Intelligent assessment system and method for F-S-D symptom group of breast cancer chemotherapy patients

Abstract

This invention belongs to the field of rehabilitation medicine technology, specifically disclosing an intelligent triage assessment system and method for F-S-D symptom clusters in breast cancer chemotherapy patients. The system includes: a multi-dimensional data acquisition module configured to acquire real-time sensory intensity data of the patient for three dimensions: fatigue, sleep disturbance, and depression, and map them into standardized symptom intensity vectors; a dynamic threshold determination module configured to determine whether to enable a fixed threshold or a dynamic threshold, and calculate a personalized dynamic threshold when enabling a dynamic threshold; a core symptom determination module configured to determine the patient's current core symptoms using a recursive decision tree; and a specialized assessment scale loading module configured to retrieve specialized assessment scales for core symptoms from the cloud for the patient to complete, and generate a structured report with severity gradients based on the completion results. This invention eliminates the need for full-scale quantitative assessment, making the testing process more targeted and reducing the workload of the assessment.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation medicine technology, and in particular to an intelligent triage assessment system and method for FSD symptom clusters in breast cancer chemotherapy patients. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] The FSD symptom cluster in breast cancer chemotherapy patients specifically refers to three interconnected symptoms that patients often experience simultaneously during or after chemotherapy: fatigue, sleep disturbance, and depression. This symptom cluster is considered one of the most common and core symptom clusters in breast cancer patients. These symptoms are interrelated and mutually causal, significantly impacting patients' quality of life, treatment adherence, and prognosis.

[0004] In existing technologies, a comprehensive assessment model using the Cancer Fatigue Scale (CFS), the Pittsburgh Sleep Quality Index (PSQI), and the Health Questionnaire-9 (PHQ-9) is often adopted to evaluate FSD symptom clusters. This model is cumbersome and repetitive. For patients who are in the interval between chemotherapy and are physically weak, this comprehensive assessment model is not only too time-consuming, but also causes severe cognitive fatigue and resistance, resulting in data distortion or a high rate of omissions.

[0005] Meanwhile, the core symptoms within the FSD symptom cluster exhibit a non-linear dynamic evolution pattern with each chemotherapy cycle. Traditional management systems lack an effective "triage-lock" logic, making it difficult to capture the core symptoms that are most troubling to patients at the current stage in real time. This results in clinical interventions often being generic and lacking specificity and precision.

[0006] Furthermore, existing intelligent assessment systems mostly use static fixed thresholds as the basis for triggering warnings or further evaluations. However, due to the significant differences in individual patient tolerance baselines and the significant time-dependent nature of chemotherapy-induced symptom fluctuations, fixed thresholds often overlook clinical abnormal fluctuations that show a significant increase relative to the individual patient's baseline but whose absolute scores have not yet reached the fixed threshold, resulting in a lag in identification. Summary of the Invention

[0007] To address the aforementioned issues, this invention proposes an intelligent triage assessment system and method for FSD symptom clusters in breast cancer chemotherapy patients. By using the entropy weight method to objectively quantify the importance of symptom dimensions, it dynamically generates personalized thresholds that fit the individual characteristics of the patient and the chemotherapy cycle, and uses a recursive decision tree to determine the patient's current core symptoms, thereby simplifying the assessment process and improving the accuracy of core symptom determination.

[0008] In some implementations, the following technical solutions are adopted:

[0009] A smart triage assessment system for FSD symptom clusters in breast cancer chemotherapy patients includes:

[0010] The multidimensional data acquisition module is configured to acquire instantaneous sensory intensity data of patients in three dimensions—fatigue, sleep disorders, and depression—in real time through a mobile interactive interface, and map the data into a standardized symptom intensity vector.

[0011] The dynamic threshold determination module is configured to determine whether to enable a fixed threshold or a dynamic threshold based on the amount of currently valid assessment data, and when enabling a dynamic threshold, calculate the weights of the three dimensions and combine them with the corrected symptom intensity vector to calculate a personalized dynamic threshold.

[0012] The core symptom determination module is configured to determine the patient's current core symptoms using a recursive decision tree based on the standardized symptom intensity vector and the currently enabled dynamic or fixed threshold.

[0013] The specialized assessment scale loading module is configured to, in response to the core symptoms, retrieve the specialized assessment scale for the core symptoms from the cloud for the patient to complete, and generate a structured report with severity gradients based on the completion results.

[0014] As a further solution, the dynamic threshold determination module determines whether to enable a fixed threshold or a dynamic threshold based on the amount of currently valid evaluation data, specifically:

[0015] Automatically store the symptom intensity vector, chemotherapy timeline information, and corresponding core scores of specific scales for each patient assessment;

[0016] If the following conditions are met: the symptom intensity vector is complete, the specific scale is filled in completely, and the time interval between two assessments is greater than the set threshold, then the assessment data is considered valid.

[0017] If the number of valid assessment data for the patient is greater than the preset value n, then a dynamic threshold is enabled; otherwise, a fixed threshold is enabled; n is a z integer greater than 2.

[0018] As a further solution, the dynamic threshold determination module calculates the weights of three dimensions when dynamic thresholding is enabled, and combines this with the corrected symptom intensity vector to calculate a personalized dynamic threshold, specifically:

[0019] The historical scores of the three dimensions of fatigue, sleep disorder, and depression in the symptom intensity vector are standardized, and the information entropy of each dimension is calculated; the dynamic weight of each dimension is calculated based on the information entropy.

