Evaluation Of Pattern Recognition Method Based On ADHD

Attention deficit hyperactivity disorder(ADHD)is a common neurodevelopmental disorder in children.The main symptoms of the patients were difficulty in concentration and hyperactivity.In recent years,pattern recognition method is widely used in disease diagnosis and nidus search,but the final results depend on the selection of classification features and classifiers.In this paper,the optimal classification model of ADHD is obtained by considering different classification features and various combinations of different classifiers.Based on this optimal classification model,we searched the nidus brain area of ADHD.The research results can provide some help for the diagnosis and treatment of ADHD.In this paper,we based on the data of ADHD-200 competition network,using Regional Homogeneity,Fractional Amplitude Of Low-frequency Fluctuations,Voxel-Mirrored Homotopic Connectivity and Degree Centrality this four characteristic brain images.The corresponding feature values of each brain region are extracted by AAL template,the feature values of each brain region are input into multiple classifiers and trained this classifiers.Finally,input the reserved test set data into the trained classifiers to evaluate the classification performance of multiple classifiers.We find that an optimized SVM proposed by crammer and singer is always stable and can be used as a stable classifier to identifyADHD neural images.The classification performance of neural network is not ideal.This paper explains the reason of this result,and puts forward some suggestions for ADHD to select features and classifiers.In order to find the brain nidus regions of ADHD,this paper proposes a voting algorithm based on the classifier with better performance,which can rank the brain regions that affect the classification accuracy.Our results show that the nidus brain areas of ADHD are mainly located in prefrontal lobe,cerebellum and caudate nucleus,which is similar to many studies on the brain areas of ADHD focus.This paper puts forward some suggestions for the feature selection and pattern recognition methods of ADHD research,and finds out the nidus brain area of ADHD,which provides some help for the diagnosis and treatment of clinical medicine of ADHD.