WO2023087917A1 - 基于多维分层漂移扩散模型的认知决策评估方法及系统 - Google Patents

基于多维分层漂移扩散模型的认知决策评估方法及系统 Download PDF

Info

Publication number
WO2023087917A1
WO2023087917A1 PCT/CN2022/120985 CN2022120985W WO2023087917A1 WO 2023087917 A1 WO2023087917 A1 WO 2023087917A1 CN 2022120985 W CN2022120985 W CN 2022120985W WO 2023087917 A1 WO2023087917 A1 WO 2023087917A1
Authority
WO
WIPO (PCT)
Prior art keywords
cognitive
distribution
decision
making
overall
Prior art date
Application number
PCT/CN2022/120985
Other languages
English (en)
French (fr)
Inventor
李诗怡
李嘉
马珠江
王晓怡
Original Assignee
北京智精灵科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京智精灵科技有限公司 filed Critical 北京智精灵科技有限公司
Publication of WO2023087917A1 publication Critical patent/WO2023087917A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients

Definitions

  • the invention relates to a cognitive decision-making evaluation method based on a multidimensional layered drift diffusion model, and also relates to a corresponding cognitive decision-making evaluation system, belonging to the field of medical care informatics.
  • Mild Cognitive Impairment is a prodromal state of Alzheimer's disease, an intermediate state between normal aging and dementia, and can be used as a "predictor" of Alzheimer's disease.
  • DDM Drift Diffusion Model
  • DDM simulates and refines the neural dynamic process of the human brain in decision-making tasks by separating the potential decision-making components contained in individual behavioral response data.
  • DDM is mainly applied to a single perception paradigm and decision paradigm to model the neural decision-making process of healthy people.
  • HDDM Hierarchical Drift Diffusion Model
  • the primary technical problem to be solved by the present invention is to provide a cognitive decision-making evaluation method based on a multi-dimensional layered drift diffusion model, which is used to realize the accurate evaluation of the impaired ability of MCI patients and promote the early evaluation of the impaired cognitive domain. Discovery, early intervention and early treatment.
  • Another technical problem to be solved by the present invention is to provide a cognitive decision-making evaluation system based on a multi-dimensional layered drift diffusion model.
  • a cognitive decision-making evaluation method based on a multidimensional hierarchical drift-diffusion model including the following steps:
  • a multidimensional hierarchical drift-diffusion model is established at the group level, and the sensitivity threshold and probability distribution of the healthy subjects in each cognitive domain are obtained through the best-fitting model and the sensitivity threshold and probability distribution of the overall metacognitive processing efficiency, as a norm for healthy people;
  • a multidimensional hierarchical drift-diffusion model is established at the individual level, and the sensitivity threshold and its probability distribution in each cognitive domain of the MCI patient and the overall The sensitivity threshold of metacognitive processing efficiency and its probability distribution are used as indicators to be measured;
  • the sensitivity threshold and the probability distribution thereof of each cognitive domain of the healthy subject and the sensitivity threshold and the probability distribution thereof of the overall metacognitive processing efficiency are obtained, specifically comprising:
  • samples are sampled from a known distribution to obtain a transition sequence
  • the transfer sequence continuously generates a new transfer sequence until the transfer sequence no longer changes and tends to a stable state
  • the posterior distribution of the model parameters is a parameter distribution at the group level
  • the parameter distribution at the group level includes the sensitivity threshold and the probability distribution of each cognitive domain of healthy subjects at the group level Sensitivity thresholds and their probability distributions for overall metacognitive processing efficiency.
  • obtaining the sensitivity threshold and probability distribution of each cognitive domain of the MCI patient and the sensitivity threshold and probability distribution of the overall metacognitive processing efficiency specifically include:
  • samples are sampled from a known distribution to obtain a transition sequence
  • the transfer sequence continuously generates a new transfer sequence until the transfer sequence no longer changes and tends to a stable state
  • the posterior distribution of the model parameters is the parameter distribution at the individual level
  • the parameter distribution at the individual level includes the sensitivity threshold and its probability distribution of each cognitive domain of each MCI patient at the individual level and the overall Sensitivity thresholds and their probability distributions for metacognitive processing efficiency.
  • comparing the relative positions of the indicators to be measured of the MCI patients in the normal model of healthy people, assessing the impairment of the MCI patients in each cognitive domain and overall metacognitive processing efficiency specifically includes :
  • the significance test is carried out on the parameter posterior density distribution, and the area of overlap between the normal model distribution of each parameter obtained by Bayesian hypothesis testing and the sample percentage extracted from the posterior density distribution of each MCI patient;
  • the MCI patients were assessed for cognitive impairment in each cognitive domain.
  • the cognitive domains at least include: sensory decision-making cognitive domains, advanced cognitive decision-making cognitive domains, and social cognitive decision-making cognitive domains.
  • the multidimensional layered drift-diffusion model is established based on the behavioral response data of subjects in different cognitive domains.
  • a cognitive decision-making evaluation system based on a multidimensional hierarchical drift-diffusion model, including a processor and a memory, the processor reads the computer program in the memory for execution Cognitive decision-making assessment methods described above.
  • the cognitive decision-making evaluation method based on the multi-dimensional hierarchical drift-diffusion model provided by the present invention adopts the posterior distribution of the sensitivity thresholds of each cognitive domain and the overall metacognitive processing efficiency calculated based on the multi-dimensional hierarchical drift-diffusion model, and Comparing the norms of MCI patients and normal people, it is possible to more accurately evaluate the relative positions of the sensitivity thresholds of each cognitive domain and the overall metacognitive processing efficiency of MCI patients in the normal norms of healthy people, and obtain the overall damage and status of MCI patients.
  • the damage status of each sub-cognitive domain realizes the accurate assessment of the impaired ability of MCI patients, and promotes the early detection, early intervention and early treatment of the decision-making ability in the damaged cognitive domain.
  • Figure 1 is a schematic diagram of the modeling process of the multidimensional layered drift-diffusion model
  • Figure 2 is a schematic diagram of the trajectory of the belief accumulation of the simulated subjects in the binary cognitive decision-making process
  • Fig. 3 is the layered structure schematic diagram of multidimensional layered drift-diffusion model
  • FIG. 4 is a flow chart of a cognitive decision-making evaluation method based on a multidimensional layered drift-diffusion model provided by an embodiment of the present invention
  • FIG. 5 is a schematic structural diagram of a cognitive decision-making evaluation system based on a multidimensional hierarchical drift-diffusion model provided by an embodiment of the present invention.
  • FIG. 1 is a schematic diagram of the modeling process of the multi-dimensional layered drift-diffusion model in an embodiment of the present invention.
  • the modeling process specifically includes the following steps:
  • S1 Select and determine the cognitive decision-making tasks and paradigms that need to be implemented, and obtain the behavioral response data of the subjects on each cognitive decision-making task.
