Method and system for identifying MCI through rules based on typical features

Abstract

The invention discloses a method for identifying MCI through rules based on typical features, and the method comprises the steps: obtaining to-be-identified detection data, and carrying out the typical feature extraction and weight combination calculation according to the detection data; and comparing the calculated weight combination with a predetermined optimal classification threshold value, and identifying whether MCI exists or not according to a comparison result. According to the method, the MCI can be quickly and accurately identified by constructing the feature threshold table, combining the feature weights and combining and calculating the optimal classification threshold, and an efficient, low-cost and easy-to-understand and easy-to-apply alternative tool is provided for MCI screening.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a method for identifying mild cognitive impairment (MCI) through rules based on typical features. Background Art

[0002] Mild cognitive impairment (MCI) is a critical transitional stage between normal aging and dementia (such as Alzheimer's disease). Patients experience measurable cognitive decline, but this decline does not yet significantly impact their ability to function. Early and accurate identification of MCI is crucial for slowing dementia progression and implementing effective interventions. Numerous studies indicate that approximately 10%-15% of MCI patients will develop dementia each year. Therefore, finding efficient and reliable methods for MCI identification has become a critical and pressing issue for the medical community.

[0003] Currently, the identification of MCI mainly relies on the following methods: 1. Neuropsychological scales: such as the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA). These scales require professional physicians to administer, which is not only time-consuming but also susceptible to subjective judgment and bias.

[0004] 2. Biomarker testing: Although detection methods such as cerebrospinal fluid Aβ / Tau detection and positron emission tomography (PET) imaging have high accuracy, the detection costs are high and some tests are invasive, which puts a physical burden on patients and is difficult to be widely used in clinical practice.

[0005] 3. Machine Learning Models: Machine learning models based on magnetic resonance imaging (MRI) / functional magnetic resonance imaging (fMRI) data can assist in the diagnosis of MCI to a certain extent. However, the training of such models is highly dependent on large-scale labeled data, resulting in poor interpretability. Clinical deployment also requires high technical barriers and hardware equipment support, which limits its widespread application. Summary of the Invention

[0006] To address the difficult-to-balance issues of existing MCI identification methods, such as high cost, complex operation, and poor interpretability, the present invention provides a method for identifying MCI through rules based on typical features. By constructing a feature threshold table, combining feature weights, and calculating the optimal classification threshold, MCI can be identified quickly and accurately, providing an efficient, low-cost, and easy-to-understand and apply alternative tool for MCI screening.

[0007] According to one aspect of the present invention, a method for identifying MCI by rules based on typical features is provided, comprising: Acquire detection data to be identified, and perform typical feature extraction and weight combination calculation based on the detection data; Comparing the calculated weight combination with a predetermined optimal classification threshold, and identifying whether it is MCI based on the comparison result; wherein the predetermined optimal classification threshold includes: Based on the collected multi-source data samples, feature threshold tables are constructed respectively. Each feature threshold table contains multiple candidate thresholds. The candidate thresholds are used to divide and mark the feature values extracted from typical features according to different candidate threshold intervals; Based on the constructed feature threshold table, the labeling results of different dimensional features in their respective candidate threshold intervals are combined to obtain a series of different weight combinations; Using weight combinations as nodes, a decision tree structure is constructed, and feature combinations and node additions are performed step by step until an ordered weight combination of leaf nodes is obtained; Based on the obtained ordered weight combination of leaf nodes, the classification threshold is calculated, and the optimal classification threshold is determined through the classification index.

[0008] As a further technical solution, the method further includes: Characteristic data were obtained using one or more of the following methods: neuropsychological scales, patient daily behavior diaries, simple digital tests, and functional near-infrared spectroscopy.

[0009] As a further technical solution, a feature threshold table is constructed, including: Extract typical features from the acquired feature data, including ACC accuracy, RT reaction time, frontopolar beta, frontopolar integral, dorsolateral beta, and dorsolateral integral; Based on the extracted typical features, candidate thresholds are divided and candidate threshold intervals are marked, and a feature threshold table is constructed for each typical feature.

