Correlation and Predictive Ability Between Sensory Characteristics and Social Interaction of Children in Autism Spectrum Disorder

doi:10.21203/rs.3.rs-965092/v1

Background: As a kind of neurodevelopmental disorder, the deficiency of social interaction and communication ability is the core symptom of autism spectrum disorder (ASD). However, ASD usually has sensory abnormalities, which may be an important influencing factor of social function. Here, the aim of the current study is to explore the abnormal sensory characteristics of ASD children and the correlation between social behavior. And further clarify the predictive effect of sensory expression on ASD social ability.

Methods: A case-control study was conducted, with children aged 3 to 10 years including ASD and typical development (TD) as subjects. We used Short Sensory Profile (SSP) questionnaire to evaluate the sensory characteristics in ASD group and TD group. And Social Responsiveness Scale (SRS) was employed to assess the social function of ASD. The differences of SSP scores between two groups of children were compared, and the correlation between SSP in ASD group and SRS was further analyzed. Furthermore, by constructing random forest classification model and support vector machine classification model, the predictive ability of each perception on social level is discussed.

Results: Significant differences were found between ASD children and TD children in the performance of each sensory field. The abnormal sensory rate of children in ASD group is as high as 91.4%, and 74% of them are accompanied by sensory disorders in multiple dimensions.It is worth noting that there is a significant negative correlation between the scores of ASD group children in all dimensions of SSP and the scores of SRS scale. Evaluation and comparison results of full-feature and 7-feature models show that the random forest model and SVM model with all-feature factors have higher sensitivity, while the random forest model with 7-feature factors has the highest specificity. Moreover, the maximum of area value under (AUC) of receiver operating characteristic (ROC) curve between the four models up to 0.859, suggest that the good prediction effect of the models.

Conclusion: Our results suggest autism children have obvious abnormalities in many sensory fields, and there is a significant correlation between this atypical sensory performance and social function. The social level of this group can be well predicted by their sensory characteristics.

Developmental Neuroscience

Neurobiology of Disease

Autism spectrum disorder

Sensory

SSP

Scocial function

Random forest

Support vector machine

Autism spectrum disorder (ASD) is a kind of complex neurodevelopmental disorder, whose main characteristics include social interaction and communication disorder, accompanied by restricted and repetitive behavior or interest^[1]. In recent years, ASD has become one of the fastest growing disease in the world. According to the results released by Centers for Disease Control and Prevention (CDC), in 2020, the prevalence of ASD is as high as 1/54^[2]. The global rate has reached 1% ~ 2%^[3]. In China, the number of patients with ASD has also increased year by year, from 2010 to 2016, it has risen to 1.8% according to the meta-analysis results^[4]. The rapid rise of ASD in the world has brought huge disease burden to families and society, and has become a major public health problem worthy of attention.

Clinical symptoms of ASD are complex and personalized, and the understanding of symptoms and clinical diagnosis of ASD has been constantly updated and revised. For a time, researchers paid more attention to the symptoms of social and communication disorders, while ignoring the abnormal feeling of many autistic patients. In fact, as early as the 1940s, Kanner had aready described the sensory defects of ASD in the article. Until recently, the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) has taken abnormal behavior of sensory stimulation as the diagnostic criteria for ASD, making sensory abnormality more closely linked with ASD^[5]. In recent years, researches focused on the field of ASD have found that sensory abnormalities occur in ASD patients of all ages, and regardless of their severity, The ratio is as high as 60–90% of individuals with ASD show sensory reactivity issues on parent or self-reports^[6].

The abnormal sensory behavior of ASD is usually manifested in many sensory fields, and the response of ASD children to sensory stimulation can be divided into three patterns: hyper-responsiveness, hypo-responsiveness and sensory seeking. Some researchers believe that sensory seeking is also a manifestation of hypo-responsiveness to stimulation, and these three patterns usually occur simultaneously in ASD individuals^[7]. Similar research has been reported by Teresa, whose SSP-based questionnaire survey found that 31% of children with neurodevelopmental disorders, including ASD, have auditory over-responsivity and 27% have tactile over-responsivity^[8]. It can be seen that the early recognition and attention of ASD patients' abnormal sensory behavior, especially the comprehensive quantification of sensory, is very necessary for evaluating the condition and behavior of ASD. Human perception is a complex system composed of multiple senses, which coordinate with each other and influence individual behavior together. However, most studies are usually limited to a single sensory, and often ignore the effect of multi-sensory integration. The current research still lacks a comprehensive exploration of the depth of feeling in all dimensions.

Researchers explored the clinical symptoms of autism, such as social deficits and restricted behaviors, or sensory abnormalities,they used to regard these symptoms as independent of each other. Recently, however, the relationship between sensory processing and these symptoms has received more attention. Sensory processing is defined as a procedure in which the brain recognizes and organizes external stimuli so as to obtain effective information from them^[9], therefore it often reflects the interaction between individuals and the environment. Gradually, more and more studies have found that the primary sensory and perception may be the basis of influencing the advanced emotional function of social interaction. Lane and colleagues analyzed the relationship between sociality and sensation in a study of 54 autistic children, and they found a clear correlation between sensory dysfunction and social skills^[10]. The abnormal sensory behavior of ASD individuals in their early life may develop into joint attention defects, and may affect their language development, eventually leading to social withdrawal and other communication obstacles^[11].

Machine learning (ML) is a kind of automatic data processing tool, which builds perfect prediction models from their own research data sets, including neural networks, decision trees, rule-based classifiers and support vector machines^[12]. This method was introduced into autism research for the initial purpose of solving the difficult diagnosis situation. In 2009, Paolo^[13] used support vector machine model to find that dyskinesia can be used as a clinical symptom for early recognition of ASD. In the subsequent research, especially in recent years, the application of machine learning algorithm in ASD field has been widely concerned. Researchers try to use the powerful data analysis function of ML and focus on various representations of ASD, so as to seek a more convenient and earlier diagnosis method. Researchers try to use the powerful data analysis function of ML and focus on various representations of ASD, so as to seek a more convenient and earlier diagnosis method. With the deepening of the understanding of ASD, the current markers are not limited to the early behavior representation. Neuroimaging, molecular biology and even gene prediction are combined with ML, providing a new means for the research of ASD^[14–16]. Although there have been many achievements in introducing ML into ASD, the vast majority of studies are still limited to difference and prediction, and there are few studies on single typical symptoms such as social interaction. The research of Alaleh et al. provides a new idea, that is, using SVM and ANN classification models at the same time to explore the influence and prediction of common sleep problems in ASD on daily behavior during the day^[17]. It can be seen that it is effective to use model construction to predict other performances with one symptom.

