An alzheimer's disease auxiliary screening system based on intelligent interaction

Abstract

The application discloses an Alzheimer's disease auxiliary screening system based on intelligent interaction, and particularly relates to the field of computer-aided diagnosis, and is used for solving the problem that it is difficult to simultaneously quantify the differences in life knowledge and professional knowledge cognitive performance in the existing screening method. Through constructing a task interaction group covering life knowledge and professional knowledge, multi-dimensional interaction operations of a testee in a task process are collected and coded to generate an interaction vector matrix. Further, through a preset task logic chain, the interaction behavior is identified by jumping points, vector residual errors are calculated, and invalid operation density statistics are performed to quantify the operation deviation. Based on the analysis result, life and professional cognitive performance indicators are generated, and a cognitive bias evolution curve of the difference between the two indicators changing over time is constructed within a set period. Finally, through identifying the continuous rising trend in the cognitive bias evolution process, auxiliary screening and early warning of asymmetric degradation of cognitive ability are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer-aided diagnosis, and more particularly, to an Alzheimer's disease auxiliary screening system based on intelligent interaction. BACKGROUND

[0002] In the early identification process of cognitive impairment, a characteristic cognitive degradation phenomenon can be observed for individuals with strong domain knowledge accumulation or engaged in professional work, that is, the degradation rhythm of life knowledge and professional knowledge is inconsistent. In the early stage of Alzheimer's disease, patients often still maintain relatively complete professional semantic structure and terminology use ability, for example, individuals proficient in organic chemistry can still correctly distinguish the structural differences between alkynes and alkenes, clearly describe the process path of substitution reaction, and even complete abstract tasks such as structure formula judgment and reaction type matching when facing terminology questions and logic analysis tasks. However, the same individuals may exhibit obvious life chain disorder in daily life, such as being unable to correctly complete sequential operations such as boiling water, cooking, and dressing, or frequently making mistakes in placing objects and walking in space in familiar environments. Unlike other brain function loss scenarios, such life common sense errors are not caused by simple memory impairment, but are caused by the preferential destruction of the medial temporal lobe and related context construction networks in the early stage of Alzheimer's disease, resulting in the first decline in behavior organization ability driven by actual life experience, while the highly consolidated and repeatedly reinforced professional knowledge system remains relatively stable.

[0003] Therefore, in practice, individuals often perform normally in standard cognitive scale tests but frequently make mistakes in life tasks, causing early identification delays. The current evaluation mechanism has not established a structural comparison method for the degradation rhythm of different knowledge systems within an individual, and it is urgent to introduce the difference between the cognitive stability of life knowledge and professional knowledge as a basis for judgment to improve the accuracy of early-stage Alzheimer's disease screening. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present application provide an Alzheimer's disease auxiliary screening system based on intelligent interaction to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0006] An Alzheimer's disease auxiliary screening system based on intelligent interaction, comprising a task construction module, an interaction extraction module, an interaction analysis module, a cognitive performance evaluation module, an evolution curve construction module, and an asymmetric degradation identification module, wherein:

[0007] The task construction module respectively constructs task interaction groups corresponding to life knowledge and professional knowledge, and presents them to the testee in turn.

[0008] The interaction extraction module collects the interaction operations of the testee in various tasks, encodes the interaction operations in each round of task, and establishes an interaction vector matrix.

[0009] The interaction analysis module performs skip point statistics and vector residual calculation on the interaction vector matrix according to the preset standard task logic chain, and counts the density of invalid operations in the interaction vector matrix.

[0010] The cognitive performance evaluation module outputs the results through the calculation and statistics of the interaction analysis module, and generates life cognitive performance indicators and professional cognitive performance indicators.

[0011] The evolution curve construction module records the change sequence of the difference between the life cognitive performance indicators and the professional cognitive performance indicators within the preset task test period, and constructs the life and professional cognitive bias evolution curve.

[0012] The asymmetric degradation identification module determines whether the testee has an asymmetric degradation trend in cognitive ability according to the cognitive bias evolution curve.

[0013] In a preferred embodiment, the task construction module respectively constructs task interaction groups corresponding to life knowledge and professional knowledge, and presents them to the testee in turn, specifically including:

[0014] According to the age, education background and professional experience of the testee, construct no less than a set number of life knowledge interaction tasks, the task content includes scene recognition, routine operation judgment and daily object use sequence reasoning;

[0015] At the same time, construct an equal amount of professional knowledge interaction tasks as the life knowledge interaction tasks, the task content includes term recognition, special process selection and scene causal judgment;

[0016] All interaction tasks are uniformly encoded and integrated into a task sequence group, and presented to the testee in turn in a segmented alternating manner.