  • cognitive decision-making tasks include at least: sensory decision-making tasks of judging the movement direction or spatial position of moving points; advanced cognitive decision-making of making choices based on the relative value of options, such as memory, reasoning, and executive control Task; a social cognitive decision-making task that learns attributes of others and evaluates social information. Therefore, according to the task content and attributes, different decision-making tasks can be labeled as tasks under the three cognitive domains of perceptual decision-making, advanced cognitive decision-making and social cognitive decision-making.
  • RT response time
  • DO decision option
  • steps S21-S24 are included:
  • the offset diffusion model is constructed by the following formula:
  • E represents the cumulative evidence balance that one option is better than another option, and provides the deviation of two options, that is, the subject's preference for an option before decision-making, when the subject's previous experience of two decisions is comparable
  • Et represents the cumulative deviation of the two options at the t-th trial (that is, the t-th specific task round);
  • t represents the time when thinking begins, including the time to encode information and make a key response ;
  • ⁇ represents the incremental evidence supporting an option in each decision time;
  • d represents the drift rate coefficient of the subject, which represents the relative amount of information about the stimulus accumulated by the individual in each time unit, which characterizes the speed of information accumulation and the measurement of individual perception sensitivity;
  • ⁇ i is the subjective value of option i, the decision threshold d( ⁇ 1 - ⁇ 2 ), represents the difference between two decision options, and represents the boundary/threshold of decision-making information collection, when the amount of accumulated information When the threshold is reached, the subject will make a decision;
  • ⁇ 2 is the
  • the model is used to simulate the example trajectory of belief accumulation of subjects in the process of binary cognitive decision-making (that is, making a decision from two options).
  • the selection probability (P) and reaction time (RT) are functions of the drift rate coefficient d, the diffusion noise coefficient ⁇ 2, and the height of the response boundary ⁇ .
  • ⁇ is the height of the evidence accumulation boundary that triggers the decision (this upper bound is the choice of option 1)
  • RT is the expected time for the end of the accumulation process (reaching either boundary ).
  • S23 Calculate parameter estimates of the offset-diffusion model, wherein the parameters of the offset-diffusion model include: the subject's decision threshold, relative starting point, drift rate, and non-decision time;
  • the input subject’s decision-making behavior response data includes decision-making time (RT) and option (DO) as reaction time (PT) and choice probability (P), according to formula 1 and formula 2, calculate the parameter estimation value of the DDM model .
  • RT decision-making time
  • DO reaction time
  • P choice probability
  • S24 Integrate the response data of subjects in multiple decision-making paradigms in different cognitive domains into the same model framework to form a multi-dimensional decision-making trial-decision-making task-cognitive domain-overall metacognitive processing efficiency from low to high Hierarchical drift-diffusion models.
  • the parameter estimation at the low-level level will be constrained by the parameter distribution at the higher-level level.
  • the level of decision-making trials-decision-making task as an example, for the observation data xi, j, k of each subject that obeys a certain likelihood function f
  • the fixed prior for the trial-level parameter ⁇ is a random variable parameterized by the task-level parameter ⁇ , from which the posterior probability formula for ⁇ is obtained:
  • x) is the likelihood function, indicating the possibility that the trial-level parameter is ⁇ and the task-level parameter is ⁇ given the observation data x
  • ⁇ ) is the given
  • ⁇ ) indicates the possibility that the trial-level parameter is equal to ⁇ when the task-level parameter is ⁇
  • P( ⁇ ) and P(x) Indicates the probability that the task-level parameter is ⁇ and the observed data is equal to x.
  • the more layers of the hierarchical model the higher the level of estimated parameters, and the more complex the posterior probability estimation formula is.
  • xi, j, k ⁇ L(ai, zi, vi, ti, sv, st, sz).
  • xi, j, k represent the subject's kth decision-making trial (Trial, Abbreviated as T), behavior observation data including decision time (RT) and option (DO);
  • L represents the likelihood function of the DDM model corresponding to each cognitive domain;
  • ai, zi, vi and ti respectively represent The decision threshold, relative starting point, drift rate, and non-decision time distribution of the i-th cognitive domain.
  • Other parameters ⁇ represent the mean; ⁇ represents the standard deviation; ak, vk, zk and tk represent the decision threshold, drift rate, relative starting point and non-decision time in the kth decision trial.
  • the Gelman-Rubin R statistic is used to evaluate the convergence of the current model. The closer the value is to 1, the better the convergence of the model is;
  • the model parameter settings are constantly adjusted to obtain a suitable multidimensional hierarchical drift-diffusion model and the corresponding parameter estimates and posterior probability distributions.
  • This multidimensional hierarchical drift-diffusion model is a multidimensional hierarchical drift-diffusion model that evaluates the cognitive decision-making of MCI patients, and is used to assist in understanding the cognitive decision-making of MCI patients. damaged condition.
  • FIG. 4 is a cognitive decision-making evaluation method based on a multi-dimensional hierarchical drift diffusion model provided by an embodiment of the present invention, specifically including steps S10-S30:
  • a multidimensional hierarchical drift-diffusion model is established at the group level to obtain the The sensitivity threshold and its probability distribution of the overall metacognitive processing efficiency, and the sensitivity threshold and its probability distribution of the overall metacognitive processing efficiency, these two distribution models are used as the norm for healthy people.
  • sensitivity thresholds and probability distributions of the healthy subjects in each cognitive domain are used as the criteria for assessing the impairment of each cognitive domain of MCI patients; Sensitivity thresholds of overall metacognitive processing efficiency and their probability distributions were used as criteria to assess the impairment of overall metacognitive processing efficiency in MCI patients.
  • a multidimensional hierarchical drift-diffusion model is established at the individual level, and the model with the best fitting degree is used to obtain the MCI patient.
  • the sensitivity threshold and its probability distribution in each cognitive domain and the sensitivity threshold and its probability distribution of the overall metacognitive processing efficiency are used as indicators to be measured.
  • the index to be measured not only includes the sensitivity threshold and its probability distribution of MCI patients in each cognitive domain, but also includes the sensitivity threshold and its probability distribution of MCI patients in the overall metacognitive processing efficiency, It is thus used to assess the impairment of MCI patients in each cognitive domain and its overall cognitive processing efficiency.
  • S30 Compare the relative positions of the indicators to be measured in MCI patients in the normal norm of healthy people, and evaluate the impairment of MCI patients in each cognitive domain and overall metacognitive processing efficiency.
  • comparing the relative positions of the indicators to be measured in MCI patients in the normal model of healthy people specifically includes steps S301-S302:
  • S301 Calculate the posterior density distribution of the normal norm of healthy people for each parameter, and the posterior density distribution of the sensitivity threshold index of MCI patients;
  • the significance test was carried out on the parameter posterior density distribution, and the overlapping area between the normal norm distribution of each parameter obtained from healthy people and the sample percentage extracted from the posterior density distribution of each MCI patient was measured by Bayesian hypothesis test.
  • the posterior density distribution of the normal norm of healthy people and the posterior density distribution of the sensitivity threshold index of MCI patients the smaller the overlapping area between the two distributions, the larger the difference between the mean values, the more damaged the MCI patients are. serious.
