A Multimodal Fusion Image Classification Method Based on Modal Distance Constraint

Abstract

The invention discloses a multimodal fusion image classification method based on modal distance constraints, comprising the following steps: the first step, obtaining rs‑fMRI data and DTI data of a plurality of subjects; the second step, for each The subjects constructed brain function network feature vectors and brain structure network feature vectors respectively; the third step, based on the Kendall tau correlation coefficient and the "overlap" mode, performed feature filtering operations on the feature vectors of the two modalities; the fourth step, in the K On the basis of the original ‑support norm, the relative distance constraints of the eigenvectors of the two modalities of the same subject before and after mapping are added to construct the objective function of the multimodal feature selection model, and the optimal eigenvectors of the two modalities are screened out ; The fifth step is to train the classifier based on the multi-core support vector machine model; input the optimal feature vectors of the two modes of the object to be tested into the trained classifier to predict its category label. The classification accuracy rate of the present invention is high.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a multimodal fusion image classification method based on modal distance constraints. Background technique [0002] In the past ten years, with the advancement of brain imaging technology, brain science research has entered a period of rapid development. Magnetic resonance imaging, as a non-invasive in vivo brain function detection technology, has rapidly become the most widely used brain imaging technology in brain science research since its birth in the 1990s by virtue of its advantages of high resolution and no radiation. Among them, magnetic resonance imaging techniques include structural magnetic resonance imaging (sMRI, structural Magnetic Resonance Imaging), functional magnetic resonance imaging (fMRI, functional Magnetic Resonance Imaging), and diffusion tensor imaging (DTI, Diffusion Tensor Imaging). Each neuroimaging technique provides a characterizatio...

AI Technical Summary

Problems solved by technology

The current method of image classification based on multi-modal fusion is to connect the features extracted from different modalities into a long feature vector for subsequent analysis, without considering the relationship between different modal data, and the classification accuracy needs to be improved

Images