Autism behavior modeling and dynamic optimization system based on digital twinning and AI

Abstract

The invention discloses an autism behavior modeling and dynamic optimization system based on digital twinning and AI, and relates to the technical field of intelligent digital diagnosis and treatment. According to the technical scheme, the digital twinborn model for reflecting the state of the autism patient in real time is constructed by collecting the multi-modal data of the autism patient, and the learning corpus is collected for training to obtain the generative AI evaluation model for evaluating the state of the autism patient; and importing digital twinborn model data of a to-be-evaluated autism patient into the generative AI evaluation model, generating an evaluation result of the current state of the patient, further generating a personalized diagnosis and treatment scheme suitable for the patient, and applying intervention measures in the diagnosis and treatment scheme to the to-be-evaluated autism patient, so that the evaluation result of the current state of the patient is obtained. And obtaining reactions of a patient under different intervention measures by using the constructed digital twinborn model, classifying the intervention measures, and feeding back and importing the intervention measures into a generative AI evaluation model to optimize a formulated personalized diagnosis and treatment scheme.

Description

Technical Field

[0001] The present invention relates to the field of intelligent digital diagnosis and treatment technology, and in particular to an autism behavior modeling and dynamic optimization system based on digital twins and AI. Background Art

[0002] Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder whose core symptoms include social communication impairments, restricted interests, and repetitive behaviors. Currently, treatments for autism primarily include behavioral therapy, language training, and social skills training. However, these approaches often lack personalization, making it difficult to precisely address each patient's unique needs. Furthermore, traditional diagnosis and treatment methods rely on the experience and subjective judgment of professional physicians, resulting in inefficiency and uneven resource allocation.

[0003] Digital twin technology offers new insights into the modeling and simulation of complex systems by creating real-time mapping and interaction between virtual models and physical entities. In the medical field, digital twin technology can be used to construct virtual models of patients, enabling comprehensive monitoring and analysis of their multimodal information, including physiological and psychological characteristics. However, the application of digital twin technology in autism diagnosis and treatment is still in its early stages of exploration, and mature methods for generating personalized treatment plans have yet to be established. Generative AI technology has made significant progress in recent years, and its application in the medical field is gaining increasing attention. Generative AI can generate new content, such as text, images, and audio, from large amounts of data, offering new possibilities for developing personalized treatment plans. For example, Professor Duan Xujun's team at the University of Electronic Science and Technology of China, using personalized brain imaging analysis technology, has used continuous theta burst stimulation (cTBS) to precisely target the amygdala and its neural circuits in children with autism, achieving significant therapeutic benefits. However, current applications of generative AI in autism diagnosis and treatment focus on single interventions, lacking a comprehensive consideration of the patient's overall condition.

[0004] In summary, although there are some AI-based digital diagnosis and treatment systems for autism in the existing technology, such as the "Autism Spectrum Disorder Screening and Auxiliary Diagnosis System" released by Tsinghua University, the system can generate early risk screening reports through AI detection and comparison technology of faces and human behaviors in images, audio and video. However, these systems mainly focus on early screening and auxiliary diagnosis, and lack support for the generation and optimization of personalized diagnosis and treatment plans. In particular, the existing technology has many shortcomings in the personalized diagnosis and treatment of autism, and there is an urgent need for a new method that can comprehensively use digital twins and generative AI technologies to achieve comprehensive assessment of autistic patients and generate personalized diagnosis and treatment plans. To this end, an autism behavior modeling and dynamic optimization system based on digital twins and AI is proposed. Summary of the Invention

[0005] The main purpose of the present invention is to provide an autism behavior modeling and dynamic optimization system based on digital twins and AI, which can effectively solve the problems in the background technology.

[0006] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0007] The autism behavior modeling and dynamic optimization system based on digital twins and AI includes the following steps:

[0008] Step 1: Collect multimodal data of autistic patients, including individual characteristic data, behavioral data, physiological data, psychological assessment data, and family environment data, and perform preprocessing operations including cleaning and normalization on the collected data;

[0009] Step 2: Using the pre-processed multimodal data, a digital twin model is constructed to reflect the status of the autism patient in real time. The digital twin model is also used to simulate the patient's response to different intervention measures.

