1. Background and significance of the study

Brain development disorders (such as autism spectrum disorder, ASD, attention deficit hyperactivity disorder, ADHD, intellectual development disorder, etc.) are neurodevelopmental diseases that seriously affect the physical and mental health of children and adolescents. These disorders have an early onset and a long course of illness, and are often accompanied by lifelong deficits in social, cognitive, and behavioral functioning, placing a heavy burden on families and society. According to statistics, the global incidence of autism spectrum disorder in children has increased significantly in recent years, and the prevalence of ASD in 8-year-old children in some areas has reached about 3%. The prevalence of ADHD in school-age children is about 5%, and the co-incidence of ADHD in ASD is as high as more than 50%. These disorders of brain development have become an important part of the global burden of disease. However, the etiological mechanisms of such diseases are extremely complex, involving multiple levels such as genetic variation, molecular pathways, abnormal brain network development, and environmental factors. Traditional research tends to focus on a single level, such as genomic analysis revealing candidate genes, neuroimaging studies revealing differences in brain structure and functional connectivity, and psychological studies characterizing cognitive and behavioral characteristics, but the research at each level is relatively fragmented, and there is a lack of a unified theoretical framework to link "bottom-up" biological data with "top-down" cognitive and behavioral symptoms. This cross-level knowledge gap makes it difficult to fully understand the pathogenic mechanisms of brain development diseases. Therefore, there is an urgent need for a theoretical model that integrates multi-scale and multi-dimensional information and can unly explain the disorders from biological basis to higher-order cognitive dysfunction, so as to deepen the understanding of disease mechanisms and guide the development of new diagnosis and treatment strategies.

The Data-Information-Knowledge-Wisdom (DIKW) model is a conceptual framework commonly used in the field of information science and cognition to describe the level of abstraction from raw data to Wisdom decision-making. However, the traditional DIKW model presents a strict hierarchical (pyramid) structure, with one-way transmission between layers, which is difficult to adapt to the complex feedback regulation characteristics and purpose-oriented behaviors of the human brain. In view of this, Prof. Yucong Duan, the applicant of this project, took the lead in proposing a new cognitive model with a "Purpose/Purpose" layer on the basis of DIKW, namely the DIKWP model. More importantly, the DIKWP model does not simply stack a new level, but connects the five elements of data, information, knowledge, wisdom, and purpose through a non-hierarchical network structure to form a cognitive network with multi-directional feedback and iterative updates. This original reticular cognitive model is an academic milestone, breaking through the limitations of the traditional linear cognitive framework and providing a new semantic structure for machine intelligence to simulate human cognition. By integrating "Purpose" into the model as an explicit level, the cognitive process not only includes the understanding of the objective world, but also the consideration of the subject's goal and purpose, so as to be closer to the essence of human thinking. The introduction of a purpose-oriented cognitive system is regarded as an innovative way to solve the "black box" problem in the field of artificial intelligence and improve the interpretability and controllability of AI systems. For brain development diseases, the cross-level semantic mapping capability of the DIKWP model has unique theoretical value: it provides us with a unified modeling framework that can integrate the abnormal representations of such diseases at different scales (from molecular and EEG data, to neural information processing, knowledge acquisition disorders, to impaired decision-making wisdom and lack of purpose) into the same semantic network for characterization and interpretation.

Specifically, in diseases such as autism, patients often exhibit abnormal sensory data processing (e.g., hyperesthesia or sluggishness), insufficient information integration ability (e.g., distracted attention, difficulty filtering stimuli), social cognitive deficits at the knowledge level (e.g., lack of understanding of the purpose of others' minds), inappropriate decision-making at the wisdom level (e.g., stereotyped behavior, lack of flexibility), and lack of social motivation at the purpose level. Traditionally, these symptoms have been explained separately by different theories, such as sensory integration dysregulation theory, theory of mind deficits, "executive function" disorder, etc., but lack a convergence perspective. The DIKWP model is expected to describe the dysfunction at the above levels in a unified language: from the abnormality of sensory input in the "data" layer to the disorder of attention/perceptual processing in the "information" layer, to the insufficient acquisition of social semantics in the "knowledge" layer, the impaired decision-making reasoning in the "Wisdom" layer, and the abnormality of motivation and purpose in the "Purpose" layer, forming a coherent causal chain and network. It can also be used as an analogy in ADHD: this model can uniformly represent how impulsive tendencies at the gene and neurotransmitter levels affect the attention control of the information layer, which in turn disrupts knowledge accumulation and task planning (Wisdom layer), leading to difficulties in maintaining intentional behavior (Purpose layer). Similarly, for intellectual developmental disabilities, the DIKWP framework is able to correlate the biological roots of cognitive delay with its performance at the level of knowledge acquisition and application. It can be seen that the DIKWP network cognitive model, as a unified modeling tool for cross-level cognitive mechanisms, can comprehensively reveal the pathological mechanisms of brain development diseases at different levels and explore the interaction between different levels, which cannot be achieved by traditional single-level research. This unique theoretical value will provide a new perspective for deepening the study of the etiology of brain developmental diseases, and provide systematic guidance for diagnosis and intervention.

While introducing the DIKWP model, this project will integrate the theory of artificial consciousness (AC) to deepen the analysis of higher-order dysfunction of brain developmental diseases. Artificial consciousness is the frontier direction in the field of artificial intelligence, which is committed to enabling machines to simulate the generation process and functional characteristics of human consciousness, including perception, emotion, willingness, and self-awareness. Professor Yucong Duan's DIKWP model is an important theoretical cornerstone of the artificial consciousness system: based on this model, he proposed the "DIKWP×DIKWP" artificial consciousness system architecture, which adds a metacognitive cycle (the second DIKWP cycle) to the basic cognitive process to realize the primary self-awareness functions such as self-monitoring, self-reflection and self-regulation. This "dual circulation" architecture is seen as a key way to build AI systems with autonomous awareness. The artificial consciousness model provides a new research tool for the problems of self-perception deficits, attentional control disorders, and abnormal motivation and willingness shown by patients with brain development disorders: we can simulate and analyze the dysfunctional mechanisms of corresponding functions in the patient's brain with the help of artificial intelligence with self-monitoring and purpose representation capabilities. For example, by artificially introducing perturbations such as "Purpose layer loss" or "abnormal attention control parameters" into the DIKWP network, the changes in the overall cognitive behavior of the artificial consciousness can be observed, so as to deduce the cause of the patient's symptoms by analogy. This method breaks through the limitations of inferring pathology only through clinical observation and passive data analysis, and can actively simulate the consciousness and cognitive processes in disease states, which is expected to reveal the internal mechanism of autism and other diseases in terms of self-awareness and higher-order cognition.

