A hybrid approach for automatic clustering and labeling of neuronal fibers in white matter of the brain

Abstract

The invention relates to a method for automatic clustering and labeling of mixed brain white matter nerve fibers, which is characterized in that: firstly, a standard brain atlas is matched to a single brain space by using a high-dimensional elastic registration method, so that a single brain can be divided into 104 anatomical regions with structural information; then use different nerve fiber bundles to extract and mark 15 kinds of nerve fiber bundles through the characteristics of different brain anatomical regions; for the other 4 kinds, the above-mentioned ones cannot be used because they are strongly entangled with peripheral nerve fiber bundles The nerve fiber bundles extracted by the method are automatically clustered and analyzed by using the method based on the similarity measure combined with Kernel-PCA and Fuzzyc-mean. Finally, a feature-based identification method is used to mark the results of the cluster analysis . Through the above method, the 19 main white matter nerve fiber bundles of the brain can be automatically extracted and labeled.

Description

technical field [0001] The invention relates to a method for automatic clustering and marking of mixed brain white matter nerve fibers, and belongs to the field of medical image processing where the image modality is diffusion tensor nuclear magnetic resonance image (DT-MRI). It is suitable for medical imaging diagnosis based on nerve fiber bundles, analysis of brain structure and function, research on brain development and neurodegenerative diseases, etc. Background technique [0002] DT-MRI diffusion tensor nuclear magnetic resonance images provide a description of the direction and structure of brain nerve fibers, and are currently the only medical imaging technology that can be used to observe brain white matter nerve fibers in vivo. The analysis of brain white matter nerve fibers based on DT-MRI not only provides a tool for direct visual observation of brain nerve fiber bundles, but also enables the quantification of brain white matter nerve fibers by combining differen...

AI Technical Summary

Problems solved by technology

At present, the diffusion tensor nuclear magnetic resonance image is processed, and the nerve fiber tracking of the DTI image is performed using DTIStudio, and the nerve fiber connected by a series of pixel points and the whole white matter nerve fiber map of the brain expressed by a series of nerve fibers can be obtained. However, thousands of nerve fibers without organization and labeling are difficult to satisfy research based on nerve fiber bundles with bioanatomical significance, making it impossible for medical researchers to recognize and identify numerous nerve fibers, and it is also impossible to integrate the brain White matter nerve fiber maps are good for medical research

Images