[0017] In a preferred embodiment, the interaction extraction module collects the interaction operations of the testee in various tasks, encodes the interaction operations in each round of task, and establishes an interaction vector matrix specifically including:

[0018] Obtain the interaction operations of the testee in each round of task, structurally decompose each interaction operation, and extract multi-dimensional interaction data including operation action semantics, action object, task code and interaction time;

[0019] Convert the multi-dimensional interaction data into interaction operation vector expression through the preset encoding mapping rule.

[0020] After each round of task completion, the interaction operation vectors are spliced into a complete interaction operation vector sequence according to the interaction operation sequence, and are labeled according to the task code;

[0021] After the completion of the set number of rounds of tasks, the interaction operation vector sequence is integrated to establish an interaction vector matrix that is continuous in time and distinguishable in tasks, and different weights are set for the columns in the interaction vector matrix.

[0022] In a preferred embodiment, the interaction operation vector expression construction process is to set an enumerable discrete coding space for each dimension of the multi-dimensional interaction data, and splice it into a structure vector according to the fixed sequence of action semantics, action object, task code and interaction time.

[0023] In a preferred embodiment, the interaction analysis module performs skip point statistics and vector residual calculation on the interaction vector matrix according to the preset standard task logic chain, and the invalid operation density in the interaction vector matrix is specifically calculated as follows:

[0024] The preset standard task logic chain defines the ideal vector template of the multi-dimensional interaction data in each round of task;

[0025] The interaction vector matrix is mapped to the standard task logic chain, the position of the task logic jump is identified in the matrix through the task code sequence in the interaction operation vector, and is marked as a skip point;

[0026] According to the column weight of the matrix, the weighted vector distance between each row of interaction operation vectors in the interaction vector matrix and the ideal vector template is calculated to obtain the vector residual of the interaction operation;

[0027] The interaction vector matrix is divided into several sectional sub-matrices according to the task rounds, and the invalid operation density of the interaction operation vectors in the sub-matrices is calculated;

[0028] The invalid operation density is obtained by calculating the proportion of the interaction time corresponding to the non-target operation interaction operation vector to the total duration of the standard task logic chain;

[0029] The non-target operation is an operation inconsistent with the ideal vector template.

[0030] In a preferred embodiment, the cognitive performance evaluation module generates life cognitive performance indicators and professional cognitive performance indicators through the calculation and statistical output results of the interaction analysis module, and specifically includes:

[0031] The skip points, vector residuals and invalid operation densities in the interaction analysis module are divided into task types according to the task codes corresponding to the interaction operation vectors, and the skip point frequency, vector residual mean and invalid operation density mean of different task types are respectively integrated and calculated.

[0032] Based on the jump point frequency, vector residual mean and invalid operation density mean of different task types, the life cognitive performance index and the professional cognitive performance index are generated by weighting comprehensive score.

[0033] In a preferred embodiment, the evolution curve building module records the change sequence of the difference between the life cognitive performance index and the professional cognitive performance index in the preset task test period, and builds the life and professional cognitive bias evolution curve, specifically comprising:

[0034] The preset task test period contains several cognitive performance index evaluations in one task test period.

[0035] After each cognitive performance index evaluation, the difference between the life cognitive performance index and the professional cognitive performance index is calculated and recorded as the cognitive bias evaluation value.

[0036] All cognitive bias evaluation values in the preset task test period are arranged in time to build a cognitive bias curve, wherein the vertical axis represents the difference between the life cognitive performance index and the professional cognitive performance index, and the horizontal axis represents the cognitive performance index evaluation times.

[0037] In a preferred embodiment, the asymmetric degradation identification module judges whether the cognitive ability of the tested person has an asymmetric degradation trend according to the cognitive bias evolution curve, specifically comprising:

[0038] When the cognitive bias evolution curve has a continuous monotonic upward trend in the preset length of the sliding window, and the growth rate exceeds the set change rate threshold, it is determined that the life cognitive performance and the professional cognitive performance of the tested person have asymmetric degradation, and a screening result warning is issued.

[0039] The technical effects and advantages of the Alzheimer's disease auxiliary screening system based on intelligent interaction are as follows:

