. Third, we reconstruct the de-noised image using the inverse shearlet transform. Finally, an anscombe inverse transform is applied to the de-noised image, which can produce the improved image. Both quantitative and visual results show that the proposed method can effectively reduce the quantum noise while enhancing the subtle details. It provides a good basis for reconstructing high quality 3D image.Reconstruction speed determines whether the 3D image can be real-time interactive and display directly. Because of the large amount of calculation and the slow calculating speed, traditional ray casting method is almost impossible to achieve real-time rendering without hardware acceleration. We propose an improved ray casting algorithm based on GPU acceleration. First, we acquire the ray ending point and ray direction by rendering volume bounding box two times and writing GPU vertex program and fragment program. Second, we determine the position of resampling points and calculate values of these resampling points through using the acceleration step sampling method and the fast compound interpolation method. Third, we stop the ray and produce the reconstructed image by using the non-transparency ahead dead method. The experimental results show that the proposed method has the advantages of low complexity and high efficiency. It can provide an effective means for the reconstruction of CT images.Finally, we propose a volume rendering algorithm of LDCT volume data based on GPU, which is combined by preprocessing algorithm and the high speed ray casting algorithm. 3D image quality and reconstruction and interactive speed show that the proposed algorithm can reconstruct the 3D image of LDCT with high speed and high quality. It provides an important means for the practical application of LDCT 3D reconstruction technology.