Brain cognitive measurement method based on deep learning feature extraction and multiple dimensionality reduction algorithms

Abstract

The invention belongs to the field of brain cognitive ability measurement, and specifically relates to a brain cognitive measurement method based on deep learning extraction features and multiple dimensionality reduction algorithms, aiming to solve high-dimensional brain image feature extraction and intelligence of cognitive ability measurement and convenience issues. The method of the present invention includes: using a feature extraction network to perform multi-channel feature extraction on the input brain image data, and obtaining a local feature vector through straightening and splicing operations; performing orthogonal projection dimensionality reduction on the obtained local feature vector; The constructed cognitive ability-local feature correspondence is used to measure the cognitive ability of the local features after dimensionality reduction, and output the measurement results. The invention realizes the automation, intelligence and convenience of the brain cognitive ability measurement; meanwhile, it has higher recognition accuracy.

Description

technical field [0001] The invention belongs to the field of brain cognitive ability measurement, and in particular relates to a brain cognitive measurement method based on deep learning feature extraction and multiple dimensionality reduction algorithms. Background technique [0002] At present, the way to measure the cognitive ability of the human brain is mainly the way of questionnaire measurement, such as examination and test, questionnaire survey and so on. This method is usually interfered by supervisory factors, including the emotional or mental state of the main tester or the testee, so it is usually difficult to obtain stable and objective evaluation results. With the advancement of brain neuroimaging technology, it has been possible to measure the structure and functional activity information of the brain in the form of high spatial and temporal resolution. However, this three-dimensional brain imaging data contains a large number of features of brain structure a...

AI Technical Summary

Problems solved by technology

However, this three-dimensional brain imaging data contains a large number of features of brain structure and function. How to use these features to measure the cognitive ability of the human brain poses a huge challenge to conventional feature-based machine learning methods: first , different from object recognition in conventional computer vision (such as recognizing object shape categories, etc.), it is impossible for humans to clearly define which features are valid features on brain imaging, especially for functional neuroimaging, humans (including doctors) cannot judge the value of features; Secondly, since 3D brain images contain a large amount of data for individuals, the screening of features becomes particularly critical.

Images