To sum up, the theoretical basis and significance of this project are as follows: taking the applicant's original DIKWP*DIKWP mesh cognitive model as the core, integrating artificial consciousness simulation methods, constructing a unified framework across the five layers of semantic space of data, information, knowledge, Wisdom and Purpose, and systematically analyzing the pathogenic mechanism of brain development diseases. This not only fills the gap in the current multi-scale research on the integration of brain science and cognitive science, but also is expected to bring a paradigm shift in clinical diagnosis and treatment: from a symptom-oriented empirical approach to precise intervention guided by cognitive mechanism models, so as to open up a new path for the diagnosis and treatment of brain developmental diseases.

2. the overall goal and technical route

Overall Objective: This project aims to construct a unified cognitive mechanism model based on the "DIKWP×DIKWP" non-hierarchical network cognitive model and artificial consciousness theory for typical brain development diseases such as autism spectrum disorder, attention deficit hyperactivity disorder, and intellectual disability, to elucidate the pathogenic mechanism of diseases across all levels of data-information-knowledge-wisdom-purpose, and then to explore new diagnostic indicators and intervention targets, and to propose integrated diagnosis and treatment strategies and technical solutions driven by artificial awareness. Specifically, it includes the following five objectives:

1. Establish DIKWP×DIKWP cognitive models to achieve unified modeling of cross-level cognitive mechanisms of brain development diseases: improve the DIKWP network cognitive model proposed by Professor Yucong Duan and expand it into a "dual circulation" DIKWP ×DIKWP architecture to characterize the whole process mechanism of patients from biological signals to cognitive purpose. Through this model, we can unify the elements involved in this project (genetic and molecular data, neural circuit information, cognitive knowledge representation, decision-making wisdom and purpose behavior), provide a unified semantic framework for disease mechanism research, and highlight the applicability and unique value of the DIKWP model in different diseases such as autism, ADHD, and intellectual disability.

2. Artificial consciousness system was used to deeply analyze the mechanism of higher-order neurological dysfunction in patients: an artificial consciousness simulation system was constructed based on the DIKWP × DIKWP architecture, so that it could have the functions of perception, self-awareness and purpose decision-making. The system was used to simulate and analyze the mechanism of higher-order dysfunctions such as self-perception deficits, attention control disorders, behavioral motivation and purpose-oriented abnormalities in patients with brain development diseases. Through the simulation of the "failure" or parameter abnormality of the corresponding modules in the artificial consciousness body, the internal causes of the above symptoms and the interaction between various cognitive levels are explored, and the explanatory power of the artificial consciousness model in the study of disease mechanism is verified.

3. To construct a multi-scale integrated model of brain developmental diseases and a DIKWP cognitive layer evaluation framework for auxiliary diagnosis and brain-like model calibration: collect and integrate the big data of patients with brain developmental diseases at different scales such as molecular, cellular, neural circuit, cognitive behavior, etc., and establish a multi-scale disease model. On this basis, the DIKWP structured cognitive layer evaluation framework was introduced, and quantitative indicators at each level (such as biomarker characteristics in the data layer, attention/perceptual function indicators in the information layer, cognitive test results in the knowledge layer, decision-making ability score in the Wisdom layer, and social motivation assessment in the Purpose layer) were designed. An intelligent auxiliary diagnosis system was developed to map the multimodal data of patients to the DIKWP framework for comprehensive evaluation, so as to improve the accuracy of early identification and classification diagnosis of diseases. At the same time, the evaluation framework is used for the calibration of brain-inspired computing models (such as neural network models or brain simulation systems) to ensure that the behavior of the model at each cognitive level is consistent with the characteristics of human diseases, so as to improve the credibility of the model in predicting disease mechanisms and efficacy.

4. A therapeutic target identification method based on cognitive interaction perturbation mapping was proposed, and an integrated diagnosis and treatment program driven by artificial consciousness was developed: the cognitive interaction perturbation mapping experiment was designed by using the established DIKWP ×DIKWP artificial consciousness model, that is, virtual intervention or perturbation was carried out for different levels of elements in the model (such as adjusting the activity of a molecular pathway, simulating and improving the intensity of a certain cognitive training, etc.), and the impact on the overall cognitive function and pathological state was observed. Through systematic intervention screening in silico, key nodes that can effectively improve the abnormal cognitive representation in the model were identified as potential therapeutic targets. Corresponding intervention strategies were developed for these targets, including small molecule drugs acting on abnormal neurotransmitter pathways, gene editing/treatment to correct pathogenic genes, and cell therapy to supplement defective neuron types. At the same time, the synergistic effect of the combined intervention was evaluated in the artificial consciousness simulation, and the closed-loop integrated diagnosis and treatment plan of "diagnosis-mechanism model-intervention response-cognitive reconstruction" was optimized. The program will take the artificial consciousness system as the core decision-making engine, intelligently recommend personalized intervention combinations according to the results of multi-layer evaluation of patients' DIKWP, and predict the efficacy. This AI-driven approach is expected to significantly accelerate target discovery and treatment design, and improve the accuracy and effectiveness of interventions for brain development diseases.

