ICA-CNN classified fMRI space activation graph smooth augmentation method

Abstract

The invention discloses an ICA-CNN classified fMRI space activation graph smooth augmentation method, and belongs to the field of biomedical signal processing. Firstly, three-dimensional space smoothing is applied to an fMRI space activation graph obtained through ICA separation, a new sample set is generated through augmentation, and then the new sample set is sent to a CNN to be classified, so that the improvement of the classification of patients and healthy people is achieved. Patient and healthy person classification is carried out on 82 tested resting state fMRI data, three different FWHM Gaussian filters are adopted to carry out space smoothing on a space activation graph separated by ICA so that three groups of sample sets are generated, and then the sample sets are sent to the CNNand are classified. Compared with the existing multi-model order data augmentation method, the method has the advantages that the classification accuracy can be improved by 5.76%, if the two methodsare combined, the classification accuracy can be improved by 21.33%, therefore, the method can independently improve the network classification performance, is also easy to combine with other augmentation methods, and remarkably improves the classification accuracy.

Description

technical field [0001] The invention belongs to the field of biomedical signal processing, in particular to an fMRI spatial activation map smoothing augmentation method for ICA-CNN classification. Background technique [0002] Convolutional neural networks (CNN) has the advantages of local perception and weight sharing. It not only performs well in tasks such as recognition, detection, and classification, but also has great achievements in smart medical care. Resting-state fMRI (rs-fMRI) data has the advantages of non-invasiveness, high spatial resolution, and easy acquisition on patient subjects. It is often used in the analysis and diagnosis of neurological disorders such as schizophrenia. Therefore, CNN with rs-fMRI data as training data will show unique advantages in the classification tasks of healthy people and patients. [0003] In view of the difficulty in collecting patient fMRI and the small amount of data, Lin Qiuhua et al. proposed an ICA-CNN classification fram...

AI Technical Summary

Problems solved by technology

However, these methods are not suitable for rs-fMRI analysis where the spatial structure cannot change arbitrarily

Images