To sum up, there are two goals of this study: a) to explore whether ASD children have more frequent sensory abnormalities than typical developmental children.b) to explore the correlation between sensory processing and social function and clinical diagnosis in ASD children.c)Explain the predictive ability of sensory characteristics to social ability.

3.1 Questionnaire survey and data analysis

3.1.1 Paticipants

The ASD group were recruited from the Research Center for Children's Developmental Behavior of Harbin Medical University. Criteria for conclusian into the ASD group was a diagnosis of ASD in a recognized clinic by a psychiatrist or clinical psychologist using the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), besides, mental development disorder, definite head trauma, nervous system diseases and major somatic diseases as well as epileptic seizure history are excluded. And the control group was recruited from kindergartens and primary schools in Harbin city. And age and sex are comparable with ASD group.

489 paticipants were included in this study and made up two groups: 266 Autism Specturm Disorder children as ASD group ,and 223 typically developing children as control group.Children in the ASD group ( 219 male, 47 female ) were between the ages of 3.02 and 10.66 years old ( mean age=5.31±0.10 years old ).And typically developing children in the control group ( 170 male, 53female ) were between the age of 3.15 and 10.99 years old ( mean age=5.32±0.12 years old ).There were no significant differences across groups in age or sex ( P>0.05 ).

3.1.2 The short sensory profile

The Short Sensory Profile (SSP) is a parent report measure of behaviors associated with abnormal responses to sensory stimuli.It is the most commonly used method to evaluate the perceptual characteristics of 3~10 years old ASD children. According to the previous study, the Short Sensory Profile has reliability of 0.90 and discriminate validity >95% in identifying children with and without sensory modulation dysfunction^[18]. The scores of the questionnaire are derived for seven sensory domains (tactile; taste/smell; movement; visual/ auditory sensitivity; under-resguardian ponsive/seeks sensation; auditory fifiltering; and low energy/weak). Every domain contatins several items, a total of 38 items. Items are scored on 1-5scale, with lower scores indicating more impairment. The scores in each domain and the total score can be divided into 3 score grades(basically normal (0),possibly abnormal (1), obviously abnormal (2))according to the corresponding score range.

3.1.3 The social responsiveness scale

The Social Responsiveness Scale (SRS)is a rating scale that has been designed to be used as both a screener, and as an aid to clinical diagnosis, for 4-18 years old ASDs^[19].The scale is composed of 65 items of 5 sub-scales (social perception, social cognition, social communication, social motivation, autistic behavior). They are all descriptions of children's daily social situations, including positive descriptions and descriptions of autism symptoms, covering behaviors in social communication, communication and repetitive and rigid behaviors of autism. The higher the score of the scale, the heavier the degree of social disorder, and the lower the score, the better the social ability.

3.1.5 Autism Diagnostic Interview-revised

The Autism Diagnostic Interview-revised (ADI-R)^[20] is a standardized and structured diagnostic tool based on interviews with parents and considered as gold standard in assessment in autism. The scale consists of professionals who conduct in-depth interviews with parents about children's daily behavior habits and other aspects and score them. The score is converted into a “diagnostic score” according to the diagnostic algorithm. ADI-R includes four dimensions: barriers to social interaction, abnormalities in communication, limited repetitive and rigid behaviors, and early developmental abnormalities. According to a 4-grade score of 0~3, the social interaction obstacle 16 items have a cut-off value of 10 points. Abnormal communication is divided into 13 items with verbal communication, with a cutoff value of 8 points, and 7 items without verbal communication, with a cutoff value of 7 points. For 8 items of repetitive stereotypic behavior, the boundary value is 3 points, for 5 items of early dysplasia, the boundary value is 3 points. The higher the score, the greater the damage in the response dimension.

3.1.6 Autism diagnostic schedule

The autism diagnostic schedule (ADOS)^[21] is a standardized and semi-structured diagnostic tool and one of the recognized gold standards for ASD diagnosis. Professional evaluators interact with children in a specific way, observe whether children have narrow interests and repetitive rigid behaviors during the game, and evaluate children's social interaction, communication ability, game and imagination ability for auxiliary diagnosis of ASD. The scale is divided into 4 modules according to the language ability of the tested children. Each item is graded as 0~2 or 0~3. Autism can be diagnosed when the respective scores and total scores in areas A and B are higher than the cut-off point.

3.2 Construction and Analysis of Model

3.2.1 Paticipants

We screened the questionnaire data of two groups of children, and selected people who accepted both SSP and SRS assessment. There are 107 children in ASD group (male =89, female =18, mean age = 7.53±0.36) and 44 children in TD group (male =41, female =3, mean age=5.85±0.17).

3.2.2 Data preprocessing

Screening the sample population who received both SSP and SRS evaluation, and eliminating the missing values. 45 children with normal SRS and 107 children with abnormal SRS were included in the model construction. Sample basic information and SSP information are extracted as characteristic variables, including age, sex and eight sensory dimensions of SSP (tactile,gustation/olfactory,motion sensitivity, low response/sensory seeking, auditory filtering, low strength, vision/auditory). With SRS results (normal=0, abnormal=1) as output variables, a two-variable prediction model is constructed.

3.2.3 Random forest model

Random forest is a special bagging method. Its principle is that decision tree is used as a model in bagging, which is composed of several classification and regression trees (Cart)^[22]. By bootstrap resampling technology, it repeatedly randomly extracts N samples from the training sample set to generate a new training sample set to train the decision tree. The random forest consists of m decision trees, and the classification results of new data are determined according to the voting scores of classification trees. CART uses Gini value as evaluation standard. The formula is Gini=1-∑(P(i)*P(i)), where P(i) is the proportion of type I samples in the data set at the current node. Random forest method can better avoid over-fitting because it uses the integration idea and samples both samples and features. In the process of building the model, the parameter ntree is tested and adjusted. The relationship between ntree value and error rate is calculated by using function in current study. Gini index is used to calculate the heterogeneity of observed values at nodes of classification tree affected by each variable, so as to compare the importance of variables. After calculating the relationship between the error rate and the number of features through cross-validation, the model of each method is constructed again by retaining the feature factors with the highest importance ranking.