  • the overall distribution refers to the overall distribution obtained according to the posterior density distribution obtained under each cognitive domain.
  • the above-mentioned method can realize accurate assessment of the impaired ability of MCI patients, and promote early detection, early intervention and early treatment of decision-making ability in the impaired cognitive domain.
  • the sensitivity threshold and probability distribution of each cognitive domain and the sensitivity threshold and probability distribution of the overall metacognitive processing efficiency are obtained in the same way, except for For healthy subjects, the sensitivity threshold and its probability distribution of each cognitive domain and the sensitivity threshold and its probability distribution of the overall metacognitive processing efficiency are used as the normal model of healthy people; for MCI patients, The obtained sensitivity threshold and its probability distribution of each cognitive domain and the sensitivity threshold and its probability distribution of the overall metacognitive processing efficiency are used as indicators to be measured.
  • S402 Obtain samples from a known distribution based on the state transition matrix to obtain a transition sequence
  • the transfer sequence is constantly generating new transfer sequences under the action of the same transfer matrix (the number of actions is 1, 2, ..., n), until the transfer sequence no longer changes and tends to a stable state;
  • the parameter distribution of the group level includes the sensitivity threshold of each cognitive domain and its probability distribution and the sensitivity threshold of the overall metacognitive processing efficiency and its probability distribution of healthy subjects on the group level
  • the parameter distribution at the individual level includes the sensitivity threshold and probability distribution of each cognitive domain of each MCI patient at the individual level and the sensitivity threshold and probability distribution of the overall metacognitive processing efficiency.
  • the present invention further provides a cognitive decision evaluation system based on a multi-dimensional layered drift diffusion model.
  • the cognitive decision assessment system includes one or more processors 51 and memory 52 .
  • the memory 52 is coupled with the processor 51 and is used to store one or more programs.
  • the one or more processors 51 Realize the cognitive decision-making evaluation method as in the above-mentioned embodiments.
  • the processor 51 is used to control the overall operation of the cognitive decision-making assessment system, so as to complete all or part of the steps of the above-mentioned cognitive decision-making assessment method.
  • the processor 51 may be a central processing unit (CPU), a graphics processing unit (GPU), a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a digital signal processing (DSP) chip, and the like.
  • Memory 52 is used to store various types of data to support the operation of the cognitive decision-making assessment system. These data, for example, may include instructions for any application or method operated on the cognitive decision-making assessment system, as well as application related data.
  • the memory 52 can be implemented by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, etc.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read-only memory
  • EPROM erasable Programmable Read Only Memory
  • PROM Programmable Read Only Memory
  • ROM Read Only Memory
  • Magnetic Memory Flash Memory
  • the cognitive decision-making evaluation system may be implemented by a computer chip or entity, or by a product with certain functions, and is used to execute the above-mentioned cognitive decision-making evaluation method, and achieve the above-mentioned method Consistent technical effects.
  • a typical embodiment is a computer.
  • a computer may be, for example, a personal computer, a laptop computer, an in-vehicle human-computer interaction device, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an e-mail device, a game console, A tablet computer, a wearable device, or a combination of any of these devices.
  • the present invention also provides a computer-readable storage medium including program instructions.
  • program instructions When the program instructions are executed by a processor, the steps of the cognitive decision-making evaluation method in any one of the above-mentioned embodiments are implemented.
  • the computer-readable storage medium can be the above-mentioned memory including program instructions, and the above-mentioned program instructions can be executed by the processor of the cognitive decision-making evaluation system to complete the above-mentioned cognitive decision-making evaluation method and achieve the same technical effect as the above-mentioned method.
  • the cognitive decision-making evaluation method and system based on the multi-dimensional layered drift diffusion model provided by the embodiment of the present invention simulates and refines the stimulus representation, information capture,
  • the dynamic process of belief accumulation and choice-making adopts a hierarchical design based on the Bayesian probability framework, integrates individual response data in different cognitive decision-making tasks, and obtains the and social cognition three) indicators of sensitivity threshold and overall metacognitive processing efficiency.
  • the present invention has the following advantages:
  • the multidimensional hierarchical drift-diffusion model can accurately simulate the decision-making process of the brain through mathematical modeling. Compared with the traditional method of analyzing behavioral response data using observation records and null hypothesis testing, the multidimensional hierarchical drift-diffusion model can be more direct. Deeply characterize and refine the dynamic neural decision-making process of stimulus representation, information capture, belief accumulation and choice-making in the brain's cognitive decision-making process;
  • the multi-dimensional hierarchical drift-diffusion model simultaneously integrates and models the individual's choice in cognitive decision-making and two response indicators when making the choice, taking into account the accuracy and speed of the subject in the cognitive task.
  • the trade-off process can more fully reflect the essential characteristics of the interaction and competition of various cognitive components in the subject's decision-making process.
  • the multidimensional hierarchical drift diffusion model integrates the behavioral responses of subjects in various cognitive decision-making tasks under different cognitive domains (perception, advanced cognition and social cognition) in the same mathematical model Data, establish a hierarchical model of decision-making trials-decision-making tasks-cognitive domains-overall metacognitive processing efficiency at the individual and group levels, and obtain the overall processing efficiency of the brain, as well as the sensitivity of each cognitive domain in the posterior
  • it can significantly reduce the calculation process and time of model fitting, and effectively improve the operating efficiency; on the other hand, it can realize the sensitivity threshold of different cognitive domains of subjects and the overall metacognition in the same ability space
  • the precise positioning of processing efficiency is convenient for comparing the impairment of different abilities of the same subject at the individual level and comparing the differences in the same cognitive decision-making ability of different subjects at the group level.
  • the MCI patients can be more accurately evaluated by comparing the norms of MCI patients with normal people Sensitivity thresholds of each cognitive domain (perception, advanced cognition, and social cognition) and the relative position of the overall metacognitive processing efficiency in the normal model of healthy people were obtained to obtain the overall impairment of MCI patients and each sub-cognitive
  • the damage status of the cognitive domain can realize the accurate assessment of the impaired ability of MCI patients, and promote the early detection, early intervention and early treatment of the decision-making ability in the impaired cognitive domain.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Pathology (AREA)
  • Databases & Information Systems (AREA)
  • Epidemiology (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

