Since the21st century, the construction of the human brain connection network has been the international frontier research topic, the two essential technologies in the process of constructing the brain network are the white matter fiber imaging and fiber tracking. The white matter fiber tracking technology, which based on diffusion-weighted magnetic resonance imaging is an important means to display the microstructure information of non-intrusive brain, and provides a new way of thinking for the study of neurophysiology, diagnosis of various neurological diseases, and the construction of the human brain connection network.Because the common diffusion tensor imaging technology has difficulty in describing the microstructure of complex fiber, a variety of high angular resolution diffusion imaging technologies gradually become a hot research in recent years. Among these, the spherical deconvolution method is a more accurate method to solve complex fiber structure at present. However, it is confronted with the disadvantages of time-consuming sampling of data, Large amount of computation and low resolution of reconstruction of brain fiber. For the characteristics of the spherical deconvolution problem, combined with non-negativity and sparsity of the distribution of fiber direction, In this paper, a sparse spherical deconvolution imaging method is developed using the compressive sensing theory, which is solved by multi-stage iterative deconvolution technique, not only improving the imaging accuracy, but also shortening the imaging time. At the same time, implementing the fiber tracking algorithm. The details are listed as follows:Firstly, as the problem of the existing spherical deconvolution model cannot stably and efficiently estimate small-angle cross structure, this paper proposes a novel reweighted sparse representation method for estimating fiber orientation distribution. A sparse representation of fiber orientation is added to spherical deconvolution model, which leads to an l1norm optimization model with sparse constraint. Finally, an iterative deconvolution algorithm of the optimization model with nonnegative constraints is presented for fiber orientation estimation. Experimental results concluded from the synthetic data, platform data and real data demonstrate that the proposed model has a higher accuracy identifying, and imaging speed is also greatly improved.Secondly, as the problem of the existing fiber tracking technology has a low tracking efficiency and cannot achieve clinical application. Based on the sparse spherical deconvolution model, this paper studies two representative tracking algorithm, which included by streamline tracking and bayesian probability tracking, then compares the experimental results by the synthetic data and real data. The experimental results show that implementing tracking algorithm on the basis of the proposed model can obtain a more accurate reconstruction results. In addition, bayesian probability tracking algorithm has a higher precision than streamline tracking algorithm, while its calculation is more complex.As a brief, this paper proposes a novel sparse spherical deconvolution imaging model, which indicates a great improvement on imaging accurancy and speed, and corresponding tracking algorithm is also more accurate and efficient than the traditional method. So far deterministic tracking method has been widely used in clinical diagnosis, while probabilistic tracking method need a long time, and is lack of clinical validation. Therefore, in future work, Taking into account the real-time tracking algorithm on high accuracy and high processing speed will become a mainstream research. With the deepening of research, Fiber tracking technology is expected to play a greater role in the clinical applications and pathological studies. |