Research On FNIRS Data Analysis And Recognition Algorithm For Amblyopia Children

In China,the number of children aged 0-14 is about 220 million at present,and the incidence of amblyopia is about 2%-3%.Amblyopia seriously affects the daily life of children,hinders the physical and mental health of children,and causes endless trouble and harm to children and their families.Modern medical research has shown that the earlier amblyopia patients receive treatment,the better their vision can be restored,and that amblyopic children under 12 years old can recover to normal vision after timely treatment.Therefore,it is very important to realize the early identification of amblyopic children.As an emerging brain functional imaging technology,Functional Near Infrared Spectroscopy(fNIRS)has the advantages of non-invasiveness,ease of use,real-time,long-term acquisition,etc.,which is especially suitable for children with amblyopia.In order to promote the research of automatic identification algorithms,analysis methods and potential biomarkers for children with amblyopia,this paper designs and completes experimental settings,data collection,data analysis,feature extraction,biomarker search,new graph structure construction,and identification of children with amblyopia.A classification algorithm based on graph structure and Graph Convolutional Network(GCN)was proposed,and experiments on recognition of amblyopic children were carried out.This article also implements an online intelligent medical auxiliary analysis and diagnosis system based on the B/S architecture,which provides doctors with auxiliary methods and data support for the clinical diagnosis of amblyopic children.The main work of this paper is as follows:(1)Construct fNIRS data sets for children with amblyopia and children with normal vision.The data set contains a total of 78 children’s experimental data,including 40 children with amblyopia and 38 children with normal vision.All data come from clinical data collected in cooperation with the hospital.The data set construction work is mainly divided into two steps:the first step is to design the experimental paradigm and collect fNIRS data.After the data is analyzed,the results are communicated with the doctor to improve the experimental design and finally determine the ideal experimental plan;the second step is preprocessing.Judge the applicability of the data and delete the collected abnormal and characteristic abnormal data,and perform spectrum analysis to design a suitable band-pass filter to filter out the noise in the data.(2)Perform fNTIRS data analysis and feature extraction and obtain a set of potential biomarkers.Carry out the experiments of complexity analysis and inter-group analysis based on the permutation entropy method,and energy feature extraction and inter-group analysis based on the wavelet packet decomposition method.The results show that the complexity and energy characteristics of each channel data in the amblyopic child group are significantly higher than normal children group.The energy feature’s amplitudes of different channels in the amblyopic child group have a large gap,while the energy feature’s distribution of each channel in the normal child group is very balanced.The results of the Two Independent Sample T-Test show that the complexity features show significant differences in more than half of the channels.This shows that complexity and energy features are a potential biomarker for identifying children with amblyopia.In this paper,the statistical features and Hjorth features are also verified for differences between groups.The experimental results prove that the margin features are significantly different.In this paper,a total of 12 distinctive features were extracted for subsequent classification and used as a set of potential biomarkers.(3)Propose a classification algorithm based on graph structure and GCN to recognize children with amblyopia.The proposed algorithm solves the problem of the failure of effective use of brain connectivity information in the existing fNIRS data classification research.The proposed algorithm defines the channel as the node of the graph structure,and the information contained in the node is the feature extracted from the corresponding channel or original data.The absolute value of the correlation coefficient between any pair of channels is defined as the edge of the graph structure.In this way,an undirected weighted functional graph structure is constructed for each object,and a threshold is set to eliminate edges with false connections.In addition,the algorithm applies GCN to the recognition field of children with amblyopia based on fNIRS data for the first time.After that,a series of amblyopia children identification experiments were designed and carried out on the constructed data set,and compared with traditional machine learning algorithms.Experimental results show that the classification performance of the proposed algorithm is far better than other algorithms,and the accuracy rate is as high as 86.40%.Based on this algorithm,an automatic recognition model of amblyopic children is generated,which provides doctors with an auxiliary diagnosis method.The classification experiment results on two public fNIRS data sets show that the classification performance of the proposed algorithm is better than the existing research,which proves the robustness of the algorithm.The proposed algorithm can be extended to the field of fNIRS data classification,has good universality and application prospects,and provides a new idea and research method for fNIRS data classification.(4)Based on the above-mentioned data analysis and classification algorithms,this article designs and implements an online intelligent medical auxiliary analysis and diagnosis system for children with amblyopia.The system provides functions such as patient fNIRS data analysis and visualization,automatic identification of children with amblyopia,and calculation of potential biomarkers.It is convenient for doctors to analyze patient data,observe potential neurophysiological characteristics and make preliminary diagnoses anytime and anywhere,and can promote data-driven diagnosis and treatment of children with amblyopia.