本发明公开了一种基于多维分层漂移扩散模型的认知决策评估方法及系统。该方法包括如下步骤：对于健康受试者，在群组水平上建立多维分层漂移扩散模型，获得健康受试者在每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，作为健康人常模；对于每一位MCI患者，在个体水平上建立多维分层漂移扩散模型，获得MCI患者在每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，作为待衡量指标。通过比较MCI患者的参数分布在健康人常模中的相对位置，从而能够对MCI患者受损能力的精准评估，促进对受损决策能力的早发现、早干预和早治疗。

Description

基于多维分层漂移扩散模型的认知决策评估方法及系统 技术领域
本发明涉及一种基于多维分层漂移扩散模型的认知决策评估方法，同时也涉及相应的认知决策评估系统，属于医疗保健信息学领域。
背景技术
个体在衰老过程中会经历一定程度的认知衰退。轻度认知障碍(Mild Cognitive Impairment，简写为MCI)是阿尔兹海默症的一种前驱状态，介于正常衰老与痴呆的中间状态，可作为阿尔兹海默症的“预报器”。随着计算机建模技术的发展，漂移扩散模型(Drift Diffusion Model，简写为DDM)逐渐兴起，并迅速被应用到心理学，尤其是认知决策任务中。DDM通过分离个体行为反应数据所包含的潜在决策成分，模拟和提炼人类大脑在决策任务中的神经动态过程。DDM主要应用于单个知觉范式、决策范式，建模健康人的神经决策过程。
为比较不同条件或情境中个体认知决策敏感性的特点，分层漂移扩散模型(Hierachical Drift Diffusion Model，简写为HDDM)应运而生。HDDM采用基于贝叶斯框架的分层设计，能够综合考虑不同水平上决策成分的特征。然而，现有的HDDM还仅限于从个体和群组水平上建模单一认知决策任务和范式中受试者的决策过程，并未扩展到聚类多种认知决策任务的情境中，无法整合评估个体不同认知域(区分为感知觉、高级认知和社会认知三类)的敏感性阈值和总体元认知加工效率。
发明内容
本发明所要解决的首要技术问题在于提供一种基于多维分层漂移扩散模型的认知决策评估方法，用于实现对MCI患者受损能力的精准评估，促进对受损认知域内决策能力的早发现、早干预和早治疗。
本发明所要解决的另一技术问题在于提供一种基于多维分层漂移扩散模型的认知决策评估系统。
为了实现上述目的，本发明采用以下的技术方案：
根据本发明实施例的第一方面，提供一种基于多维分层漂移扩散模型的认知决策评估方法，包括以下步骤：
对于健康受试者，在群组水平上建立多维分层漂移扩散模型，通过拟合优度最佳的模型，获得所述健康受试者在每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，作为健康人常模；
对于每一位MCI患者，在个体水平上建立多维分层漂移扩散模型，通过拟合优度最佳的模型，获得所述MCI患者在每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，作为待衡量指标；
比较所述MCI患者的待衡量指标在所述健康人常模中的相对位置，评估所述MCI患者在各认知域和总体元认知处理效率上的受损情况。
其中较优地，获得所述健康受试者每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，具体包括：
利用马尔科夫链蒙特卡罗方法，采用Metropolis采样算法，根据每个认知域对应的目标概率密度函数，输入多维分层漂移扩散模型中的所有参数的联合后验分布作为抽样样本的初始状态，生成马尔科夫链状态转移矩阵；
基于状态转移矩阵从已知分布中采样得到样本，以得到转移序列；
所述转移序列在相同的转移矩阵作用下，不断生成新的转移序列，直至转移序列不再发生变化趋于稳定状态；
当所述转移序列在第n步收敛稳定时，对符合平稳分布的对应样本集进行蒙特卡罗模拟求和，得到模型参数的后验分布；其中，n为转移序列趋于稳定状态时的作用次数；
其中，所述模型参数的后验分布为群组水平的参数分布，所述群组水平的参数分布包括健康受试者在群组水平上的每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布。
其中较优地，获得所述MCI患者每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布具体包括：
利用马尔科夫链蒙特卡罗方法，采用Metropolis采样算法，根据每个认知域对应的目标概率密度函数，输入多维分层漂移扩散模型中的所有参数的联合后验分布作为抽样样本的初始状态，生成马尔科夫链状态转移矩阵；
基于状态转移矩阵从已知分布中采样得到样本，以得到转移序列；
所述转移序列在相同的转移矩阵作用下，不断生成新的转移序列，直至转移序列不再发生变化趋于稳定状态；
当所述转移序列在第n步收敛稳定时，对符合平稳分布的对应样本集进行蒙特卡罗模拟求和，得到模型参数的后验分布；其中，n为转移序列趋于稳定状态时的作用次数；
其中，所述模型参数的后验分布为个体水平的参数分布，所述个体水平的参数分布包括每一位MCI患者在个体水平上的每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布。
其中较优地，比较所述MCI患者的待衡量指标在所述健康人常模中的相对位置，评估所述MCI患者在各认知域和总体元认知处理效率上的受损情况具体包括：
对参数后验密度分布进行显著性检验，通过贝叶斯假设检验衡量得到的各个参数的健康人常模分布与每一位MCI患者的后验密度分布中提取的样本百分比之间重叠的面积；
根据两个分布之间的重叠面积评估MCI患者在每一个认知域敏感性阈值及其总体元认知加工效率上的受损情况。
其中较优地，所述健康人常模分布与所述MCI患者的后验密度分布之间的重叠面积越小，均值的差距越大，则所述MCI患者的认知受损越严重。
其中较优地，通过比较所述MCI患者的每个认知域的敏感性阈值及其概率分布，在所述健康受试者的每个认知域的敏感性阈值及其概率分布中的相对位置，评估所述MCI患者在每个认知域上的认知受损情况。
其中较优地，通过比较所述MCI患者的总体元认知处理效率的敏感性阈值及其概率分布，在所述健康受试者的总体元认知处理效率的敏感性阈值及其概率分布中的相对位置，评估所述MCI患者在总体元认知处理效率上的认知受损情况。
其中较优地，所述认知域至少包括：感知觉决策认知域、高级认知决策认知域和社会认知决策认知域。
其中较优地，所述多维分层漂移扩散模型基于受试者在不同认知域 下的行为反应数据建立。
根据本发明实施例的第二方面，提供一种基于多维分层漂移扩散模型的认知决策评估系统，包括处理器和存储器，所述处理器读取所述存储器中的计算机程序，用于执行上述的认知决策评估方法。
本发明所提供的基于多维分层漂移扩散模型的认知决策评估方法，采用基于多维分层漂移扩散模型计算得到的各认知域敏感性阈值和总体元认知处理效率的后验分布，将MCI患者和正常人常模对比，能够更为准确地评估MCI患者各认知域敏感性阈值和总体元认知加工效率在健康人常模中的相对位置，获得MCI患者的整体受损情况和各个子认知域的受损状况，实现对MCI患者受损能力的精准评估，促进对受损认知域内决策能力的早发现、早干预和早治疗。
附图说明
图1为多维分层漂移扩散模型的建模过程示意图；
图2为模拟受试者在二元认知决策过程中信念累积的轨迹示意图；
图3为多维分层漂移扩散模型的分层结构示意图；
图4为本发明实施例提供的基于多维分层漂移扩散模型的认知决策评估方法的流程图；
图5为本发明实施例提供的基于多维分层漂移扩散模型的认知决策评估系统的结构示意图。
具体实施方式
