Background

Parkinson's disease (PD) is the second most common neurodegenerative disease in the world after Alzheimer's disease, which is characterized by a large loss of dopaminergic neurons in the substantia nigra of the midbrain, resulting in tremors, bradykinesia and other movement disorder symptoms. With the aging of the population, the number of Parkinson's disease patients is growing rapidly, and the number of Parkinson's disease patients in China has exceeded 3 million, and it is expected to exceed 5 million in 2030 (about half of the total number of patients in the world). The prevalence of Parkinson's disease in the elderly population over 65 years old in China is about 1.7%, showing a significant upward trend with age. The high disability rate and irreversible progression of the disease place a heavy burden on patients' families and social healthcare systems.

At the same time, the convergence of brain science and artificial intelligence has become a frontier field of global scientific and technological competition. National strategies such as the "China Brain Plan" clearly list brain-computer interface as a key development direction in the future, and brain-computer fusion technology is regarded as an important way to overcome neurological diseases such as paralysis and Parkinson's disease. At present, the treatment of Parkinson's disease is still mainly based on symptomatic remission, and there is a lack of fundamental intervention methods. In particular, the pathology of the neural circuit of the disease is complex, and how to adjust the brain function and alleviate the symptoms through external technical means is one of the major scientific challenges.

In the existing treatments, drug therapies (e.g., levodopa replacement) are effective in controlling symptoms in the early stages, but long-term use is less effective and produces significant side effects; Surgical therapies such as Deep Brain Stimulation (DBS) can significantly alleviate motor symptoms such as tremor in late Parkinson's disease, and are currently recognized as one of the effective interventions. However, none of these therapies have stopped disease progression and have limited improvement in the cause. In particular, traditional DBS usually adopts an open-loop continuous stimulation mode, which stimulates the target nuclei of the patient's brain for a long time under fixed parameters. This "constant stimulation" strategy cannot be adapted to the patient's real-time state, and often leads to overstimulation or understimulation, which induces side effects such as speech disorder and impulse control disorder. At the same time, the open loop stimulation cannot cope with the intraday fluctuations of Parkinson's symptoms, and there is still room for improvement in clinical efficacy. It is important to note that the neural mechanisms underlying why DBS can alleviate symptoms have not been fully elucidated. Some studies have compared brain connectivity before and after DBS treatment through imaging, and found that long-term stimulation can lead to the development of brain network topological reorganization in the direction of a healthy state. These suggests that external electrical stimulation may induce changes in the plasticity of the brain's neural circuits, thereby improving function. However, there is still a lack of systematic research on the specific brain regions and synaptic regulatory mechanisms involved in this remodeling process. This project aims at this scientific gap and uses Parkinson's disease as an example to deeply explore the phenomenon of brain-computer interface-induced brain circuit remodeling.

In order to overcome the shortcomings of traditional loop-opening stimulation, researchers proposed the concept of adaptive deep brain stimulation (adaptive DBS) (aDBS), which dynamically adjusts stimulation parameters on demand based on closed-loop feedback, in order to improve symptoms and reduce side effects. Preliminary studies suggest that aDBS has the potential to control Parkinson's symptoms more precisely; However, there are still many challenges in this field, such as the limitations of the reliability of effective biofeedback indicators (brain signaling biomarkers), as well as the optimization complexity and clinical applicability of stimulus control strategies. According to the consensus of leading experts, aDBS needs to be deeply integrated with emerging technologies (such as artificial intelligence) to achieve large-scale application. However, the current knowledge and verification in this direction is still insufficient, and in-depth research is still needed to break through the bottleneck.

It is worth mentioning that at present, some companies have launched DBS devices with EEG signal perception functions (such as Medtronic's Percept PC), and start-ups such as Neuralink have tried to develop brain-computer interface chips with high channel count. However, these initial explorations are still far from truly intelligent closed-loop regulation, and they do not yet have the ability to perform high-level semantic understanding and purpose decision-making on neural signals. The artificial consciousness closed-loop scheme proposed in this project will be innovative at this level, giving brain-computer interfaces "self-awareness" like regulatory wisdom.

