A system and method for early screening of autism

Abstract

An autism early screening system and method belong to autism screening. The system comprises: a data acquisition module for acquiring video images of parent-child interaction; the video images comprise: voice data and image data; a display module for displaying a scoring interface, the scoring interface comprising: video images, a plurality of characteristic behaviors of children and a scoring column of the characteristic behaviors; an input module for receiving an evaluation value input by an evaluator in the scoring column of the characteristic behaviors according to the video images; a classification module for inputting the evaluation value of each characteristic behavior received by the input module into a pre-trained early prediction model to obtain an autism risk evaluation result output by the early prediction model. After the parent and the child perform the parent-child interaction, the evaluator, i.e. a professional, observes and quantifies the behavior of the child according to the recorded video content, and then outputs a risk result of whether the child has autism through the early prediction model. The whole screening process is suitable for a remote medical scene.

Description

Technical Field

[0001] This invention relates to the field of autism screening, and more specifically to an early autism screening system and method. Background Technology

[0002] Autism spectrum disorder, also known as autism, is a lifelong neurodevelopmental disorder that occurs in early childhood. Many autistic symptoms may persist throughout life, significantly impacting the quality of life for the autistic individual and their family members. Furthermore, the treatment and care of an autistic individual present a severe challenge to the entire family. Statistics show that the global prevalence of autism is approximately 1%, while in my country it is about 7%, making it the leading cause of mental disabilities among children in my country.

[0003] However, my country still faces a severe shortage of resources in early screening and diagnosis, early intervention, and related research for autism. Furthermore, current clinical practice remains heavily reliant on screening and diagnostic tools developed in Western countries. Because these tools do not adequately consider cultural differences between Chinese and Western children in areas such as language expression, social interaction patterns, and behavioral norms, systemic biases can occur when using them to screen Chinese children. At the same time, the distribution of autism screening and diagnostic resources across different provinces in my country is extremely uneven, which further leads to a large number of children missing the optimal intervention window.

[0004] Therefore, designing an early autism screening tool suitable for children in my country is an urgent technical problem to be solved. Summary of the Invention

[0005] The main technical problem solved by this invention is to conduct accurate and convenient early screening for childhood autism.

[0006] According to a first aspect, one embodiment provides an early autism screening system, the system comprising: The data acquisition module is used to acquire video images of parent-child interactions; the video images include: audio data and image data; The display module is used to display the scoring interface, which includes: the video image, multiple characteristic behaviors of the child, and a scoring column for the characteristic behaviors; The input module is used to receive the evaluation value entered by the evaluator in the scoring column of the characteristic behavior based on the video image; The classification module is used to input the evaluation values ​​of each feature behavior received by the input module into the pre-trained early prediction model to obtain the autism risk assessment result output by the early prediction model.

[0007] In one embodiment, the early prediction model is trained as follows: Select a portion of the sample set from the low-risk, medium-risk, and high-risk sample sets as the validation set, and use the remaining sample set as the training set; The training set is input into the deep learning model, and the deep learning model is trained to obtain an initial early prediction model; The validation set is input into the deep learning model to validate the deep learning model and obtain the confidence level of the deep learning model. When the confidence level is lower than the preset confidence level, the model parameters of the deep learning model are adjusted, and training continues with the training set until the confidence level is higher than the preset confidence level.

[0008] In one embodiment, the video images of parent-child interaction include: video images of children and parents playing games; wherein the games include one or more of bubble games, cooking games, pretend games, and building games.

[0009] In one embodiment, the duration of parent-child interaction in the video images is between 12 and 15 minutes.

[0010] In one embodiment, the plurality of characteristic behaviors include multiple of the following: spontaneous language or vocalization, pointing gestures, gestures and body language, eye contact, facial expressions directed toward others, request / demand behavior, social initiation behavior, initiation of shared attention behavior, social communication behavior, and behavior expressing pleasure in social interaction.

[0011] In one embodiment, the evaluation values ​​of the characteristic behaviors in the training set and the evaluation values ​​of the characteristic behaviors of the child are obtained by the rater using the same rating system.

[0012] In one embodiment, the display module is further configured to display the autism risk assessment results.