[0010] As a further technical solution, constructing a feature threshold table also includes: For each typical feature, multiple candidate thresholds are set, and candidate threshold intervals are formed between adjacent candidate thresholds. Each candidate threshold interval corresponds to a marking symbol, thereby forming a feature judgment system including six feature threshold tables.

[0011] As a further technical solution, based on the constructed feature threshold table, the labeling results of different dimensional features in their respective candidate threshold intervals are combined, including: According to the set rules and order, each feature labeling result is gradually merged two by two or multiple times to generate a series of different weight combinations.

[0012] As a further technical solution, a decision tree structure is constructed with weight combinations as nodes, and feature combinations and node additions are performed level by level until an ordered weight combination of leaf nodes is obtained, including: In the process of combining features, each newly generated weight combination is treated as a child node of the parent node and arranged according to the order and logical relationship of the combination; Continue to combine features and add nodes until no new combinations can be made. At this time, the nodes at the end of the decision tree are leaf nodes, and the weight combinations represented by these leaf nodes are the final ordered weight combinations.

[0013] As a further technical solution, based on the obtained ordered weight combination of leaf nodes, the classification threshold is calculated, and the optimal classification threshold is determined by the classification index, including: Collect patient data with known MCI status as training samples, and use the ordered weight combination corresponding to each sample as input data; Using maximum likelihood estimation or cross-validation methods, calculations and optimizations are performed on these ordered weight combination data based on threshold sets to calculate classification indicators at different thresholds; The threshold that makes the classification index reach the optimal value is selected as the final optimal classification threshold.

[0014] As a further technical solution, the method further includes: When the patient is determined to be suspected MCI based on the comparison result of the calculated weight combination and the predetermined optimal classification threshold, an MCI identification report is generated.

[0015] According to one aspect of the present invention, a regularized MCI recognition system based on a combination of typical features is provided, comprising: The first main module is used to obtain detection data to be identified, and perform typical feature extraction and weight combination calculation based on the detection data; The second main module is configured to compare the calculated weight combination with a predetermined optimal classification threshold, and identify whether the patient is MCI based on the comparison result; wherein the predetermined optimal classification threshold includes: Based on the collected multi-source data samples, feature threshold tables are constructed respectively. Each feature threshold table contains multiple candidate thresholds. The candidate thresholds are used to divide and mark the feature values extracted from typical features according to different candidate threshold intervals; Based on the constructed feature threshold table, the labeling results of different dimensional features in their respective candidate threshold intervals are combined to obtain a series of different weight combinations; Using weight combinations as nodes, a decision tree structure is constructed, and feature combinations and node additions are performed step by step until an ordered weight combination of leaf nodes is obtained; Based on the obtained ordered weight combination of leaf nodes, the classification threshold is calculated, and the optimal classification threshold is determined through the classification index.

[0016] According to one aspect of the present invention, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium stores computer instructions, wherein the computer instructions enable the computer to execute the method for identifying MCI by rules based on typical features.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. Improved Accuracy: By constructing a feature threshold table to finely divide features, combining the weighted combination of gradually merged features, and calculating the optimal classification threshold, it can more accurately capture the characteristic differences of MCI patients. Compared with traditional methods, it greatly improves the accuracy of MCI identification and effectively reduces the rates of misdiagnosis and missed diagnosis.

[0018] 2. Flexible and adjustable: The candidate thresholds in the feature threshold table can be flexibly adjusted according to different clinical needs and sample data characteristics, and the construction method of the weight combination can also be optimized, making this method adaptable to diverse application scenarios and more versatile and scalable.

[0019] 3. Clear logic and easy explanation: The entire recognition process is based on clear rules and steps. From feature threshold division to weight combination, and then to the calculation of the optimal classification threshold, each link has clear logic and basis, which makes it easier for doctors to understand and explain the diagnostic process to patients, enhancing the credibility of the diagnostic results.

[0020] 4. Cost and efficiency advantages: This method mainly relies on routine data collection and rule-based calculations. It does not require complex and expensive detection equipment and large amounts of labeled data. The operation process is relatively simple. While reducing diagnostic costs, it can quickly complete MCI identification and is suitable for widespread promotion and application in primary medical institutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief introduction will be given below to the drawings used in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 A flowchart of a method for identifying MCI using rules based on typical features provided by an embodiment of the present invention.

