Visualization Of Brain Fiber Clustering Based On Sequential Framework

With the development of magnetic resonance imaging technology,high angular resolution diffusion imaging(HARDI)provide a promising way of estimating the neural fiber pathways in the human brain non-invasively.However,it is difficult to analyze the vast number of resulting tracts quantitatively.How to visualize the resulting tracts is a very important issue in clinical studies.Brain fiber clustering technology clusters white matter fiber pathways into bundles that are consistent with the neuroanatomy.It could enhance the perception of fiber structure,which is an important means of visual analysis for brain fibersSince classical clustering techniques cannot be applied to high-dimensional brain fibers data directly and have the disadvantage of being sensitive to parameter and large calculation.This paper improves brain fiber similarity measurement algorithm based on the structural characteristics of brain fibers and presents a fiber clustering algorithm based on fast density peak search.For the problem of high complexity,this paper presents a fiber clustering algorithm based on sequential clustering framework.The detail work is listed as follows:Firstly,similarity measurement algorithm is improved for brain fibers.Because classic fiber similarity measurement algorithm does not consider the similarity of the overall morphology of the fiber,this paper uses the dynamic time warping algorithm to measure similarity of brain fibers.By stretching and shrinking the two fibers to nonlinear integration into the same length,we calculate the distance for the optimal warping path as similarity distance.Secondly,this paper presents a fiber clustering algorithm based on fast density peak search,since the classic clustering algorithm's design is mainly based on point data which does not consider the spatial distribution of brain fiber bundles.The algorithm firstly calculates the density radius with random sampling and fiber's density and minimum distance,moreover selects the cluster center in a decision graph by the user,finally realizes the brain fiber clusteringThirdly,this paper presents a fiber clustering algorithm based on sequential clustering framework.In the case of using similarities as the sole information for clustering,most of the processing time is devoted to the computation of these similarities.Aiming at this problems,this paper uses parametric models to represent the clusters.The parametric models represent all fibers in the cluster to calculate similarity.At last,we implemented the prototype system of brain fiber visualization based on QT to satisfy the user requirement.