[0010] Step 3: Collect learning data and use the learning data to train a generative AI assessment model for autism status assessment. The learning data includes the assessment conclusions of past senior supervisors and the industry standards used as the basis for the assessment, as well as the diagnosis and treatment plans developed in accordance with the industry standards based on the assessment conclusions.

[0011] Step 4: Importing the digital twin model data of the autism patient to be evaluated into the generative AI evaluation model, and using the trained generative AI evaluation model to generate an evaluation result of the current status of the autism patient to be evaluated;

[0012] Step 5: Based on the obtained assessment results, use the trained generative AI assessment model to generate a personalized diagnosis and treatment plan suitable for the autistic patient to be assessed. The personalized diagnosis and treatment plan is an intervention measure that combines one or more of behavioral therapy, language training, and social skills training.

[0013] Autism behavior modeling and dynamic optimization system based on digital twins and AI, including:

[0014] Multimodal data acquisition module, used to collect multimodal data of autistic patients, including individual characteristic data, behavioral data, physiological data, psychological assessment data, and family environment data, and perform preprocessing operations on the collected data, including cleaning and normalization;

[0015] A digital twin model construction module uses preprocessed multimodal data to construct a digital twin model that reflects the status of autistic patients in real time. The digital twin model is also used to simulate the patient's response to different intervention measures.

[0016] A learning corpus acquisition module is used to collect learning corpus, wherein the learning corpus includes the evaluation conclusions of past senior supervisors and the industry standards based on which the evaluations were made, as well as the diagnosis and treatment plans formulated in accordance with the industry standards based on the evaluation conclusions;

[0017] An evaluation model construction module, for using the learning corpus to perform training to obtain a generative AI evaluation model for assessing the status of autism patients;

[0018] An evaluation result generation module, configured to import the digital twin model data of the autism patient to be evaluated into the generative AI evaluation model, and generate an evaluation result of the current status of the autism patient to be evaluated using the trained generative AI evaluation model;

[0019] A diagnosis and treatment plan generation module is used to generate a personalized diagnosis and treatment plan suitable for the autistic patient to be evaluated based on the obtained evaluation results using the trained generative AI evaluation model, wherein the personalized diagnosis and treatment plan is an intervention measure that combines one or more of behavioral therapy, language training and social skills training.

[0020] The system further comprises:

[0021] A data update module is used to dynamically update the digital twin model and the learning corpus, and use the updated data to retrain and optimize the generative AI evaluation model.

[0022] The system further comprises:

[0023] An intervention effect acquisition module is used to apply the intervention measures in the obtained personalized diagnosis and treatment plan to the autism patient to be evaluated, and use the constructed digital twin model to obtain the patient's response to different intervention measures;

[0024] An intervention effectiveness assessment module is used to classify interventions administered to autistic patients as interventions with positive effects, interventions with negative effects, and interventions with no clear effect, based on the patients' responses;

[0025] The treatment plan optimization module is used to import the classification results of intervention terms into the generative AI evaluation model to optimize the formulated personalized treatment plan.

[0026] The system further includes a memory, a processor, and a computer program stored in the memory and executable on the processor.

[0027] Furthermore, the method also includes: dynamically updating the digital twin model and the learning corpus, and using the updated data to retrain and optimize the generative AI evaluation model.

[0028] Furthermore, the method further includes: applying the intervention measures in the obtained personalized diagnosis and treatment plan to the autism patient to be evaluated, and using the constructed digital twin model to obtain the patient's response to different intervention measures;

[0029] The interventions administered to the autism patients were categorized as those with positive effects, those with negative effects, and those with no clear effect, based on the patients' responses.

[0030] The classification results of intervention wording are imported into the generative AI evaluation model to optimize the personalized diagnosis and treatment plan.