In order to further improve the intelligence level of closed-loop brain-computer regulation, we introduce the DIKWP artificial consciousness model proposed by Professor Yucong Duan as a theoretical basis. The DIKWP model adds the highest level of "Purpose" to the classical DIKW (Data, Information, Knowledge, Wisdom) cognitive level, forming a five-layer cognitive system of "Data-Information-Knowledge-Wisdom-Purpose". This extension emphasizes the central role of decision-making goals in the cognitive process of agents, and realizes two-way feedback and iterative update of semantic information at all layers through the establishment of a network structure. With the DIKWP model, AI systems can be clearly motivated, making every step of the decision-making process traceable and explainable. Compared with traditional black-box algorithms, this embedded "purpose"-oriented artificial consciousness framework is expected to make closed-loop neuromodulation more intelligent, adaptive, safe and controllable. For example, artificial consciousness can understand the purpose of "relieving the patient's tremor" and can actively adjust the stimulation parameters; And with interpretable semantic feedback, doctors understand the rationale for each adjustment. This cognitive model of human-computer integration will greatly improve the effectiveness and credibility of brain-computer interface intervention.

It should be pointed out that there is no mature "artificial consciousness + implantable brain-computer" device at home and abroad for direct reference, and this project will carry out research under the condition of no pre-clinical equipment basis, with theoretical and algorithm innovation as the core driving force, and explore a new path of brain-computer fusion intervention. Although this process from 0 to 1 is challenging, it will also fully reflect the original innovation value of the project and provide important support for the realization of the goals of key projects.

Scientific issues

In response to the above background and challenges, this project intends to delve into the following key scientific questions:

• Mechanism of Brain Circuit Remodeling: How Does Long-term Implantable Brain-Computer Fusion Intervention Affect and Reshape the Neural Circuit Function in Parkinson's Disease-Related Brain Regions? In particular, how does exogenous electrical stimulation induce plasticity changes in neurotransmitter regulation (at the microscopic level) and brain network connectivity (at the macroscopic level) to achieve improvement in motor symptoms? At present, little is known about this mechanism, and quantitative studies need to be elucidated.

• Functional Semantic Modeling and Interaction Mapping: How to Construct a Semantic Model of Brain Region Functions and Realize Goal-Driven Semantic Interaction Mapping of Brain Regions? That is, can the patterns of neural activity in various functional areas of the brain be described with interpretable semantic features and mapped to therapeutic goals (e.g., tremor relief, improved motor performance)? How can the activities of different brain regions synergize with the semantic representation corresponding to specific motor functions or symptoms to guide the artificial consciousness to carry out targeted intervention and regulation?

• Closed-loop control of artificial consciousness: How does an agent based on the DIKWP artificial consciousness model form a closed-loop interaction with the brain of patients with Parkinson's disease? Specifically, how can artificial consciousness obtain state information from the cranial nerve signals recorded by implanted electrodes, and obtain the optimal stimulation parameter scheme through the multi-level inference feedback of "data-information-knowledge-wisdom-purpose"? How does this closed-loop system achieve adaptive learning, adjust stimulation strategies based on long-term effects, and ensure the stability and safety of the system in complex and changing physiological environments?

• Multi-objective Response and Predictability: Is the Impact of Artificial Conscious Intervention on Brain Functional Behavior Predictable and Quantifiable? In the face of multiple symptom targets of Parkinson's disease (such as simultaneous control of tremor and muscle rigidity), how can closed-loop intervention achieve multi-objective synergistic regulation, and what are its action pathways and interaction mechanisms? Is it possible to establish models or indicators to predict the evolution of symptoms after artificial consciousness intervention, evaluate potential side effects, and optimize intervention strategies accordingly, so that the system can take care of the main symptoms without introducing new functional impairments?

• Chip-based implementation and system integration: How to turn the DIKWP artificial consciousness model into an implantable neurointervention chip and system prototype? In the absence of a ready-made clinical device foundation, what engineering and technical difficulties (such as real-time signal processing, low power consumption, high security, etc.) need to be overcome in order to organically combine artificial awareness algorithms with implantable hardware? How to design the entire system integration scheme to take into account the practical requirements such as volume and power consumption limitations, wireless communication, and safety isolation, and ensure that the developed prototype has the basis for further preclinical research?

Research objectives and content

The overall goal of this project is to elucidate the remodeling mechanism of the cranial neural circuit in Parkinson's disease under long-term implantable brain-computer fusion intervention, and to develop an intelligent closed-loop neuromodulation system based on the DIKWP artificial consciousness model. Focusing on this goal, the research content of the project mainly includes the following five aspects:

1. Functional semantic modeling of brain regions and goal-driven interaction mapping of brain regions: Functional semantic modeling is carried out for key brain regions related to motor regulation in Parkinson's disease, such as the thalamus, motor cortex, and basal ganglion nuclei in the basal ganglia circuit. By collecting and analyzing neural signals and behavioral data from these brain regions, the semantics of the physiological functions they represent (e.g., motor purpose or pathological state corresponding to different oscillation frequency bands) can be extracted. On this basis, the methodology of the DIKWP model was introduced to establish the semantic interaction mapping relationship between the semantic representations of each brain region and the specific motor function goals. This will shed light on how activity patterns in different brain regions correlate to the semantic goals of motor control, and how this association can be used to achieve goal-driven regulation of neural circuits. For example, patients with Parkinson's disease have abnormally increased activity in the low-β band (approximately 13–20 Hz) of the subthalamic nucleus (STN) and are thought to be biomarkers of bradykinesia and muscle stiffness in dopamine-deficient states. We assign similar neural signal patterns to corresponding semantic labels (such as "motor inhibition signals") to establish the mapping relationship between pathological states and functional goals, and guide artificial consciousness to carry out targeted regulation.

