Brain Decoding Of Kanizsa Illusion Image Based On FMRI

In recent years,with the continuous development of cognitive neuroscience and cognitive psychology,people can use neuroimaging techniques,mainly functional magnetic resonance imaging(fMRI)technology to observe brain activation patterns under different cognitive tasks and decode their corresponding cognitive states.This process is called "reading the brain." As one of the main sources of people's information from the outside world,visual decoding of visual information has become a research hotspot in current brain science.Studies have shown that early visual cortex has a strong stimuli response to simple geometric images,and the processing of illusory contours and illusion graphics is more complicated than simple geometric images,such as the shape perception process of Kanizsa triangular illusion.Not only related to early visual cortical processing,but also to more advanced brain regions such as the V4 region,it is a visual processing process with low-level and high-level feedforward feedback links.Therefore,illusory brain decoding research is also a challenging target in the field of brain science.This paper studies the brain function pattern and signal response corresponding to the brain information coding of Kanizsa illusion image through fMRI technology,detects the functional region and neural mechanism of brain coding illusion image.Based on the research of illusion coding mechanism,the pattern recognition method is used to realize the visual perceptual classification and decoding of illusory images and non-illusion images.Our main work includes the following two parts:First,this study designed two forms of stimulation: Illusory Contour and NonIllusory Contour.By collecting the magnetic resonance data of the two kinds of stimulation images subjects viewed,the visual region voxel response characteristics were used to select the voxels highly correlated with the stimulation,and the difference between the activation of the brain coding illusion image and the non-illusion image functional region was studied.We assigned the brain to the brodmann brain partition template and selected five regions of interest related to visual information processing for processing,namely brodmann17,18,19,20,and 21 regions.The experimental results show that the illusory images have stronger activation intensity and wider activation range than the non-illusion images in the five visual regions.The results of this study demonstrate that contour integration and object representation are identical in the brain coding region,but more of these voxel points are involved in contour integration.Secondly,based on the study of brain function patterns,the illusion signals and nonillusion signals are classified and decoded by using Support Vector Machine(SVM)and Naive Bayesian pattern recognition methods.The method of leaving one method and K fold(we use 5 fold)cross-validation are used for processing analysis.The experimental results show that the classification accuracy rate for the brodmann 17,18,and 19 regions can reach more than 80%,and the classification accuracy rate for the brodmann 20 and 21 regions can reach more than 70%.The results prove that the visual perception classification of illusory images and non-illusion images can be realized by the traditional pattern recognition method by analyzing the neuron signals.