下面结合附图和具体实施例对本发明的技术内容进行详细具体的说明。
图1为本发明实施例中，多维分层漂移扩散模型的建模过程示意图。该建模过程具体包括以下步骤：
S1：选择和确定需要实施的认知决策任务和范式，获取受试者在各个认知决策任务上的行为反应数据。
具体的，在本发明实施例中，认知决策任务至少包括：判断动点运动方向或空间位置的感知觉决策任务；基于选项相对价值做出选择的记忆、推理、执行控制的高级认知决策任务；学习他人属性、评估社会信息的社会认知决策任务。由此，根据任务内容及属性，可以将不同的决策任务标记为感知觉决策、高级认知决策和社会认知决策这三个认知域 下的任务。
同时，获取受试者在各个认知决策任务上的行为反应数据具体包括：获取受试者在不同认知决策任务中的反应选择及做出决策所需要的反应时间。因此，在本发明实施例中，第i个认知域下第j个任务范式中的第k个决策试次的决策时间(response time，简写为RT)和选项(decision option，简写为DO)被编码进观测数据xi，j，k中，即(RT，DO)～xi，j，k。i＝1，2或3。
S2：对受试者在不同认知决策任务上的行为反应数据进行拟合，建立多维分层漂移扩散模型。
具体的，包括步骤S21～S24：
S21：针对受试者在单个试次中的行为反应观测数据构建偏移扩散模型。
该偏移扩散模型通过如下公式进行构建：
E t+1＝E t+Δ
Δ～N(d(μ 1-μ 2)，σ 2)         (1)
其中，E表示一个选项优于另一个选项的累积证据平衡，提供了两个选项的偏差，即决策前受试者对某个选项的偏好，当受试者对两种决策的已有经验相同时E＝0；Et表示在第t个试次(即第t个具体的任务回合)时，两个选项的累积偏差；t表示思考开始的时间，包括对信息编码与做出按键反应的时间；Δ表示在每个决策时间中支持一个选项的增量证据；d表示受试者的漂移率系数，表示在每个时间单位内个体积累的关于刺激的相对信息量，表征了信息积累的速度和个体感知敏感性的测量；μi是选项i的主观价值，决策阈值d(μ 1-μ 2)，代表两个决策选项之间的差异，表征决策信息收集的边界/阈值，当累积信息量达到该阈值时受试者即做出决策；σ 2是扩散噪声系数，表示证据积累和/或比较系统中的噪声；N()表示正态分布，指决策轮次间的增量证据在具有特定方差(σ2)的决策阈值(μ1-μ2)周围呈正态分布Δ～(N)((d(μ1-μ2)，σ2)。
S52：分析计算受试者的选择概率和预期响应的反应时间；
具体的，通过以下方程式进行分析计算：
Figure PCTCN2022120985-appb-000001
如图2所示，该模型用以模拟受试者在二元认知决策(即从两个选项中做出决策)过程中信念累积的示例轨迹。选择概率(P)和反应时间(RT)是漂移率系数d、扩散噪声系数σ2和响应边界θ的高度的函数。其中θ是触发决策的证据积累边界的高度(此上限为对选项1的选择)，p(ch＝1)是达到上限而非下限的概率，RT是累积过程结束的预期时间(到达任一边界)。
S23：计算偏移扩散模型的参数估计值，其中，偏移扩散模型的参数包括：受试者的决策阈值、相对起始点、漂移率和非决策时间；
具体的，输入受试者的决策行为反应数据包括决策时间(RT)和选项(DO)作为反应时间(PT)和选择概率(P)，根据公式1和公式2，计算DDM模型的参数估计值。
S24：将受试者在不同认知域内多个决策范式中的反应数据整合到同一模型框架中，形成由低至高的决策试次-决策任务-认知域-总体元认知加工效率的多维分层漂移扩散模型。
其中，在多维分层漂移扩散模型中，低层次水平的参数估计会受到更高层次水平参数分布的约束。以决策试次-决策任务这一层次举例，对于服从某种似然函数f|θ分布的每一个受试者的观测数据xi，j，k而言，假设该受试者在特定认知决策任务(感知觉、高级认知和社会认知三类)中决策轮次的参数θk在具有特定方差(σ 2)的平均值(μ)周围呈正态分布(N)(λ＝(μ，σ))，而这些任务水平参数又是从给定超先验(hyper-prior)G0的数据中估计得到的，因此可以得到这样的生成性描述：
μ，σ～G0()；θk～N(μ，σ 2)；xi，j，k～f(θk)。
对试次水平参数θ的固定先验是由任务水平参数λ参数化的随机变量，据此得到θ的后验概率公式：
Figure PCTCN2022120985-appb-000002
其中，P(θ，λ|x)为似然函数，表示在给定观测数据x的情况下试次水平参数为θ、任务水平参数为λ的可能性，P(x|θ)为给定试次水平参数 θ的情况下观测数据等于x的可能性，P(θ|λ)表示任务水平参数为λ的情况下试次水平参数等于θ的可能性，P(λ)和P(x)表示任务水平参数为λ、观测数据等于x的概率。总体而言，当分层模型的层次越多，估计参数的层次越高，其后验概率估算公式越复杂。
如图3所示，xi，j，k～L(ai，zi，vi，ti，sv，st，sz)。其中，xi，j，k表示该受试者在第i个认知域(Field，简写为F)的第j个范式任务下(Paradigm，简写为P)的第k个决策试次(Trial，简写为T)中，包括决策时间(RT)和选项(DO)在内的行为观测数据；L表示由每个认知域对应的DDM模型的似然函数；ai、zi、vi和ti分别表示第i个认知域的决策阈值、相对起始点、漂移率和非决策时间分布。其他参数μ表示均值；σ表示标准差；ak、vk、zk和tk代表在第k个决策试次中决策阈值、漂移率、相对起始点和非决策时间。
S3：对所述多维分层漂移扩散模型进行优化，以得到拟合优度最佳的模型。
通过评估拟合优度，找到收敛性达到预定值的多维分层漂移扩散模型，当拟合效果不佳时，不断调整模型参数设定从而得到合适的模型。
具体的，采用Gelman-Rubin R统计量来评估当前模型的收敛性，其值越接近1，说明模型收敛性越好；
计算当前模型的偏差信息准则，在考虑模型复杂度的情况下评估模型的拟合优度，其值越小说明模型拟合越好；
或者，进行后验预测检验，从估计出的模型中抽取新的数据集，对比这些模拟数据集与实际观测数据的一致性，以评估当前模型对观测数据的拟合效果；
当模型拟合优度不佳时，不断调整模型参数设定从而得到合适的多维分层漂移扩散模型以及相应的参数估计值和后验概率分布。
由此，通过上述方式可建立多维分层漂移扩散模型，该多维分层漂移扩散模型即为对MCI患者的认知决策进行评估的多维分层漂移扩散模型，用于辅助了解MCI患者的认知受损情况。
请参照图4所示，为本发明实施例提供的一种基于多维分层漂移扩散模型的认知决策评估方法，具体包括步骤S10～S30：
S10：利用第一实施例得到的多维分层漂移扩散模型，建立健康人常 模。
具体的，通过第一实施例中所述的建模方法，对于所有的健康受试者，在群组水平上建立多维分层漂移扩散模型，获得所述健康受试者在每个认知域的敏感性阈值及其概率分布，以及总体元认知处理效率的敏感性阈值及其概率分布，将此两个分布模型作为健康人常模。
其中，可以理解的是，所述健康受试者在每个认知域的敏感性阈值及其概率分布用于作为评估MCI患者各认知域受损情况的标准；所述健康受试者在总体元认知处理效率的敏感性阈值及其概率分布用于作为评估MCI患者的总体元认知处理效率受损情况的标准。
S20：获取每一位MCI患者的待衡量指标。
具体的，通过第一实施例中所述的建模方法，对于每一位MCI患者，在个体水平上建立多维分层漂移扩散模型，通过拟合优度最佳的模型，获得所述MCI患者在每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，作为待衡量指标。
其中，可以理解的是，该待衡量指标既包括MCI患者在每个认知域的敏感性阈值及其概率分布，还包括MCI患者在总体元认知处理效率的敏感性阈值及其概率分布，从而用于评估MCI患者在各认知域及其总体认知处理效率上的受损情况。