2. DIKWP Model-Driven Closed-loop Inference Feedback and Adaptive Neural Regulation: Design an artificial consciousness control algorithm to embed the DIKWP model into the closed-loop brain-computer system. Specifically, a set of artificial consciousness decision-making engines was developed to enable the extraction of "data-information-knowledge-wisdom" multi-level features from the neural signals recorded by the implanted electrodes, and combined with the "Purpose" module at the top level (e.g., for the purpose of alleviating specific symptoms) to evaluate and inferentially feedback the current pathological state of Parkinson's disease. In terms of signal decoding, deep learning and other methods will be explored to extract reliable information and knowledge features from raw neural data. In terms of decision optimization, adaptive algorithms such as reinforcement learning are introduced to enable artificial consciousness to continuously improve stimulus strategies based on closed-loop feedback. The artificial consciousness engine will generate adaptive stimulation parameters (such as current intensity, frequency, pulse pattern, etc.) in real time, and feed the stimulation signals back to the target brain region through the implanted device, forming a closed-loop regulatory mechanism. At the same time, the self-learning mechanism is studied, so that the artificial consciousness system can continuously adjust internal knowledge and strategies according to the feedback of efficacy, so as to achieve closed-loop learning optimized over time. This part will produce a set of interpretable closed-loop control algorithms for artificial consciousness, which can realize intelligent and adaptive control of neurostimulation and ensure that every step of decision-making is evidence-based.

3. Multi-layer network modeling of brain regions and simulation of artificial consciousness intervention and regulation: Constructing a brain model of Parkinson's disease covering the microscopic neuron and macroscopic brain network levels. On the one hand, neurotransmitter dynamics (such as the modulation of dopamine in the basal ganglia circuit) and synaptic plasticity rules are simulated at the microscopic level to characterize the abnormal activity pattern of neural circuits in Parkinson's disease. On the other hand, at the macro level, the topology and network dynamics model of brain region connectivity were established to reproduce the differences between normal and pathological brain functional networks. In this multi-level model, the artificial consciousness regulation module is introduced to simulate the intervention process of DIKWP artificial consciousness in the brain network: how the neural activity in the model responds and reorganizes when the artificial consciousness emits a specific stimulus pattern. Through simulation experiments, we will observe how long-term artificial consciousness intervention can lead to the remodeling of network structure and function, such as the change of synchronous oscillation mode, the improvement of network efficiency, or the suppression of abnormal rhythms. During the simulation, we will test a variety of closed-loop stimulus paradigms (e.g., periodic stimulus, event-triggered stimulus, etc.) to quantify changes in network synchronization, discharge frequency distribution, information transfer efficiency, and other indicators to identify key features of loop remodeling. In particular, we looked at the presence of changes such as a decrease in the intensity of abnormal β oscillations, an increase in the efficiency of effective connections, and assessed the association of these changes with symptom improvement. This study will provide a quantitative basis for understanding how AI intervention induces brain network plasticity, and reveal the mechanism of circuit remodeling behind the improvement of Parkinson's disease symptoms.

4. Non-human primate simulation verification and multi-objective response analysis: Non-human primate (such as Parkinson's disease monkey model) experiments or high-fidelity simulation platform (Parkinson's disease monkey model can be obtained by neurotoxin MPTP induction) to verify and predict the artificial consciousness intervention strategy. Design experimental tasks in which a closed-loop stimulation system driven by artificial consciousness intervenes in the motor regulation processes of the primate model, such as inhibiting limb tremor or improving gait motor coordination in real time. During the experiment, multi-dimensional outcome indicators (multi-objective) were monitored, including changes in motor symptom scores, physiological signal characteristics, and cognitive/affective side effects. By comparing and analyzing animal behavior and neural activity before and after artificial consciousness intervention, we evaluated the response effect of the closed-loop system to different symptom targets and their interaction pathways. Focus on the predictability of the effect of artificial consciousness intervention: can the simulation model predict the trajectory of symptoms after intervention? For multi-objective control (e.g., simultaneous control of tremor and muscle rigidity), can artificial consciousness coordinate and optimize each indicator without compromising the other? This section will verify the effectiveness and safety of artificial consciousness intervention in complex real-life physiological environments, and provide a basis for subsequent clinical applications.