[0013] In one embodiment, the data acquisition module is further configured to acquire information parameters of the child; wherein, the information parameters include one or more of the following: age, gender, height, weight, chest circumference, and head circumference; the display module also displays the child's information parameters when displaying the scoring interface; and / or, The data acquisition module is also used to obtain information parameters of the parents and family income parameters; wherein, the information parameters include one or more of the following: age, gender, height, weight, chest circumference, and head circumference; the display module also displays the information parameters of the parents and family income parameters when displaying the scoring interface.

[0014] According to a second aspect, one embodiment provides a method for early screening of autism, the method comprising: Collect video images of parent-child interactions; the video images include: audio data and image data; Receive the evaluation value entered by the evaluator in the scoring column of the characteristic behavior based on the video image; The received evaluation values ​​of each characteristic behavior are input into a pre-trained early prediction model to obtain the autism risk assessment results output by the early prediction model.

[0015] In one embodiment, the multiple characteristic behaviors include multiple of the following: spontaneous language or vocalization, pointing gestures, gestures and body language, eye contact, facial expressions directed at others, request / demand behavior, social initiation behavior, initiation of shared attention behavior, social communication behavior, and behavior expressing pleasure in social interaction.

[0016] According to the autism early screening system and method described in the above embodiments, children are generally relaxed when playing with their parents. Therefore, the captured video images can accurately reflect the child's true state. This significantly improves the accuracy of assessors' (professionals) evaluation of the child's characteristic behaviors based on the captured video images. These evaluation values ​​are then input into a pre-trained early prediction model, which classifies the data to obtain an accurate assessment result—that is, whether the child is at risk of autism. Because the person playing with the child is changed from a professional to the parents, the assessment cost can be effectively controlled. Since professionals can remotely assess and score the child's characteristic behaviors in the video images and output the assessment results through the pre-trained early prediction model, timely and accurate medical assessments can be obtained even in areas with limited medical resources. Attached Figure Description

[0017] Figure 1 This is a structural block diagram of an early autism screening system according to one embodiment; Figure 2 This is a flowchart illustrating an early autism screening method according to one embodiment. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0019] Furthermore, the features, operations, or characteristics described in the specification can be combined in any suitable manner to form various embodiments. At the same time, the steps or actions in the method description can be rearranged or adjusted in a manner obvious to those skilled in the art. Therefore, the various orders in the specification and drawings are only for the clear description of a particular embodiment and do not imply a necessary order, unless otherwise stated that a particular order must be followed.

[0020] The serial numbers assigned to components in this document, such as "first" and "second," are used only to distinguish the described objects and have no sequential or technical meaning. The terms "connection" and "linkage" used in this invention, unless otherwise specified, include both direct and indirect connections (linkages).

[0021] Autism spectrum disorder, also known as autism, is a lifelong neurodevelopmental disorder that occurs in early childhood. Many autistic symptoms may persist throughout life, significantly impacting the quality of life for the autistic individual and their family members. Furthermore, the treatment and care of an autistic individual present a severe challenge to the entire family. Statistics show that the global prevalence of autism is approximately 1%, while in my country it is about 7%, making it the leading cause of mental disabilities among children in my country.

[0022] However, my country still faces a severe shortage of resources in early screening and diagnosis, early intervention, and related research for autism. Furthermore, current clinical practice remains heavily reliant on screening and diagnostic tools developed in Western countries. Because the scales and assessment criteria of these tools are based on Western populations, they do not adequately consider cultural differences between Chinese and Western children in areas such as language expression, social interaction patterns, and behavioral norms. Therefore, using these tools to screen Chinese children can lead to systemic bias. At the same time, the distribution of autism screening and diagnostic resources across my country is extremely uneven, further causing many children to miss the optimal intervention window.

[0023] Based on this, this application proposes an early autism screening system. After recording video content of parent-child interactions, an assessor (professional) observes, scores, and quantifies each characteristic behavior of the child based on the recorded video content. The scores of each characteristic behavior are then input into a pre-trained early prediction model, which directly outputs the result indicating whether the child is at risk of autism. The entire screening process can accurately and conveniently screen for children's symptoms without relying on screening and diagnostic tools developed in Western countries. Moreover, this screening process is also applicable to telemedicine scenarios, thereby breaking down the barriers of significantly uneven distribution of medical resources in my country.