3.2.4 Support vector machine model

Support Vector Machine (SVM) is a classification model based on the principle of structural risk minimization and VC dimension theory, which seeks the best balance between the learning ability and learning Accuracy of the model to obtain the best generalization ability^[23]. SVM algorithm can better identify and classify small samples and nonlinear high-dimensional samples. When the low-dimensional data in the model is directly mapped to the high-dimensional, the number of dimensions will explode. Therefore, it is necessary to introduce Kernal Function to transform the original features and improve the dimension of the original features to solve the problem of linear inseparability and improve the prediction accuracy of the model.

In this study, the Gaussian kernel function with the best performance is used as the kernel function of SVM model. Firstly, 13 feature factors are substituted into SVM model for training. Secondly, according to the importance of random forest model, the former feature factors are selected to reduce the dimension of feature factors, and then substituted into SVM model for training again.

3.2.5 Model construction and verification

In order to avoid the structural risks in the process of model construction and verify the generalization ability of prediction model, K-CV cross-validation method is adopted, and 9/10 of random samples are used as model construction data and 1/10 as model verification data. The separation of the two groups of data can better verify the model construction effect, and 10% discount cross-validation is performed five times.

The ntree function (number of decision trees generated) of the stochastic forest model with 10 characteristic factors is adjusted to 200, and the error rate can be relatively stable. The gamma and c functions of SVM model with 10 characteristic factors are both set to 1, and the error size and generalization effect of the model are balanced by the more centered function. The random forest model and SVM model are constructed respectively, and the prediction effect of the model is evaluated by the sensitivity and specificity of the model to the test set prediction and the AUC under the ROC curve.

3.6 Data dimension reduction

The varImpPlot function in random forest is used to rank the importance of all factors, reduce the dimension of characteristic factors, select the characteristic factors that have great influence on the prediction results, and reconstruct the prediction model. According to the cross-validation results of the relationship between error rate and characteristic factors, when the number of characteristic factors is about 7, the error rate is low and stable. The order of importance from high to low is: low response/sensory seeking, total score of multiple senses, age, auditory filtering, low strength, vision/auditory, touch, gustatory/olfactory, motion sensitivity, and gender. Therefore, the first seven characteristic factors are selected to construct the random forest model and SVM model again, and the prediction results are compared with those of the ten characteristic prediction model.

3.3 Procedures

The entire study was approved by the Medical Ethics Committee of Harbin Medical University. All the tested children received informed consent signed by their guardians before entering the group.

To obtain information regarding children’s sensory function in the context of everyday life, the parents of all of the participents completed the SSP. In order to evaluate the behavior and social situation of autistic children, Of the 223 children in the ASD group ,there are 113 were given SRS, In addition, due to the ADI-R and ADOS interviews need a long time, we would arrange with parents to conduct the interviews on non-working days or children's rest days. Parents was arranged to conduct ADI-R interviews in a quiet conversation room for more than one hour, while children would completed ADOS assessment in another room. In our study ,there were 90 ASD children were completed ADI-R, and 94 ASD children were given ADOS.

After the questionnaires completed, the researchers checked whether there were any errors or omissions to ensure the completeness and correctness of the questionnaires. And then according to the corresponding algorithm, the questionnaire evaluation results are obtained. ADI-R, and ADOS are evaluated by professional researchers, and then gave out the diagnostic result. Finally, according to the collected two parts of information as 10 characteristic variables and social level as output variables, the ASD social prediction model is constructed by using random forest algorithm and support vector machine algorithm Then, the importance of the characteristic factors is sorted by the importance sorting function in the random forest model, and the dimension of the characteristic factors is reduced. The feature factors after dimension reduction are brought into the machine learning algorithm again, and a more concise social prediction model is constructed.

3.4 Statistical analysis

The original data were analyzed by SSPS17.0. The Shapiro-Wilk method was used to test the normality of the scores of each scale. The Chi-square test was used to determine whether there was gender difference between ASD group and control group. And we used Mann-Whitney U method to test whether there was age, SSP scores difference between both groups. Chi-square test was used to analyze whether there was any difference between ASD group and control group in scores grade of short sensory profile(SSP). We also used Spearman rank correlation to analyze the correlation between SSP score and clinical scale in case group. The random forest model and support vector machine model are constructed and verified by R3.5.3.

4.1 Analysis of differences in sensory characteristics between groups

According to Mann-Whitney test, the scores of ASD group in seven sub-fields of touch, motor sensitivity, low response/sensation seeking, auditory filtering, low strength, vison/auditory and the total scores of multiple senses were significantly lower than those of typical development group, and the difference was statistically significant (P<0.05) (see Table 1).χ² test showed that the proportion of normal grades in ASD group was lower than that in TD group in all sensory fields. The ratio of abnormal grades (possibly abnormal and obviously abnormal) was higher than that of TD group (P<0.05) (see Table 1).

Table 1

Comparison of SSP Score Characteristics between ASD Group and Healthy Control (Mid (P25,P75)/n(%))
		ASD(n=266)	TD(n=223)	Z/χ2	P
Tactile	Mid(P25,P75)	33(30,35)	34(31,35)	-2.73	0.005
	nomal grade(0)	213(80.4%)	197(88.3%)	5.719	0.017
	Abnomal grade(1+2)	52(19.6%)	26(11.7%)	5.719	0.017
Gustation/Olfaction	Mid(P25,P75)	18(15,20)	19(17,20)	-4.361	<0.001
	nomal grade(0)	199(74.8%)	202(90.6%)	20.445	<0.001
	Abnomal grade(1+2)	67(25.2%)	21(9.4%)	20.445	<0.001
Movement sensitivity	Mid(P25,P75)	13(11,15)	14(12,15)	-3.168	0.002
	nomal grade(0)	157(59.2%)	152(68.5%)	4.431	0.035
	Abnomal grade(1+2)	108(40.8%)	70(31.5%)	4.431	0.035
Hypo-sensitivity/Sensory seeking	Mid(P25,P75)	28(26,31)	31(27,33)	-5.087	<0.001
	nomal grade(0)	70(68.3%)	176(79.6%)	7.942	0.005
	Abnomal grade(1+2)	84(31.7%)	45(20.4%)	7.942	0.005
Auditory filtering	Mid(P25,P75)	19(16,22)	24(21,28)	-10.995	<0.001
	nomal grade(0)	59(22.2%)	144(65.2%)	91.723	<0.001
	Abnomal grade(1+2)	207(77.8%)	77(34.8%)	91.723	<0.001
Low strength	Mid(P25,P75)	22(18,27)	28.5(25,30)	-10.452	<0.001
	nomal grade(0)	90(33.8%)	165(74.3%)	79.515	<0.001
	Abnomal grade(1+2)	176(66.2%)	57(25.7%)	79.515	<0.001
Vision/Auditory	Mid(P25,P75)	20(17,22)	21(19,23.25)	-3.426	0.001
	nomal grade(0)	180(67.7%)	170(76.9%)	5.113	0.024
	Abnomal grade(1+2)	86(32.3%)	51(23.1%)	5.113	0.024
Total score of multiple senses	Mid(P25,P75)	152(138,164)	167(155,178)	-9.534	<0.001
	nomal grade(0)	118(44.9%%)	171(78.8%)	57.150	<0.001
	Abnomal grade(1+2)	145(55.1%)	46(21.2%)	57.150	<0.001
ASD, autism spectrum disorders;TD, typically developing control;0 means that the SSP questionnaire score is within the normal range, 1 means that the score is within the possible abnormal range, and 2 means that the score is obviously abnormal; %,The percentage of SSP score in the corresponding level to the total number of people

