Graph model-based brain function registration method

Abstract

The invention discloses a brain function registration method based on a graph model, and the method comprises the following steps: mapping high-dimensional brain function image data to a two-dimensional time sequence matrix by taking a brain function activity signal of a subject in a specific cognitive function state as input and taking a brain graph model as a basis; constructing a graph convolutional neural network model to distinguish different cognitive function states, and utilizing a meta analysis method to generate a brain activation distribution prior graph to assist in predicting a brain function activation mode of each subject specificity; combining the two sides to map brain function image data of each subject to a shared representation space suitable for a large-scale group, and finally achieving the accurate brain function alignment between individuals. According to the method, the effect dose of statistical test on a group can be enhanced, the number of tested samples required in brain cognitive function research is reduced, the clinical research cost is saved, and meanwhile, the graph representation information generated in the shared representation space can also be used for accurately predicting the tested brain function state and behavioral indexes.

Description

technical field [0001] The invention relates to the fields of medical imaging and deep learning, in particular to a brain function registration method based on a graphical model. Background technique [0002] In functional magnetic resonance imaging (FMRI) studies, group analysis using brain functional imaging data from multiple subjects has taken up an increasing proportion. On the one hand, group analysis based on multiple subjects can effectively verify the universality and effectiveness of the research results on different subjects, and at the same time, it can also improve the effect size of statistical analysis in brain functional imaging analysis [1]. On the other hand, due to the different anatomical structures and functional area positioning of different subjects, it is necessary to register the brain function image data of different subjects, for example, all subjects are registered to the image template in the same standard space, and further The brain function i...

AI Technical Summary

Problems solved by technology

However, due to the geometric distortion of brain functional images and the difference in contrast between gray and white matter tissues, the cross-modal registration from brain functional images to brain structural images faces great challenges[2]

In addition, due to the possible differences in the anatomical position, size, and shape of the brain functional areas of different subjects, the correspondence between the anatomical structure of the brain and the functional areas of the brain may not be perfect. Not completely consistent [3], this difference in brain anatomy and brain functional areas has been confirmed in many studies [4]

Therefore, although the registered brain functional image data obtained by the registration method based on structural morphology can achieve accurate correspondence between subjects in terms of brain anatomy, it is difficult to achieve accurate correspondence between subjects in terms of brain function representations (such as specific cognitive function states). There may still be deviations in brain functional areas and brain activation patterns, etc.), and it is impossible to achieve functional alignment in the strict sense among subjects, which affects the effect size of statistical analysis, especially in the case of language, working memory, etc. Differences in higher cognitive function, often functional alignment of brain imaging data is more important than anatomical alignment

Images