[0024] Please refer to Figure 1 , Figure 1 This is a structural block diagram of an early autism screening system according to one embodiment. The system 10 includes: a data acquisition module 101, a display module 102, an input module 103, and a classification module 104.

[0025] The data acquisition module 101 is used to acquire video images of parent-child interactions; the video images include: audio data and image data.

[0026] It should be noted that the data acquisition module 101 may include a communication module for acquiring video images of parent-child interactions from external devices. The data acquisition module 101 may also include a camera for directly acquiring video images of parent-child interactions.

[0027] In some embodiments, video images of parent-child interaction include: video images of children and parents playing games; wherein the games include one or more of bubble games, cooking games, pretend games, and building games.

[0028] In practical applications, games that allow children and parents to interact can be bubble games, cooking games, pretend play, and building games; they can also be bubble games, cooking games, and pretend play; they can also be bubble games and cooking games; or they can be bubble games, cooking games, pretend play, or building games.

[0029] In some embodiments, the duration of parent-child interaction in the video images is between 12 and 15 minutes. It should be noted that the duration of parent-child interaction can be 12 minutes, 13 minutes, 14 minutes, or 15 minutes.

[0030] In practical applications, when children and parents interact with bubble games, cooking games, pretend play, and building games, the interaction time for bubble games can be 2-3 minutes, for cooking games it can be 4-5 minutes, and for pretend play it can be 6-7 minutes.

[0031] In practical applications, when children and parents interact with bubble games, cooking games, and pretend play, the interaction time for bubble games can be 4-5 minutes, for cooking games it can be 3-4 minutes, and for pretend play it can be 5-6 minutes.

[0032] In practical applications, when children and parents interact with bubble games, cooking games, and building games, the interaction time for bubble games can be 5-6 minutes, for cooking games it can be 4-5 minutes, and for building games it can be 3-4 minutes.

[0033] In practical applications, when children and parents interact through building games, cooking games, and pretend play, the interaction time for building games can be 3-4 minutes, for cooking games it can be 4-5 minutes, and for pretend play it is 5-6 minutes.

[0034] In practical applications, when children and parents interact with bubble games, pretend play, and building games, the interaction time for bubble games can be 5-6 minutes, the interaction time for pretend play can be 3-4 minutes, and the interaction time for building games can be 4-5 minutes.

[0035] In practical applications, when children and parents interact with each other through bubble games and cooking games, the interaction time for bubble games can be 5-6 minutes, and for cooking games, it can be 7-9 minutes. When children and parents interact through pretend play and building games, the interaction time for pretend play can be 6-7 minutes, and for building games, it can be 6-8 minutes. Similarly, when children and parents interact through bubble games and pretend play, the interaction time for bubble games can be 7-8 minutes, and for pretend play, it can be 5-7 minutes. When children and parents interact through bubble games and building games, the interaction time for bubble games can be 8-9 minutes, and for building games, it can be 4-6 minutes. When children and parents interact through cooking games and pretend play, the interaction time for cooking games can be 5-6 minutes, and for pretend play, it can be 7-9 minutes. Finally, when children and parents interact through cooking games and building games, the interaction time for cooking games can be 7-9 minutes, and for building games, it can be 5-6 minutes.

[0036] In some embodiments, the data acquisition module 101 is further configured to acquire information parameters of the child; wherein the information parameters of the child include one or more of age, gender, height, weight, chest circumference, and head circumference; and / or, the data acquisition module 101 is further configured to acquire information parameters of the parents and family income parameters; wherein the information parameters of the parents include one or more of age, gender, height, weight, chest circumference, and head circumference.

[0037] In practical applications, in order to improve the accuracy of the screening system, the data acquisition module 101 can acquire information parameters of the child, parents and family income parameters while acquiring voice data and image data; it can also acquire information parameters of the child while acquiring voice data and image data; and it can also acquire information parameters of the parents and family income parameters while acquiring voice data and image data.

[0038] It should be noted that when obtaining information parameters about children, one can obtain the child's age, gender, height, weight, chest circumference, and head circumference.

[0039] It should be noted that when obtaining information parameters for a child, it is also possible to obtain five of the total information parameters. For example, obtaining the child's age, gender, height, weight, and chest circumference; or obtaining the child's age, gender, height, weight, and head circumference; or obtaining the child's age, gender, height, chest circumference, and head circumference; or obtaining the child's age, gender, weight, chest circumference, and head circumference; or obtaining the child's age, height, weight, chest circumference, and head circumference. Of course, the specific parameter selection can also be adjusted according to the actual situation.