According to the nature of each topic in SSP questionnaire, we divided all items into three modes: sensitivity, dullness and seeking, and described the number of ASD people with these three abnormalities respectively. The number of people with only one abnormal pattern was 95 (hypersensitivity =84(31.6%), insensitivity =7(2.63%), sensory seeking =4(1.5%)). In addition, 69 people had two kinds of sensory abnormalities at the same time (sensitivity and unresponsiveness =46, sensitivity and seeking =20, dullness and seeking =3), and 59 people showed all three kinds of abnormalities at the same time.

In addition, we selected five basic sensory dimensions in SSP questionnaire, and compared the number of abnormal people in ASD group. The proportion of ASD patients with only one or more sensory disorders was analyzed. The results showed that the abnormal proportion of each dimension was different (tactile abnormality =52(19.5%), taste/smell =67(25.2%), motion sensitivity =108(40.6%), auditory filtering =207(77.8%), visual/auditory =86(32.3%)). In addition, in sensory disorders individuals of ASD group, 26% only had a single sensory abnormality, and 74% showed two or more sensory defects at the same time.

4.2 Correlation analysis between SSP score and behavior scale score in ASD group

Spearman correlation analysis showed that SSP gustatory/olfactory and motor sensitivity scores were negatively correlated with ADI-R social function scores, tactile scores were negatively correlated with nonverbal communication scores (P<0.05); and auditory filtering score was negatively correlated with ADOS communication and social interaction score (P<0.05) (see table 4) .Except that the correlation of movement sensitivity was not statistically significant, other sensory domains of SSP had obvious negative correlation with SRS (P<0.05) (see Table 2).

Table 2 Spearman correlation analysis betwwen SSP score and SRS/ADIR/ADOS score（r）

	SRS(n=113)						ADI-R(n=90)		ADOS(n=94)
	social perception	social cognation	social communication	social motivation	autism behavior	total score	non-vobal communication	social function	communication	social interaction
Tactile	-0.186	-0.226^*	-0.281^**	-0.221^*	-0.110	-0.276^**	-0.267^*	-0.161	0.134	-0.179
Gustation/Olfaction	-0.228^*	-0.177	-0.189^*	-0.184	-0.264^**	-0.229^*	-0.153	-0.207^*	0.229^*	0.044
Movement sensitivity	-0.120	-0.109	-0.109	-0.178	-0.026	-0.141	-0.134	-0.212^*	0.100	-0.073
Hypo-sensitivity/Sensory seeking	-0.346^**	-0.409^**	-0.435^**	-0.391^**	-0.474^**	-0.528^**	-0.154	-0.154	0.048	-0.182
Auditory filtering	-0.312	-0.263^**	-0.444^*	-0.423^**	-0.274^**	-0.456^**	-0.138	-0.166	-0.085	-0.253^*
Low strength	-0.165	-0.173	-0.217^*	-0.081	-0.000	-0.189^*	0.162	0.072	-0.017	-0.058
Vision/Auditory	-0.189^*	-0.138	-0.194^*	-0.238^*	-0.264^**	-0.258^**	-0.072	-0.188	0.075	-0.030
Total score of multiple senses	-0.379^**	-0.352^**	-0.449^**	-0.360^**	-0.345^**	-0.485^**	-0.112	-0.168	0.059	-0.150

* P < 0.05;** P < 0.01; SRS, Social Responsiveness Scale; ADI-R, Autism Diagnostic Interview-Revised;ADOS,Autism diagnostic schedule

4.3 Analysis, evaluation and comparison of random forest model SVM model

The area under curve (AUC) of receiver operating characteristic curve (ROC), sensitivity and specificity were selected to evaluate the prediction effect of the prediction model. The four models are evaluated in different dimensions and compared. Both the stochastic forest model and SVM model with all feature factors have higher sensitivity, specificity and AUC than the model with data dimension reduction. The random forest model with all characteristic factors has the highest sensitivity, up to 0.916, which proves that it has high recognition ability to judge social defects through snsory abnormalities; The stochastic forest model with 7 characteristic factors has the highest specificity, reaching 0.944.(see Table 3)

Table 3

Comparison of invalidation result of prediction models
model category	sensitivity	specificity	AUC
10-factors Random forest	0.916	0.778	0.828
7- factors Random forest	0.711	0.944	0.827
10-factors SVM	0.850	0.822	0.859
7- factors SVM	0.822	0.841	0.848

ROC of the four prediction models is shown in Fig. 1, and the four curves are close to the upper left corner, and AUC is higher than 0.8. the prediction effect of the models is good (0.8 ~ 0.9). Among them, the SVM prediction model with all feature factors has the best AUC, and the prediction authenticity and generalization effect of the model are the best.(see Table 3 and figure 1)

5.1 Abnomal sensory processing of ASD

As we all know, sensory system contains many dimensions such as touch, oflactory, taste, auditory and vision, and even sensory abnormality has many manifestations, which makes it difficult to distinguish and evaluate the abnormal sensory behavior of ASD individuals. The parental assessment questionnaire used in this study accurately grasped the sensory characteristics of ASD and TD children in different classification. In the seven sensory domains and multisensory integration included in SSP questionnaire, the degree and proportion of sensory abnormality in ASD group were more significant compared with TD group. Our results were consistent with Rogers who have found that 95% of the sample of children with ASD demonstrated some degree of sensory processing dysfunction on the SSP total score^[24]. ASD individuals' sensory problem has many implications in behavior. For example, they may hate physical contact, fear light, and extremely sensitive to tiny sounds, etc. These sensory challenges, which exist in ASD childhood, may cause difficulties for children to develop normal communication and social ability.