[0040] The cognitive ability asymmetric degeneration recognition method provided by the application, through a task construction module, sets a task interaction group of life knowledge and professional knowledge respectively, effectively covering the task response behavior of different cognitive types; an interaction extraction module structures and codes the operation behavior of the testee, constructs a standardized interaction vector matrix, and ensures that the behavior data has comparability and analyzability; an interaction analysis module combines a standard task logic chain to perform jump point recognition, residual error evaluation and invalid operation density statistics, and realizes multidimensional evaluation of the interaction operation logic consistency and task execution deviation; a cognitive performance evaluation module generates life and professional dimension cognitive performance indicators based on the analysis results, has clear dimension division and response interpretation power; an evolution curve construction module models the time sequence of the difference between the two types of indicators, forming a quantifiable cognitive deviation evolution trend; and an asymmetric degeneration recognition module determines whether there is inconsistency between the life and professional dimensions in the cognitive ability degradation based on the trend curve change, can identify the structural cognitive ability decline signal in the early stage, and improves the screening accuracy and preclinical recognition rate. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 FIG. 1 is a schematic diagram of an Alzheimer's disease auxiliary screening system based on intelligent interaction according to the application. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0043] Embodiment 1, Figure 1 The application provides an Alzheimer's disease auxiliary screening system based on intelligent interaction, which comprises a task construction module, an interaction extraction module, an interaction analysis module, a cognitive performance evaluation module, an evolution curve construction module and an asymmetric degeneration recognition module, wherein:

[0044] The task construction module respectively constructs a task interaction group corresponding to life knowledge and professional knowledge, and presents them to the testee in turn;

[0045] The interaction extraction module collects the interaction operations of the testee in various tasks, codes the interaction operations in each round of task, and establishes an interaction vector matrix;

[0046] The interaction analysis module performs jump point statistics and vector residual error calculation on the interaction vector matrix according to the preset standard task logic chain, and counts the invalid operation density in the interaction vector matrix;

[0047] The cognitive performance evaluation module generates the life cognitive performance indicators and the professional cognitive performance indicators through the calculation and statistical output of the interactive analysis module;

[0048] The evolution curve construction module records the change sequence of the difference between the life cognitive performance indicators and the professional cognitive performance indicators within the preset task test period, and constructs the life and professional cognitive bias evolution curve;

[0049] The asymmetric degradation identification module determines whether the cognitive ability of the testee has an asymmetric degradation trend according to the cognitive bias evolution curve.

[0050] The task construction module respectively constructs the task interaction groups corresponding to the life knowledge and the professional knowledge, and presents them to the testee in turn.

[0051] The basic information of the testee is collected and structured. The information includes but is not limited to the age, education years, education type (such as science and engineering, medicine, language, etc.), career history (post name, working years, professional field), etc. of the testee. These basic parameters are used as the basis for individual adaptation of task construction, and are directly used as the basis for content design and presentation order of interactive tasks. After completing the information collection, the testee is grouped and constructed, and no less than six rounds of task interaction units are constructed by default, each round of task consisting of one life knowledge task and one professional knowledge task, alternating presentation, and the total number of tasks is no less than twelve.

[0052] In the construction process of the life knowledge task interaction unit, real life situations are explicitly used as the knowledge base, and interactive task content is constructed through graphics, text, voice or embedded operation interface. The task types include scene recognition tasks, routine operation judgment tasks and daily object use sequence reasoning tasks. Among them, the scene recognition task presents common life environments such as bedroom, kitchen, bank business hall and subway station through images, and requires the testee to complete the recognition or matching operation from multiple environment options. The routine operation judgment task is based on the logic of life program behavior to establish operation judgment content, such as identifying laundry process, purchasing medicine process, and kettle use process, etc., and uses step fill-in or operation path selection to interact. The daily object use sequence reasoning task presents multiple life tools or objects (such as tableware, condiments, and electrical components) through a mixed way of graphics and text, and requires the user to complete the correct use sequence arrangement. Each type of task is equipped with a combination of interference items to simulate the cognitive situation of information redundancy or logic dislocation in real life, and the number of interference items is dynamically adjusted according to the task round, with an example value of no less than two per round.

[0053] The professional knowledge type task interaction unit is constructed based on the professional category identified in the testee's professional background data, and personalized task generation is performed by calling the task templates in the professional task library through knowledge extraction rules. The task types include three types of term recognition tasks, special process selection tasks and scenario causal judgment tasks. The term recognition task randomly mixes the professional terms related to the occupation and the unrelated terms in the unified option list, and requires the user to complete the term classification and interpretation matching operation. The special process selection task presents the professional task operation steps through the flowchart module, and presets the logical errors or process rearrangement items, and the testee needs to complete the error correction or sorting adjustment. The scenario causal judgment task sets the combination of key node events in the typical professional workflow, and requires the user to judge the logical and causal relationship of each event. The options of all professional type tasks meet the condition of unique logical correctness, ensuring the stability and discriminant validity of task scoring.

[0054] All life knowledge type and professional knowledge type tasks need to be structured and packaged by unified coding rules to form a task metadata structure. The structure includes task identification code, type field (life / professional), task logic label, preset operation node number, target task chain length and standard answer field. Then, the task construction module integrates all task structure bodies in sequence to generate a complete task sequence group. The presentation strategy of the task sequence group adopts a segmented alternating execution mechanism, that is, each life type task and its corresponding professional type task appear alternately to avoid the aggregation of task types affecting the cognitive test results.