[0023] Figure 2 A schematic diagram of the process of determining the optimal classification threshold provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] It should be noted that: In the existing technology, regularized recognition methods based on typical cognitive characteristics (such as delayed memory recall, decreased executive function, and reduced language fluency) have not been fully studied and applied.

[0025] This type of regularized recognition method has many potential advantages: 1. Interpretability: By adopting clinically validated feature thresholds, it can be directly linked to the diagnostic criteria for MCI, providing a clear and explicit basis for interpretation of the diagnostic results, making them easier for doctors and patients to understand.

[0026] 2. Efficiency: This method does not require complex calculations and data analysis, is simple and quick to operate, and is very suitable for primary medical institutions to carry out preliminary screening for MCI.

[0027] 3. Scalability: It can flexibly integrate multi-source data, including neuropsychological scale scores, patients' daily behavior logs, simple digital test results, etc., to comprehensively assess the patient's cognitive status from multiple dimensions and improve the accuracy and reliability of diagnosis.

[0028] Based on this, the present invention proposes a regularized MCI identification method based on typical feature combinations. By defining multi-dimensional feature criteria and constructing a logical decision tree, specifically constructing a feature threshold table, merging feature weight combinations and calculating the optimal classification threshold, it effectively solves the difficult-to-balance problems of high cost, complex operation, poor interpretability and other existing MCI identification methods, and provides an efficient, low-cost, easy-to-understand and easy-to-use alternative tool for MCI screening.

[0029] It should be pointed out that the "combination" and "merger" referred to in the present invention both refer to the merging of the weights of different features. For example, the weights of feature 1 are A1, A2, A3, and the weights of feature 2 are B1, B2. Then, the ordered combination of the weights of feature 1 and feature 2 refers to the arrangement and combination of A1, A2, A3 and B1, B2. This operation can be regarded as the combination or merger of feature 1 and feature 2 in the specification of the present invention.

[0030] The terms "including" and "having" and any variations thereof in the description and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions, for example, a process, method, system, product or apparatus that includes a series of steps or units is not necessarily limited to the steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0031] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, the technical features in the various embodiments or single embodiments provided by the present invention are arbitrarily combined with each other to form a new technical solution. This combination is not restricted by the sequence of steps and / or structural composition mode, but must be based on the ability of ordinary technicians in this field to implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that this combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0032] The embodiment of the present invention provides a method for identifying MCI by rules based on typical features, such as Figure 1 As shown, first, the detection data to be identified is obtained, and typical features are extracted and weight combination calculations are performed based on the detection data; then, the calculated weight combination is compared with a predetermined optimal classification threshold, and whether it is MCI is identified based on the comparison result.

[0033] Obtain the test data to be identified, specifically: collect multi-source data from patients, including but not limited to neuropsychological scales (such as the subscores of the MoCA scale), patients' daily behavior logs (such as the number of times they forget items, the frequency of getting lost, etc.), simple digital test results (such as the forward / reverse score of the digit span test), and functional near-infrared spectroscopy (fNIRS) test data.

[0034] Typical feature extraction and weighted combination calculation are performed based on the test data. Specifically, the numerical values of each typical feature are extracted from the test data to be identified. The candidate threshold intervals for each numerical value are searched in a pre-established feature threshold table, and the corresponding markers for each numerical value within the candidate threshold intervals are determined. Then, according to the weighted combination rules, the markers corresponding to the numerical values are combined to obtain a final weighted combination. This final weighted combination is compared with a pre-determined optimal classification threshold. If the threshold condition is met, the patient is classified as a suspected MCI patient; otherwise, the patient is classified as a non-suspected MCI patient.

[0035] like Figure 2 As shown, the predetermination of the optimal classification threshold includes the following steps: Step 1: construct feature threshold tables based on the collected multi-source data samples. Each feature threshold table contains multiple candidate thresholds. The candidate thresholds are used to divide and mark the feature values extracted from typical features according to different candidate threshold intervals.