5. Develop a digital cognitive rehabilitation system and a simulated consciousness training platform for patients with brain development disorders, and verify the effectiveness in the preclinical model: Based on the above theoretical and technical achievements, a digital rehabilitation training system for patients with brain development disorders is developed. The system will include digital cognitive rehabilitation modules (such as a series of interactive training games or virtual reality scenarios designed for attention, memory, social cognition, etc.) and simulated awareness training modules (i.e., virtual partners or coaches driven by artificial consciousness with personalized interaction and situational guidance functions). Patients repeatedly practice their weak cognitive skills (e.g., social dialogue, self-emotion recognition, concentration exercises, etc.) by interacting with simulated consciousness bodies in a safe and controlled digital environment. The system uses the DIKWP framework to evaluate the progress of patient training at multiple levels, and adaptively adjusts the training content and difficulty to achieve personalized rehabilitation. We will first validate the efficacy and safety of the system in preclinical models, such as testing the effects of certain trained stimuli on neuroplasticity in animals in autism models, or recruiting a small number of children with developmental disabilities for a preliminary feasibility trial. Referring to existing research, virtual reality combined with AI has been used to improve the social skills of people with autism and has proven to be effective; Digital play therapies such as EndeavorRx have also been approved by the FDA for improving attention in children with ADHD. Therefore, the cognitive rehabilitation platform developed in this project is expected to enter clinical application after validation, providing children with a new choice of digital therapeutics.

Technical route: This project will be promoted in accordance with the process of mechanism model construction→ artificial consciousness simulation→ multi-scale integration → target discovery → scheme research and development→ system development and verification, and each module will be connected and gradually deepened to form a closed-loop research path (see figure omitted). Firstly, at the theoretical level, the framework of DIKWP × DIKWP unified cognitive model was established to define the elements at each level and their interaction relationships. Based on this, an artificial consciousness simulation platform was developed to realize the digital reproduction of human cognitive processes. Subsequently, a large number of multimodal data related to brain development diseases (including genome/proteome data, brain imaging and electrophysiological data, neuropsychological assessment results, etc.) were collected to build a multi-scale database of diseases, and mapped into the DIKWP model for comprehensive analysis. In the simulation platform, the model parameters are flexibly adjusted and computational experiments are carried out: on the one hand, the functional deficits at each level under the disease state are simulated to verify the explanatory power of the model, and on the other hand, various virtual interventions are introduced to evaluate their improvement effect on the pathological state in the model (i.e., cognitive perturbation experiments). By comparing the effects of different perturbations, the most potential intervention methods and corresponding biological targets were screened. Next, a specific diagnosis and treatment plan was designed based on the screening results, and the artificial awareness system was used to optimize and iterate the program—the AI agent will continuously adjust the intervention combination based on the simulation feedback until the best cognitive improvement effect is obtained. Based on the optimization plan, we will enter the application development stage: develop intelligent auxiliary diagnosis software (embedded in the existing diagnosis process of the hospital to realize the automatic analysis and risk assessment of multi-level indicators of patients) and a digital rehabilitation training platform (rehabilitation app or VR system for patients to use in the clinical or home environment). Finally, animal model experiments and small-scale clinical trials were used to validate and evaluate the proposed protocol and system, collect evidence of efficacy and safety, and improve technical details. The whole technical route emphasizes human-computer interaction and iterative improvement: the model results will guide the biological experiment, and the experimental data will in turn correct the model parameters; Similarly, the trial feedback of the diagnosis and treatment system will also be used to further improve the artificial awareness algorithm and cognitive model, forming a closed loop of continuous evolution. Through the above route, we will deliver a complete solution covering "mechanism analysis, intelligent diagnosis, target screening, innovative therapy, and rehabilitation training" at the end of the project period, providing full-chain support from theory to application for brain development diseases.

3. Sub-module research content

Module 1: DIKWP× The construction of DIKWP model and the unified modeling of cross-level cognitive mechanisms of brain development diseases

Research content: This module constructs the applicant's original DIKWP × DIKWP non-hierarchical cognitive model for the core pathological mechanism of brain development diseases, and is used to model different levels of cognitive impairment mechanisms in a unified manner. Specific work includes:

1) Formal definition of DIKWP × DIKWP model: On the basis of the original DIKWP mesh model, the "dual circulation" architecture is introduced, and the basic cognitive process and metacognitive process are combined to form the DIKWP × DIKWP model. We will define the nodes and relationships of the model in a mathematically and computer-executable form, such as using directed graphs or tensor networks to represent the five elements and two-way feedback connections, so as to ensure that the model is strictly logically self-consistent and computable.

2) Disease cognitive representation mapping: For the three types of autism, ADHD and intellectual disability, the typical abnormal manifestations of autism, information, knowledge, wisdom and purpose are summarized based on literature and expert knowledge, and these abnormalities are anchored to the corresponding nodes or connections of the model. For example, sensory hypersensitivity in autism can be mapped to abnormal input signal gain in the data layer, imperfect semantic network in the knowledge layer corresponding to social comprehension disorders, and insufficient activation of nodes in the purpose layer due to low social motivation. Through this mapping, a model characterization of the disease is established.

3) Cross-level mechanism hypothesis testing: The constructed model is used to verify several important cross-level mechanism hypotheses in the simulation environment. For example, to test "how information filtering disorder leads to impaired knowledge acquisition": to simulate the decline of attention control (information layer) function, and to observe changes in the learning performance of new concepts in the knowledge layer, so as to explain the association between attention deficit and academic difficulties in ADHD; In another example, the reverse inhibitory effect of the reduction of social motivation on the perceptual-knowledge layer in the Purpose layer was simulated to explain the vicious circle of sensory overload and social avoidance in autism. These simulations will produce quantitative data that can be used to validate the plausibility of the model.

4) Model optimization and unified theory: According to the simulation results and experimental feedback, the model parameters and structure are continuously modified to make it possible to explain the multi-level characteristics of multiple diseases at the same time, and extract the common mechanisms and differences across diseases. This module is expected to form the final version of the DIKWP × DIKWP unified cognitive mechanism model, which will serve as the cornerstone of all subsequent research work.