The high incidence of sensory abnormality has been generally agreed by researchers. However, there are still more features behind this appearance, which are worthy of further consideration. The current results show that although ASD children have amazing damage in many senses, the abnormal proportion of each sense is different, the high abnormal rate of auditory filtering has attracted our attention. In the process of communication, the exposed environment often faces the simultaneous input of various sounds, which requires the ability to extract and receive useful information from complex background sounds, which is also an important manifestation of auditory function^[25]. In this study, most parents indicated that ASD children could not concentrate and do things normally in noisy environment. This shows that it is very difficult for these individuals to filter out useless noises. In the past literature, some researchers pointed out that hearing filtration disorder is very high in ASD patients^[26]. Groen has investigated the level of disyllabic words extracted from various auditory backgrounds by individuals with autism, and found that these people need a higher signal-to-noise ratio than the control group in order to perceive the sounds in pink noise^[27]. Sensory functions and symptoms are mutually mapped. The general anomaly of ASD in a certain dimension just reveals the reasons of their puzzling behaviors on weekdays. Therefore, the key to improve the poor performance is to start with those patients who have sensory injuries, which is likely to save more prognosis.

The environment in which human beings live is complex and changeable, the source of sensory stimulation is never single, and the sensory system is also very complex. In our investigation, the vast majority of ASD individuals not only have one sensory abnormality, but also are not limited to sensitivity or dullness, but a variety of them coexist. All kinds of sensory abnormalities that co-exist among ASD individuals may play their own "roles" and produce various behavioral effects. For example, children with both tactile and auditory filtering difficulties are likely to resist the approach of others and often fail to respond to their calls. This is likely to cause great obstacles to the smooth communication, which is likely to aggravate the symptoms. In addition to the single influence, when external environmental stimuli are applied to human body, they are often mixed with a variety of sensory information, which requires individuals to have the ability to process these inputs at the same time. Just as daily conversation involves the rapid integration of auditory and visual information to create a whole and dynamic perception. Failure to integrate these sensory cues in time will lead to defects in speech perception^[28–29]. In the questionnaire evaluation, we found that up to 30% of ASD children have visual/auditory and auditory filtering abnormalities. A previous study by Noel found that ASD children's audio-visual time combination window is four times larger than that of typical developing children, which indicates that ASD children are completely deficient in audio-visual acuity^[30]. The sensory disturbance of ASD in more than one dimension has made them difficult to cope with, however, the change of reaction mode is an important reason for their abnormal behavior. Sensory hypersensitivity and unresponsiveness are the criteria of ASD diagnosis. Besides these two models, we also evaluated over-seeking. Among the people with only one abnormal reaction (42%), hypersensitivity still accounted for the majority (37%), the proportion of people with multiple reactions also can not be ignored (57%). The results of high sensitivity are consistent with Taylor's research^[31]. These different reaction modes bring ASD's sensory processing to two extremes, extreme excitement or negativity. The lack of proper feelings may also make them infatuated with the pleasure brought by excessively seeking certain feelings, and more likely lead to rigid behavior. A great deal of research has been done on the causes of different patterns, which may be affected by cortical excitability, abnormal receptor function at molecular level and genes^[32–34]. Cross-disciplinary anomalies and complicated reaction patterns make it difficult to distinguish ASD's sensory behavior and seriously affect their other skills.

5.2 ASD sensory abnormality predicts social disorder

Information integration across multiple sensory organs is the foundation of cognition. As social animals, interaction with each other is the most essential feature of human beings, which requires them to give full play to the function of sensory system in order to obtain more valuable information and carry out social activities when facing the world composed of chaotic information^[35]. However, aberrant responses of ASD individuals to sensory stimuli may contribute to aggravate social disorders of autism. From the statistical point of view, the current study analyzes the relationship between the scores of sensory questionnaire and the scores of clinical scale reflecting autism behavior. The results show that there is a negative correlation between the sensory score in SSP and SRS score, as well as the scores of subscales for evaluating social ability in ADI-R and ADOS, which further proves that with the aggravation of the abnormal degree of perception, the social defects of autism are more obvious. And the state of each sensory field has different degrees of correlation with social interaction.

As one of the important ways to explore with the external environment as early as infancy, tactile is very important to the development of individuals. Tactile processing dysfunction is not only related to the emotion of ASD individuals, but also may cause social disorders. Touch, especially social contact expressing friendly nature, such as hug, is very necessary in the process of communication. Moreover, gentle emotional touch plays an important role in relieving individual anxiety and improving social level^[36]. However, autistic people often show excessive defense to mild touch stimulation, but low response ability to painful stimulation^[37–38], which contributes to the failure of detecting social clues. The current study found that SSP tactile score is negatively correlated with ADI-R non-verbal communication score, which is similar to Foss-Feig's^[39] research conclusion. It can be seen that the abnormality of tactile sensation can affect the nonverbal communication ability of ASD, which is the basis of social communication^[40]. The relationship between tactile and ASD social symptoms is also reflected in the fact that tactile information is crucial in distinguishing the relationship between self and others, and the distinction of self -others is the basis of social cognition^[41]. For example, as described in SRS questionnaire, autistic children may not be able to understand the interrelation of things (such as causes and causes) like their peers, which is the prompt of social cognitive defects and ultimately affects the overall social ability of autism.

Besides touch, the influence of vision and auditory on social behavior can not be ignored. "Listening" and "observing" are two indispensable "ties" to communicate with others. When these feelings appear obstacles, it is likely to affect norm al social interaction. Our results suggest that the abnormality of audio-visual perception is inversely related to social function from three perspectives: social perception, social communication and social motivation. Emotion recognition is an important part of social cognition, including understanding facial expressions and emotion rhythm^[42–43], which just depend on the intact visual-auditory ability. Up to 80% of environmental information is transmitted to the brain through vision^[44], and visual processing is related to social behaviors such as joint attention imitating others^[45–46]. However, the current study found that these abilities of ASD group were weak by SRS scale and correlated with abnormal visual performance. Most ASD individuals have difficulty in looking at their faces, and they often pay more attention to the local and grasp the whole, which makes them tired of dealing with complex and subtle social stimuli. For example, ASD children usually pay too much attention to other people's lip movements in the process of dialogue with people, instead of understanding social cues such as emotions through comprehensive observation of facial expressions^[47]. Moreover, conversational communication must be the most important way in the social process. Besides facial expressions, language elements play an important role in the transmission of emotions and attitudes. But due to the auditory abnormality, ASD children's excessive obsession with certain language features (such as voice or intonation) during communication makes it difficult for them to filter out socially irrelevant information from other people's words, and then integrating social clues-such as language information expressing emotions. And excessive hearing sensitivity may cause children's excessive fear of a certain voice, further limiting their social progress. Our SSP assessment found that ASD children are often difficult to concentrate in noisy environments, emphasizing the obstacles in their auditory filtering and auditory attention. And this weakness is more obvious in the recognition of human voice^[48], which may be an important explanation for autistic individuals' lack of attention preference for social stimulation^[49].