[0031] Furthermore, in step 2, the process of building the digital twin model includes the following steps:

[0032] Classify and integrate the pre-processed multimodal data to obtain the patient's behavioral data set, physiological data set, psychological assessment data set, and home environment data set.

[0033] Using a standardized behavioral assessment scale to evaluate the patient's behavioral data, obtain the patient's social interaction and behavioral performance, as well as the patient's repetitive behavior assessment results, wherein the repetitive behaviors include stereotyped behaviors, compulsive behaviors, and ritualistic behaviors;

[0034] Real-time collection of physiological data from patients to reflect their physiological state and emotional response, analysis of the data using machine learning algorithms, and identification of physiological characteristics associated with autism;

[0035] Use standardized cognitive ability testing tools to obtain assessment results of patients' cognitive level and intellectual development, as well as emotional state and behavioral problems;

[0036] Questionnaires are used to understand the patient's family environment, including family members' education level, family atmosphere, and parent-child interaction. Video recordings or on-site observations are used to analyze the interaction patterns between the patient and family members and assess the effectiveness of the family support system.

[0037] The acquired multimodal data are fused to build a digital twin model of the patient.

[0038] Furthermore, the principles for optimizing the formulation of personalized diagnosis and treatment plans are as follows:

[0039] Reduce the frequency of intervention wording that has little effect;

[0040] Avoid generating interventions that have negative consequences;

[0041] Increase the frequency of interventions that have a positive impact.

[0042] Furthermore, the intervention measures with positive effects are intervention measures applied when the patient experiences one of the following conditions, wherein the condition one includes at least one of improved social communication skills, reduced stereotyped behaviors, expanded interests, improved adaptive skills, improved self-care ability, improved learning ability, reduced behavioral problems, and improved emotional regulation ability;

[0043] The intervention measures with negative effects are intervention measures applied when the patient experiences the following two conditions, wherein the two conditions include at least one of reduced social communication ability, increased stereotyped behavior, narrowed interest range, reduced adaptive skills, reduced self-care ability, reduced learning ability, increased behavioral problems, and reduced emotional regulation ability;

[0044] The intervention wording of the insignificant effect is the intervention measure applied when both the first and second situations occur.

[0045] The present invention has the following beneficial effects:

[0046] Compared with the existing technology, by collecting multimodal data of autistic patients, including individual characteristic data, behavioral data, physiological data, psychological assessment data and family environment data, the collected data is preprocessed including cleaning and normalization, and the preprocessed multimodal data is used to build a digital twin model for reflecting the status of autistic patients in real time, collect learning corpus, use the learning corpus to train to obtain a generative AI assessment model for autistic patient status assessment, import the digital twin model data of the autistic patient to be assessed into the generative AI assessment model, use the trained generative AI assessment model to generate an assessment result of the current status of the autistic patient to be assessed, and based on the obtained assessment result, use the trained generative AI assessment model to generate an individualized model suitable for the autistic patient to be assessed. The method comprises the following steps: a) developing a personalized diagnosis and treatment plan, applying the obtained intervention measures in the personalized diagnosis and treatment plan to the autistic patients to be evaluated; b) using the constructed digital twin model to obtain the patients' responses under different intervention measures; c) classifying the intervention measures applied to the autistic patients to be evaluated into intervention measures with positive effects, intervention measures with negative effects, and intervention wordings with insignificant effects according to the patients' responses; c) importing the classification results of the intervention wordings into the generative AI evaluation model to optimize the formulated personalized diagnosis and treatment plan; d) dynamically updating the digital twin model and the learning corpus; and retraining and optimizing the generative AI evaluation model with the updated data; and comprehensively utilizing digital twin and generative AI technologies to achieve comprehensive evaluation of autistic patients and formulation of personalized diagnosis and treatment plans, which is helpful to improve the status of autistic patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1This is a schematic diagram of the overall structure of the autism behavior modeling and dynamic optimization system based on digital twins and AI in the present invention;

[0048] Figure 2 This is a schematic diagram of the structure of the autism behavior modeling and dynamic optimization system based on digital twins and AI of the present invention;

[0049] Figure 3 Schematic diagram of the optimization process of the personalized diagnosis and treatment plan in the technical solution of the present invention. DETAILED DESCRIPTION

[0050] The present invention will be further described below in conjunction with specific embodiments. The accompanying drawings are for illustrative purposes only and represent only schematic diagrams rather than actual drawings. They should not be understood as limiting the present invention. In order to better illustrate the specific embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product.