[0040] It should be noted that when obtaining information parameters for a child, it is also possible to obtain four of the total information parameters. For example, obtaining the child's age, gender, height, and weight; obtaining the child's age, gender, height, and chest circumference; obtaining the child's age, gender, height, and head circumference; obtaining the child's age, height, weight, and chest circumference; obtaining the child's gender, height, weight, and chest circumference; obtaining the child's age, gender, height, and head circumference; of course, the specific parameter selection can also be adjusted according to the actual situation.

[0041] It should be noted that when obtaining information parameters for a child, it is also possible to obtain three of the total information parameters. For example, obtaining the child's age, gender, and height; obtaining the child's age, gender, and weight; obtaining the child's age, gender, and chest circumference; obtaining the child's age, gender, and head circumference; obtaining the child's age, height, and weight; obtaining the child's age, height, and chest circumference; of course, the specific parameter selection can also be adjusted according to the actual situation.

[0042] It should be noted that when obtaining information parameters for a child, it is also possible to obtain two of the total information parameters. For example, obtaining the child's age and gender; obtaining the child's age and height; obtaining the child's age and weight; obtaining the child's age and chest circumference; obtaining the child's age and head circumference; obtaining the child's height and weight. Of course, the specific parameter selection can also be adjusted according to the actual situation.

[0043] It should be noted that when obtaining information parameters about children, one can also obtain the child's age, gender, height, weight, chest circumference, or head circumference.

[0044] It should be noted that when obtaining information parameters about parents, you can obtain their age, gender, height, weight, chest circumference, and head circumference.

[0045] It should be noted that when obtaining information parameters from parents, it is also possible to obtain five of the total information parameters. For example, obtaining the parents' age, gender, height, weight, and chest circumference; or obtaining the parents' age, gender, height, weight, and head circumference; or obtaining the parents' age, gender, height, chest circumference, and head circumference; or obtaining the parents' age, gender, weight, chest circumference, and head circumference; or obtaining the parents' age, height, weight, chest circumference, and head circumference. Of course, the specific parameter selection can also be adjusted according to the actual situation.

[0046] It should be noted that when obtaining information parameters from parents, it is also possible to obtain four of the total information parameters. For example, obtaining the parents' age, gender, height, and weight; obtaining the parents' age, gender, height, and chest circumference; obtaining the parents' age, gender, height, and head circumference; obtaining the parents' age, height, weight, and chest circumference; obtaining the parents' gender, height, weight, and chest circumference; obtaining the parents' age, gender, height, and head circumference; of course, the specific parameter selection can also be adjusted according to the actual situation.

[0047] It should be noted that when obtaining information parameters about parents, it is also possible to obtain three of the total information parameters. For example, obtaining the parents' age, gender, and height; obtaining the parents' age, gender, and weight; obtaining the parents' age, gender, and chest circumference; obtaining the parents' age, gender, and head circumference; obtaining the parents' age, height, and weight; obtaining the parents' age, height, and chest circumference; of course, the specific parameter selection can also be adjusted according to the actual situation.

[0048] It should be noted that when retrieving parents' information parameters, it is also possible to retrieve two of the total information parameters. For example, retrieving parents' age and gender; retrieving parents' age and height; retrieving parents' age and weight; retrieving parents' age and chest circumference; retrieving parents' age and head circumference; retrieving parents' height and weight. Of course, the specific parameter selection can also be adjusted according to the actual situation.

[0049] It should be noted that when obtaining information parameters about parents, you can also obtain their age, gender, height, weight, chest circumference, or head circumference.

[0050] Display module 102 is used to display the rating interface, which includes: video images, multiple characteristic behaviors of the child, and rating columns for the characteristic behaviors.

[0051] It should be noted that the display module 102 can be an electronic display, such as a touch screen or an electronic screen with buttons.

[0052] It should be noted that the rating column can be an input box, allowing the assessor (professional) to enter the score corresponding to each of the child's characteristic behaviors, i.e., the assessment value. Alternatively, the rating column can be an option box with multiple scores, allowing the assessor (professional) to select the score corresponding to each of the child's characteristic behaviors, i.e., the assessment value.