Anorexia, picky eaters and difficulty in eating are common in autistic children, which implies abnormal oflaction and taste. However, another important finding of this study is that this abnormality not only affects ASD's eating behavior, but also contributes to social difficulties. In the past, it was thought that terrestrial mammals usually use their sense of smell to identify their peers' emotions and engage in social activities. Recently, more evidence shows that human beings have similar phenomena, and sociochemical signals can transmit emotions such as aggression, happiness and fear^[50]. Abnormal oflaction may may affect the correct detection of these signals. In the correlation analysis, we found that smell/taste expression is not only related to communication ability, but also leads to the occurrence of autism behavior. The definition of autism behavior in SRS scale includes unacceptable changes in daily habits, unusual perceptual interests, and being considered weird or unusual by other children. For ASD individuals, these performances not only interfere with their emotions, but also seem to be incompatible with others and become an important obstacle to integrating into social environment. As Lahera's research found, smell perception has a great relationship with emotional behavior and cognitive ability^[51].

The total score of SSP questionnaire is the sum of scores obtained by subjects in all sensory dimensions, which is a comprehensive evaluation of their sensory performance and reflects the overall perceptual function. The obvious correlation between the total score and SRS scale gives us an important hint: besides analyzing individual feelings, the influence of multi-sensory integration function on social level is also very significant. In the face of complex social environment, individuals often need to receive multiple sensory stimuli at the same time, which requires multiple senses to be mobilized at the same time to cope with a large amount of rushing information. Audio-visual integration needs a lot of employment in the process of communication, in order to fully analyze the information expressed by each other's facial movements, gestures and voice output. In this study, we found that some ASD children have visual and auditory difficulties at the same time, which may make it difficult for them to relate other people's language and facial expressions, which also explains that many autistic children have obstacles in understanding lip language. And will affect their language development, thus developing into interpersonal frustration^[52]. Even some children in the sample have tactile, gustatory/olfactory and audio-visual problems at the same time, all of these senses are helpful to identify emotions, and the sensory integration disorder of ASD children is undoubtedly a great destruction of their social skills.

5.3 Predictive effect of social function by model

The severity of social function is an important factor affecting the quality of life and prognosis of ASD group. The evaluation methods of ASD communication ability in clinic usually include questionnaire survey, psychological measurement and evaluation by appraisers during the game^[53–55]. However, these methods are usually influenced by subjective factors more or less. In recent years, with the development of computer technology, big data analysis methods are widely used in scientific research, and machine learning is increasingly appearing in ASD field. At present, machine learning algorithms are involved in many researches, including finding effective biomarkers for diagnosis, classifying ASD by various phenotypes and predicting risk genes^[56–58]. In this study, this method is innovatively used to judge how much sensory behavior can predict social function. Our results significantly prove that both random forest and SVM models, including 10 features and 7 features, can reliably and stably distinguish ASD social interaction (AUC>0.8). And the 10-feature model has higher sensitivity, while the 7-feature model shows higher specificity. This result provides a new perspective for future research. SRS scale, as an output factor in our model, is widely used in clinical practice, and it is one of the necessary tools to correctly grasp ASD diagnosis and children's social assessment. The characteristic factors used in the model construction all come from the scores of each sensory field in SSP questionnaire, as well as demographic characteristics such as gender and age. Our findings provide evidence for predicting the possibility of diagnosis and socialization through sensory expression.

We rank the contribution of characteristic factors to the model from low to high by the method of data dimension reduction, and find that the top three are low response/sensory seeking, total score of multi-feeling and age. It can be seen that these factors play a higher role in predicting social function. However, the importance of scores in each sensory sub-field ranks behind the total score. This may also suggest that the influence of feeling on social interaction may be the result of the synthesis of various senses, and the total effect is far greater than the single effect. The influence of cross-multisensory fusion on individual behavior has attracted wide attention as a separate field, especially for autistic people. In order to improve adaptive behavior, multi-sensory-based interventions and treatments have also been carried out in ASD^[59]. Electrophysiological and neuroimaging studies have found that neural circuits and brain structure are the factors that affect multisensory integration, which provides a basis for biological research between sensation and social interaction^[60]. The predictive effect of demographic characteristics shows that age is the robust factor of the model, while the importance of gender is the least among many characteristics. The predictive effect of demographic characteristics shows that age is the robust factor of the model, while the importance of gender is the least among many characteristics.

Limitations and Future Research

The limitation of this article lies in the fact that our method of evaluating children's sensory behavior only depends on the way parents report, so there is a lack of measurement of children's subjective feelings of stimulation. Future research should further use psychometrics and other more objective methods to evaluate the sensory performance of ASD children. In addition, although we have revealed the relationship between sensation and ASD symptoms, the biological reasons and causal relationship between them are still unclear. We need to devote ourselves to exploring the neurobiological mechanism in order to expect a deeper understanding of the mechanism. In the construction of the model, this research innovatively links feeling and social interaction, and further clarifies the existence of the opposite relationship between them. However, it is only limited to the predictive function, and the proof of causal relationship is still somewhat inadequate.

Among the autistic children, abnormal sensation is very prominent, and often the same individual shows multidimensional sensory defects. Moreover, disability in this aspect is significantly related to social barriers, which can be used as an important predictor. More attention than ASD feeling is of great significance.

Ethics approval and consent to participate

This study was approved by the ethics review committee of Harbin Medical University (Approval no. 2013004). All participants gave written consent.

Consent for publication

No applicable.

Availability of data and materials

The datasets generated and/or analyzed during the current study are not publicly available due to specifications on data availability within ethics approval.

Competing interests

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding

This study was funded by Natural Science Foundation of China (81874270).

Authors' contributions

JHZ conceived of the study, performed the data acquisition and analysis, drafted and revised the manuscript. LLF and YZ responsible for personnel recruitment and questionnaire distribution, YZ and YTL jointly contrust the model and analyzed the results. WX completed the distribution of questionnaires and the contact and communication between institutions and parents. XXL Conduct questionnaire recovery and quality check. JW and LIW conceived of the study, and helped to received to revise the manuscript, and are responsible for the overall quality control of the research . All authors read and approved the final manuscript.