[0051] The specific implementation process of the technical solution of the present invention includes the following steps:

[0052] Step 1: Collect multimodal data of autistic patients, including individual characteristic data, behavioral data, physiological data, psychological assessment data, and family environment data, and perform preprocessing operations including cleaning and normalization on the collected data.

[0053] It should be noted that behavioral data may include social interaction behaviors, stereotyped behaviors, etc.; physiological data may include brain waves, heart rate, etc.; psychological assessment data may include cognitive ability test results, emotional state assessment, etc.; individual characteristic data may include data such as age, symptom severity, development level, etc.

[0054] Step 2: Use the preprocessed multimodal data to build a digital twin model that reflects the status of autism patients in real time. The digital twin model is also used to simulate the patient's response to different intervention measures. Specifically, the process of building the digital twin model includes the following steps:

[0055] Classify and integrate the pre-processed multimodal data to obtain the patient's behavioral data set, physiological data set, psychological assessment data set, and home environment data set.

[0056] Use standardized behavioral assessment scales to evaluate the patient's behavioral data, including social interactions and behavioral performance, as well as the patient's repetitive behavior assessment results. Repetitive behaviors include stereotyped behaviors, compulsive behaviors, and ritualistic behaviors.

[0057] Real-time collection of physiological data from patients to reflect their physiological state and emotional response, analysis of the data using machine learning algorithms, and identification of physiological characteristics associated with autism;

[0058] Use standardized cognitive ability testing tools to obtain assessment results of patients' cognitive level and intellectual development, as well as emotional state and behavioral problems;

[0059] Questionnaires are used to understand the patient's family environment, including family members' education level, family atmosphere, and parent-child interaction. Video recordings or on-site observations are used to analyze the interaction patterns between the patient and family members and assess the effectiveness of the family support system.

[0060] The acquired multimodal data are fused to build a digital twin model of the patient.

[0061] The behavioral data evaluation method is as follows:

[0062] Standardized scale assessment: Use standardized behavioral assessment scales, such as the Autism Diagnostic Observation Scale (ADOS) and the Autism Diagnostic Interview Rating-Revised (ADIR). The ADOS observes the patient's social interactions and behavioral performance through a series of standardized activities, while the ADIR obtains detailed behavioral information from the patient through interviews with parents or caregivers.

[0063] Repetitive Behavior Scale Assessment: The Revised Version of the Repetitive Behavior Scale (RBSR) is used to assess the patient's repetitive behaviors, including stereotyped behaviors, compulsive behaviors, ritualistic behaviors, etc. This scale assesses the patient's daily behavior through reports from parents or caregivers.

[0064] The physiological data evaluation method is as follows:

[0065] Physiological monitoring equipment: Wearable devices or physiological monitoring instruments, such as electroencephalogram (EEG) and heart rate variability (HRV) monitors, are used to collect patients' physiological data in real time. This data can reflect the patient's physiological state and emotional response.

[0066] Data Analysis: Machine learning algorithms analyze physiological data to identify physiological characteristics associated with autism. For example, EEG data can be used to analyze brain activity patterns, while HRV data can reflect the patient's autonomic nervous system function.

[0067] The psychological assessment data method is as follows:

[0068] Cognitive ability testing: Standardized cognitive ability testing tools, such as the Wechsler Intelligence Scale for Children (WISC) or the Draw a Person test (DAP), are used to assess the patient's cognitive level and intellectual development.