5. Prototype of DIKWP Artificial Consciousness Chip-driven Implantable Neural Intervention System: Based on the above theoretical and model research, a prototype of artificial consciousness closed-loop neural intervention technology was developed. Development of a dedicated DIKWP Artificial Consciousness Processing Unit (ACPU). The artificial consciousness decision-making algorithm is implemented on the hardware platform to meet the stringent requirements of implantable applications for real-time performance and low power consumption. The implantable neurostimulation system is designed, including modules such as multi-channel electrode array and cortical implant interface, and the ACPU chip is embedded in the body device to realize the recording of nerve signals targeting the brain region and the precise stimulation output. At the system control level, the semantic plasticity control mechanism of electrode parameters is introduced, which allows the stimulation parameters to be dynamically adjusted with the update of artificial consciousness knowledge, so as to ensure that the stimulation mode after long-term implantation has the ability to adapt to evolve. At the same time, a closed-loop monitoring module of explainable algorithms at the target level was constructed to interpret and review the safety of artificial consciousness decisions (for example, the semantic basis for each stimulus adjustment was given to avoid abnormal output). In addition, we will develop a basic software system (similar to the Artificial Consciousness Operating System, ACOS) of the artificial consciousness chip, which is used to coordinate the operation of each functional module of the chip and the flow of semantic information, and embed a security monitoring mechanism to ensure that the system behavior is controllable and monitorable. Bench tests and preliminary animal experiments were conducted to verify the functional feasibility and safety of the prototype system. The development of this prototype will lay the foundation for the subsequent development of clinical brain-computer fusion devices, and reflect the exemplary value of this project in promoting technological progress with algorithm and architecture innovation in the absence of off-the-shelf equipment.

Key technical difficulties and solutions

In the process of realizing this project, a number of key technical difficulties need to be overcome. For each difficulty, we propose the following solutions:

• Semantic problem of brain signals: The nerve signals in the brain of patients with Parkinson's disease are complex and changeable, and how to convert massive and noisy neural data into semantic information with clear physiological significance is a major challenge. For example, multi-channel EEG/local field potential signals are often mixed with EMG artifacts and environmental noise, and advanced signal processing and pattern recognition techniques must be applied to extract useful information. It is difficult to directly extract the correspondence between the oscillation frequency and discharge pattern of different brain regions and the motor purpose or symptom state. Solution: Using multimodal data fusion and semantic modeling technology, the hierarchical cognitive framework of the DIKWP model is used to abstract brain signals step by step (extracting information and knowledge from raw data, and then rising to Wisdom and Purpose). Specifically, through time-frequency analysis, machine learning classification and other methods, stable feature patterns were extracted from the original neural signals. Combined with the knowledge graph related to Parkinson's disease, these patterns were assigned physiological semantic labels to construct a functional semantic dictionary of brain regions. Then, with the goal of improving specific symptoms, the mapping relationship between brain signal features and treatment goals was aligned to form a "signal pattern-semantics-target" correspondence framework. In this way, the artificial consciousness system can read the "semantic signals" of the brain and achieve a high-level understanding of the state of the brain.

• Stability and adaptability of closed-loop control: Closed-loop neuromodulation based on artificial consciousness requires algorithms to analyze brain states in real time and dynamically adjust stimuli in milliseconds. In a complex and volatile physiological environment, the algorithm must remain stable without out-of-control oscillations or long decision delays. How to ensure the stable convergence and rapid response of the long-term operation of the closed-loop system is a major technical difficulty. Solution: The "dual circulation" architecture of DIKWP × DIKWP is introduced to add a metacognitive monitoring loop in addition to the main cycle of artificial consciousness decision-making. The main circulation evaluates the brain state and stimulates decision-making in real time, and the meta-circulation supervises and reflects on the output of the main circulation, discovers abnormalities in time and adjusts internal parameters, so as to achieve stable control under self-monitoring. At the same time, by setting up the reward and punishment mechanism and introducing the reinforcement learning algorithm, the artificial consciousness can continuously optimize the stimulation strategy in the closed-loop interaction. Safety constraints are imposed in the algorithm, such as limiting the rate and amplitude of change in stimulus parameters to avoid violent oscillations. In addition, asynchronous parallel processing is adopted in the system implementation to improve the decision-making throughput and ensure real-time performance. A multi-pronged approach ensures that the closed-loop control not only has adaptive optimization capabilities, but also always maintains a safe and stable working range.