Expected Results: Complete the construction of DIKWP×DIKWP model and disease mechanism mapping, publish 1~2 model-related theoretical papers, and form a systematic theoretical explanation of the cross-layer cognitive mechanism of brain development diseases. Several key pathological hypotheses were verified in the model to provide a basis for the formulation of follow-up intervention strategies. The model will be presented in the form of a map or software, visually demonstrating the mechanism of disease association from the biological level to the cognitive level, providing an intuitive tool for scientific research and teaching.

Module 2: Analysis of higher-order neurological dysfunction mechanisms based on artificial consciousness theory

Research content: This module uses the DIKWP × DIKWP model established in module 1 to construct an artificial consciousness system and conduct an in-depth analysis of higher-order dysfunction in patients with brain development diseases. Main research steps: 1) Artificial consciousness system development: design and implement an artificial consciousness agent (software agent) based on DIKWP ×DIKWP architecture. The agent consists of two parts: the basic cognitive cycle (realizing the perception, processing and decision-making of environmental data) and the metacognitive cycle (monitoring and regulating one's own cognitive state). We will endow the agent with human-like physiological parameters (such as virtual receptive field, memory capacity, reward mechanism, etc.) and emotion and purpose representation mechanisms, so that it has a similar perceptual-emotional-voluntary function as humans. 2) Normal cognitive baseline training: Artificial consciousness agents are trained to complete a series of cognitive tasks in a virtual environment through reinforcement learning and other techniques as a baseline for normal functioning. For example, it is trained to focus on goal-oriented tasks to ensure that the agent has a certain ability to control attention and maintain purpose. Train them to interact with virtual social individuals to develop self-other-awareness and social behavior patterns. Once the training converges, the agent will exhibit a relatively stable functional state in all layers of the DIKWP, close to the expected human function. 3) Disease state simulation: Targeted perturbation of artificial consciousness agents to simulate higher-order dysfunction of patients. Specifically, self-perception deficit simulation - weakens the performance of the self-monitoring module in the metacognitive loop, and makes the agent slow to perceive changes in his own state (error, success, etc.), so as to simulate the self-awareness disorder of autistic patients; Attention deficit simulation: introducing noise or reducing attention selectivity in the information layer to observe the agent's ADHD-like transient attention and distracted behavior; Motivational Willingness Disorder Simulation: Reducing the agent's sensitivity to task reward or goal adherence at the Purpose layer, resulting in a state of insufficient motivation and poor behavioral purposefulness in the artificial consciousness body, similar to the lack of social motivation in autism or poor persistence in ADHD. By placing agents in these perturbed states to perform tasks, we will collect data on their behavioral performance and internal state and compare them with the results of cognitive assessments in real patients. 4) Mechanism analysis: Based on the results of simulation experiments, the key internal mechanisms that cause specific higher-order dysfunctions are extracted. For example, we may find that when the information filtering (attention control) is weakened, the representation confusion of the agent knowledge layer increases, and the decision-making error rate increases, which confirms the supporting role of attention on high-level cognition. In another example, when the driving force of the Purpose layer is reduced, the agent is more likely to abort the task when encountering difficulties, which illustrates how the lack of motivation leads to insufficient behavioral persistence. These findings will be used to explain the intrinsic causal chain of the patient's symptoms and provide clues to interventions (e.g., whether enhancement of a certain layer of function compensates for the deficit). 5) Model validation: Some mechanisms will be selected to be verified in real patients or animal models, such as using brain imaging technology to observe whether the abnormal activation of brain information filtering related networks in ADHD patients during attention tasks is consistent with simulation prediction. The validation results will further feed back the correction of the surrogate model.

Expected Results: Completed the prototype development of the artificial consciousness agent system, and successfully simulated the behavioral characteristics of typical higher-order dysfunctions such as autism and ADHD. He has produced a mechanism explanation model for the dysfunction of brain development diseases in terms of self-awareness, attention, and willingness, and has published more than one academic paper in authoritative journals at home and abroad. This work will demonstrate for the first time that artificial consciousness systems can be used to study disease mechanisms, deepen our understanding of the nature of these higher-order cognitive impairments, and lay the foundation for the development of innovative therapies.

Module 3: Multi-scale Modeling of Brain Development Diseases and Construction of DIKWP Cognitive Level Assessment Framework

Research content: This module focuses on connecting multi-scale biological data of brain development diseases with cognitive models, establishing multi-scale disease models from molecular to in vivo levels, introducing the DIKWP structured cognitive layer evaluation system, and developing intelligent auxiliary diagnosis and model calibration tools. The main contents are as follows: 1) Multi-scale data integration and disease modeling: Collect multi-level data related to brain development disorders in project partner hospitals and public databases, including: genetic and epigenetic data (pathogenic gene variation, copy number variation, DNA methylation, etc.), molecular and cellular level data (blood and cerebrospinal fluid biochemical markers, neurotransmitter levels, histopathology, etc.), neural circuit level data (functional magnetic resonance imaging fMRI connection mode, EEG/ magnetoencephalography, brain network atlas, etc.) and behavioral and cognitive assessment data (scales, questionnaires, neuropsychological test scores, etc.). Multi-modal data fusion technology is used to map data of different scales to the same patient object, and a multi-scale feature vector or knowledge graph of disease is established. If necessary, the dimensionality reduction and clustering methods of machine learning are applied to extract the key variables that can represent the main pathological features from the massive features. 2) DIKWP cognitive layer evaluation index extraction: According to the DIKWP framework, the above multi-scale features were mapped to five cognitive levels. For example, gene/cell abnormalities are classified as "data layer" lesions, simple perceptual function and attention indicators are classified as "information layer", language ability and social cognition test results are classified as "knowledge layer", complex decision-making and problem-solving ability assessment are classified as "wisdom layer", and motivation, interest, and goal-related scale results are classified as "purpose layer". A composite score or set of indicators is designed for each level to quantify the patient's degree of functional integrity at that level. We will develop the DIKWP multi-level cognitive function assessment scale, which integrates subjective and objective data to profile patients in an all-round way. 3) Development of intelligent auxiliary diagnosis system: Based on the above evaluation framework, a machine learning model is established to output diagnostic suggestions after inputting the patient's multi-layer feature vectors. For example, training a classification model to identify ASD, ADHD, or intellectual disability categories, or further classify subtypes; Train regression models to predict disease severity scores, and more. In particular, we pay attention to the contribution of the combination of indicators at each level of DIKWP to diagnosis, hoping to discover cross-level diagnostic feature patterns and improve the diagnostic accuracy and ability to distinguish comorbidities (such as ASD with ADHD). In addition, a visual interface was developed to visually display the five-layer DIKWP score radar chart of patients to clinicians to assist clinical decision-making. 4) Calibration and verification of brain-inspired models: select the current international advanced brain computing models or simulation platforms (such as large neural networks to simulate the cognitive process of the human brain), and apply our DIKWP evaluation framework to its output to evaluate the cognitive performance of the model. If the DIKWP spectrum of the model is systematically biased from that of real patients, we will adjust the model structure or parameters (such as adding feedback connections at the Purpose layer) to make it closer to the real pathology. This process can also be seen as a validation of the DIKWP theory: a successful calibration will prove the universality and validity of the DIKWP framework.