[0053] In some embodiments, the multiple characteristic behaviors include multiple of the following: spontaneous language or vocalization, pointing gestures, gestures and body language, eye contact, facial expressions directed toward others, request / demand behavior, social initiation behavior, initiation of shared attention behavior, social communication behavior, and behavior expressing pleasure in social interaction.

[0054] It should be noted that spontaneous language or vocalization refers to a child's behavior of speaking to or talking to their parents during interactive play.

[0055] It should be noted that pointing refers to the behavior of a child pointing to an object during interactive play with their parents.

[0056] It should be noted that gestures and body language represent behaviors such as clapping and gesturing with hands during interactive games with parents.

[0057] It should be noted that eye contact behavior refers to the behavior of children making eye contact and interacting with their parents during play.

[0058] It should be noted that facial expressions directed towards others represent the facial expressions and actions of children when facing their parents during interactive games, such as smiling, frustrated, sad, calm, and fearful facial expressions.

[0059] It should be noted that demand / request behavior represents a child's expressive behavior of asking or requesting their parents to do something during interactive play.

[0060] It should be noted that social initiation behavior refers to children's proactive behavior in initiating interactions with their parents during play.

[0061] It should be noted that initiating joint attention behavior refers to the behavior of children that attracts the attention of their parents during interactive games.

[0062] It should be noted that social communication behavior refers to the behavior of children communicating with their parents during interactive games.

[0063] It should be noted that the expression of pleasure in social interaction refers to the behavior of children expressing happiness during interactive games with their parents.

[0064] In some embodiments, the display module 102 displays not only the data collected by the data acquisition module 101, but also the data obtained by the data acquisition module 101. For example, when the data acquisition module 101 acquires the child's information parameters, the display module 102 will display the child's information parameters on the rating display interface. When the data acquisition module 101 acquires the parents' information parameters and family income parameters, the display module 102 will display the parents' information parameters and family income parameters on the rating display interface. When the data acquisition module 101 acquires the child's information parameters, the parents' information parameters, and the family income parameters, the display module 102 will display the child's information parameters, the parents' information parameters, and the family income parameters on the rating display interface. This enriches the diversity of the data.

[0065] The input module 103 is used to receive the evaluation value entered by the evaluator in the scoring column of the characteristic behavior based on the video image.

[0066] It should be noted that the assessors enter evaluation values ​​for multiple characteristic behaviors of the child in the rating column. That is, the assessors enter evaluation values ​​for the child's spontaneous language or vocalization, pointing gestures, hand gestures and body language, eye contact, facial expressions directed at others, requests / demands, social initiation, initiation of shared attention, social communication, and expressions of pleasure in social interactions. Of course, assessors may also enter evaluation values ​​for only some of the child's characteristic behaviors in the rating column.

[0067] The classification module 104 is used to input the evaluation values ​​of each feature behavior received by the input module 103 into the pre-trained early prediction model to obtain the autism risk assessment results output by the early prediction model.

[0068] It should be noted that the autism risk assessment results output by the early predictive models are either a result indicating that the child has autism risk or a result indicating that the child has no autism risk.

[0069] In some embodiments, after the early prediction model outputs an autism risk assessment result, the display module 102 can display the output autism risk assessment result. For example, the display module 102 may show that the child has autism risk, or it may show that the child has no autism risk.

[0070] In some embodiments, the early prediction model is trained as follows: Select a portion of the sample set from the low-risk, medium-risk, and high-risk sample sets as the validation set, and use the remaining sample set as the training set; The training set is input into the deep learning model to train the deep learning model and obtain the initial early prediction model. The validation set is input into the deep learning model to validate the model and obtain its confidence score. If the confidence score is lower than the preset confidence score, the model parameters are adjusted and the model is trained again using the training set until the confidence score is higher than the preset confidence score.

[0071] In practical applications, 20% of the sample set can be used as the validation set, and the remaining 80% can be used as the training set. That is to say, the validation set contains low-risk, medium-risk, and high-risk sample sets; the training set contains low-risk, medium-risk, and high-risk sample sets.

[0072] It should be noted that both the validation set and the training set contain assessment values ​​for multiple characteristic behaviors of multiple children, as well as labels indicating whether the children are at risk of autism.