• Multi-scale brain model accuracy: Accurate multi-scale brain models are required to simulate the effects of artificial consciousness intervention on brain circuits. However, the brain spans multiple levels from a single neuron to a brain region network, with a large number of parameters and a wide range of distributions, and large individual differences (such as synaptic connection strength, dopamine concentration, etc., which are difficult to accurately measure), and it is extremely challenging to establish a model that can be close to real physiology. Solution: Adopt the method of hierarchical modeling and step-by-step verification to reduce complexity. Firstly, a local neural circuit model was constructed based on the literature and experimental data to simulate the microscopic pathology of Parkinson's disease (such as the imbalance of excitation/inhibition of the basal ganglia pathway caused by dopamine deficiency), and to verify whether the model can reproduce the known abnormal discharge patterns. Then, the local model was integrated into a macroscopic brain network model, and the anatomical connection data was fused to reproduce the interactive behavior of multiple brain regions of the brain, and the model parameters were adjusted through the electrophysiological data of the existing Parkinson's animal model, so that the output of the model was consistent with the characteristics of the real brain network. On the built model, the artificial consciousness algorithm was introduced for simulation intervention, and the key parameters or mechanism assumptions in the model were continuously revised by comparing the simulation results with experimental observations, so as to improve the credibility of the model prediction. At the same time, the sensitivity analysis of the model parameters is carried out to quantify the uncertainty range to ensure the robustness of the simulation conclusions. Through such cyclic iterations, the real situation is gradually approximated, so that the model is not only biologically rational, but also can provide guidance for algorithm optimization.

• Artificial consciousness chip and implantation technology difficulties: The implementation of complex artificial consciousness algorithm into an implantable hardware system needs to solve a series of engineering problems such as limited computing resources, power consumption control, heat dissipation, and biocompatibility. At present, there is no special chip or implant device for artificial consciousness that can be directly used, and everything needs to be developed from scratch. Solution: Adopt a combination of ASIC ASIC design or FPGA prototyping. First, build a prototype on the FPGA to verify the function and performance of the artificial consciousness algorithm, and then design the ASIC after the algorithm is mature to reduce the risk of repeated tape-out. The chip architecture is designed to optimize parallel computing and storage access, and use fixed-point computing and model compression to reduce power consumption. The package design is designed in collaboration with the medical engineering team to develop a stable and reliable implantable electrode interface using highly biocompatible materials and miniaturized packaging to ensure long-term implantation safety. In addition, the tissue reactions that may be triggered by long-term electrode implantation (such as signal attenuation due to gliosis) should be fully considered, and mitigation measures should be taken in terms of materials and structures. Through rigorous fatigue testing and environmental simulation, it ensures the reliability of the hardware in the body for long-term operation. In case of lagging behind the chip R&D progress, the preparatory plan is to use in vitro equipment to simulate the chip function in the short term to ensure that the overall research does not stop.

• System security and algorithm controllability: There is a safety risk for "artificial consciousness" to directly intervene in the patient's brain, and it is necessary to ensure that the system decision-making is reliable and there is no dangerous output. Especially in the absence of unprecedented equipment to refer to, how to prove that the algorithm is safe and effective is critical to the success of the project. Solution: Introduce explainability constraints and hierarchical monitoring mechanisms at the algorithm level. Every step of the decision is accompanied by an explainable basis, which is reviewed by the monitoring module, eliminating completely agnostic black box operations. At the same time, multiple levels of safety protection are set up, such as triggering an emergency shutdown or switching to a safe mode (e.g., resuming a preset fixed stimulus) when abnormal brain activity is detected or the algorithm output is out of bounds. During the implementation of the project, experiments were carried out step by step: a large number of offline data analysis and simulation were carried out to verify the algorithm behavior, then small-scale animal experiments were carried out to evaluate the safety, and finally the experimental scope was expanded with the support of sufficient evidence. The project team will also work closely with ethics experts and clinicians to develop detailed risk plans and emergency response procedures. In the event of an unforeseen situation, activate the plan immediately to minimize the risk. Through the above measures, the safety and controllability of the closed-loop system of artificial consciousness can be ensured to the greatest extent, and a reliable foundation will be laid for possible clinical trials in the future.

