A task-FMRI-guided method for deep clustering of brain white matter fibers

Abstract

The invention belongs to a clustering method of brain white matter fibers, in particular to a task-fMRI-guided deep clustering method of brain white matter fibers, using task-fMRI data to obtain white matter fiber functional information, and using DTI data to obtain white matter fiber structure information , combine the two to represent a single white matter fiber, and cluster it as input to a convolutional autoencoder embedded in the cluster. The present invention extracts the average task-fMRI signal on the fiber track to represent the fiber, and combines the structural information from the DTI data to further limit and optimize the clustering result structurally. The two kinds of information are jointly input into the cluster-embedded convolutional autoencoder (CAEEC) to generate clustered fiber bundles. During CAEEC training, the reconstruction-oriented loss, clustering-oriented loss and sparse regularization term are combined The training process is optimized; the convolutional autoencoder embedded with clustering can better extract the hierarchical structure of fibers and preserve the local characteristics of the data in the feature space.

Description

technical field [0001] The invention belongs to the field of medical image processing and deep learning, and in particular relates to a method for deep clustering of brain white matter fibers guided by task-fMRI. Background technique [0002] In recent years, with the development of Diffusion Tensor Imaging (DTI) technology, researchers can use Tractography to infer the trajectory of nerve fibers in the brain. However, the inferred individual fiber trajectories are difficult to exploit. Compared with a single fiber, a fiber bundle (a group of fibers) is of great significance for improving people's perception of fiber structure and function, as well as fiber bundle-based diagnosis and treatment. For example, we can compare Alzheimer's and normal groups in a bundle Differences in the characteristics of fibers for disease diagnosis. Therefore, the researchers aimed to partition the dense fiber trajectories of the whole brain into fiber tracts with internal integrity, namely f...

AI Technical Summary

Problems solved by technology

However, the functional implications of the fiber bundles obtained from these fiber clusters are unclear.

Images