[0073] It should be noted that the deep learning model can be a convolutional neural network, for example, the convolutional neural network can be a YOLOv8 network.

[0074] In some embodiments, the evaluation values ​​of the characteristic behaviors in the training set and the evaluation values ​​of the child's characteristic behaviors are obtained by the rater using the same rating system.

[0075] It should be noted that each characteristic behavior has a corresponding scoring system.

[0076] The following explanation uses a scoring system for eye contact behavior as an example.

[0077] There is no eye contact of any kind, and they often actively avoid eye contact, such as by turning away, pushing away, or closing their eyes. 6 points. Occasionally, they use ambiguous or inappropriate eye contact to initiate, terminate, or adjust social interactions, so I'd give them a 5 out of 10. Frequently uses ambiguous or inappropriate eye contact to initiate, terminate, or regulate social interactions, score 4. There is occasional direct eye contact, which is quite appropriate; however, this eye contact lacks flexibility and subtle changes, so I would give it a score of 3. There is frequent and appropriate direct eye contact, but this eye contact lacks flexibility and subtle changes, so I would give it a score of 2. Occasional and appropriate eye contact, with subtle changes in the eyes, blending in with gestures and body language, earns 1 point; Appropriate and subtle eye contact, integrated with gestures and body language, score 0.

[0078] This application improves the accuracy of assessments by refining the scoring system for each characteristic behavior to accurately capture the subtle differences in children's characteristic behaviors.

[0079] This application provides an early autism screening system, comprising: a data acquisition module for acquiring video images of parent-child interactions; the video images include: audio data and image data; a display module for displaying a scoring interface, which includes: video images, multiple characteristic behaviors of the child, and scoring columns for the characteristic behaviors; an input module for receiving assessment values ​​entered by assessors in the scoring columns for the characteristic behaviors based on the video images; and a classification module for inputting the assessment values ​​of each characteristic behavior received by the input module into a pre-trained early prediction model to obtain an autism risk assessment result output by the early prediction model. When parents and children play together, the child is relatively relaxed, so the acquired video images can accurately reflect the child's true state. This greatly improves the accuracy of assessors' (i.e., professionals) assessments of the child's characteristic behaviors based on the acquired video images. After inputting these assessment values ​​into the pre-trained early prediction model, the model classifies the data to obtain an accurate assessment result, i.e., whether the child is at risk of autism. Because the people who play with the children in the early stages are replaced by parents instead of professionals, the assessment costs can be effectively controlled; because professionals can remotely assess and score the children's characteristic behaviors in video images and output assessment results through pre-trained early prediction models, timely and accurate medical assessments can be obtained even if the children's area has relatively scarce medical resources.

[0080] Please refer to Figure 2 , Figure 2 This is a flowchart illustrating an early autism screening method according to one embodiment. The method includes steps S201-S203.

[0081] Step S201: Collect video images of parent-child interaction; the video images include: audio data and image data.

[0082] Step S202: Receive the evaluation value entered by the evaluator in the scoring column of the characteristic behavior based on the video image.

[0083] Step S203: Input the received evaluation values ​​of each characteristic behavior into the pre-trained early prediction model to obtain the autism risk assessment results output by the early prediction model.

[0084] In some embodiments, the multiple characteristic behaviors include multiple of the following: spontaneous language or vocalization, pointing gestures, gestures and body language, eye contact, facial expressions directed toward others, request / demand behavior, social initiation behavior, initiation of shared attention behavior, social communication behavior, and behavior expressing pleasure in social interaction.

[0085] This application provides a method for early screening of autism. The method includes acquiring video images of parent-child interactions; receiving assessment values ​​input by assessors in the scoring columns of characteristic behaviors based on the video images and a scoring system; and inputting the received assessment values ​​of each characteristic behavior into a pre-trained early prediction model to obtain an autism risk assessment result output by the early prediction model. When parents play with their children, the children are relatively relaxed, so the acquired video images can accurately reflect the child's true state. This greatly improves the accuracy of assessors' (i.e., professionals) assessments of the child's characteristic behaviors based on the acquired video images. After inputting these assessment values ​​into the pre-trained early prediction model, the model classifies and obtains an accurate assessment result, i.e., whether the child is at risk of autism. Because the person playing with the child in the early stages is replaced by the parents instead of professionals, the assessment cost can be effectively controlled. Since professionals can remotely assess and score the child's characteristic behaviors in the video images and output the assessment result through the pre-trained early prediction model, timely and accurate medical assessments can be obtained even in areas with limited medical resources.