Expected Results: Establish a multimodal database containing ≥ 100 patients and a hierarchical evaluation system of DIKWP. One prototype system for intelligent diagnosis of brain development diseases has been developed, and its diagnostic accuracy is improved compared with existing methods (for example, the diagnostic accuracy of ASD is increased by no less than 5 percentage points). He has written 1 paper on multi-scale data fusion diagnosis. The applicability of the DIKWP framework in the simulation system was verified through the evaluation and calibration of the brain-inspired model, and the relevant results were published in interdisciplinary journals. The results of this module will provide a directly applicable auxiliary diagnostic tool for clinical practice, and provide an objective quantitative means for the evaluation of follow-up intervention strategies.

Module 4: Therapeutic target discovery based on cognitive interaction perturbation mapping and research on artificial consciousness-driven diagnosis and treatment protocols

Research content: This module aims to use the aforementioned cognitive models and artificial awareness platforms to discover innovative drugs and other intervention targets, and to construct AI-driven integrated diagnosis and treatment protocols. The core idea is to explore potential interventions through trial and error in a simulated environment, and to screen for the best options by simulating their impact on cognitive function. The main research steps are: 1) Cognitive interaction perturbation mapping experiment: in the artificial consciousness simulation platform, the biophysical and cognitive parameters that can be adjusted (such as neuronal excitability, synaptic plasticity parameters, corresponding to drug effects) are introduced into the artificial consciousness simulation platform; Gene expression levels correspond to gene therapy; Specific cognitive training intensity corresponds to rehabilitation therapy, etc.). Using the method of high-throughput computational experiments, these parameters were adjusted one by one or in combination, and the impact on the performance indicators of each layer of the agent DIKWP and the overall task performance were recorded, and the "intervention-effect" mapping network was drawn. Particular attention is paid to changes in parameters that significantly improve the agent's cognitive deficit (the abnormality simulated in Module 2) and locate their corresponding biological implications. For example, the discovery that increasing levels of a certain neurotransmitter can alleviate simulated attention impairment suggests that this neurotransmitter pathway may be a drug target for ADHD. 2) Screening of candidate therapeutic targets: Comprehensively mapping the results to screen out the key intervention nodes at multiple levels: at the molecular level, there may be several high-impact genes or proteins; At the cellular circuit level, it may be the pattern of activity of a particular brain region or neural circuit; At the cognitive level, it may be some kind of training paradigm. Combined with literature search and expert knowledge, the biological rationality and feasibility of these nodes were evaluated, and some of the most promising novel targets were selected. For example, if it is found that modulating prefrontal-hippocampal circuit connectivity significantly improves learning ability in the model, then this circuit is used as an intervention target for electrical stimulation or neuromodulation. We plan to select 2-3 targets in each disease for further validation. 3) Formulation and optimization of intervention strategies: design specific intervention strategies for each target screened. For example, for molecular targets, molecular docking and virtual screening are used to find potential small molecule drugs, or topical use of existing drugs is considered for indications (drug retargeting); For gene targets, design gene therapy protocols (e.g., vector delivery RNA interference or CRISPR correction); For cell/loop targets, optogenetics, transcranial magnetic stimulation and other technical means should be considered for regulation. For purely cognitive targets, develop intensive training or behavioral therapy protocols. In the artificial awareness platform, a more refined virtual patient model (taking into account individual differences) is introduced to simulate the effects and synergies of the above interventions. If multiple targets require joint intervention, we will apply reinforcement learning algorithms to allow AI to autonomously experiment with the combination and sequence of interventions to optimize efficacy and reduce side effects. The result is a personalized diagnosis and treatment plan for different subtypes of patients, including specific drug/non-drug intervention combinations and procedures. 4) Experimental validation: select some representative intervention programs and carry out validation experiments in vitro and animal models. For example, to test whether the effects of a new drug on social behavior and brain functional connectivity are consistent with model predictions in a mouse model of autism; Specific brain stimulation interventions were administered to ADHD model mice, and attention improvement was observed. Some safer interventions (e.g., cognitive training regimens) will also be evaluated in preclinical human trials: a small sample of patients will be recruited to observe changes in DIKWP before and after the intervention, compared to model expectations. The results of the validation are used to adjust and refine the diagnosis and treatment plan. 5) Closed-loop feedback mechanism: The verification data is fed back to the AI-driven decision-making system, and the intervention decision-making algorithm is constantly updated to make it gradually self-improve in the real world, and finally form a closed-loop treatment system of artificial consciousness that can be used in clinical practice.

Expected Results: Propose a variety of innovative therapeutic targets and intervention strategies, and apply for no less than 2 related invention patents. At least 1 new drug target (e.g., animal experiments have demonstrated that improving the function of a certain protein can improve autistic behavior), 1 new physical/digital intervention method (e.g., VR scenario training combined with transcranial electrical stimulation to improve attention), and 1~2 papers have been published. Integrate the above results to form a white paper or guideline manuscript of AI-driven diagnosis and treatment plan for brain development diseases, covering disease identification, target selection, personalized intervention process, etc., to lay the foundation for subsequent clinical trials and application promotion. The results of this module will connect the "last mile" from mechanism to intervention, bringing practical new therapies to patients.