Acknowledgements

The authors express special thanks to all the children and their parents who participated. This study was funded by Natural Science Foundation of China (81874270).

American Psychiatric Association, 2013. Diagnostic and Statistical Manual of Mental Disorders, 5th ed. American Psychiatric Association, Washington, D.C.
Maenner M J, Shaw K A, Baio J, Washington A et al. Prevalence of Autism Spectrum Disorder Among Children Aged 8 Years - Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2016. Morbidity and mortality weekly report Surveillance summaries (Washington, DC : 2002), 2020, 69(4).
WHO. Autism spectrum disorders [M]. Geneva: World Health Organization, 2019.
SHI H F，ZHANG J X，ZHANG R，et al．Prevalence of autism spectrum disorders in children aged 0-6 years in China：a meta-analysis［J］．Journal of Peking University（Health Sciences）,2017,49（5）：798-806.
Kanner L. Autistic disturbances of affective contact. Acta Paedopsychiatr. 1968;35(4):100-36.
Ausderau K, Sideris J, Furlong M, Little LM, Bulluck J, Baranek GT. National survey of sensory features in children with ASD: factor structure of the sensory experience questionnaire (3.0). J Autism Dev Disord. 2014;44(4):915-925.
Baranek GT, David FJ, Poe MD, Stone WL, Watson LR. Sensory Experiences Questionnaire: discriminating sensory features in young children with autism, developmental delays, and typical development. J Child Psychol Psychiatry. 2006 Jun;47(6):591-601.
Tavassoli T, Brandes-Aitken A, Chu R, et al. Sensory over-responsivity: parent report, direct assessment measures, and neural architecture. Mol Autism. 2019;10:4.
Haberberger RV, Barry C, Dominguez N, Matusica D. Human Dorsal Root Ganglia. Front Cell Neurosci. 2019;13:271.
Lane AE, Young RL, Baker AE, Angley MT. Sensory processing subtypes in autism: association with adaptive behavior. J Autism Dev Disord. 2010;40(1):112-122.
Baranek GT, Watson LR, Boyd BA, Poe MD, David FJ, McGuire L. Hyporesponsiveness to social and nonsocial sensory stimuli in children with autism, children with developmental delays, and typically developing children. Dev Psychopathol. 2013;25(2):307-320.
Rahman MM, Usman OL, Muniyandi RC, Sahran S, Mohamed S, Razak RA. A Review of Machine Learning Methods of Feature Selection and Classification for Autism Spectrum Disorder. Brain Sci. 2020;10(12):949. Published 2020 Dec 7.
Perego P, Forti S, Crippa A, Valli A, Reni G. Reach and throw movement analysis with support vector machines in early diagnosis of autism. Annu Int Conf IEEE Eng Med Biol Soc. 2009;2009:2555-2558.
Nogay HS, Adeli H. Machine learning (ML) for the diagnosis of autism spectrum disorder (ASD) using brain imaging [published online ahead of print, 2020 Aug 31]. Rev Neurosci.
Zhan Y, Wei J, Liang J, et al. Diagnostic Classification for Human Autism and Obsessive-Compulsive Disorder Based on Machine Learning From a Primate Genetic Model. Am J Psychiatry. 2021;178(1):65-76.
Osredkar J, Gosar D, Maček J, et al. Urinary Markers of Oxidative Stress in Children with Autism Spectrum Disorder (ASD). Antioxidants (Basel). 2019;8(6):187.
Alivar A, Carlson C, Suliman A, et al. Smart bed based daytime behavior prediction in Children with autism spectrum disorder - A Pilot Study. Med Eng Phys. 2020;83:15-25.
MCINTOSH D N , MILLER L , SHYU V ,et al. Development and validation of the short sensory profile. Sensory Profile User’s Manual, 1999,59–73.
CONSTANTINO J N ，Gruber C P .Social responsiveness scale（SRS）manual[M].USA：Western Psychological Services,2005.
Guo Yanqing, Yang Xiaoling, Jing Liu, Jia Meixiang. Study on diagnostic validity and reliability of autism diagnostic interview scale (revised) [J]. Chinese Journal of Psychiatry, 2002(01):45-48.
Tang yilang, Guo Yanqing, Catherine E.Rice, Wang Yufeng, Joseph F.Cubells. introduction of autism diagnostic observation scale, one of the gold standards for autism diagnosis [J]. international journal of psychiatry, 2010,37(01):38-40.
Rahman R, Matlock K, Ghosh S, Pal R. Heterogeneity Aware Random Forest for Drug Sensitivity Prediction. Scientific reports, 2017, 7(1):11347.
Wu Qing, Wang Wan, Wang Lingzhi. Exponential smoothing support vector classifier. Journal of Xi 'an University of Posts and Telecommunications. 2014, 19(04):9-14.
Rogers SJ, Hepburn S, Wehner E. Parent reports of sensory symptoms in toddlers with autism and those with other developmental disorders. J Autism Dev Disord. 2003;33(6):631-642.
Ashburner J, Ziviani J, Rodger S (2008) Sensory processing and classroom emotional behavioural, and education outcomes in children with autism spectrum disorder. American Journal of Occupational Therapy 62: 564–573.
Lane A, Young RL, Baker AEZ, Angley M (2010) Sensory processing subtypes in autism: Association with adaptive behavior. Journal of Autism and Developmental Disorders 40: 112–122.
Groen WB, van Orsouw L, Huurne Nt, et al. Intact spectral but abnormal temporal processing of auditory stimuli in autism. J Autism Dev Disord. 2009;39(5):742-750.
Stevenson, R. A., Siemann, J. K., Schneider, B. C., Eberly, H. E., Woynaroski, T. G., Camarata, S. M., Wallace, M. T., 2014. Multisensory temporal integration in autism spectrum disorders. Journal of Neuroscience, 34(3), 691-697.
Stevenson, R. A., Segers, M., Ncube, B. L., Black, K. R., Bebko, J. M., Ferber, S., Barense, M. D., 2018. The cascading influence of multisensory processing on speech perception in autism. Autism, 22(5), 609-624.
Noel JP, Lytle M, Cascio C, Wallace MT. Disrupted integration of exteroceptive and interoceptive signaling in autism spectrum disorder. Autism Res. 2018;11(1):194-205. Taylor E,