[0055] The interaction extraction module collects the interaction operations of the testee in various tasks, encodes the interaction operations in each round of tasks, and establishes an interaction vector matrix.

[0056] The interaction data flow of the testee in the execution process of each round of tasks is obtained. The interaction data should include the operation actions performed by the testee in each interaction task, and the record should have the basic information of time stamp, operation content, target response object, task identification, etc. All interaction operations need to be collected and archived immediately, and the collection frequency should meet the response time requirement of the task granularity, and the recommended interval is controlled within 50ms to ensure high-precision behavior capture.

[0057] The interaction behavior data of the testee in each round of task is first subjected to a structured decomposition operation. Specifically, each interaction behavior record needs to be decomposed into four basic dimensions: operation action semantics, action object identification, task code identification, and interaction time parameter. The operation action semantics is a descriptive classification reflecting the nature of the behavior, such as "click", "slide", "long press", etc., which can be standardized and classified through the preset behavior type dictionary of task control logic; the action object identification is the specific interactive element label pointed to by the operation, such as image, option, text box, etc., which is extracted in combination with the identification mapping of the task interface elements; the task code identification is uniquely identified according to the task presentation sequence, such as 0101 for the first round of life task and 0203 for the third round of professional task; and the interaction time parameter is the trigger time point of the operation, recorded as the relative timestamp after the start of the task. After completing the extraction of the basic dimensions, the interaction operation vector construction is performed. To ensure the discreteness and comparability of the vector expression, a discrete coding space is defined for each of the above dimensions. The construction of the coding space is based on the preset enumeration rules, and to enhance the semantic relevance and effectiveness of subsequent vector distance calculation, operation actions and objects with similar semantics are assigned adjacent or close coding identifiers to preserve their semantic relevance. For example: in the action semantics dimension, "click" is assigned a coding value of 01, "slide" is assigned a coding value of 02, and "long press" is assigned a coding value of 03; in the action object identification dimension, a corresponding discrete mapping relationship is established according to the fixed number of task elements in the interface layout; the task code dimension directly inherits the sequential number defined during task setting; and the interaction time dimension is interval coded according to time periods, and so on. After completing the coding of each dimension, the vectors are generated by concatenating them in the fixed order of "action semantics - action object - task code - time" to serve as the vector expression of each interaction operation.

[0058] The following is an example: a certain life knowledge class interaction operation "click the wall switch to turn on the light" is coded as a structure vector [01, 12, 0103, 1], where 01 represents "click" action, 12 represents "wall switch" object, 0103 is the current task code, and 1 represents that the operation occurs in the time interval coded as 1; another professional knowledge class operation "drag the reagent bottle to the experiment table" is coded as [05, 21, 0205, 2], and the corresponding coding meanings are "drag", "reagent bottle", "task code 1105", and "time interval code 2". The above structure vector, as the basic unit of the interaction vector matrix, is directly used for subsequent vector distance calculation and jump point recognition operations, ensuring uniform expression form and engineering solvability.

[0059] After each round of task is completed, the interaction operation vectors generated by all interaction operations in the current round are sequentially spliced in the actual operation order to form a complete interaction operation vector sequence. This sequence has a time sequence characteristic and is used to analyze the logical consistency of the operation behavior and the task progress. At the same time, the current task code field is appended to each vector in the sequence as a marker.

[0060] After all the tasks of the set number of rounds are completed, the interaction operation vector sequence of each round of task is integrated into a unified interaction vector matrix. The matrix is arranged in time continuity, the row represents the interaction operation order, and the column corresponds to the four dimensions and the additional code information. Based on the different importance of different dimensions in the evaluation process, different weights are assigned to the columns in the interaction vector matrix. For example, in the subsequent jump point identification and task consistency comparison, the "task code" dimension and the "operation action semantics" dimension have higher discriminant contribution, and the default weight values can be set to 0.35 and 0.30; while the "interaction time" and "action object" dimensions are relatively secondary, and the weight values are set to 0.20 and 0.15. The weight setting adopts normalized configuration, the sum is 1, and the correlation of each dimension to the cognitive decline index in the training data is adjusted and optimized.

[0061] The interaction analysis module performs jump point statistics and vector residual calculation on the interaction vector matrix according to the preset standard task logic chain, and counts the density of invalid operations in the interaction vector matrix.