S30：比较MCI患者的待衡量指标在健康人常模中的相对位置，评估MCI患者在各认知域和总体元认知处理效率上的受损情况。
其中，比较MCI患者的待衡量指标在健康人常模中的相对位置具体包括步骤S301～S302：
S301：计算各个参数的健康人常模的后验密度分布，以及MCI患者的敏感性阈值指标的后验密度分布；
S302：根据步骤301中得到的健康人常模的后验密度分布与MCI患者的敏感性阈值指标的后验密度分布，提取样本百分比之间重叠的面积；
对参数后验密度分布进行显著性检验，通过贝叶斯假设检验衡量得到的各个参数的健康人常模分布与每一位MCI患者的后验密度分布中提取的样本百分比之间的重叠面积。
S303：根据重叠面积评估MCI患者在各认知域的敏感性阈值及其总体元认知加工效率上的受损情况。
其中，健康人常模的后验密度分布与MCI患者的敏感性阈值指标的后验密度分布，这两个分布之间的重叠面积越小，均值的差距越大，表示MCI患者的受损越严重。
可以理解的是，针对各认知域而言，需要比较的是同一个认知域下，健康人常模的后验密度分布与MCI患者的敏感性阈值指标的后验密度分布之间的重叠面积；针对总体元认知加工效率而言，需要比较的是两个整体分布之间的重叠面积。所述整体分布是指：根据各认知域下得到的后验密度分布，所得到的总体分布。
由此，通过上述方法能够实现对MCI患者受损能力的精准评估，促进对受损认知域内决策能力的早发现、早干预和早治疗。
在上述实施例中，对于任一受试者而言，获得每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布的方式相同，只是对于健康受试者而言，所获得的每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布作为健康人常模；对于MCI患者而言，所获得的每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布作为待衡量指标。
下面对如何获得每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布进行说明，具体包括步骤S401～S404：
S401：利用马尔科夫链蒙特卡罗方法，采用Metropolis采样算法，根据每个认知域对应的目标概率密度函数，输入多维分层漂移扩散模型中的所有参数的联合后验分布作为抽样样本的初始状态，生成马尔科夫链状态转移矩阵；
S402：基于状态转移矩阵从已知分布中采样得到样本，以得到转移序列；
S403：所述转移序列在相同的转移矩阵作用下(作用次数为1，2，…，n)，不断生成新的转移序列，直至转移序列不再发生变化趋于稳定状态；
S404：当所述转移序列在第n步收敛稳定时，对符合平稳分布的对应样本集进行蒙特卡罗模拟求和，得到模型参数的后验分布；其中，n为转移序列趋于稳定状态时的作用次数。
其中，所述群组水平的参数分布包括健康受试者在群组水平上的每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，所述个体水平的参数分布包括每一位MCI患者在个体水平上的每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布。
在上述认知决策评估方法的基础上，本发明进一步提供一种基于多维分层漂移扩散模型的认知决策评估系统。如图5所示，该认知决策评估系统包括一个或多个处理器51和存储器52。其中，存储器52与处理器51耦接，用于存储一个或多个程序，当所述一个或多个程序被所述一个或多个处理器51执行，使得所述一个或多个处理器51实现如上述实施例中的认知决策评估方法。
其中，处理器51用于控制该认知决策评估系统的整体操作，以完成上述认知决策评估方法的全部或部分步骤。该处理器51可以是中央处理器(CPU)、图形处理器(GPU)、现场可编程逻辑门阵列(FPGA)、专用集成电路(ASIC)、数字信号处理(DSP)芯片等。存储器52用于存储各种类型的数据以支持在该认知决策评估系统的操作，这些数据例如可以包括用于在该认知决策评估系统上操作的任何应用程序或方法的指令，以及应用程序相关的数据。该存储器52可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，例如静态随机存取存储器(SRAM)、电可擦除可编程只读存储器(EEPROM)、可擦除可编程只读存储器(EPROM)、可编程只读存储器(PROM)、只读存储器(ROM)、磁存储器、快闪存储器等。
在一个示例性实施例中，该认知决策评估系统具体可以由计算机芯片或实体实现，或者由具有某种功能的产品来实现，用于执行上述的认知决策评估方法，并达到如上述方法一致的技术效果。一种典型的实施例为计算机。具体地说，计算机例如可以为个人计算机、膝上型计算机、车载人机交互设备、蜂窝电话、相机电话、智能电话、个人数字助理、媒体播放器、导航设备、电子邮件设备、游戏控制台、平板计算机、可穿戴设备或者这些设备中的任何设备的组合。
在另一个示例性实施例中，本发明还提供一种包括程序指令的计算机可读存储介质，该程序指令被处理器执行时实现上述任意一个实施例 中的认知决策评估方法的步骤。例如，该计算机可读存储介质可以为上述包括程序指令的存储器，上述程序指令可由认知决策评估系统的处理器执行以完成上述的认知决策评估方法，并达到如上述方法一致的技术效果。
综上所述，本发明实施例提供的基于多维分层漂移扩散模型的认知决策评估方法及系统，通过数学建模的方式，模拟和提炼大脑进行认知决策过程中刺激表征、信息捕获、信念积累和做出选择的动态过程，采用基于贝叶斯概率框架的分层设计，整合个体在不同认知决策任务中的反应数据，并获取个体在不同认知域(感知觉、高级认知和社会认知三类)的敏感性阈值的指标和总体元认知处理效率。
与现有技术相比较，本发明具有以下优点：
1)多维分层漂移扩散模型通过数学建模的手段能够精确模拟大脑的决策过程，相比于传统的采用观察记录和零假设检验分析行为反应数据的方式，多维分层漂移扩散模型能够更直接地表征和提炼大脑进行认知决策过程中刺激表征、信息捕获、信念积累和做出选择的动态神经决策过程；
2)多维分层漂移扩散模型同时整合和建模了个体在认知决策中的抉择以及做出该选择的反应时两种反应指标，综合考虑了认知任务中受试者准确性和速度的权衡过程，能够更为充分的反映受试者决策加工过程中各种认知成分相互影响和竞争的本质特征。
3)多维分层漂移扩散模型在同一个数学模型内整合受试者在不同的认知域(感知觉、高级认知和社会认知三类)下的多种认知决策任务中的行为反应数据，在个体和群组两个水平上建立决策试次-决策任务-认知域-总体元认知加工效率的分层模型，获得大脑整体加工效率，以及后验的各个认知域的敏感性阈值水平，一方面能够显著减少模型拟合的计算过程和时间，有效提升运行效率，另一方面能够在同一个能力空间内实现对受试者不同认知域敏感性阈值和总体元认知加工效率的精准定位，便于在个体水平对比同一受试者不同能力的受损情况和在群体水平上对比不同受试者相同认知决策能力的差异。
4)采用基于多维分层漂移扩散模型计算得到的各认知域敏感性阈值和总体元认知处理效率的后验分布，将MCI患者和正常人常模对比， 能够更为准确地评估MCI患者各认知域(感知觉、高级认知和社会认知三类)敏感性阈值和总体元认知加工效率在健康人常模中的相对位置，获得MCI患者的整体受损情况和各个子认知域的受损状况，实现对MCI患者受损能力的精准评估，促进对受损认知域内决策能力的早发现、早干预和早治疗。
上面对本发明所提供的基于多维分层漂移扩散模型的认知决策评估方法及系统进行了详细的说明。对本领域的一般技术人员而言，在不背离本发明实质内容的前提下对它所做的任何显而易见的改动，都将构成对本发明专利权的侵犯，将承担相应的法律责任。

