Brain fiber rapid clustering method based on continuous clustering framework

Abstract

The invention provides a brain fiber rapid clustering method based on a continuous clustering framework, which comprises the following steps 1) a fiber cluster parameter model is established, which comprises a fiber cluster center, density, intra-group distance and number of fibers; 2) a certain number of fibers are read, and an algorithm initialized parameter model is searched through a density peak value; 3) a new fiber is read, and similarity distance calculation is carried out on the fiber cluster represented by the parameter model, and the new fibers are distributed to a cluster or placedin a cache container according to the similarity distance; 4) step 2) is repeated; and whether the parameter model is updated or not through the threshold value of the cache container is determined until all the fibers are classified and processed; 5) finally, density peak value searching clustering is carried out on all the parameter models, and the brain fiber clustering process is ended. According to the invention, a fiber cluster is represented through a parameter model, and the fiber similarity calculation is carried out on all the fibers in the cluster in the form of a parameter model,and finally clustering is carried out on all the parameter models.

Description

technical field [0001] The invention relates to the field of brain science, is a kind of machine learning technology, and is mainly used for analyzing brain nerve fibers, especially an unsupervised clustering method. Background technique [0002] In recent years, the development and maturity of MRI technology represented by diffusion tensor imaging (DTI) and high angular resolution diffusion imaging (HARDI) has brought good prospects for non-invasive detection of human brain white matter nerve fiber structure. However, such methods usually generate huge and incomprehensible fiber data sets, and how to perform accurate visual analysis of this fiber collection has been a very important issue in clinical research. Brain nerve fiber clustering technology clusters fibers with similar structures into fiber bundles that conform to anatomical knowledge, thereby improving people's perception of fiber structure. It is an important means in the visual analysis of brain fibers. [0003...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problem that the existing brain fiber clustering algorithm consumes too much time in the process of calculatin

Images