[0062] After completing the construction of the interaction vector matrix and setting the column weights, the interaction analysis phase is entered, and first, the task structure is standardized modeled to construct a standard task logic chain. The logic chain connects the interaction actions expected to be completed by each round of task in the ideal order by predefining the task execution order, the standard operation steps to be performed in each round of task, and the interaction time set, forms a task standard path, and indicates that all task option designs meet the correctness condition of unique logic. Each task standard path will be indexed with the corresponding task round, corresponding to the task code in the interaction data, ensuring that each round of task can be independently mapped to its corresponding standard path. To facilitate quantitative comparison, each operation in the standard task path is converted into an ideal vector template, which completely corresponds to the interaction operation vector in structure. The ideal vector template is composed of four dimensions, including expected operation action, corresponding object, standard task code, and recommended operation time period, and is filled with ideal operation data of each step in the standard task execution process to construct a multi-row standard structure vector set.

[0063] Subsequently, the interaction vector matrix is mapped to the standard task logic chain, and the operation sequence of each round of task is compared with the standard path sequence in the vertical direction. Specifically, according to the task code field in the interaction operation vector, the matrix is divided by round, and the operation sequence in the corresponding task segment is extracted in time sequence. Then, the expected operation sequence in the standard task logic chain is compared. In the case of skipping, missing, out-of-sequence or redundant operations, the task execution sequence in the interaction vector will have a breakpoint inconsistent with the standard chain. Such a breakpoint is defined as a "jump point". The identification of the jump point uses the task code sequence analysis algorithm to detect whether there is a deviation from the expected sequence in the continuous vector sequence. Once found, the jump point is marked. The following is a specific example:

[0064] The standard task logic chain is the life knowledge sub-task with task codes executed in sequence: 0201→0202→0203→0204→0205, which represents the normal execution path from "kitchen appliance identification" to "bathroom product sequence judgment". In the interaction vector sequence actually executed by a certain testee, the extracted task code sequence is: 0201→0202→0204→0203→0205. In the above execution sequence, the jump behavior from 0202 to 0204 violates the linear progression requirement of the standard logic chain, and 0204 is identified as a jump point that is executed in advance. At the same time, the reverse sequence behavior from 0204 back to 0203 also constitutes a jump point, and 0203 is regarded as a reverse jump point.

[0065] Based on the identification of the jump point, the execution accuracy of the interaction operation is evaluated, and the vector distance between each row of interaction operation vector and its corresponding ideal vector template is calculated. Since the contribution weights of each dimension in cognitive performance are different, the weighted Euclidean distance is used as the weighting vector distance calculation method. The calculation of the weight is based on the preset weight combination in the interaction extraction module.

[0066] The interaction vector matrix is divided into multiple sections according to the task round, and the invalid operation density of the sub-matrix formed by each task section is counted. In each sub-matrix, all interaction operation vectors are traversed to determine whether they are non-target operations. The determination criterion is that if any dimension of the interaction operation vector is completely inconsistent with the ideal vector template (i.e. not included in the code, which is completely irrelevant operation semantics and irrelevant operation object), it is determined as a non-target operation. The total time consumption of non-target operations (accumulated according to the operation time stamp difference) is counted, and the proportion of the non-target operation time consumption to the total time length of the standard task logic chain is taken as the invalid operation density of the sub-matrix. For example, if the standard execution time of a task is 60 seconds, and the actual detected non-target operation time consumption is 24 seconds, then the invalid operation density is 0.4. This density index reflects the interference behavior and deviation degree of the tester in the task execution, and judges whether there are behaviors such as "repeated groping" and "invalid attempts" in the execution of the task. If the behavior density is significantly higher (much higher than the average), it indicates that there is cognitive hesitation or understanding obstacle in this stage.

[0067] The cognitive performance evaluation module generates life cognitive performance indicators and professional cognitive performance indicators through the calculation and statistical output results of the interaction analysis module.

[0068] After completing the skip point recognition, vector residual calculation and invalid operation density statistics of the interaction vector matrix, the cognitive performance evaluation stage is entered. This stage aims to quantify the cognitive level of the tester in life knowledge and professional knowledge related tasks according to the behavior analysis results of the tester under different task types, and to generate cognitive performance indicators with strong comparability and clear interpretability. In the execution process, first, the three indicators generated in the interaction analysis module, i.e. skip point, vector residual and invalid operation density, are divided into task type dimensions according to the task code information embedded in the interaction operation vector. The specific operation is to divide the interaction operation vectors into two sets of life knowledge type tasks and professional knowledge type tasks according to the task code prefix or identification bit, and to respectively count and aggregate the skip point frequency, vector residual value and invalid operation density value corresponding to each task set.

[0069] The statistical method of skip point frequency is to count the number of interaction vector entries that occur skip points in each task type set, and divide it by the total number of operation vectors of the task type to form the skip point frequency index. For example, if a certain type of professional task contains 100 interaction operation vectors, and 28 of them occur skip point behavior, then the skip point frequency of this type of task is 0.28. This value is used to reflect the frequency of the destruction of the logical structure in the operation sequence, and indirectly measures the performance of the understanding of the task order and the stability of the cognitive process.