[0069] Emotional status assessment: Emotional assessment scales, such as the Children's Behavior and Emotion Checklist (CBCL), are used to assess the patient's emotional state and behavioral problems.

[0070] The home environment data method is as follows:

[0071] Family Environment Questionnaire: This questionnaire is used to understand the patient's family environment, including family members' education level, family atmosphere, parent-child interaction, etc. These factors have a significant impact on the patient's recovery and behavioral performance.

[0072] Family interaction observation: Through video recording or on-site observation, the interaction patterns between patients and family members are analyzed to evaluate the effectiveness of the family support system.

[0073] Comprehensive Assessment and Digital Twin Model

[0074] Data fusion: Multimodal data is integrated to build a digital twin model of the patient. This model can reflect the patient's condition in real time and simulate the patient's response to different interventions.

[0075] Generative AI-assisted assessment: Generative AI technology is used to analyze digital twin models and generate personalized assessment reports. AI can quickly generate detailed assessment results based on individual patient characteristics, such as age, symptom severity, and developmental level.

[0076] Dynamic monitoring and feedback

[0077] Real-time monitoring: Wearable devices and mobile apps monitor patients’ training progress and physiological status in real time. Generative AI can use this data to adjust treatment plans and provide feedback to patients and parents.

[0078] Regular evaluation: Conduct comprehensive evaluations of patients regularly and update the digital twin model to ensure the continued effectiveness and personalization of diagnosis and treatment plans.

[0079] The steps of multimodal data fusion are as follows:

[0080] Data preprocessing

[0081] Data collection: Acquire multimodal data from different sources, such as through API interfaces, sensors, file uploads, etc.

[0082] Data cleaning: remove noise, fill missing values, unify timestamps, etc. to ensure data quality.

[0083] Standardization processing: Convert data in different formats into a unified standard format to facilitate subsequent processing.

[0084] Feature extraction

[0085] Image data: Use convolutional neural networks (CNN) to extract image features.

[0086] Text data: Use pre-trained language models (such as the Transformer architecture) to extract text features.

[0087] Audio data: Use long short-term memory networks (LSTMs) or other sequence models to process audio data.

[0088] Data alignment

[0089] Alignment operation: Establishing correspondence between data of different modalities, such as image-text alignment, speech-text alignment, etc.

[0090] Similarity metrics: Calculate the similarity between data of different modalities, such as cosine similarity, to quantify the relationship between them.

[0091] Feature fusion

[0092] Feature-level fusion: Before multimodal data is fed into the model, features from different modalities are fused together. Common methods include concatenation, addition, multiplication, and bilinear fusion.

[0093] Model-level fusion: The feature information of different modalities is integrated through models such as neural networks to achieve cross-modal information interaction.

[0094] Decision-level fusion: Fusion of the independent decision results of each modality, such as voting method, weighted average method, etc.

[0095] Fusion method selection

[0096] Joint fusion method: Features of different modalities are projected into a shared semantic subspace for fusion.

[0097] Collaborative fusion method: Integrate information from different modalities through cross-modal similarity method or hierarchical spatial fusion method.

[0098] Large model-based fusion: Use a large multimodal model (such as BLIP2) for end-to-end fusion to directly convert the input multimodal data into a unified representation.

[0099] Step 3: Collect learning corpus and use it to train a generative AI assessment model for autism status assessment. The learning corpus includes the assessment conclusions of past senior supervisors, the industry standards used as the basis for the assessment, and the diagnosis and treatment plans formulated according to the industry standards based on the assessment conclusions.

[0100] Step 4: Import the digital twin model data of the autism patient to be evaluated into the generative AI evaluation model, and use the trained generative AI evaluation model to generate an evaluation result of the current status of the autism patient to be evaluated.

[0101] Step 5: Based on the obtained assessment results, use the trained generative AI assessment model to generate a personalized diagnosis and treatment plan suitable for the autistic patient to be assessed. The personalized diagnosis and treatment plan is an intervention measure that combines one or more of behavioral therapy, language training, and social skills training.