Module 5: Development and effectiveness verification of digital cognitive rehabilitation system and simulated consciousness training platform

Research content: Based on the intervention strategies proposed in Module 4, this module is oriented to clinical applications, develops a comprehensive platform for digital cognitive rehabilitation and artificial awareness training, and verifies its effectiveness and safety through preclinical studies. The main tasks are as follows:

1) Overall system architecture design: According to the rehabilitation needs of patients with brain development diseases, the functional modules and processes of the digital rehabilitation system are designed. The system will take the form of a client application or a virtual reality platform, including: an evaluation module (collecting multi-level cognitive data of patients and accessing the evaluation framework of module 3), a training module (a training game or task set for different cognitive functions), an artificial awareness counselor module (an intelligent agent monitors patient performance in real time, gives feedback and guidance), and a data recording and analysis module (records training data for efficacy evaluation and strategy adjustment). Determine system hardware requirements (e.g., VR headsets, brain-computer interface optional components, etc.) and data security specifications.

2) Rehabilitation training content development: A series of training content is designed according to each cognitive level of DIKWP. For example, data/information layer training includes sensory integration training, attention concentration exercises (such as target finding stimulation games); Knowledge-level training includes language comprehension and social cognitive training (role-playing games, talking to virtual characters and practicing social skills); Wisdom layer training includes problem-solving and situational decision simulation (completing tasks in a virtual campus environment to train planning and execution functions); Purpose-level training focuses on enhancing motivation and self-awareness (recognizing emotions through virtual characters, training patients to recognize their own emotions and set goals). Each training session will be set with multiple difficulty levels and personalized parameters to suit patients of different ability levels, and will be dynamically adjusted under the control of AI coaches.

3) Artificial Awareness Counselor Development: Use the artificial awareness agent of module 2 to create the image of "virtual rehabilitation therapist". The agent is implanted in the rehabilitation system and assumes a variety of roles: for example, when the patient is engaged in social training, the agent plays the role of a dialogue partner to present different expressions and reactions; In attention training, the agent monitors the patient's concentration and gives prompts or rewards in a timely manner; In Purpose training, the agent discusses goals with the patient to enhance their autonomy. The agent will evaluate the patient's status based on the patient's real-time performance (obtained through motion capture, physiological sensing, etc.), and decide the next interaction according to preset rules or reinforcement learning strategies (such as increasing the difficulty of the challenge or switching training strategies), forming a closed-loop human-computer interaction training process.

4) Software integration and testing: develop the above modules into software prototypes and integrate them into a unified platform. Functional testing and user experience optimization were carried out, and rehabilitation therapists and some patients were invited to participate in trial feedback to improve the interactive interface and process. Ensure the system is easy to use, fun, and motivating to improve patient compliance.

5) Effectiveness verification: select appropriate animal models and human volunteers to carry out preclinical trials. For animals (e.g., genetically engineered autism model mice), a customized training device was used to provide some digital-like training elements (such as perceptual stimuli and reward mechanisms) to observe the effects of long-term training on their behavior and brain plasticity, and to verify the biological effects of digital cognitive stimuli. For humans, a small-sample, single-group controlled trial will be carried out in collaboration with clinicians: several children with autism or ADHD will be recruited, and a digital rehabilitation system will be added to the standard treatment for a period of time, and the changes in DIKWP indicators and clinical symptoms (such as improvement of attention test scores, changes in social behavior rating scores, etc.) will be compared before and after the intervention. At the same time, adverse events and technical issues are closely monitored to assess system safety and stability. If conditions permit, a control group can also be set up for a preliminary study of randomized controlled trials. The validation results will be analyzed using statistical methods to determine the achievement of the preliminary efficacy of the system.

Expected Results: Completed the development of one set of "Digital Cognitive Rehabilitation and Simulation Awareness Training Platform for Brain Development Disorders", and obtained the relevant qualifications for computer software copyright or medical device registration. Previous clinical studies have been projected to demonstrate a positive effect of the system in improving target cognitive function (e.g., a significant increase in the average attention score after training, or an improvement in social interaction behavior ratings) with no significant adverse effects. We will write a report on the results of system development and experiments, and publish one paper in a journal in the field of rehabilitation medicine or digital medicine. In addition, based on the validation results, the system was improved and upgraded to prepare for the next large-scale clinical trial. This module will ultimately deliver an innovative rehabilitation tool with clinical potential to improve the prognosis of patients with brain development disorders in combination with traditional treatments.

4. feasibility analysis

Technical and theoretical feasibility: This project is built on a solid foundation of preliminary research and a mature theoretical framework. The applicant, Professor Yucong Duan, has been deeply involved in the field of cognitive computing and artificial intelligence for many years, and is the original proposer of the DIKWP model and an internationally recognized leader in this field. The DIKWP model and its extension (including the "dual circulation" artificial consciousness architecture) have obtained 114 authorized invention patents at home and abroad, forming a complete intellectual property system, which fully proves its scientific innovation and feasibility. In particular, the DIKWP mesh cognitive model described in the patent has been verified by many cases, which can provide clear cognitive levels and transformation rules for large-scale AI systems, and provide an executable semantic framework for the development of artificial consciousness systems. These preliminary results show that it is technically feasible to construct an artificial awareness and cognitive assessment system with DIKWP as the core. In addition, the team has developed the DIKWP artificial consciousness prototype system, which has won the "Best Poster Award" at a recent academic conference. For example, we have successfully implemented the "Purpose-Driven DIKWP Physiological Artificial Consciousness Prototype", which initially demonstrates the effect of embedding the Purpose layer into an artificial intelligence body to produce self-regulating behavior. These advances have laid the foundation for the construction of more complex artificial consciousness simulations and cognitive models for this project.