[0070] The calculation method of the vector residual mean is to perform mean calculation on the residual values of all interaction operation vectors in the task type set, and the result represents the average deviation of the operation behavior from the ideal path in the type of task. Since the residual value has been standardized by the aforementioned weighted vector distance calculation method, the mean value has the feasibility of cross-task type comparison. This index can be regarded as the overall reflection of operation accuracy and understanding correctness.

[0071] The calculation method of the invalid operation density mean is to first divide the sub-matrix according to the task round, obtain the invalid operation density value of each sub-matrix, and then take the mean value of all sub-matrix density values according to the task type to obtain the average invalid operation density index of life type task and professional type task. This index reflects the concentration, target direction and comprehensive understanding of the task purpose of the tested person in the task execution process.

[0072] After the task type division and statistics of the above three indexes are completed, the cognitive performance in different types of tasks can be comprehensively evaluated based on the weighted scoring strategy. In order to ensure the controllability and adaptability of the score, the default scoring weights of the three indexes are preset as follows: the skip point frequency weight is 0.4, the vector residual mean weight is 0.35, and the invalid operation density mean weight is 0.25. With this configuration, the three indexes are brought into the comprehensive scoring function (such as weighted summation) according to the normalized numerical values, and the comprehensive score under each task type is calculated. The comprehensive score is the final output value of the life cognitive performance index and the professional cognitive performance index. The lower the score, the closer the cognitive behavior to the standard task path, the more accurate the operation, and the smaller the deviation. The index is a floating point value in the standardized [0, 1] interval, which can be used for cross-time comparison and population screening evaluation. In actual application, if the life cognitive performance index of the tested person is significantly lower than the professional cognitive performance index, it may indicate that there is a risk of degradation in the ability to handle routine transactions; otherwise, it may reflect a decline in the mastery of professional knowledge.

[0073] The evolution curve construction module records the change sequence of the difference between the life cognitive performance index and the professional cognitive performance index within the preset task test period, and constructs the life and professional cognitive deviation evolution curve.

[0074] After the generation of the life cognitive performance indicators and professional cognitive performance indicators, to further monitor their trends in a set time period, and to identify their potential deviation relationship and evolution characteristics, a cognitive deviation evolution curve needs to be constructed. This step takes the task test period as the basic unit, maps the cognitive performance difference in each round of test as time series data, and realizes the quantitative tracking of the dynamic difference between life and professional cognitive performance. First, in the test scheme setting stage, the basic structure of the task test period is preset. The period is composed of several rounds of interactive tasks, ensuring that not less than a set number of cognitive performance indicator evaluation operations are performed within the period. In the example setting, a task test period can be defined as including 10 rounds of task evaluation, each round of evaluation consisting of a group of interactive tasks, and corresponding to a group of life cognitive performance indicators and professional cognitive performance indicators.

[0075] After each round of evaluation is completed and the corresponding life and professional cognitive performance indicators are obtained, the difference between the two is calculated immediately, defined as the "cognitive deviation evaluation value" of the round of evaluation. The difference is expressed in absolute value form, or the positive and negative signs are retained according to the definition of the indicators to reflect the direction of deviation. For example, if the results of a round of evaluation are: life cognitive performance indicators are 0.32, professional cognitive performance indicators are 0.24, then the cognitive deviation evaluation value is +0.08; if not, then it is -0.08. The sign of the difference can be used to determine which type of cognitive ability is more significantly impaired.

[0076] After completing the calculation of each round of cognitive deviation evaluation value, the system numbers all the evaluation values in chronological order according to the order in the task test period, and forms a sequence of ordered data based on this. Each item in the sequence corresponds to the degree of cognitive deviation at a task evaluation point. Then, the sequence is plotted as a curve to construct the cognitive deviation evolution curve. The horizontal axis of the curve is the number of evaluations within the task test period (such as the first round, the second round, …, the tenth round), and the vertical axis is the cognitive deviation evaluation value corresponding to each evaluation. This curve is used to visually express the trend of the difference between life and professional cognitive performance in the entire test period. The trend of the curve can reveal the volatility, growth, periodicity and potential inflection point of the deviation. To ensure the engineering applicability of the curve, the time span of the horizontal axis should be ensured to be equidistant, and the evaluation value source should be consistent to avoid introducing abnormal values due to uneven number of tasks or fluctuations in evaluation methods. To enhance readability and analysis ability, key points (such as the maximum deviation point, the deviation direction reversal point, etc.) can also be marked in the curve, and sliding window smoothing processing can be performed on local fluctuations to identify trend changes rather than occasional fluctuations.