[0036] In this step, multi-source data samples are collected, including but not limited to neuropsychological scales, patients' daily behavior logs, simple digital test results, and functional near-infrared spectroscopy data.

[0037] Furthermore, typical features are extracted based on the collected data, such as memory features, executive function features, language fluency features, etc. Preferably, the extracted typical features may include ACC accuracy, RT reaction time, frontopolar beta, frontopolar integral, dorsolateral beta, and dorsolateral integral.

[0038] Furthermore, for each typical feature extracted, a feature threshold table is constructed. Each feature threshold table contains multiple candidate thresholds, and the feature values are divided and marked according to different candidate threshold intervals.

[0039] Step 2: Based on the constructed feature threshold table, the labeling results of different dimensional features in their respective candidate threshold intervals are combined to obtain a series of different weight combinations.

[0040] Based on the constructed feature threshold table, the labeling results of different features are combined. For example, the ACC accuracy rate labeled as A1 and the RT reaction time labeled as B1 are combined to obtain the new weighted combination G6(A1B1). Following a specific rule and sequence, the various feature labeling results are gradually combined in pairs or multiple combinations to generate a series of different weighted combinations.

[0041] It should be noted that the certain rules and order described here refer to grouping multiple features once, combining the features in each group in an orderly manner to obtain the combined weights of each group, and then grouping the multiple features in each group after the first grouping a second time, combining the features in each group after the second grouping in an orderly manner to obtain the combined weights of each group at the next level, and so on, grouping and orderly combining within the groups step by step until no further feature combination can be performed, and finally forming a number of combined weights with different hierarchical structures.

[0042] For example, when extracting ACC accuracy, RT reaction time, frontopolar beta, frontopolar integral, dorsolateral beta, and dorsolateral integral, it is preferable to initially group ACC accuracy and RT reaction time into one group, and frontopolar beta, frontopolar integral, dorsolateral beta, and dorsolateral integral into another group. For the combinations of ACC accuracy (weights A1, A2, and A3) and RT reaction time (weights B1 and B2), six different weight combinations can be obtained by sequentially combining the weights: A1B1, A2B1, A3B1, A1B2, A2B2, and A3B2. Similarly, another group with four features—frontopolar beta (weights C1, C2, C3), frontopolar integral (weights D1, D2, D3), dorsolateral beta (weights E1, E2, E3), and dorsolateral integral (weights F1, F2, F3)—can be further grouped to obtain two subgroups each containing two features, such as C and D in one group and E and F in another. For each subgroup, nine different weight combinations are formed. The weight combinations of the two subgroups are then combined in an ordered manner, resulting in 45 different weight combinations. Finally, the combined weights obtained from the first two groups are combined again in an ordered manner to form the final weight combination.

[0043] Step 3: Use the weight combination as a node to build a decision tree structure, perform feature combination and node addition step by step until the ordered weight combination of the leaf node is obtained.

[0044] By continuously merging and combining features, a decision tree-like structure is constructed. Nodes are added based on the series of different weight combinations obtained in step 2, ultimately resulting in ordered weight combinations at the leaf nodes. These ordered weight combinations represent the final state under different feature combinations and are used for subsequent classification decisions.

[0045] Step 4: Calculate the classification threshold based on the obtained ordered weight combination of leaf nodes, and determine the optimal classification threshold through the classification index.

[0046] Based on the resulting ordered weight combinations, statistical analysis or other appropriate algorithms are used to calculate the optimal classification threshold. By continuously adjusting and optimizing this threshold, the classification of whether a patient has MCI based on this threshold is achieved with the highest accuracy and reliability.

[0047] Furthermore, based on the calculated optimal classification threshold, the patient's characteristic data is judged, and the MCI identification result is output to clearly inform the patient whether he or she is a suspected MCI patient and provide corresponding diagnostic suggestions.

[0048] As a preferred embodiment, the embodiment of the present invention illustrates the specific process of identifying MCI through rules based on typical features, which mainly includes the data collection stage, the feature threshold table construction stage, the step-by-step merging of features to obtain weight combination stage, the ordered weight combination stage of obtaining leaf nodes, the optimal classification threshold calculation stage and the result output stage.

