Image classification method based on semi-multi-mode fusion feature reduction frame

Abstract

The present invention discloses an image classification method based on a semi-multi-mode fusion feature reduction frame. Namely, a relation between a brain function and structure network feature vectors is maintained to better capture complementation information of a plurality of mode data and allow disease classification accuracy to be further improved. The method provided by the invention fullyexcavates reference information provided by structure network data, selects valid feature data from a brain function network feature vector matrix, and adds a new constraint on the basis of a known K-support normal form to remain distances of different mode feature data. Analysis and experiment results show that the image classification method based on the semi-multi-mode fusion feature reductionframe is better than known KSN and NF-KSN methods in the prior art. Consistency network connection selected in the invention comprehensively considers correlation of the structure and the function network layer with the diseases and is not limited in the function network layer, and the selected consistency network connection taken as disease biomarkers is higher in reliability.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image classification method based on a semi-multimodal fusion feature reduction framework. Background technique [0002] In recent years, brain imaging techniques, including structural magnetic resonance imaging (sMRI), diffusion tensor imaging (DTI), and functional magnetic resonance imaging (fMRI), have demonstrated their importance in the clinical medical research process for revealing disease progression. meaning. [0003] Among them, the use of machine learning algorithms such as multivariate pattern classification analysis (MVPA) to distinguish healthy and diseased brains is one of the hotspots in current brain imaging research. At present, most of the relevant research work is carried out around the multi-modal multi-task method, which makes full use of the complementary information of multiple modal data and achieves good classification results. At the same ...

AI Technical Summary

Problems solved by technology

In general, functional brain networks can be analyzed using between-group statistical comparisons and multivariate modeling methods, but these methods are not suitable for detecting disease-related brain structural networks

Images