Brain reaction comparison method based on low-order multielement generalized linear model

Abstract

The invention provides a brain reaction comparison method based on a low-order multielement generalized linear model. The brain reaction comparison method comprises the following steps: firstly, proposing a low-order multielement generalized linear model for combined modeling on reactions of all brain voxels, introducing parameter estimation of the model as an optimization problem into an optimization function with a penalty term to ensure time-space smoothness of a hemodynamic response function (HRF) and sparsity of a recognized brain area, meanwhile, establishing an effective optimization algorithm to estimate brain activities of group, and finally disposing a rapid selection strategy to achieve rapid recognition on penalty parameters and brain areas of different reactions. By adopting the brain reaction comparison method provided by the invention, not only are differences of different areas and stimulation types HRFs flexibly depicted, but also 'borrowing' of information among different voxels can be achieved, the recognition accuracy of brain areas is increased, meanwhile, compared with a conventional method, the method has less parameters and high calculation efficiency, and anovel method is provided for studying brain activities.

Description

technical field [0001] The invention relates to the technical field of biomedical image analysis, in particular to the comparison of brain responses to two stimuli based on functional magnetic resonance imaging data Background technique [0002] Functional magnetic resonance imaging (fMRI) is a technique that measures brain activity by detecting changes in blood oxygen levels in blood vessels. It is widely used in the study of the human brain due to its non-invasiveness and high spatial resolution. . Each fMRI time series typically has hundreds of time points ranging from 0.5 s to 2 s, indicating the activity of a single brain voxel over time. Since human brain activity changes with time, regions, subjects and input stimuli, fMRI data evoked by multiple stimuli are not only large in number but also very complex and contain huge noise. [0003] In fMRI experiments using experimental stimulus sequences to elicit human brain activity, the research focus is usually based on th...

AI Technical Summary

Problems solved by technology

First, the subject’s HRF estimator has large variability due to the low signal-to-noise ratio (SNR) of the fMRI data, resulting in unstable feature quantities

Second, tests for one HRF feature may fail to identify differences in other HRF features, for example, t-tests comparing HRF heights often fail to detect differences in HRF shape

Third, detection of HRF features is essentially single-voxel analysis, ignoring the spatial nature of fMRI data

Images