[0102] Step 6: Apply the intervention measures in the obtained personalized diagnosis and treatment plan to the autism patients to be evaluated, and use the constructed digital twin model to obtain the patient's response to different intervention measures.

[0103] Step 7: Based on the patient's response, the interventions administered to the autism patient to be evaluated are classified as interventions with positive effects, interventions with negative effects, and interventions with no significant effect. Among them, interventions with positive effects are interventions administered when the patient experiences one of the following conditions, where the condition one includes at least one of improved social communication skills, reduced stereotyped behaviors, expanded interests, improved adaptive skills, improved self-care ability, improved learning ability, reduced behavioral problems, and improved emotional regulation ability;

[0104] Interventions with negative effects are those that are used when patients experience two of the following conditions, including at least one of decreased social communication ability, increased stereotyped behavior, narrowed interest range, decreased adaptive skills, decreased self-care ability, decreased learning ability, increased behavioral problems, and decreased emotional regulation ability;

[0105] The intervention wording for the insignificant effect is the intervention applied when both Situation 1 and Situation 2 occur.

[0106] To determine whether the intervention measures for autism patients are effective, we can evaluate them from the following aspects:

[0107] Improvement of core symptoms

[0108] Social communication skills: Standardized scales (such as the Autism Diagnostic Observation Schedule (ADOS)) are used to assess whether the patient's social interaction and communication skills have improved. For example, whether the patient is able to communicate with others more proactively and whether they can better understand others' emotions and intentions.

[0109] Stereotyped behaviors and restricted interests: Use a repetitive behavior scale (eg, RBSR) to assess whether the patient's stereotyped behaviors have decreased and their interests have broadened.

[0110] Adaptive skills development

[0111] Self-care ability: Adaptive behavior assessment scales (such as the Vineland Adaptive Behavior Scale (VABS)) are used to assess whether the patient's self-care ability in daily life, such as dressing, eating, and personal hygiene, has improved.

[0112] Learning ability: Assess the patient's progress in cognitive, language, and learning skills, such as improvement in intelligence quotient (IQ).

[0113] Reduction in behavioral problems

[0114] Frequency and severity of problem behaviors: Behavioral assessment tools (such as the ABC Checklist for Aberrant Behavior) are used to record whether the frequency and severity of the patient's problem behaviors (such as aggressive behavior, self-harm behavior, etc.) have decreased.

[0115] Emotional regulation: This assesses whether the patient's emotional stability has improved and whether they are better able to cope with stress and frustration.

[0116] Long-term effects of interventions

[0117] Long-term follow-up studies: Long-term follow-up studies are used to assess the lasting effects of interventions, for example, whether early intensive behavioral intervention (EIBI) continues to provide significant benefits in adulthood.

[0118] Education and employment status: Assess whether the patient is able to better adapt to the school environment after completing the intervention and whether he or she is able to achieve a certain degree of independent living and employment in adulthood.

[0119] Comprehensive assessment by a multidisciplinary team

[0120] Multidisciplinary collaboration: Comprehensively assess the patient's progress by combining the assessment opinions of multidisciplinary professionals such as speech therapists, occupational therapists, and psychologists.

[0121] Parent and Teacher Feedback: Feedback from parents and teachers is collected to understand whether the patient's performance in the home and school settings has improved.

[0122] Data-based evaluation methods

[0123] Randomized controlled trials (RCTs): These are trials that evaluate the effectiveness of different interventions. For example, interventions like Applied Behavior Analysis (ABA) and the Early Childhood Development Denver Model (ESDM) have performed well in RCTs.

[0124] Meta-analysis: A meta-analysis combines and evaluates the results of multiple studies to determine the overall effectiveness of different interventions.

[0125] Consideration of individual differences

[0126] Individualized assessment: taking into account the impact of each patient's baseline level, age, intervention intensity and other factors on the intervention effect.