Technical route

The project is progressing gradually according to the technical path of "theoretical research→ simulation verification→ prototype development", and each research task is connected with each other and iteratively optimized to form a closed-loop R&D process. Firstly, the semantic representation of brain function in Parkinson's disease was obtained through semantic modeling of brain regions, which laid a knowledge foundation for subsequent intervention. On this basis, a closed-loop control algorithm driven by DIKWP artificial consciousness was developed to endow the system with the ability of independent analysis and decision-making. Then, the algorithm was simulated and tested in a multi-level brain network model, and the brain circuit effect of artificial consciousness intervention was predicted and evaluated, and the algorithm parameters and model assumptions were optimized accordingly. Next, the optimized artificial consciousness closed-loop system was introduced into the non-human primate experimental platform for verification, the system performance and safety were investigated in the real biological system, and the experimental results were used to improve the model and algorithm. Finally, the development and integration of artificial consciousness chip and implantable intervention system were completed, and the overall system function was verified through bench tests and animal experiments. The whole technical route guides practice with theory and corrects theory with experiments, gradually reduces research risks, ensures the coherence of the results of each stage of the project, and finally achieves the expected goals of the project.

Phased task scheduling

In order to ensure the orderly implementation of the project, we will proceed gradually in stages, and the specific plan is as follows:

1. Phase 1 (Year 1-2): Focus on basic theories and key models. Completed the functional semantic modeling of the brain region of Parkinson's disease, and constructed the semantic database of the brain region (including the functional semantic definition of the relevant brain region of Parkinson's disease) and the preliminary interaction mapping model of the brain region (corresponding to research content 1). A prototype of a closed-loop control algorithm for artificial consciousness was developed in parallel to realize the neural signal decoding and stimulation parameter generation of typical scenarios (such as simulated tremor) in a simulated environment (corresponding to research content 2). At the end of the stage, the preliminary results of the semantic model of the cerebral circuit of Parkinson's disease and the decision-making algorithm of artificial consciousness were produced, and one paper was written and the invention patent was applied for. The first stage lays the theoretical and methodological foundation of the project.

2. Phase 2 (Year 3–4): Focus on simulation validation and system optimization. The artificial consciousness algorithm developed in the first stage was embedded in the multi-level brain network model, and large-scale simulation experiments were carried out on the computer to analyze the impact of artificial consciousness intervention on the dynamic characteristics of the brain circuit, verify the hypothesis of brain circuit remodeling, and improve the algorithm accordingly (corresponding to the research content 3). At the same time, preliminary validation was carried out on the non-human primate experimental platform: the MPTP-induced Parkinson's disease monkey model was selected, the artificial consciousness closed-loop stimulation system was introduced to test the effect of real-time control of tremor and other symptoms, and multi-objective response data were collected (corresponding to the research content 4). Combined with the simulation and experimental results, the parameters and models of the artificial consciousness algorithm are optimized, so that the system has good robustness and generalization. In the second stage, it is expected to publish 2 research papers, forming an in-depth understanding of the brain-computer fusion remodeling mechanism and a set of optimized closed-loop control algorithms, so as to prepare for the development of prototypes.

3. Phase 3 (Year 5): Focus on system integration and prototyping. The DIKWP artificial consciousness chip was developed and the prototype of the implantable neural intervention system was built, including electrode array, signal acquisition/stimulation module, wireless communication and power management (corresponding to research content 5). Complete the chip tape-out or FPGA implementation, integrate the software and hardware systems, and debug repeatedly. The performance of the prototype system was then evaluated in animal experiments: focusing on long-term implantation stability, biocompatibility, and the efficacy and safety of closed-loop stimulation for symptom improvement. According to the test results, the system design is improved and finalized, and the principle prototype and technical report of the artificial consciousness brain-computer fusion intervention system are finally delivered. In the third stage, it is planned to publish 1-2 papers, apply for 1-2 related patents, and comprehensively summarize the project results. The completion of this phase marks the achievement of the project goals and lays a solid foundation for subsequent preclinical research and application translation.

Composition of the research team

This project is composed of multidisciplinary experts with rich experience in the intersection of artificial intelligence and brain science, and the team has the comprehensive R&D capabilities required to complete this project:

• Project Leader: Prof. Yucong Duan (Supported by the Active Medicine Committee of the World Association of Artificial Consciousness) – Project Leader, the proposer of the DIKWP Theory of Artificial Consciousness. As the first inventor, he has obtained more than 100 authorized invention patents at home and abroad, and has a deep accumulation in the interdisciplinary research of artificial intelligence and brain science. Professor Duan is responsible for the overall program design, the overall planning of the research route, and the guidance of the direction of artificial consciousness algorithms.

• Experts in Artificial Intelligence and Conscious Computing – The core members of the team include many professors and researchers with profound attainments in cognitive computing and basic theories of artificial intelligence, who have been engaged in the research of machine learning, large models and artificial consciousness models for a long time, and are proficient in the principles and implementation of the DIKWP framework. In the project, he is responsible for the research and development of artificial consciousness algorithms, the semantic decoding of brain signals and the design of the overall system architecture, and is the main undertaker of algorithm innovation.

