The Image Processing Of Diffusion Tensor Imaging For The Image Guided Neurosurgery System

Image Guided Neurosurgery system (IGNS) is a kind of medical instrument which integrates many modern technologies, such as medical image processing, biomedical engineering and even computed spatial locating methods. It has played an important role in ensuring the safety of operation and avoiding the risks of clinical surgeries. With the wide application of IGNS, the demand for precise and clinical information is greater than ever. Since functional MRI can provide surgeries with more useful functional information, it has initiated a new beginning of IGNS.As a kind of functional MRI, diffusion tensor imaging (DTI) can easily discern white matter (WM) and grey matter (GM) in brain. DTI based fiber tractography (FT) is a unique technique that can non-invasively identify specific white matter tracts in the brain in vivo. DTI-FT can directly illustrate the relation between lesions and surrounding fibers in 3D visualization and enrich the necessary information for IGNS.If the nerve pathways of patients were hurt during neurosurgeries, this would lead to unrecoverable hemiparalysis. Therefore, the protection of nerve pathways is vital and significant. Since the DTI-FT can provide direct assistance for clinical surgeries, the development of DTI-FT in IGNS has aroused the interest of many researches.Currently, there are two existing problems for the application of DTI in IGNS. The first problem is the distortion correction of multiple-sequence DTI images. The second problem is the development of DTI-FT algorithm in IGNS.For the two mentioned problems, our study mainly includes two aspects:1. The first is the distortion correction of multiple-sequence DTI images.The distortion of DTI is always caused by eddy current (EC) and magnetic susceptibility artifacts (MSA). In DTI images, the geometric distortion along phase coding is prominent; while the geometric along the phase coding is inconspicuous. Geometric distortion of DTI images always results in inner brain tissues shift and brain contour deformation and it will certainly lead to the uncertainty of DTI and DTI-FT in the planning of neurosurgeries. Therefore, the distortion correction of DTI is necessary in IGNS.Different methods have been used to calibrate the distortion of DTI images. The distortion correction of DTI images ever resort to the improvement of hardware and pulse sequence compensation. However, these methods cannot fully calibrate the geometric distortions due to the eddy current (EC), and they are also difficult to implement in clinical applications. In our study, the method of image processing using non-linear registration aligning distorted image with reference image is used to calibrate geometric distortion. Based on the segmentation of 3D brain images of an anatomical image and a corresponding distorted DTI image, the two segmented 3D brain images are registered by multi-resolution B-spline registration for deducing the optimal 3D space transformation. Finally, multiple-sequence DTI images are calibrated with the optimal 3D space transformation. Compared with the single B-spline registration, the experimental results prove that the proposed method is more robust and it can promote the geometric fidelity of DTI in IGNS.2. The second is the development of DTI-FT.Generally, DTI-FT algorithms can be mainly classified into two groups. The first group is the deterministic tracking and the second group is the probabilistic tracking. Admittedly, the deterministic tracking has prevalent advantages over the probabilistic tracking for providing determinate information in clinical applications. At present, the algorithm of FT implemented in IGNS is usually the fiber assignment by continuous tracking (FACT). The FACT is fastest, simplest and low cost in computation time. However, it is difficult to display crossing fibers in human brain.For the advantages of FACT, the implementation of FACT in IGNS is firstly accomplished and validated by clinical data; then a new tracking method based on the two-tensor model is presented for fiber crossing. For the acquisition limitation of DTI, many voxels clearly contain more than one fiber bundle which have the characteristic of planer anisotropy. Based on the two-tensor model, the planer tensor can be decomposed into two tensors. The two tensors can be deduced for two principal vectors for tracking directions. At the planer and non-planer voxels, an improved tracking strategy is built for fiber propagation. Compared with the FACT and extended streamline tractography (XST), our method is more robust for the feasibility of displaying crossing fibers and thus, it is superior to the FACT and XST.In conclusion, this paper represents scientific and effective approaches for the two problems of DTI in IGNS. Related with the study, five papers were accepted or published (one indexed by El in English and four in Chinese) and one paper was submitted (one indexed by SCI in English).