Claims (10)

  1. 一种基于多维分层漂移扩散模型的认知决策评估方法，其特征在于包括以下步骤：
    对于健康受试者，在群组水平上建立多维分层漂移扩散模型，通过拟合优度最佳的模型，获得所述健康受试者在每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，作为健康人常模；
    对于每一位MCI患者，在个体水平上建立多维分层漂移扩散模型，通过拟合优度最佳的模型，获得所述MCI患者在每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，作为待衡量指标；
    比较所述MCI患者的待衡量指标在所述健康人常模中的相对位置，评估所述MCI患者在各认知域和总体元认知处理效率上的受损情况。
  2. 如权利要求1所述的认知决策评估方法，其特征在于，获得所述健康受试者每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，具体包括：
    利用马尔科夫链蒙特卡罗方法，采用Metropolis采样算法，根据每个认知域对应的目标概率密度函数，输入多维分层漂移扩散模型中的所有参数的联合后验分布作为抽样样本的初始状态，生成马尔科夫链状态转移矩阵；
    基于状态转移矩阵从已知分布中采样得到样本，以得到转移序列；
    所述转移序列在相同的转移矩阵作用下，不断生成新的转移序列，直至转移序列不再发生变化趋于稳定状态；
    当所述转移序列在第n步收敛稳定时，对符合平稳分布的对应样本集进行蒙特卡罗模拟求和，得到模型参数的后验分布；其中，n为转移序列趋于稳定状态时的作用次数；
    其中，所述模型参数的后验分布为群组水平的参数分布，所述群组水平的参数分布包括健康受试者在群组水平上的每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布。
  3. 如权利要求2所述的认知决策评估方法，其特征在于，获得所述 MCI患者每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布，具体包括：
    利用马尔科夫链蒙特卡罗方法，采用Metropolis采样算法，根据每个认知域对应的目标概率密度函数，输入多维分层漂移扩散模型中的所有参数的联合后验分布作为抽样样本的初始状态，生成马尔科夫链状态转移矩阵；
    基于状态转移矩阵从已知分布中采样得到样本，以得到转移序列；
    所述转移序列在相同的转移矩阵作用下，不断生成新的转移序列，直至转移序列不再发生变化趋于稳定状态；
    当所述转移序列在第n步收敛稳定时，对符合平稳分布的对应样本集进行蒙特卡罗模拟求和，得到模型参数的后验分布；其中，n为转移序列趋于稳定状态时的作用次数；
    其中，所述模型参数的后验分布为个体水平的参数分布，所述个体水平的参数分布包括每一位MCI患者在个体水平上的每个认知域的敏感性阈值及其概率分布和总体元认知处理效率的敏感性阈值及其概率分布。
  4. 如权利要求3所述的认知决策评估方法，其特征在于，比较所述MCI患者的待衡量指标在所述健康人常模中的相对位置，评估所述MCI患者在各认知域和总体元认知处理效率上的受损情况，具体包括：
    对参数后验密度分布进行显著性检验，通过贝叶斯假设检验衡量得到的各个参数的健康人常模分布与每一位MCI患者的后验密度分布中提取的样本百分比之间重叠的面积；
    根据两个分布之间的重叠面积评估MCI患者在每一个认知域敏感性阈值及其总体元认知加工效率上的受损情况。
  5. 如权利要求4所述的认知决策评估方法，其特征在于：
    所述健康人常模分布与所述MCI患者的后验密度分布之间的重叠面积越小，均值的差距越大，则所述MCI患者的认知受损越严重。
  6. 如权利要求4所述的认知决策评估方法，其特征在于：
    通过比较所述MCI患者的每个认知域的敏感性阈值及其概率分布，在所述健康受试者的每个认知域的敏感性阈值及其概率分布中的相对位置，评估所述MCI患者在每个认知域上的认知受损情况。
  7. 如权利要求4所述的认知决策评估方法，其特征在于：
    通过比较所述MCI患者的总体元认知处理效率的敏感性阈值及其概率分布，在所述健康受试者的总体元认知处理效率的敏感性阈值及其概率分布中的相对位置，评估所述MCI患者在总体元认知处理效率上的认知受损情况。
  8. 如权利要求1所述的认知决策评估方法，其特征在于所述认知域至少包括：感知觉决策认知域、高级认知决策认知域和社会认知决策认知域。
  9. 如权利要求8所述的认知决策评估方法，其特征在于：
    所述多维分层漂移扩散模型基于受试者在不同认知域下的行为反应数据建立。
  10. 一种基于多维分层漂移扩散模型的认知决策评估系统，其特征在于包括处理器和存储器，所述处理器读取所述存储器中的计算机程序，用于执行权利要求1所述的认知决策评估方法。
PCT/CN2022/120985 2021-11-17 2022-09-23 基于多维分层漂移扩散模型的认知决策评估方法及系统 WO2023087917A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111365418.9 2021-11-17
CN202111365418.9A CN113793685B (zh) 2021-11-17 2021-11-17 基于多维分层漂移扩散模型的认知决策评估方法及系统