[0086] This document describes various exemplary embodiments with reference to them. However, those skilled in the art will recognize that changes and modifications can be made to the exemplary embodiments without departing from the scope of this document. For example, various operational steps and components for performing operational steps can be implemented in different ways depending on the specific application or considering any number of cost functions associated with the operation of the system (e.g., one or more steps can be deleted, modified, or combined with other steps).

[0087] While the principles herein have been illustrated in various embodiments, numerous modifications to the structures, arrangements, proportions, elements, materials, and components, particularly suited to specific environments and operational requirements, may be used without departing from the principles and scope of this disclosure. These modifications and other alterations or alterations will be included within the scope of this document. Those skilled in the art will recognize that many changes can be made to the details of the above embodiments without departing from the fundamental principles of the invention.

Claims

1. An early screening system for autism, characterized in that, The system includes: The data acquisition module is used to acquire video images of parent-child interactions; the video images include: audio data and image data; The display module is used to display the scoring interface, which includes: the video image, multiple characteristic behaviors of the child, and a scoring column for the characteristic behaviors; The input module is used to receive the evaluation value entered by the evaluator in the scoring column of the characteristic behavior based on the video image; The classification module is used to input the evaluation values ​​of each feature behavior received by the input module into the pre-trained early prediction model to obtain the autism risk assessment result output by the early prediction model.

2. The early screening system as described in claim 1, characterized in that, The early prediction model was trained in the following manner: Select a portion of the sample set from the low-risk, medium-risk, and high-risk sample sets as the validation set, and use the remaining sample set as the training set; The training set is input into the deep learning model, and the deep learning model is trained to obtain an initial early prediction model; The validation set is input into the deep learning model to validate the deep learning model and obtain the confidence level of the deep learning model. When the confidence level is lower than the preset confidence level, the model parameters of the deep learning model are adjusted, and training continues with the training set until the confidence level is higher than the preset confidence level.

3. The early screening system as described in claim 1, characterized in that, The video images of parent-child interaction include: video images of children and parents playing games; wherein the games include one or more of bubble games, cooking games, pretend games, and building games.

4. The early screening system as described in claim 1 or 3, characterized in that, The duration of parent-child interaction in the video images is between 12 and 15 minutes.

5. The early screening system as described in claim 1, characterized in that, The aforementioned multiple characteristic behaviors include: spontaneous language or vocalization, pointing gestures, gestures and body language, eye contact, facial expressions directed at others, request / request behavior, social initiation behavior, initiation of shared attention behavior, social communication behavior, and behaviors expressing pleasure in social interactions.

6. The early screening system as described in claim 2, characterized in that, The evaluation values ​​of the characteristic behaviors in the training set and the evaluation values ​​of the characteristic behaviors of the children are obtained by the raters using the same rating system.

7. The early screening system as described in claim 1, characterized in that, The display module is also used to display the autism risk assessment results.

8. The early screening system as described in claim 1, characterized in that, The data acquisition module is also used to acquire information parameters of the child; wherein, the information parameters include one or more of the following: age, gender, height, weight, chest circumference, and head circumference; the display module also displays the child's information parameters when displaying the scoring interface; and / or, The data acquisition module is also used to obtain information parameters of the parents and family income parameters; wherein, the information parameters include one or more of the following: age, gender, height, weight, chest circumference, and head circumference; the display module also displays the information parameters of the parents and family income parameters when displaying the scoring interface.

9. A method for early screening of autism, characterized in that, The method includes: Collect video images of parent-child interactions; the video images include: audio data and image data; Receive the evaluation value entered by the evaluator in the scoring column of the characteristic behavior based on the video image; The received evaluation values ​​of each characteristic behavior are input into a pre-trained early prediction model to obtain the autism risk assessment results output by the early prediction model.

10. The early screening method as described in claim 9, characterized in that, Multiple characteristic behaviors include: spontaneous language or vocalization, pointing gestures, gestures and body language, eye contact, facial expressions directed at others, request / request behavior, social initiation behavior, initiation of shared attention behavior, social communication behavior, and behaviors expressing pleasure in social interactions.

Images