Research On Pattern Classification Of Autism Spectrum Disorder Related To Diagnosis Of Attention Deficit Hyperactivity Disorder

[Objective]Autism Spectrum Disorder(ASD)and Attention Deficit Hyperactivity Disorder(ADHD)are common neurodevelopmental disorders.Clinically,there are many overlaps between the symptoms of the two diseases,which bring great troubles to clinical diagnosis.In order to explore the heterogeneity of the two diseases,we will select indicators from neuropsychology and/or neuroimaging,and use the Support Vector Machine(SVM)algorithm to conduct research in high-function ASD and ADHD to explore the significance of different indicators in the differentiation of the two diseases.This study is divided into two parts: Study 1 analyzes the significance of joint empathy and executive function indicators for the identification of the two diseases from the perspective of psychological indicators,and tests the accuracy of classification models.The second study discusses the significance of Regional Homogeneity(Re Ho)index in the differentiation of ASD and ADHD from the perspective of biological indicators,so as to provide direction for the differentiation of the two diseases.[Method]Study 1: Research on Pattern Classification of ASD and ADHD Based on Sympathy and Executive Function IndexesFrom January 2018 to June 2019,30 children with high-function ASD,33 children with ADHD and 39 children with Typical Development(TD)recruited in the community during the same period were enrolled.All the participants in the study completed the following neuropsychological scale tests: The Griffith Empathy Measure Parent Ratios(GEM-PR)scale was used to evaluate the empathy ability of all the participants.The Behavioral Rating Inventory of Executive Function(BRIEF)evaluates all subjects’ executive functions,the Autism Diagnostic Observation Schedule,ADOS)evaluates the severity of core symptoms in children with ASD,and Chinese Version of Swan-Sonolan and Pelham,Version Ivscale,SNAP-IV evaluates the severity of core symptoms in children with ADHD.The empathy and executive function differences of the three groups of subjects were compared and analyzed,and Pearson correlation analysis was made with their core symptoms respectively.Finally,the classification features were extracted by using the Support Vector Machine(SVM)algorithm,and the accuracy of the classification model was tested.Study 2: Research on the Model Classification of ASD and ADHD by Re Ho Value Combined with Neuropsychological IndicatorsFrom January 2018 to June 2019,33 high-function ASD children,35high-function ADHD children and 30 TD children recruited in the community during the same period were enrolled in the group.The neuropsychological test and behavioral data processing process are the same as study 1.All participants were required to avoid taking neuropsychiatric drugs within one week before completing functional magnetic resonance scanning,and to stay awake and avoid head movement during scanning.Then the collected brain image data are preprocessed to Re Ho index on dparsfa(http://www.restfmri.net/forum/DPARSF)platform,and then the differences of Re Ho value of the three groups of subjects are compared and analyzed on SPM12 platform,which are taken as Re Ho classification characteristics,the voxel values of the extracted difference masses are respectively Spearman correlated with the category labels,and SVM algorithm is used to respectively test the classification accuracy of single Re Ho index for the three groups of subjects and the accuracy of constructing model discrimination type in joint behavioral indexes.[Results]Study 1: Research on Pattern Classification of ASD and ADHD Based on Sympathy and Executive Function Indexes1.Comparing and analyzing the differences of empathy and executive function indexes of the three groups of subjects,the results show that: the total score of GEM scale [ASD group:(0.67±0.64),ADHD group:(1.00±0.79),TD group:(0.98±0.73)]and the scores of all factors(F=3.595～10.363,all P<0.05),BRIEF scale total score(ASD group:(62.79±7.45),ADHD group(59.47±8.77)and TD group(49.08±7.91)had significant differences in scores of all factors(F=6.557～33.205,P<0.05).The scores of empathy and executive function in ASD group and ADHD group were generally higher than those in TD children(all P<0.05).There was significant difference in cognitive empathy scores between ASD and ADHD groups on GEM scale(P<0.05).In BRIEF scale conversion(P<0.05),emotion control(P<0.05),inhibition(P<0.05)and working memory(P<0.05),no significant difference was found between GEM scale and BRIEF scale.2.The difference results of the three groups of subjects are respectively related to the respective core symptoms of ASD and ADHD,the results show that: ASD group’s BRIEF scale transition(r= 0.422,P<0.05),emotion control(r=0.625,P<0.05)were positively correlated with ADOS scale communication+social interaction item scores;ADHD group’s response inhibition(r=0.534,P<0.05)and working memory(r=0.529,P<0.05)in BRIEF scale are positively correlated with SNAP-IV scale’s inattention+hyperactivity/impulsion score.No correlation was found between GEM total score and sub-scale score and corresponding items in ADOS and SNAP-IV scales.3.When BRIEF sub-scale index was used as classification feature,the classification accuracy rate of the three groups was 58.82%.The classification accuracy rate was 54.24% when GEM scale was used alone.When switching and emotional control related to ASD core symptoms were selected as characteristic indicators,the classification accuracy of the three groups was 51.96%.When reaction inhibition and working memory related to ADHD core symptoms were selected as characteristic indicators,the classification accuracy of the three groups of subjects was 56.86%.4.Compared with the classification using a single scale index,the classification accuracy of the three groups is the highest,reaching 62.75%,when combining all sub-component indexes of BRIEF scale and GEM scale.Study 2: Research on the Model Classification of ASD and ADHD by Re Ho Value Combined with Neuropsychological Indicators1.Comparative analysis of three groups of subjects Re Ho characteristic difference shows: there were differences among the three groups in Supra Marginal＿R,Parietal＿Sup＿L,Parietal＿Inf＿R and Cerebelum＿Crus2＿L(P < 0.05).The distribution of Re Ho relative value in this mass region is TD>ASD>ADHD,and the difference is statistically significant(P<0.05).2.Further extracting voxel values of the three groups of subjects’ differential masses as characteristics,it was found that the maximum classification Accuracy(ACC)and Area Under Curve(AUC)of brain resting state local consistency index(Re Ho)were 60.74% and 0.6671 respectively.3.Compared with the classification using resting state local consistency(Re Ho)features alone,the maximum ACC and AUC in combination with Re Ho index and specific scale features are 73.53% and 0.7943 respectively.[Conclusion]1.ASD and ADHD children show different damage characteristics in empathyand executive function.Compared with single scale index,the classificationaccuracy of joint empathy and executive function index is higher,which ismore helpful to distinguish the two diseases.2.ASD and ADHD have abnormal local consistency in multiple brain regions.The two diseases can be distinguished to some extent by using Re Ho indexalone.3.Compared with Re Ho index classification alone,the classification accuracyis improved when combining specific scale with Re Ho index,which is morehelpful for clinical auxiliary diagnosis.