From the perspective of multidisciplinary development, the key technical routes on which this project is based have made important breakthroughs in their respective fields, and have the conditions for integrated innovation. On the one hand, the fields of brain science and cognitive science have accumulated a large number of multi-scale data and theoretical models for autism, ADHD and other diseases, such as candidate genes found in genome association studies, functional network abnormalities revealed by brain imaging studies, and executive function deficit hypotheses proposed by cognitive psychology. On the other hand, recent advances in artificial intelligence and data science have enabled us to integrate and analyze these heterogeneous data – deep learning and network analytics techniques can be used to mine potential correlational patterns in multidimensional data and build models of complex systems. In recent years, the research on artificial consciousness has developed rapidly in the world, and some preliminary models have been able to simulate human perception and some cognitive functions. Virtual reality and human-computer interaction technologies are also becoming more and more mature, and have been applied to the social training of children with autism and have achieved results. It can be said that the DIKWP cognitive model, artificial consciousness simulation, multimodal data fusion, intelligent decision-making and digital rehabilitation involved in this project have a good foundation in their respective fields, and our team has the conditions to organically combine them to produce a "1+1>2" synergistic effect. In particular, the introduction of the DIKWP framework is expected to become a link to integrate knowledge from various fields, connect biology and cognitive disciplines, reduce the uncertainty of research, and improve the probability of success.

Research team and supporting conditions: This project is undertaken by a multidisciplinary high-level team, and the core members have rich research experience and outstanding achievements in artificial intelligence, brain science, biomedical engineering, etc. The project leader, Professor Yucong Duan, is also an academician of the International Academy of Advanced Technology and Engineering and the chairman of the World Association of Artificial Consciousness, and has made great achievements in the research of basic theories and cognitive models of artificial intelligence. Many of the project backbones come from the fields of neurobiology, psychiatry, clinical and computer science, and have participated in national key R&D programs and major projects, and have a deep understanding of the mechanism of brain development diseases and AI medical technology. The team has all the key technologies needed to carry out this project, including cognitive model construction, machine learning, big data processing, brain imaging analysis, animal experiments and clinical trial design. In terms of research conditions, the relying unit has established scientific research platforms such as "Artificial Consciousness and Cognitive Intelligence Laboratory" and "Multimodal Brain Imaging Center", equipped with high-performance computing clusters, virtual reality equipment, neural signal recording analyzers and other hardware, which can fully meet the needs of model simulation and data processing. We have partnerships with hospitals and rehabilitation centers to obtain clinical samples and conduct preclinical studies. At the same time, existing patents and software achievements will be seamlessly used in this project, speeding up the research and development process. In addition, the project team has an extensive academic cooperation network at home and abroad, and can consult experts in related fields at any time to carry out collaborative research. All these ensure the feasibility and efficient progress of the project implementation.

Risk analysis and response: This project is cutting-edge and innovative, involving multidisciplinary interdisciplinarity, and potential challenges include: the complexity of the model may bring difficulty to analysis, some hypotheses are difficult to verify, and the regulatory and ethical considerations in the process of clinical translation. In this regard, we have formulated a plan: in the model design, adhere to modular, gradual development, if the global model is too complex and difficult to converge, we will simplify the sub-model and gradually superimpose; In terms of verification, if a hypothesis cannot be verified by conventional experiments, we will seek alternative experimental paradigms or use cooperative team resources. For clinical application approval and ethics, we will communicate with regulatory authorities and ethics experts early in the development process to ensure that the product design meets specifications and strictly adheres to the protection of subject rights and interests during the validation phase. It is worth mentioning that because this project conducts a large number of simulation experiments in a virtual environment, the risk and cost of real-world experiments can be reduced to a large extent. On the whole, the project is bold in innovation but well-thought-out in its plans, with controllable risks and multiple response paths, which has high implementation feasibility.

5. Phased achievements and assessment indicators

In order to ensure the smooth implementation of the project, we plan to divide the entire research plan into three phases, and the key tasks and assessment indicators of each stage are as follows:

• Phase 1 (initial stage of the project, 1~2 years): model construction and basic platform construction. Complete the construction and optimization of the DIKWP×DIKWP unified cognitive model, build a basic platform for artificial consciousness simulation, and realize the simulation of normal people with at least one typical cognitive task; Establish a multimodal database of brain development diseases and a hierarchical evaluation index system of DIKWP; The prototype of intelligent auxiliary diagnosis algorithm was developed. Assessment indicators: submit DIKWP× DIKWP model formal description and software module, the model can reproduce the basic cognitive behavior of normal people; The database covers multi-layer data of no less than 100 samples; The accuracy of the auxiliary diagnostic model for the classification of patients in the training set ≥ 85%. At least 2 papers published in the stage (including 1 SCI journal paper); The stage summary report passed the expert review.

• Phase 2 (mid-project, 3~4 years): mechanism validation and target screening. The artificial consciousness platform was used to successfully simulate the key cognitive impairment phenomena of patients with autism and ADHD, and the list of candidate intervention targets was screened through multiple sets of simulation experiments. Complete the prototype development and preliminary test of the intelligent diagnosis system to improve the accuracy of diagnosis; Cell/animal level validation experiments were conducted for at least 2 novel therapeutic targets; A prototype of the digital rehabilitation training platform was developed and trialled in a small sample of patients. Assessment indicators: the coincidence rate of patient simulation behavior output by artificial consciousness model with real patient characteristics ≥70% (qualitative indicators are evaluated by experts); List no less than 5 candidate intervention targets with clear biological significance, and form an intervention plan design description for each target; The classification accuracy of the auxiliary diagnostic system on the validation set was >5% higher than that of Phase 1. Complete the initial validation of at least 1 intervention in an animal model and observe changes in efficacy in the expected direction; The digital rehabilitation prototype system has complete functions, no major software defects, and positive feedback from small-scale trials. Published 2~3 papers and applied for more than 1 Chinese invention patent; The mid-term inspection passed.