[0049] During the data collection phase, the data from the neuropsychological scale test, the simple digital test, and the fNIRS test were used as the test data to be identified, as follows: 1. Neuropsychological Assessment: Professionally trained medical staff will administer the MoCA assessment according to standard operating procedures, recording each subscore in detail. During the test, ensure a quiet and comfortable environment to avoid external interference that could affect the patient's test results.

[0050] 2. Simple Digit Test: This test involves testing the patient's digit span, including both forward and reverse digit memory. During the test, a sequence of digits is verbally recited to the patient at a rate of one digit per second. The patient is asked to repeat the sequence in the correct order, and the length of the longest sequence the patient can accurately repeat is recorded.

[0051] 3. fNIRS test: Use functional near-infrared spectroscopy equipment for testing. Let the patient sit in a comfortable chair, keep quiet and relaxed, and fix the probe array of the fNIRS device on the patient's forehead according to the standard position to ensure that the probe is in close contact with the skin. During the test, the patient can perform simple cognitive tasks, such as digital memory, word association, etc. At the same time, the device continuously collects data on changes in hemoglobin oxygenation and deoxyhemoglobin concentration in the patient's prefrontal cortex area. The sampling frequency is set to 10 Hz, and each test lasts for 5 minutes. Record the original data during the test, including the optical density change values ​​at different time points and different channels.

[0052] In the stage of building the feature threshold table, the specific steps are as follows: Nback task feature threshold table construction: Based on the back task score (ACC) greater than 0.7 (out of 1), it is labeled A1; scores greater than or equal to 0.5 and less than or equal to 0.7 are labeled A2; and scores less than 0.5 are labeled A3. Furthermore, based on RT (reaction time), scores less than 740 are labeled B1; scores greater than or equal to 740 are labeled B2. This constructs the nback feature threshold table.

[0053] FNIRS detection feature threshold table construction: Based on the frontopolar beta score, scores greater than or equal to 0.095 are labeled C1; scores greater than 0 but less than 0.095 are labeled C2, and scores less than or equal to 0 are labeled C3. Based on the frontopolar integral score, scores greater than or equal to 10 are labeled D1, scores greater than 0 but less than 10 are labeled D2, and scores less than or equal to 0 are labeled D3. Based on the dorsolateral beta score, scores greater than or equal to 0.095 are labeled E1, scores greater than 0 but less than 0.095 are labeled E2, and scores less than or equal to 0 are labeled E3. Based on the dorsolateral integral score, scores greater than or equal to 10 are labeled F1; scores greater than 0 but less than 10 are labeled F2, and scores less than or equal to 0 are labeled F3.

[0054] The fNIRS detection feature labeling results of the above different dimensions are integrated into the fNIRS detection feature threshold table, and together with other feature threshold tables, they constitute a complete feature judgment system.

[0055] The constructed feature threshold tables are shown in Table 1 and Table 2, where Table 1 shows the divided candidate threshold intervals, and Table 2 shows the marking symbols of each candidate threshold interval, that is, the weight.

[0056] Table 1 Candidate threshold intervals .

[0057] Table 2 Marking symbols for each candidate threshold interval .

[0058] Gradually merge features to obtain the weight combination stage. The specific process is as follows: Based on the existing feature combinations, the labeling results from the fNIRS feature threshold table are combined. For example, A1 for ACC accuracy and B3 for RT reaction time are combined with D1, the global brain activation intensity marker from the fNIRS feature, to obtain the new weighted combination G15 (A1B3D1). Alternatively, C2 for verbal fluency, E2, the specific brain activation asymmetry marker from the fNIRS feature, and A3 for ACC accuracy are combined to obtain the weighted combination G16 (C2E2A3). Following the established merging rules, various feature labeling results, including fNIRS features, are continuously merged to further enrich the diversity of weighted combinations.

[0059] Obtain the ordered weight combination stage of the leaf nodes. The specific process is as follows: 1. Build a decision tree structure with weight combinations as nodes. During the feature merging process, each newly generated weight combination is treated as a child node of the parent node and arranged according to the order of merging and logical relationship.