[0127] Parent involvement: Assess the involvement and synchronization of parents in the intervention process, as parents' active participation has an important impact on the effectiveness of the intervention.

[0128] Through the above-mentioned multi-dimensional evaluation method, the effectiveness of intervention measures for autism patients can be determined more comprehensively.

[0129] Step 8: Import the classification results of the intervention wording into the generative AI evaluation model to optimize the formulated personalized diagnosis and treatment plan. Specifically, the optimization principles for the formulation of personalized diagnosis and treatment plans are:

[0130] Reduce the frequency of intervention wording that has little effect;

[0131] Avoid generating interventions that have negative consequences;

[0132] Increase the frequency of interventions that have a positive impact.

[0133] Step 9: Dynamically update the digital twin model and learning corpus, and use the updated data to retrain and optimize the generative AI evaluation model.

[0134] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. Autism behavior modeling and dynamic optimization system based on digital twins and AI, characterized by: The following steps are involved: Step 1: Collect multimodal data of autistic patients, including individual characteristic data, behavioral data, physiological data, psychological assessment data, and family environment data, and perform preprocessing operations including cleaning and normalization on the collected data; Step 2: Using the pre-processed multimodal data, a digital twin model is constructed to reflect the status of the autism patient in real time. The digital twin model is also used to simulate the patient's response to different intervention measures. Step 3: Collect learning data and use the learning data to train a generative AI assessment model for autism status assessment. The learning data includes the assessment conclusions of past senior supervisors and the industry standards used as the basis for the assessment, as well as the diagnosis and treatment plans developed in accordance with the industry standards based on the assessment conclusions. Step 4: Importing the digital twin model data of the autism patient to be evaluated into the generative AI evaluation model, and using the trained generative AI evaluation model to generate an evaluation result of the current status of the autism patient to be evaluated; Step 5: Based on the obtained assessment results, use the trained generative AI assessment model to generate a personalized diagnosis and treatment plan suitable for the autistic patient to be assessed. The personalized diagnosis and treatment plan is an intervention measure that combines one or more of behavioral therapy, language training, and social skills training.

2. The autism behavior modeling and dynamic optimization system based on digital twins and AI according to claim 1 is characterized in that: The method also includes: dynamically updating the digital twin model and the learning corpus, and using the updated data to retrain and optimize the generative AI evaluation model.

3. The autism behavior modeling and dynamic optimization system based on digital twins and AI according to claim 1 is characterized in that: The method further includes: applying the intervention measures in the obtained personalized diagnosis and treatment plan to the autism patient to be evaluated, and using the constructed digital twin model to obtain the patient's response to different intervention measures; The interventions administered to the autism patients were categorized as those with positive effects, those with negative effects, and those with no clear effect, based on the patients' responses. The classification results of intervention wording are imported into the generative AI evaluation model to optimize the personalized diagnosis and treatment plan.

4. The autism behavior modeling and dynamic optimization system based on digital twins and AI according to claim 1 is characterized in that: In step 2, the process of building a digital twin model includes the following steps: Classify and integrate the pre-processed multimodal data to obtain the patient's behavioral data set, physiological data set, psychological assessment data set, and home environment data set. Using a standardized behavioral assessment scale to evaluate the patient's behavioral data, obtain the patient's social interaction and behavioral performance, as well as the patient's repetitive behavior assessment results, wherein the repetitive behaviors include stereotyped behaviors, compulsive behaviors, and ritualistic behaviors; Real-time collection of physiological data from patients to reflect their physiological state and emotional response, analysis of the data using machine learning algorithms, and identification of physiological characteristics associated with autism; Use standardized cognitive ability testing tools to obtain assessment results of patients' cognitive level and intellectual development, as well as emotional state and behavioral problems; Questionnaires are used to understand the patient's family environment, including family members' education level, family atmosphere, and parent-child interaction. Video recordings or on-site observations are used to analyze the interaction patterns between the patient and family members and assess the effectiveness of the family support system. The acquired multimodal data are fused to build a digital twin model of the patient.