• Phase 3 (late project, 5th year): Comprehensive integration and validation evaluation. Integrate the results of each module and improve the decision-making system for diagnosis and treatment driven by artificial awareness; For the most promising 1~2 intervention programs, in-depth animal experiment verification (including multi-index evaluation of behavior and pathophysiology) was completed, and statistically significant efficacy improvement results were obtained. Conduct controlled trials of digital rehabilitation systems in target patient populations to evaluate their clinical benefits; Organize and form research reports and technical specifications on the mechanism and diagnosis and treatment strategies of brain development diseases, and plan clinical transformation. Assessment indicators: The integrated diagnosis and treatment decision-making system can give personalized intervention suggestions based on patient data, and the effectiveness of the recommendations is verified in the simulation test (compared with the unoptimized program, the cognitive improvement predicted by the model is increased by >20%); In animal experiments, the core symptom indicators of the intervention group were significantly improved compared with the model group (P<0.05). In the rehabilitation system trial with no less than 10 patients, the response rate (the proportion of symptom scale score improvement reaching the predetermined standard) was not less than 70%, and no serious adverse events were reported; Form 1 copy of the general project report and 1 copy of the standardized technical manual; Published more than 2 papers (including 1 SCI Zone 2 or above paper) and applied for 1 related international patent; All the acceptance indicators of the project were achieved, and the final acceptance was passed.

6. Clinical transformation and promotion paths

The ultimate goal of this project is to apply the research results to clinical practice for the benefit of the majority of patients with developmental disabilities. To this end, we have developed a clear clinical translation and promotion plan:

First of all, in terms of diagnostic evaluation technology, the DIKWP multi-layer evaluation framework and intelligent diagnostic system developed in this project will enter the clinical verification stage in the later stage of the project. Once its effectiveness is verified to be superior to existing diagnostic methods, we will collaborate with children's hospitals and psychiatric hospitals to conduct multi-center clinical trials to collect data from larger samples to verify stability. At the same time, it is preparing to apply for national medical device approval, and the diagnostic system will be registered and certified as an auxiliary diagnostic software (Software as a Medical Device, SaMD). At the promotion level, we plan to promote this new assessment system to psychiatrists and clinical psychologists through academic conferences and continuing education, and to develop user manuals and training materials to ensure that clinicians can master its use. Relying on the team's influence in the domestic artificial intelligence medical field, we will actively strive to incorporate relevant indicators into clinical diagnosis and treatment guidelines or expert consensus, so as to improve industry recognition.

Second, in terms of therapeutic targets and intervention programs, we will cooperate with biopharmaceutical companies and clinical research institutions to advance the drug development or clinical trial process for the selected new drugs or gene targets. If it is a new indication (retargeting drug) of an existing marketed drug, the design of a phase II clinical validation trial will be initiated; For new drugs with a new mechanism, we will consider optimizing the drug lead compound, pharmacological and toxicological evaluation, and then applying for IND approval. Gene and cell therapy targets are further validated in larger animal models in collaboration with gene therapy companies or stem cell centers, and then applied for clinical trials after maturity. We plan to bring at least one new therapy into clinical trials within 3-5 years of the end of the program. At the same time, non-invasive physical and digital intervention programs (e.g., brain stimulation, digital rehabilitation training) will enter clinical translation earlier. We will complete the preliminary clinical trial data during the project period and apply for approval from the relevant competent authorities to market the digital rehabilitation system as a medical device or digital therapeutics. The success of Akili's approval of play therapies shows that regulators are embracing this form of digital intervention. We will also refer to its path to ensure that the product meets security and privacy specifications and is approved for registration.

Thirdly, in terms of the promotion and application of the digital rehabilitation platform, we plan to adopt a multi-channel strategy. In addition to being used as a prescription rehabilitation tool in the medical system, we will also cooperate with special education schools and rehabilitation training institutions to promote the system, so that more children with developmental disabilities can use the system in their daily training. In order to lower the barrier to use, we will develop a version that adapts to a variety of terminals (tablets, PCs, VR devices), and launch a cloud service model for access in under-resourced areas. The project team will hold a demonstration training course to train a group of therapists and teachers who have mastered the digital rehabilitation system as "seed users", so as to expand the application from point to point. Considering the large base of children with developmental disabilities in China and the significant regional differences, we also plan to cooperate with public welfare foundations to carry out charitable donations or pilot applications in some underdeveloped areas to verify the large-scale feasibility and effect of this system in a real environment. Once it proves to have significant social benefits, we will actively propose to the National Health Commission and other competent authorities to include digital rehabilitation in the scope of national rehabilitation plans and medical insurance payments, and further promote its popularization.

Finally, in terms of industrial transformation and continuous R&D, relying on the results of this project, we have the conditions to incubate specialized high-tech enterprises or cooperate deeply with existing medical AI companies to build scientific research prototypes into stable commercial products. The team's rich patent reserves and technical secrets will play a key role in the industrialization process. We will formulate a clear intellectual property sharing and licensing policy to protect the interests of our partners and promote a win-win situation for industry, academia and research. After industrialization, more resources will be invested through the market mechanism to continuously improve product functions and upgrade the version according to clinical feedback. For example, we will continue to expand the training content library, optimize artificial intelligence algorithms, and increase support for other related diseases (such as depression and anxiety with developmental disabilities) to expand the applicability of products. In terms of promoting internationalization, we plan to cooperate with well-known foreign research institutions and autism organizations to promote this achievement to the world. In terms of academic communication, through publishing high-level papers and participating in international conferences, we will improve China's influence in the intersection of artificial consciousness and brain diseases, and export the "China Plan".

In short, while completing the scientific research goals, this project will closely focus on clinical needs and open up a one-kilometer road from laboratory to clinic. Through the above-mentioned multi-level and multi-path transformation and promotion measures, we strive to make our theoretical and technical achievements truly applied to the diagnosis and treatment of patients with autism and ADHD within a few years after the end of the project, improve the diagnostic efficiency and intervention effect, reduce the burden on patients' families and society, and produce significant social and economic benefits. We believe that the successful implementation and transformation of this project will contribute to the solution of the global problem of brain development diseases, and lead a new direction for the deep integration of artificial intelligence and brain science.