• Neuroscience & Clinical Medicine Specialists – including neurobiologists and clinical neurologists familiar with Parkinson's disease and movement disorders. One team member is the chief physician of the Department of Neurology of a tertiary hospital, with rich experience in the clinical diagnosis and treatment of Parkinson's disease; The other is a researcher at the Institute of Neuroscience, specializing in Parkinson's animal models and deep brain stimulation mechanisms. They are responsible for providing guidance on disease mechanisms, primate experimental design, and interpretation of the biological significance of brain circuit remodeling, ensuring that research is closely related to clinical practice.

• Brain-Computer Interface and Biomedical Engineering Specialists – The team includes engineering experts with extensive experience in brain-computer interfaces and implantable medical devices, including a professor of electrical engineering and a senior engineer in biomedical engineering. They are involved in implantable electrode design, chip hardware development, wireless communication and energy supply solutions, and other tasks to ensure that artificial awareness algorithms can be implemented on real hardware and meet safety specifications and engineering constraints.

• Young scientific research backbone – A number of young teachers, postdoctoral fellows and doctoral students with doctoral degrees are the backbone of the project, with expertise in data analysis, software development, integrated circuit design, animal experiment technology, etc. They will implement the research plan in each topic direction, promote the progress of the project on a daily basis, and grow into compound innovative talents through project training.

• Collaboration and Division of Labor – This project will rely on the superior resources of the supporting units, and cooperate with top universities and research institutions in China to tackle key problems. For example, it plans to cooperate with a university brain science research institute to carry out primate experiments, and it will provide a professional primate center platform and technical support; Cooperate with the neurology department of a large hospital to evaluate clinical indicators and demonstrate safety. Each unit has clear responsibilities, complements each other's advantages, and ensures efficient cooperation through regular discussions and docking. Cross-unit collaborative innovation will provide strong support for the solution of project problems.

Overall, the project team has a team of about 15 people, including 5 senior researchers, 3 associate senior staff, and a number of postdoctoral and graduate students, covering multidisciplinary professionals such as artificial intelligence, neuroscience, and biomedical engineering. This complementary and well-organized team will provide a strong guarantee for the smooth implementation of the project.

Risk assessment and contingency plans

This project is highly innovative, but also accompanied by certain technical risks. We have a comprehensive plan for potential risks:

• The effect of the algorithm is not as expected: The closed-loop algorithm of artificial consciousness may not achieve the expected effect in the actual complex brain signal environment, and even make mistakes in decision-making. Planning: Develop multiple sets of algorithm options (such as signal decoding and control strategies based on different models) in advance, and select the best performers through simulation and small-sample experiments. Once an algorithm does not work well in subsequent experiments, it will quickly switch to an alternative. At the same time, expert advisors are introduced to regularly evaluate the output of the algorithm, and the results are manually verified to prevent systematic deviations, and the algorithm parameters are continuously adjusted according to the feedback until they are stable and reliable.

• Models do not match reality: Simulation models may not fully reproduce the response of real brains to stimuli, resulting in predictions that are inconsistent with actual experiments. Planning: Adopt the "modeling-experiment" iterative method to continuously revise the model with new experimental data. If the key parameters of the model are found to be inaccurate, the model structure or parameter distribution should be adjusted immediately based on the experimental results. When necessary, data-driven black-box models are introduced to replace the parts that are difficult to model and improve the overall simulation accuracy. If certain hypotheses (e.g., specific plasticity mechanisms) are falsified, the theoretical framework should be updated in time to ensure that the research direction is flexibly adjusted and not confined to the original model.

• Delays in hardware development or insufficient performance: Technical difficulties may be encountered during the development of artificial awareness chips or implantable devices, resulting in lagging progress or performance substandard (such as excessive power consumption and excessive size). Preliminary plan: Adopt modular parallel R&D, give priority to the verification of key functional modules, and then integrate and optimize to shorten the R&D path. Develop a detailed R&D milestone plan, closely monitor each node, and identify and resolve issues in a timely manner. If the ASIC development is not completed in time, we will consider replacing it with FPGAs or existing wearable devices in the short term to complete the experimental verification and continue to optimize the ASIC design. In terms of resource allocation, funds and manpower are reserved to meet the sudden needs of technical research and ensure that hardware research and development are generally controllable.