5. The autism behavior modeling and dynamic optimization system based on digital twins and AI according to claim 3 is characterized in that: The principles for optimizing personalized diagnosis and treatment plans are: Reduce the frequency of intervention wording that has little effect; Avoid generating interventions that have negative consequences; Increase the frequency of interventions that have a positive impact.

6. The autism behavior modeling and dynamic optimization system based on digital twins and AI according to claim 5 is characterized in that: The intervention measures with positive effects are intervention measures applied when the patient experiences one of the following conditions, wherein the condition one includes at least one of improved social communication ability, reduced stereotyped behavior, expanded interest range, improved adaptive skills, improved self-care ability, improved learning ability, reduced behavioral problems, and improved emotional regulation ability; The intervention measures with negative effects are intervention measures applied when the patient experiences the following two conditions, wherein the two conditions include at least one of reduced social communication ability, increased stereotyped behavior, narrowed interest range, reduced adaptive skills, reduced self-care ability, reduced learning ability, increased behavioral problems, and reduced emotional regulation ability; The intervention wording of the insignificant effect is the intervention measure applied when both the first and second situations occur.

7. Autism behavior modeling and dynamic optimization system based on digital twins and AI, characterized by: The system is used to implement the steps of the autism behavior modeling and dynamic optimization system based on digital twins and AI according to any one of claims 1 to 6, including: Multimodal data acquisition module, used to collect multimodal data of autistic patients, including individual characteristic data, behavioral data, physiological data, psychological assessment data, and family environment data, and perform preprocessing operations on the collected data, including cleaning and normalization; A digital twin model construction module uses preprocessed multimodal data to construct a digital twin model that reflects the status of an autism patient in real time, wherein the digital twin model is also used to simulate the patient's response to different intervention measures; A learning corpus acquisition module is used to collect learning corpus, wherein the learning corpus includes the evaluation conclusions of past senior supervisors and the industry standards based on which the evaluations were made, as well as the diagnosis and treatment plans formulated in accordance with the industry standards based on the evaluation conclusions; An evaluation model construction module, for using the learning corpus to perform training to obtain a generative AI evaluation model for assessing the status of autism patients; An evaluation result generation module, configured to import the digital twin model data of the autism patient to be evaluated into the generative AI evaluation model, and generate an evaluation result of the current status of the autism patient to be evaluated using the trained generative AI evaluation model; A diagnosis and treatment plan generation module is used to generate a personalized diagnosis and treatment plan suitable for the autistic patient to be evaluated based on the obtained evaluation results using the trained generative AI evaluation model, wherein the personalized diagnosis and treatment plan is an intervention measure that combines one or more of behavioral therapy, language training, and social skills training.

8. The autism behavior modeling and dynamic optimization system based on digital twins and AI according to claim 7 is characterized in that: The system further comprises: A data update module is used to dynamically update the digital twin model and the learning corpus, and use the updated data to retrain and optimize the generative AI evaluation model.

9. The autism behavior modeling and dynamic optimization system based on digital twins and AI according to claim 7 is characterized in that: The system further comprises: An intervention effect acquisition module is used to apply the intervention measures in the obtained personalized diagnosis and treatment plan to the autism patient to be evaluated, and use the constructed digital twin model to obtain the patient's response to different intervention measures; An intervention effectiveness assessment module is used to classify interventions administered to autistic patients as interventions with positive effects, interventions with negative effects, and interventions with no clear effect, based on the patients' responses; The treatment plan optimization module is used to import the classification results of intervention terms into the generative AI evaluation model to optimize the formulated personalized treatment plan.

10. The autism behavior modeling and dynamic optimization system based on digital twins and AI according to claim 7, characterized in that: The system also includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein, when the processor executes the program, it is capable of implementing the steps of the autism behavior modeling and dynamic optimization system based on digital twins and AI as described in any one of claims 1 to 6.

Images