Dan Xia

Geometric Calibration of a Micro-CT System and Performance for Insect Imaging

Micro-CT with a high spatial resolution in combination with computer-based-reconstruction techniques is considered a powerful tool for morphological study of insects. The quality of CT images crucially depends on the precise knowledge of the scan geometry of the micro-CT system. In this paper, we have proposed a method to calculate the deviation of rotating axis for compensating deficiency of existing methods. A practical application of this geometric calibration method of the micro-CT system for insect imaging is presented. We have performed the computer-simulation study and experimental study with our prototype micro-CT system. The results demonstrate that the proposed technique is accurate and robust. In addition, we have evaluated the imaging characteristics of the detector in terms of modulation-transfer function (MTF). Finally, insect imaging performance and image reconstruction from data acquired with different energies are presented.

Investigation of BPF algorithm in cone-beam CT with 2D general trajectories

A mathematical derivation was conducted to illustrate that exact 3D image reconstruction could be achieved for z-homogeneous phantoms from data acquired with 2D general trajectories using the back projection filtration (BPF) algorithm. The conclusion was verified by computer simulation and experimental result with a circular scanning trajectory. Furthermore, the effect of the non-uniform degree along z-axis of the phantoms on the accuracy of the 3D reconstruction by BPF algorithm was investigated by numerical simulation with a gradual-phantom and a disk-phantom. The preliminary result showed that the performance of BPF algorithm improved with the z-axis homogeneity of the scanned object.

FAST CONE-BEAM CT IMAGE RECONSTRUCTION BASED ON BPF ALGORITHM: APPLICATION TO ORTHO-CT

Multi-GPUs accelerated BPF algorithm is developed for improving computational efficiency. Three major acceleration techniques are introduced: (1) dividing reconstructed volume into subsets vertically which reduces the computational cost on the boundary term between parallel chords; (2) transposed method is used to avoid low efficiency of access in global memory which is caused by different chords selection; (3) optimized memory allocation scheme is adopted. Experimental data are used to evaluate the image quality and reconstruction time. It takes only 4.118 s to reconstruct a volume image of 512 × 512 × 512 with 360 projection data of 512 × 512 on dual NVIDIA Tesla C2070 cards. Added the time consuming on data read, transfer and storage part, the complete reconstruction process could be finished in less than 9 s. In particular, BPF-based ROI-reconstruction for cone beam Ortho-CT shows promising application prospect.

Phase-contrast Imaging Simulation Based on a Micro-CT System

Propagation-based phase-contrast imaging was simulated based on paraxial Fresnel-Kirchoff diffraction integral and spherical wave illumination. Under a developed micro-CT system parameters, the effects of focal-spot size and imaging geometry on phase-contrast imaging have been investigated using a 2 mm thickness polystyrene edge phantom. An equivalent mono-energy was used to substitute the polychromatic spectrum of the micro-focus x-ray source. To consider effects of focal-spot size and detector resolution, the obtained phase-contrast image with an ideal point source was convolved with source intensity distribution and point spread function of detector. Simulations show reasonable influences of the two parameters and are in good agreement with experimental results.

Iterative reconstruction method for irregular object

Conventional CT reconstruction algorithm, filtered-backprojection (FBP), requires the projection data acquired within the angular range from 0 to 2pi. However, for an irregular object, the data acquired at some projection views may be not available due to the detector saturation. In this work, we proposed a method to optimize the current patterns for obtaining two sets of complementary attenuation data and to employ an iterative method based on equation and inequality to improve the reconstruction quality. The simulation results obtained from the projection data, which are generated by use of the GEANT toolkit, show that the uses of optimal current scanning method and the corresponding iterative reconstruction method can yield the accuracy images with lower noise level.

Region of interest three-dimensional visualization in clinical human CT images sequence

Medical image analysis is more than just algorithms. Visualization of the original image data and processed results, interaction with the data, as well as the data themselves are also important. It has been widely used for demonstrating lesions or their localization in the musculoskeletal system, vascular system, respiratory and alimentary systems. Region of interest (ROI) imaging visualization techniques in computed tomography (CT), which can visualize an ROI image from the CT sequence data set of the ROI, can be used not only for reducing imaging-time but also for potentially increasing clinical analysis accuracy of the particular area images. VC++6.0 with visualization toolkit (VTK) are adopted to reconstruct the 3D images using the 2D CT image sequence in DICOM format. The ROI images visualization by use of the marching cubes (MC) algorithm from the clinical human CT data. The experimental results show that this method is robust, easy to implement and excellent image quality evaluation. Because the proposed method requires minimum visualization data for clinical analysis, it can reduce imaging-time, and has excellent image quality evaluation. These methods can assist doctors to make better and more accurate diagnosis.

Real-Time 3D Space Coordinate Acquisition of Medical Visualization Data

Medical visualization refers to the techniques and processes used to create images of the human body for clinical purposes or medical science including the study of normal anatomy and physiology. The visualization of medical images data sets is to reconstruct 3D images with the 2D slice images so as to reveal the 3D configuration of organs through human visual system. Visual C++ are used to reconstruct 3D images using the CT slice sequence. The key algorithms and human CT 3D visualization results are given in this paper. The coordinates can be acquired by the mouse clicking in the 3D space, by which to realize the point coordinate acquisition of the 3D medicine images. The visualization of medical images can provide us with more information and means of visual interactive for simulated operations and assistant diagnosis. The technique can realize the real time interaction quantitative measurement of three-dimensional CT image.