Brain disease classification method based on 3D attention convolution and self-supervised learning

Abstract

The invention discloses a brain disease classification method based on 3D attention convolution and self-supervised learning, and belongs to the field of brain science research. The method specifically comprises the following steps: acquiring resting-state fMRI data and preprocessing; constructing functional connection data based on the fMRI whole brain voxels; dividing a data set; and performing brain disease classification based on attention convolution and self-supervised learning. According to the method, spatial features of whole-brain voxels are extracted from fMRI data by using 3DCNN of an attention mechanism, meanwhile, more meaningful characterization is mined by using self-supervised learning, and finally, combined training is performed on a classification task and a self-supervised auxiliary task to optimize parameters. The method provided by the invention can better explore the spatial information of the brain and mine the implicit features of the data, so that the classification effect is improved, and the method is reasonable and reliable.

Description

technical field [0001] The invention belongs to the field of brain science research, in particular, the invention relates to a brain disease classification method based on 3D attention convolution and self-supervised learning. Background technique [0002] The human brain is a highly complex system of the human body, containing a large number of neuron cells. Through the interaction between multiple neurons, neuron clusters or multiple brain regions, the human brain can complete various complex tasks. The structure and function of the human brain are extremely complex, far beyond our current cognitive capabilities. Therefore, it is undoubtedly very meaningful to explore and understand the working mechanism of the human brain and unravel the mystery of the brain. The classification of brain diseases based on fMRI data is an important research direction in the field of brain science. Analysis of fMRI data is helpful for understanding autism spectrum disorder (Autism spectrum...

AI Technical Summary

Problems solved by technology

However, one-dimensional data cannot reflect the spatial relationship of voxel data, while 3DCNN can handle spatial data

However, human brain data has the characteristics of small sample size and high feature dimension. 3DCNN cannot take into account the importance of each voxel, and it is difficult to dig out internal information from high-dimensional data. The classification of voxel data based on 3DCNN needs to be further improved.

Images