[0077] The cognitive bias evolution curve, as a basic data structure for subsequent asymmetric degeneration identification, has the following technical values: 1. It can be used to evaluate whether the cognitive ability difference is expanded with the advancement of task testing; 2. It can provide dynamic judgment basis for the direction of cognitive ability deviation; 3. It can provide interpretable quantitative clues for the construction of individualized intervention strategies or early degeneration screening. Especially in population screening or continuous tracking, if the curve shows a stable upward trend or a sharp fluctuation feature, further analysis mechanism needs to be introduced to judge whether it belongs to the asymmetric cognitive degeneration performance.

[0078] The asymmetric degeneration identification module judges whether the cognitive ability of the tested person has an asymmetric degeneration trend according to the cognitive bias evolution curve.

[0079] After the cognitive bias evolution curve is constructed, a fixed number of continuous evaluation values are selected from the first evaluation value to form a sliding window sequence. The length of the sliding window should be determined in combination with the total number of evaluations within the task testing period and the trend sensitivity required for identification, and the recommended value is 5 to 8 consecutive evaluation results within the period. Taking 6 as an example, the cognitive bias evaluation values of the last 6 rounds are taken as a subsequence each time the calculation is performed, which is used as the basis for judgment.

[0080] In each sliding window, the cognitive bias evaluation values in the window are sequentially compared to see whether they strictly satisfy the monotone increasing relationship, that is, for any subsequence, the two adjacent values satisfy that the latter value is greater than the former value. If it is satisfied, it is judged whether the overall growth rate of the cognitive bias values in the window exceeds the preset change rate threshold. The growth rate is defined as the difference between the evaluation value at the end of the sliding window and the evaluation value at the beginning divided by the absolute value of the starting value. If the starting value is close to 0, a lower limit value (such as 0.05) is introduced to avoid extreme amplification. The threshold value is usually set to be between 15% and 30%, combined with the test benchmark of the normal cognitive population. For example, if the change rate threshold is set to 0.2 and the growth rate of the first and last values of the current window is 0.27, it satisfies the trend significance judgment standard. Once the sliding window satisfies the dual judgment conditions of continuous monotone increase and change rate exceeding the threshold, it is determined that the current tested person has an asymmetric degeneration trend between the life cognitive performance and the professional cognitive performance. At this time, a screening result warning should be generated, and the abnormal label should be attached to the current task testing period for subsequent report generation, medical intervention suggestion or further cognitive testing reference.

[0081] The above formulas are dimensionless values, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the latest real situation. The preset parameters and threshold values in the formula are set by the person skilled in the art according to the actual situation.

[0082] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above-described embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center through a wired (for example, infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0083] Those of ordinary skill in the art can realize that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0084] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device, and module can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0085] In several embodiments provided in the present application, it should be understood that the disclosed system, device, and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed ones can be indirect coupling or communication connection through some interfaces, devices, or modules, which can be electrical, mechanical, or other forms.

[0086] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment.

[0087] In addition, the functional modules in each embodiment of the present application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0088] If the functions are realized in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0089] The above is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0090] Finally: the above is only the preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application, should be included in the protection scope of the present application.

Claims

1. An Alzheimer's disease assisted screening system based on intelligent interaction, characterized in that, It includes a task construction module, an interaction extraction module, an interaction analysis module, a cognitive performance evaluation module, an evolution curve construction module, and an asymmetric degradation recognition module, among which: The task building module constructs task interaction groups corresponding to life knowledge and professional knowledge, and presents them to the test subjects in sequence. The interaction extraction module collects the interactive operations of the test subjects in various tasks, encodes the interactive operations in each round of tasks, and establishes an interaction vector matrix; The interaction analysis module performs jump point statistics and vector residual calculation on the interaction vector matrix according to the preset standard task logic chain, and counts the density of invalid operations in the interaction vector matrix. The cognitive performance assessment module generates life cognitive performance indicators and professional cognitive performance indicators through the calculation and statistical output results of the interactive analysis module. The evolution curve construction module records the sequence of differences between life cognitive performance indicators and professional cognitive performance indicators within a preset task testing period, and constructs an evolution curve of the deviation between life and professional cognition. The asymmetric degradation identification module determines whether the test subject's cognitive ability exhibits an asymmetric degradation trend based on the cognitive bias evolution curve; The interaction extraction module collects the test subjects' interactive operations in various tasks, encodes the interactive operations in each round of tasks, and establishes an interaction vector matrix, specifically including: Acquire the interactive operations of the test subjects during each round of tasks, decompose each interactive operation in a structured manner, and extract multi-dimensional interactive data containing operation semantics, action objects, task codes and interaction time; Multidimensional interactive data is converted into interactive operation vector representations through preset encoding mapping rules; After each round of tasks is completed, the interaction operation vectors are concatenated into a complete sequence of interaction operation vectors according to the order of interaction operations, and then marked according to the task code; After completing the task with a set number of rounds, integrate the sequence of interactive operation vectors, establish a temporally continuous and task-distinguishable interactive vector matrix, and set different weights for the columns in the interactive vector matrix that represent the dimensions of interactive data. The interaction analysis module performs jump point statistics and vector residual calculation on the interaction vector matrix according to the preset standard task logic chain, and specifically includes the following: A pre-defined standard task logic chain is established. This logic chain connects the expected interactive actions to be completed in each round of tasks in an ideal order by pre-defining the task execution order, the standard operation steps to be executed in each round of tasks, and the set of interaction times, forming a standard task path and defining an ideal vector template for multi-dimensional interactive data in each round of tasks. The interaction vector matrix is ​​mapped to the standard task logic chain. The position of the task logic jump is identified vertically in the interaction vector matrix by the task encoding order in the interaction operation vector and marked as the jump point. Based on the column weights of the interaction vector matrix, calculate the distance between the weighted vector of each row of the interaction operation vector in the interaction vector matrix and the ideal vector template, and obtain the vector residual of the interaction operation. The interaction vector matrix is ​​divided into several sub-matrices according to the task rounds, and the invalid operation density is statistically analyzed for the interaction operation vectors of the sub-matrices. The invalid operation density is obtained by calculating the proportion of the interaction time corresponding to the interaction operation vector of the non-target operation to the total time of the standard task logic chain; The non-target operation is any interaction operation that is inconsistent with the ideal vector template.

