Behavior And Neuroimgaing Study On Visual Recognition Enhancement Training

Visual object recognition is a process of detecting,separating and recognizing target objects in a simple or complex background.The visual object recognition ability varies greatly among individuals,but behavioral training can effectively improve the object recognition ability.Existing behavioral studies under laboratory conditions have shown that short-term training can improve the recognition rate of ordinary subjects on simple geometric shapes,ordinary objects and faces.Different physical attributes of visual objects have different effects on the training methods and training differences of visual object recognition.For simple object recognition,it takes subjects only a few hours to acquire the ability to recognize geometric shapes.For objects in the real world,the improvement of visual object recognition ability often requires a long period of training,which can eventually lead to the acquisition of expert skills.For example,it often takes several years for students of imaging department to grow into experienced imaging physicians.For real world object recognition,how to break the traditional cognitive curve and shorten the learning time is still a challenging problem.The source of behavior is the brain.Our brains are not fixed tissues but malleable and constantly adapt to a dynamic world.Affected by key factors such as sensory input,experience shaping,pathology and nutrition,the brain's function and structure are constantly variable.Robust visual recognition behavior often corresponds to a stable neural basis.Neuroimaging technology allows for real-time,non-invasive,wide-field observations of the brain.In particular,in the past 30 years,magnetic resonance technology has become one of the most commonly used methods of brain detection and been widely used in the study of brain plasticity.Many studies believe that: the free fluctuation activity information of the brain in the resting state contains the set of experience at all previous moments,which is the concentrated expression of the trait information of the subjects.Therefore,a large number of studies use the activity state of the brain in the resting state to predict the behavior of the subjects without intervention or after intervention.This paper focuses on the enhancement of visual recognition of complex objects in the real world by conducting behavioral and neuroimaging studies.The main work and innovations are as follows:First,in order to solve the question of the enhancement of visual recognition of complex objects in the real world,psychology and cognitive neuroscience provide a large number of training paradigms.According to the actual demand of the scientific research project,in the face of the recognition of SAR images of class X ground chariots,this article optimized and implemented a method for strengthening recognition of real world object.In the face of the training for SAR image recognition,this article designed the quantitative training platform of SAR image recognition with efficient learning paradigm and achieved the fast training to recognize the SAR images.Specifically speaking,on the one hand,this paper summarized the effective training paradigm of visual recognition through a systematic and complete literature review.Then this paper proposed the optimal training paradigm the minimum-feedback learning paradigm with bipolar feedback based on the attributes of practical problems and designed the learning tasks for SAR image based on this paradigm.On the other hand,this paper designed a multi-location Free Response Receiver Operating Characteristic algorithm that can comprehensively measure the training personnel's SAR image recognition ability.This algorithm fused the commonly used multi-dimensional decision index(specificity,sensitivity and confidence level of personnel)into a representation index,which solved the key problem of quantitative evaluation of visual object recognition ability.The analysis results based on the simulation data show that the algorithm is effective and highly sensitive.Then,in this paper 12 ordinary college students without radar background of basic radar knowledge was recruited to train using the feedback-learning paradigm.The subjects was tested and trained for 20 days using the platform.The statistical analysis results show that the subjects' behavior has been significantly improved and achieved the approximate expert level after training.The results proved the effectiveness and efficiency of the training method.Secondly,the object recognition ability prediction research based on resting-state functional magnetic resonance imaging data is carried out in this paper,and the prediction relationship between central representation and behavioral representation is established.According to the actual needs of the research project,52 imaging interns were recruited from the imaging department of the first affiliated hospital of Xi 'an jiaotong University to collect behavioral data and resting-state functional magnetic resonance imaging data.The common measurement indexes of resting-state f MRI were selected to establish the feature vectors and predicted the visual object recognition ability of each subject.Then Multi-voxel Pattern Analysis(MVPA)was used to establish the correlation between central representation and behavioral representation.Specific as follows: in this paper three dimension was calculated of resting-state f MRI including Degree Centrality,Regional Homogeneity and the Amplitude of Low Frequency Fluctuation.The three dimensions were combined into a fourth index and the four indexes are used as the prediction characteristics.The paper used the searchlight to extract the feature and then principal component analysis was used to reduce feature dimension.The SVR was trained based on the quantitative evaluation score.The leave-one-out validation was used and the accuracy between the predicted value and the actual value was quantitatively analyzed by Pearson correlation.Then multiple comparative statistical tests were used to identify brain regions with statistical significance.Our data analysis showed that the prediction efficiency of DC was the highest and the prediction efficiency of Left Middle Frontal Gyrus was the highest.