[0060] 2. Continue merging features and adding nodes until no new mergers can be performed. At this time, the nodes at the end of the decision tree are leaf nodes, and the weight combinations represented by these leaf nodes are the final ordered weight combinations.

[0061] The final ordered weight combination and the corresponding interval range are shown in Table 3.

[0062] Table 3 Final ordered weight combinations and corresponding interval ranges .

[0063] In Table 3, G represents the combination of features A and B, for example, A1B1 represents G6, A1B2 represents G5, and so on. J represents the combination of CDEF, where the combination of C and D is represented as H, the combination of E and F is represented as I, and the combination of H and I is further represented as J. Therefore, after obtaining the patient's feature values ABCDEF, A and B can be combined to determine G, C and D to determine H, E and F to determine I, and then H and I to determine J. Finally, the final weight combination and its corresponding numerical range can be determined by combining G and J.

[0064] It should be noted that in Table 3, the values increase from 1. The values themselves have no special meaning, but are human-defined rules that are considered when designing feature mappings. For example, we want the weights corresponding to samples that meet the A1B2 features to be as small as possible.

[0065] Calculate the optimal classification threshold stage. The specific process is as follows: 1. Collect a certain amount of patient data with known MCI status as training samples, and use the ordered weight combination corresponding to each sample as input data.

[0066] 2. Use algorithms such as maximum likelihood estimation and cross-validation to perform calculations and optimizations on these ordered weighted combinations based on a set of thresholds (as shown in Table 3). By repeatedly trying different thresholds, calculate metrics such as accuracy and recall for the classification at each threshold.

[0067] 3. Select the threshold that makes the classification index reach the best (such as the highest accuracy and a high recall rate) as the final optimal classification threshold.

[0068] Table 4 Threshold range table .

[0069] Table 4 shows the threshold range for each feature, allowing us to determine a reasonable threshold within this range. For example, the ACC range is between 0.7 and 0.8, with values taken every 0.05, i.e., 0.7\0.75\...\0.8. During the optimization process, the optimal threshold is selected from these values.

[0070] In the result output stage, the specific process is as follows: 1. Process the characteristic data of the patient to be diagnosed according to the above process to obtain the corresponding ordered weight combination.

[0071] 2. Compare the ordered weight combination with the calculated optimal classification threshold. If the threshold condition is met, the patient is judged as a suspected MCI patient; otherwise, the patient is judged as a non-suspected MCI patient.

[0072] 3. Generate a detailed MCI identification report, including the patient's basic information, test results, feature thresholds, weight combination process, final judgment, and corresponding diagnostic recommendations. The report is provided to the patient and their family in written or electronic form. Medical staff will explain the report in detail, inform the patient of follow-up precautions, and provide recommendations for further examination or treatment.

[0073] The implementation of each embodiment of the present invention is based on programmed processing performed by a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of each embodiment of the present invention are packaged into various modules. Based on this reality, and in addition to the aforementioned embodiments, an embodiment of the present invention provides a regularized MCI identification system based on typical feature combinations. This system is used to implement the regularized MCI identification method based on typical feature combinations described in the aforementioned method embodiments.

[0074] The system includes: a first main module for acquiring detection data to be identified, extracting typical features and calculating weight combinations based on the detection data; a second main module for comparing the calculated weight combinations with a predetermined optimal classification threshold, and identifying whether the condition is MCI based on the comparison result; wherein the predetermined optimal classification threshold includes: Based on the collected multi-source data samples, feature threshold tables are constructed respectively. Each feature threshold table contains multiple candidate thresholds. The candidate thresholds are used to divide and mark the feature values extracted from typical features according to different candidate threshold intervals; based on the constructed feature threshold table, the marking results of different dimensional features in their respective candidate threshold intervals are combined to obtain a series of different weight combinations; with the weight combination as the node, a decision tree structure is constructed, and feature combination and node addition are performed step by step until the ordered weight combination of leaf nodes is obtained; based on the obtained ordered weight combination of leaf nodes, the classification threshold is calculated, and the optimal classification threshold is determined through the classification index.