• Uncertainty in animal experiments: Experimental results in primate models of Parkinson's disease may vary greatly from person to person, or with unforeseen safety issues (e.g., unexpected conditions in animals). Preliminary plan: Strictly follow the ethical requirements of animal experiments, first conduct low-dose and safe tests on a small number of experimental animals, observe their behavior and physiological changes, and then gradually expand the sample and stimulation intensity after confirming safety. Establish an emergency plan, such as stopping the stimulation immediately and intervening to properly treat the animal if it is found to be experiencing severe discomfort or abnormal behavior. At the same time, prepare alternative protocols: if primate experiments are blocked, consider turning to rodents for partial validation, or utilizing ex vivo brain slice/organoid experiments to obtain validation data to reduce reliance on a single model. Through multi-channel data acquisition, ensure that key scientific questions are supported by validation.

• Risk of multidisciplinary collaboration: The project involves multiple fields such as artificial intelligence, neuromedicine, and electronic engineering, and the team members have different backgrounds, making it difficult to communicate collaboratively and coordinate schedules. Plan: Establish a standardized project management and communication mechanism. Hold regular plenary meetings and symposia to unify technical roadmaps and terminology. The project leader will take the lead, and the person in charge of each sub-project will be responsible for daily coordination, and a special person will be designated to track the progress. Introduce project management tools to manage tasks on Gantt charts, identify delays and adjust resources in a timely manner. If there is a change in team members, quickly use the back-up talent or recruit new members to fill the staff. Through the above measures, the project team can be operated efficiently and collaboratively, and the risk of delays caused by poor communication will be reduced.

Expected outcomes and translatable application value

Through the implementation of the project, we expect to achieve the following results and create significant application value:

• Theoretical breakthrough: elucidate the mechanism of long-term implantable brain-computer interface on the remodeling of brain circuits in Parkinson's disease, and enrich the scientific understanding of neuroplasticity in movement disorders. This theoretical progress will fill the international gap in the research on brain-computer fusion-induced brain plasticity, and is expected to be published in top journals in the field of brain science or neural engineering, providing a solid foundation for subsequent related research and theoretical guidance for the formulation of more effective intervention strategies.

• Innovation of artificial consciousness model: The DIKWP artificial consciousness model was successfully applied to the field of neuromodulation, and an artificial consciousness algorithm framework for brain-computer interface was proposed. The framework realizes the closed-loop control of the whole process of AI from "perception-cognition-decision-purpose", and can be traced and explained for each step of operation. This not only verifies the effectiveness of the DIKWP model in complex biological systems, but also provides a new idea for solving the current "black box" problem of AI systems and improving the interpretability of AI decisions, which can be further extended to other AI medical scenarios and promote the development of AI towards a higher level of autonomy and security.

• Key technologies and prototype systems: The principle prototype of the implantable neural intervention system driven by the DIKWP artificial consciousness chip was developed. The prototype has the ability to autonomously perceive the pathological state and dynamically adjust the stimulus, which is expected to achieve effective relief of Parkinson's symptoms in animal experiments. It is expected to apply for more than 3 domestic and foreign invention patents, form independent intellectual property rights, and lay a technical foundation for the development of a new generation of intelligent brain pacemaker in China. This move will enhance China's independent innovation strength in the field of high-end medical devices, reduce dependence on imported equipment, and inject impetus into the development of related industries.

• Papers, patents, and talent training: During the implementation of the project, it is expected to publish more than 5 papers in international authoritative journals (at least 3 of which are included in SCI) to enhance China's academic influence in the interdisciplinary field of artificial intelligence and brain-computer interface; At the same time, it has applied for 3-5 domestic and foreign invention patents to ensure the intellectual property protection of core technologies. In addition, through the practice of the project, a number of young scientific research talents such as doctoral students and postdoctoral fellows will be trained, and an interdisciplinary team proficient in brain-computer integration and artificial intelligence will be trained, providing a valuable talent reserve for the sustainable development of this field in China.

• Application prospects: The results of this project have important potential value in clinical and industrial fields. For patients with Parkinson's disease, the intelligent closed-loop stimulation system based on artificial consciousness is expected to provide a more refined and efficient symptom management method, reduce the dosage of drugs and side effects while reducing motor symptoms, and improve the quality of life of patients. This technology can also be extended to other neurological diseases, such as essential tremor, epilepsy, depression and other diseases that require brain stimulation treatment, so as to achieve a safer and more personalized treatment plan. From an industrial point of view, the new paradigm of "artificial consciousness + brain-computer interface" explored in this project will lead the new direction of medical AI and neuromodulation technology, and is expected to give birth to a new generation of high-end medical device products, forming China's first-mover advantage in this field. The research results will also lay a solid foundation for follow-up clinical translational research (such as human clinical trials), ultimately provide cutting-edge scientific and technological support for the construction of "Healthy China", and promote the coordinated development of brain science and brain-like intelligence in China.