2. The Alzheimer's disease assisted screening system based on intelligent interaction according to claim 1, characterized in that, The task construction module constructs task interaction groups corresponding to life knowledge and professional knowledge, and presents them to the test subject in sequence, including: Based on the age, educational background and professional experience of the test subjects, construct interactive tasks related to life knowledge with no fewer than a set number of rounds. The tasks include scene recognition, judgment of routine operations and reasoning about the order of use of everyday items. At the same time, we will build a number of professional knowledge-based interactive tasks that are equal to the number of life knowledge-based interactive tasks. The tasks include terminology identification, selection of special processes, and causal judgment within a scenario. All interactive tasks are uniformly coded and integrated into task sequence groups, and presented to the test subjects in a segmented and alternating manner.

3. The Alzheimer's disease assisted screening system based on intelligent interaction according to claim 1, characterized in that, The process of constructing the interactive operation vector representation involves setting an enumerable discrete encoding space for each dimension of the multidimensional interactive data, and concatenating them into a structure vector according to a fixed order of action semantics, action object, task encoding, and interaction time.

4. The Alzheimer's disease assisted screening system based on intelligent interaction according to claim 1, characterized in that, The cognitive performance assessment module generates life cognitive performance indicators and professional cognitive performance indicators through the calculation and statistical output results of the interactive analysis module. These indicators specifically include: The jump points, vector residuals, and invalid operation densities in the interaction analysis module are divided into task types according to the task codes corresponding to their respective interaction operation vectors. The jump point frequency, mean vector residual, and mean invalid operation density of different task types are then integrated and calculated. Based on the jump point frequency, mean vector residual, and mean invalid operation density of different task types, life cognitive performance indicators and professional cognitive performance indicators are generated through a weighted comprehensive scoring method.

5. The Alzheimer's disease assisted screening system based on intelligent interaction according to claim 1, characterized in that, The evolution curve construction module records the sequence of differences between life cognitive performance indicators and professional cognitive performance indicators within a preset task testing period, and constructs the evolution curve of life and professional cognitive deviation, specifically including: A pre-defined task testing cycle is established, and one task testing cycle includes several assessments of cognitive performance indicators. After each cognitive performance indicator assessment, the difference between the life cognitive performance indicator and the professional cognitive performance indicator is calculated and recorded as the cognitive bias assessment value. All cognitive bias assessment values ​​within the preset task testing period are arranged by time to construct a cognitive bias evolution curve. The vertical axis of the curve represents the difference between life and professional cognitive performance indicators, and the horizontal axis represents the number of assessments of cognitive performance indicators.

6. The Alzheimer's disease assisted screening system based on intelligent interaction according to claim 5, characterized in that, The asymmetric degradation identification module determines whether the test subject's cognitive ability exhibits an asymmetric degradation trend based on the cognitive bias evolution curve, specifically including: When the cognitive bias evolution curve is within a preset sliding window, and the cognitive bias assessment value continues to show a monotonically increasing trend, and the increase exceeds the set change rate threshold, it is determined that there is an asymmetric degradation between the test subject's life cognitive performance and professional cognitive performance, and a screening result warning is issued.

Images