[0075] The regularized MCI recognition system based on typical feature combinations provided by the embodiments of the present invention addresses the difficult-to-balance issues of existing MCI recognition methods, such as high cost, complex operation, and poor interpretability. By adopting the aforementioned modules, by constructing a feature threshold table, merging feature weight combinations, and calculating the optimal classification threshold, MCI can be identified quickly and accurately, providing an efficient, low-cost, and easy-to-understand and easy-to-use alternative tool for MCI screening.

[0076] It should be noted that the system embodiments provided by the present invention are not only used to implement the methods in the above-mentioned method embodiments, but also used to implement the methods in other method embodiments provided by the present invention. The only difference lies in the setting of corresponding functional modules, and the principles thereof are basically the same as the principles of the above-mentioned system embodiments provided by the present invention. As long as those skilled in the art refer to the specific technical solutions in other method embodiments on the basis of the above-mentioned system embodiments, obtain corresponding technical means and technical solutions composed of these technical means by combining technical features, and on the premise of ensuring the practicality of the technical solutions, improve the modules in the above-mentioned system embodiments to obtain corresponding system class embodiments for implementing the methods in other method class embodiments.

[0077] Based on the same inventive concept as the aforementioned embodiment, an embodiment of the present invention further provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions cause the computer to execute the method for identifying MCI based on typical features and rules, including: Acquire detection data to be identified, and perform typical feature extraction and weight combination calculation based on the detection data; Comparing the calculated weight combination with a predetermined optimal classification threshold, and identifying whether it is MCI based on the comparison result; wherein the predetermined optimal classification threshold includes: Based on the collected multi-source data samples, feature threshold tables are constructed respectively. Each feature threshold table contains multiple candidate thresholds. The candidate thresholds are used to divide and mark the feature values extracted from typical features according to different candidate threshold intervals; Based on the constructed feature threshold table, the labeling results of different dimensional features in their respective candidate threshold intervals are combined to obtain a series of different weight combinations; Using weight combinations as nodes, a decision tree structure is constructed, and feature combinations and node additions are performed step by step until an ordered weight combination of leaf nodes is obtained; Based on the obtained ordered weight combination of leaf nodes, the classification threshold is calculated, and the optimal classification threshold is determined through the classification index.

[0078] In summary, the method for identifying MCI using rules based on typical features of the present invention has the following advantages: 1. Improved Accuracy: By constructing a feature threshold table to finely divide features, combining the weighted combination of gradually merged features, and calculating the optimal classification threshold, the system can more accurately capture the characteristic differences of MCI patients. Compared with traditional methods, it greatly improves the accuracy of MCI identification and effectively reduces the rates of misdiagnosis and missed diagnosis.

[0079] 2. Flexible and adjustable: The candidate thresholds in the feature threshold table can be flexibly adjusted according to different clinical needs and sample data characteristics, and the construction method of the weight combination can also be optimized, making this method adaptable to diverse application scenarios and more versatile and scalable.

[0080] 3. Clear logic and easy explanation: The entire recognition process is based on clear rules and steps. From feature threshold division to weight combination, and then to the calculation of the optimal classification threshold, each link has clear logic and basis, which makes it easier for doctors to understand and explain the diagnostic process to patients, enhancing the credibility of the diagnostic results.

[0081] 4. Cost and efficiency advantages: This method mainly relies on routine data collection and rule-based calculations. It does not require complex and expensive testing equipment or large amounts of labeled data. The operation process is relatively simple. While reducing diagnostic costs, it can quickly complete MCI identification and is suitable for widespread promotion and application in primary medical institutions.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying MCI using rules based on typical features, characterized in that: include: Acquire detection data to be identified, and perform typical feature extraction and weight combination calculation based on the detection data; Comparing the calculated weight combination with a predetermined optimal classification threshold, and identifying whether it is MCI based on the comparison result; wherein the predetermined optimal classification threshold includes: Based on the collected multi-source data samples, feature threshold tables are constructed respectively. Each feature threshold table contains multiple candidate thresholds. The candidate thresholds are used to divide and mark the feature values extracted from typical features according to different candidate threshold intervals; Based on the constructed feature threshold table, the labeling results of different dimensional features in their respective candidate threshold intervals are combined to obtain a series of different weight combinations; Using weight combinations as nodes, a decision tree structure is constructed, and feature combinations and node additions are performed step by step until an ordered weight combination of leaf nodes is obtained; Based on the obtained ordered weight combination of leaf nodes, the classification threshold is calculated, and the optimal classification threshold is determined through the classification index.

