Partial least square correlation-based fMRI image clustering system and method

Abstract

The invention provides an fMRI image clustering system and method based on partial least squares correlation. The method comprises the steps that each fMRI image is preprocessed; calculating function connection features on each preprocessed fMRI image to obtain a function connection feature matrix; preprocessing the scale data to obtain a scale feature matrix; calculating brain characterization features on the functional connection feature matrix and the scale feature matrix by using a partial least square correlation method; and clustering the fMRI image to be clustered by using the brain representation feature matrix. According to the method, the fMRI image is clustered by calculating the brain characterization features of the fMRI image features and the scale features through partial least squares correlation, a stable clustering result can be obtained, and the method has the advantages of being wide in applicability, high in clustering feature interpretability and high in clustering result stability.

Description

technical field [0001] The invention belongs to the technical field of data processing, and further relates to a resting-state functional magnetic resonance imaging (functional Magnetic Resonance Imaging) image clustering system and method based on partial least squares correlation in the technical field of data mining. The present invention can be applied to classify and process resting state functional nuclear magnetic resonance images. Background technique [0002] Blood oxygen level dependent BOLD (Blood Oxygen Level Dependent) functional magnetic resonance imaging resting-state fMRI can show the changes in deoxyhemoglobin concentration caused by neurometabolism to indirectly reflect the spontaneous fluctuations of neuronal signals. Cerebral neural activity causes hemodynamic responses that result in increased regional blood flow and volume, resulting in an increase in the ratio of oxyhemoglobin to deoxyhemoglobin, which in turn leads to a decrease in local magnetic inho...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to address the above-mentioned deficiencies in the prior art, and propose a fMRI image clustering system and method based on partial least squares correlation, aiming to solve the limitations of the prior art in fMRI image clustering, Problems with poor interpretability and low stability

[0006] The overall idea of ​​realizing the object of the present invention is: in order to solve the problem of the large limitation of the calculation of the high-dimensional features in the existing fMRI image calculation, the partial least squares correlation calculation used in the present invention can directly calculate the high-dimensional data as input, Avoid the existing fMRI univariate correlation calculation to calculate each feature one by one and need to perform multiple comparison correction, and also avoid the requirement that the input data be normal, the relationship between the variables is linear and cannot have a high degree of multi-collinearity. Applicable Broader

In order to solve the problem of poor interpretability of existing fMRI image clustering, the present invention calculates brain representation features by integrating functional connectivity features and scale features, and the obtained brain representation features are fusion features of brain function information and scale information , which is highly interpretable

In order to solve the problem of low stability of existing fMRI image clustering, the present invention clusters fMRI images using interpretable brain representation features. This feature is not only a comprehensive reflection of high-dimensional information in fMRI images, but also reflects The information of the scale feature corresponding to the image, so using this feature for clustering can obtain more stable results

Images