Publications (1)

Publication Number Publication Date
WO2023087917A1 true WO2023087917A1 (zh) 2023-05-25

Family

ID=78877382

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/120985 WO2023087917A1 (zh) 2021-11-17 2022-09-23 基于多维分层漂移扩散模型的认知决策评估方法及系统

Country Status (2)

Country Link
CN (1) CN113793685B (zh)
WO (1) WO2023087917A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113793685B (zh) * 2021-11-17 2022-03-25 北京智精灵科技有限公司 基于多维分层漂移扩散模型的认知决策评估方法及系统
CN116401377B (zh) * 2023-04-04 2024-01-23 华中师范大学 一种基于扩散概率分布的时序知识图谱推理方法
CN117172627B (zh) * 2023-11-03 2024-02-27 腾讯科技（深圳）有限公司 一种业务执行方法、装置、设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110084359A (zh) * 2019-03-20 2019-08-02 山东师范大学 动态直觉模糊认知图构建方法、时间序列预测方法及系统
CN111967495A (zh) * 2020-07-06 2020-11-20 中国科学院计算技术研究所 一种分类识别模型构建方法
WO2021053580A1 (en) * 2019-09-17 2021-03-25 Neuroinova, Lda Method and device for tracking cognitive performance in mild cognitive impairment
CN113380407A (zh) * 2021-05-30 2021-09-10 重庆医科大学 构建认知障碍智能预测方法
CN113793685A (zh) * 2021-11-17 2021-12-14 北京智精灵科技有限公司 基于多维分层漂移扩散模型的认知决策评估方法及系统

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5671056B2 (ja) * 2009-12-16 2015-02-18 コーニンクレッカ フィリップス エヌ ヴェ ユニバーサル医療デバイスドライバアダプタ
CN102982493A (zh) * 2012-11-06 2013-03-20 殷淑琴 社区认知能力损害老年群体的综合干预技术
CN105808970B (zh) * 2016-05-09 2018-04-06 南京智精灵教育科技有限公司 一种在线认知评估方法
CN106339796B (zh) * 2016-08-17 2019-11-15 国家海洋局北海环境监测中心 一种海洋溢油风险评估方法
CN112840314A (zh) * 2018-08-02 2021-05-25 布莱特临床研究有限公司 对正在运行中的临床试验进行动态数据监测和实时优化的系统、方法及实施过程
CN111009323B (zh) * 2019-11-12 2023-11-10 河北工业大学 一种基于knn-ann的硬脑膜下血肿损伤预测方法
CN112037001A (zh) * 2020-09-03 2020-12-04 云账户技术(天津)有限公司 打款风险预测模型训练方法、打款风险预测方法及其装置
CN113467465B (zh) * 2021-07-22 2023-08-04 福州大学 一种面向机器人系统的人在回路决策建模与控制方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110084359A (zh) * 2019-03-20 2019-08-02 山东师范大学 动态直觉模糊认知图构建方法、时间序列预测方法及系统
WO2021053580A1 (en) * 2019-09-17 2021-03-25 Neuroinova, Lda Method and device for tracking cognitive performance in mild cognitive impairment
CN111967495A (zh) * 2020-07-06 2020-11-20 中国科学院计算技术研究所 一种分类识别模型构建方法
CN113380407A (zh) * 2021-05-30 2021-09-10 重庆医科大学 构建认知障碍智能预测方法
CN113793685A (zh) * 2021-11-17 2021-12-14 北京智精灵科技有限公司 基于多维分层漂移扩散模型的认知决策评估方法及系统

Also Published As

Publication number Publication date
CN113793685B (zh) 2022-03-25
CN113793685A (zh) 2021-12-14

Similar Documents

Publication Publication Date Title
WO2023087917A1 (zh) 基于多维分层漂移扩散模型的认知决策评估方法及系统
US9861308B2 (en) Method and system for monitoring stress conditions
JP7299427B2 (ja) 機械学習に基づいて心理指標データ及び身体指標データを学習することにより、精神健康を予測し、精神健康ソリューションを提供する方法及びそれを用いた精神健康評価装置
CN109243604B (zh) 一种基于神经网络算法的川崎病风险评估模型的构建方法及构建系统
CN115699204A (zh) 基于多个机器学习模型的临床预测因子
CN111862020A (zh) 一种眼前节生理年龄预测方法、装置、服务器及存储介质
CN111403039A (zh) 一种动态平衡评估方法、装置、设备及介质
CN116739037A (zh) 具有人格特征的人格模型构建方法及装置
CN115188475A (zh) 一种狼疮肾炎患者风险预测方法
CN112382382B (zh) 一种代价敏感的集成学习分类方法及系统
Pedroto et al. Predicting age of onset in TTR-FAP patients with genealogical features
Arias-Londoño et al. Complexity analysis of pathological voices by means of hidden markov entropy measurements
Kristensen et al. Is whole-brain functional connectivity a neuromarker of sustained attention? Comment on Rosenberg & al.(2016)
Kavya et al. Heart Disease Prediction Using Logistic Regression
Nieto Ramos et al. Bayesian inference for fitting cardiac models to experiments: estimating parameter distributions using Hamiltonian Monte Carlo and approximate Bayesian computation
CN114068012B (zh) 一种面向认知决策的多维分层漂移扩散模型建模方法
Gulumbe et al. Predicting diabetes mellitus using artificial neural network through a simulation study
CN117672495B (zh) 基于人工智能的房颤合并冠心病患者远期死亡率预测方法
CN117409978B (zh) 一种疾病预测模型构建方法、系统、装置及可读存储介质
Musto et al. On a Survival Gradient Boosting, Neural Network and Cox PH Based Approach to Predicting Dementia Diagnosis Risk on ADNI
Kounali et al. M110. PREDICTING RISK OF PSYCHOSIS IN A GENERAL POPULATION SAMPLE
Gill et al. Establishing a Machine Learning-Based Random Forest Classifier to Estimate Autism Risk
CN113966999A (zh) 一种索拉非尼耐药性预测方法、装置及可存储介质
Qin Research on hypertension prediction model based on XGBoost algorithm
CN117012289A (zh) 一种利用胁迫响应生物标志物作为预警指标构建模型方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22894455

Country of ref document: EP

Kind code of ref document: A1