2. The method for identifying MCI by rules based on typical features according to claim 1, characterized in that: The method further comprises: Characteristic data were obtained using one or more of the following methods: neuropsychological scales, patient daily behavior diaries, simple digital tests, and functional near-infrared spectroscopy.

3. The method for identifying MCI by rules based on typical features according to claim 2, characterized in that: Construct a feature threshold table, including: Extract typical features from the acquired feature data, including ACC accuracy, RT reaction time, frontopolar beta, frontopolar integral, dorsolateral beta, and dorsolateral integral; Based on the extracted typical features, candidate thresholds are divided and candidate threshold intervals are marked, and a feature threshold table is constructed for each typical feature.

4. The method for identifying MCI by rules based on typical features according to claim 3, characterized in that: Constructing a feature threshold table also includes: For each typical feature, multiple candidate thresholds are set, and candidate threshold intervals are formed between adjacent candidate thresholds. Each candidate threshold interval corresponds to a marking symbol, thereby forming a feature judgment system including six feature threshold tables.

5. The method for identifying MCI by rules based on typical features according to claim 1, characterized in that: Based on the constructed feature threshold table, the labeling results of different dimensional features in their respective candidate threshold intervals are combined, including: According to the set rules and order, each feature labeling result is gradually merged two by two or multiple times to generate a series of different weight combinations.

6. The method for identifying MCI by rules based on typical features according to claim 1, characterized in that: With weight combinations as nodes, a decision tree structure is constructed, and feature combinations and node additions are performed step by step until an ordered weight combination of leaf nodes is obtained, including: In the process of combining features, each newly generated weight combination is treated as a child node of the parent node and arranged according to the order and logical relationship of the combination; Continue to combine features and add nodes until no new combinations can be made. At this time, the nodes at the end of the decision tree are leaf nodes, and the weight combinations represented by these leaf nodes are the final ordered weight combinations.

7. The method for identifying MCI by rules based on typical features according to claim 1, characterized in that: Based on the obtained ordered weight combination of leaf nodes, the classification threshold is calculated, and the optimal classification threshold is determined through classification indicators, including: Collect patient data with known MCI status as training samples, and use the ordered weight combination corresponding to each sample as input data; Using maximum likelihood estimation or cross-validation methods, calculations and optimizations are performed on these ordered weight combination data based on threshold sets to calculate classification indicators at different thresholds; The threshold that makes the classification index reach the optimal value is selected as the final optimal classification threshold.

8. The method for identifying MCI by rules based on typical features according to claim 1, characterized in that: The method further comprises: When the patient is determined to be suspected MCI based on the comparison result of the calculated weight combination and the predetermined optimal classification threshold, an MCI identification report is generated.

9. A regularized MCI recognition system based on a combination of typical features, characterized by: include: The first main module is used to obtain detection data to be identified, and perform typical feature extraction and weight combination calculation based on the detection data; The second main module is configured to compare the calculated weight combination with a predetermined optimal classification threshold, and identify whether the patient is MCI based on the comparison result; wherein the predetermined optimal classification threshold includes: Based on the collected multi-source data samples, feature threshold tables are constructed respectively. Each feature threshold table contains multiple candidate thresholds. The candidate thresholds are used to divide and mark the feature values extracted from typical features according to different candidate threshold intervals; Based on the constructed feature threshold table, the labeling results of different dimensional features in their respective candidate threshold intervals are combined to obtain a series of different weight combinations; Using weight combinations as nodes, a decision tree structure is constructed, and feature combinations and node additions are performed step by step until an ordered weight combination of leaf nodes is obtained; Based on the obtained ordered weight combination of leaf nodes, the classification threshold is calculated, and the optimal classification threshold is determined through the classification index.

10. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the method for identifying MCI by rules based on typical features according to any one of claims 1 to 8.

Images