Holt R, Tavassoli T, Ashwin C, Baron-Cohen S. Revised scored Sensory Perception Quotient reveals sensory hypersensitivity in women with autism. Mol Autism. 2020;11(1):18.
Takarae Y, Sweeney J. Neural Hyperexcitability in Autism Spectrum Disorders. Brain Sci. 2017;7(10):129.
Coghlan, S.; Horder, J.; Inkster, B.; Mendez, M.A.; Murphy, D.G.; Nutt, D.J. GABA system dysfunction in autism and related disorders: From synapse to symptoms. Neurosci. Biobehav. Rev. 2012, 36, 2044–2055.
Loke, Y.J.; Hannan, A.J.; Craig, J.M. The Role of Epigenetic Change in Autism Spectrum Disorders.Front. Neurol. 2015, 6.
Baum SH, Stevenson RA, Wallace MT. Behavioral, perceptual, and neural alterations in sensory and multisensory function in autism spectrum disorder. Prog Neurobiol. 2015;134:140-160.
Butts, J., 2001. Outcomes of comfort touch in institutionalized elderly female residents Geriatric. Nursing (Lond.) 22 (4), 180–184.
Pelc K, Cheron G, Dan B. Behavior and neuropsychiatric manifestations in Angelman syndrome. Neuropsychiatr Dis Treat. (2008) 4:577–84.
Rattaz C, Dubois A, Michelon C, Viellard M, Poinso F, Baghadadli A. How do children with autism spectrum disorders express pain? A comparison with developmentally delayed and typically developing children. Pain. (2013) 154:2007–13.
Foss-Feig JH, Heacock JL, Cascio CJ. TACTILE RESPONSIVENESS PATTERNS AND THEIR ASSOCIATION WITH CORE FEATURES IN AUTISM SPECTRUM DISORDERS. Res Autism Spectr Disord. 2012;6(1):337-344.
Burgoon J, Guerrero L, Floyd K: Nonverbal Communication. Boston, Allyn & Bacon, 2009.
Deschrijver E, Wiersema JR, Brass M. Action-based touch observation in adults with high functioning autism: Can compromised self-other distinction abilities link social and sensory everyday problems?. Soc Cogn Affect Neurosci. 2017;12(2):273-282.
Bänziger, T., Grandjean, D., Scherer, K.R., 2009. Emotion recognition from expressions in face, voice, and body: the Multimodal Emotion Recognition Test (MERT). Emotion 9, 691–704 (Washington, D.C.).
Adolphs, R., 2003. Cognitive neuroscience of human social behaviour. Nat. Rev. Neurosci., 4, 165–178.
Haupt C, Huber AB. How axons see their way--axonal guidance in the visual system. Front Biosci. 2008;13:3136–3149.
Iacoboni M. Imitation, empathy, and mirror neurons. Annu Rev Psychol. 2009;60:653–670.
Leekam S, Baron-Cohen S, Perrett D, Milders M, Brown S. Eyedirection detection: a dissociation between geometric and joint attention skills in autism. Br J Dev Psychol. 1997;15:77–95.
Foxe, J.J., Molholm, S., Del Bene, V.A., Frey, H.P., Russo, N.N., Blanco, D., Ross, L.A.,2015. Severe multisensory speech integration deficits in high-functioning school[1]aged children with autism spectrum disorder (ASD) and their resolution during early adolescence. Cerebral Cortex 25 (2), 298–312.
Porges SW, Macellaio M, Stanfill SD, McCue K, Lewis GF, Harden ER, Handelman M, Denver J, Bazhenova OV, Heilman KJ. Respiratory sinus arrhythmia and auditory processing in autism: modifiable deficits of an integrated social engagement system? Int J Psychophysiol. 2013 Jun;88(3):261-70.
Tillmann J, Tuomainen J, Swettenham J. The Effect of Visual Perceptual Load on Auditory Awareness of Social vs. Non-social Stimuli in Individuals with Autism. J Autism Dev Disord. 2020 Jul 1.
Endevelt-Shapira Y, Perl O, Ravia A, et al. Altered responses to social chemosignals in autism spectrum disorder. Nat Neurosci. 2018;21(1):111-119.
Lahera, G., Ruiz-Murugarren, S., Fernández-Liria, A., Saiz-Ruiz, J., Buck, B.E., Penn, D.L.,2016. Relationship between olfactory function and social cognition in euthymic bipolar patients. CNS Spectr. 21 (01), 53–59.
Foxe, J.J., Molholm, S., Del Bene, V.A., Frey, H.P., Russo, N.N., Blanco, D., Ross, L.A.,2015. Severe multisensory speech integration deficits in high-functioning schoolaged children with autism spectrum disorder (ASD) and their resolution during early adolescence. Cerebral Cortex 25 (2), 298–312.
Najafabadi MG, Sheikh M, Hemayattalab R, Memari AH, Aderyani MR, Hafizi S. The effect of SPARK on social and motor skills of children with autism. Pediatr Neonatol. 2018;59(5):481-487.
Kidd SA, Berry-Kravis E, Choo TH, et al. Improving the Diagnosis of Autism Spectrum Disorder in Fragile X Syndrome by Adapting the Social Communication Questionnaire and the Social Responsiveness Scale-2. J Autism Dev Disord. 2020;50(9):3276-3295.
Morrison KE, Pinkham AE, Kelsven S, Ludwig K, Penn DL, Sasson NJ. Psychometric Evaluation of Social Cognitive Measures for Adults with Autism. Autism Res. 2019;12(5):766-778.
Chen Q, Qiao Y, Xu XJ, You X, Tao Y. Urine Organic Acids as Potential Biomarkers for Autism-Spectrum Disorder in Chinese Children [published correction appears in Front Cell Neurosci. 2019 Aug 22;13:388]. Front Cell Neurosci. 2019;13:150.
Cheng H, Yu J, Xu L, Li J. Power spectrum of spontaneous cerebral homodynamic oscillation shows a distinct pattern in autism spectrum disorder. Biomed Opt Express. 2019;10(3):1383-1392.
Wang J, Wang L. Prediction and prioritization of autism-associated long non-coding RNAs using gene expression and sequence features. BMC Bioinformatics. 2020;21(1):505.
Case-Smith J, Weaver LL, Fristad MA. A systematic review of sensory processing interventions for children with autism spectrum disorders. Autism. 2015;19(2):133-148.
Siemann JK, Veenstra-VanderWeele J, Wallace MT. Approaches to Understanding Multisensory Dysfunction in Autism Spectrum Disorder. Autism Res. 2020;13(9):1430-1449.

Correlation and Predictive Ability Between Sensory Characteristics and Social Interaction of Children in Autism Spectrum Disorder

Status:

Version 1

Abstract

Figures

Introduction

Methods