[0020] Using the patient's historical symptom intensity vector and chemotherapy time sequence information as input, a pre-trained symptom time sequence correction model is used to output the corrected symptom intensity vector.

[0021] Based on the dynamic weights and the corrected symptom intensity vector, a personalized dynamic threshold is calculated.

[0022] As a further approach, a personalized dynamic threshold is calculated based on the dynamic weights and the corrected symptom intensity vector, specifically as follows:

[0023] ;

[0024] ;

[0025] in, For personalized dynamic thresholds, This represents the maximum value of the dynamic weights in the three dimensions. This is an adjustment coefficient used to control the magnitude of the impact of dynamic weights on the threshold. As the baseline value, As a preset scoring benchmark, The corrected symptom intensity vector The mean.

[0026] As a further approach, a recursive decision tree is used to determine the patient's current core symptoms, specifically:

[0027] If one or more dimensions of the symptom intensity vector have a score component that is greater than or equal to the current enabled threshold, then the one or more dimensions will be used as the patient's current core symptom.

[0028] If the score components of all dimensions in the symptom intensity vector are less than the current activation threshold, then the dimension corresponding to the highest score component is selected as the patient's current core symptom.

[0029] If two or three dimensions have equal scores and are the highest-scoring items, an interactive branch is triggered, allowing the patient to select the current core symptom.

[0030] As a further solution, the special assessment scale loading module calls the special assessment scale corresponding to the core symptoms from the cloud and automatically filters out duplicate assessment items between multiple scales.

[0031] As a further solution, it also includes: an adaptive intervention strategy push and early warning module, which is configured to determine the corresponding damage coefficient based on the scores of each sub-dimensional of the special assessment scale and automatically generate intervention suggestions; when self-injury risk indicators are identified, an encrypted high-risk early warning package is immediately generated and pushed to the medical terminal to enable remote monitoring.

[0032] As a further solution, it also includes: an efficacy verification module, which is configured to compare the core symptom assessment results corresponding to different chemotherapy times with the results of assessing ordinary questionnaire data using network analysis. If the accuracy meets the predetermined requirements, the current assessment logic will continue to be used; otherwise, the parameters will be adjusted until the requirements are met.

[0033] In other embodiments, the following technical solutions are adopted:

[0034] A smart triage assessment method for FSD symptom clusters in breast cancer chemotherapy patients includes:

[0035] The system acquires real-time sensory intensity data of patients in three dimensions: fatigue, sleep disorders, and depression through a mobile interactive interface, and maps the data into a standardized symptom intensity vector.

[0036] Based on the amount of currently valid assessment data, determine whether to enable a fixed threshold or a dynamic threshold, and when enabling a dynamic threshold, calculate the weights of the three dimensions, and combine them with the corrected symptom intensity vector to calculate a personalized dynamic threshold.

[0037] Based on the standardized symptom intensity vector and the currently enabled dynamic or fixed threshold, a recursive decision tree is used to determine the patient's current core symptoms.

[0038] In response to the core symptoms, a specific assessment scale for the core symptoms is retrieved from the cloud for the patient to complete, and a structured report with severity gradients is generated based on the completion results.

[0039] In other embodiments, the following technical solutions are adopted:

[0040] A terminal device includes a processor and a memory, the processor being used to implement instructions; the memory being used to store multiple instructions adapted to be loaded and executed by the processor to perform the above-described intelligent triage assessment method for FSD symptom clusters in breast cancer chemotherapy patients.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] (1) The present invention uses a digital slider based on the logic of the Visual Analog Scale (VAS) to map the patient’s subjective feelings into a standardized symptom intensity vector; based on the current corresponding threshold, the current unique core symptom is locked through the decision tree logic, and then the patient is diverted to the specific scale corresponding to the core symptom according to the core symptom; no full quantitative assessment is required, the testing process is more targeted, the assessment workload can be reduced, and the assessment efficiency and accuracy can be improved.

[0043] (2) This invention uses a threshold to determine whether a patient's symptoms have reached a critical level requiring unconditional priority treatment. When the score of a certain symptom dimension is greater than or equal to the threshold, it indicates that the symptom is a core symptom. The system considers the symptom to be of high risk and must be treated first. When the amount of effective assessment data is small, this invention uses a fixed threshold as the system's starting benchmark and safety net. When the amount of effective assessment data reaches a certain value, a dynamic threshold can be adaptively determined. Through the dynamic threshold, the differences between patients are fully considered, and those cases that, although not reaching an absolutely high score, have deteriorated drastically relative to the patient's own baseline can be captured, thereby discovering latent risks that are trending towards deterioration earlier and avoiding clinical omissions caused by fixed thresholds.

[0044] (3) Based on the patient’s historical assessment data, the present invention uses the entropy weight method to determine the dynamic weight of each dimension to characterize its influence on the determination of core symptoms; the LSTM model is used to correct the symptom intensity vector, and the dynamic threshold is calculated based on the dynamic weight of each dimension and the corrected symptom intensity vector; the dynamic threshold can keenly capture those hidden risks that have low absolute scores but have deteriorated drastically relative to the patient’s own baseline, thereby effectively solving the problem of ignoring the huge differences in the individual patient’s tolerance baseline and the time dependence of symptom fluctuation caused by the traditional static fixed threshold “one-size-fits-all” approach, thus improving the accuracy of core symptom determination.

[0045] Other features and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0046] Figure 1 This is a schematic diagram illustrating the process of determining whether to enable a fixed threshold or a dynamic threshold in an embodiment of the present invention.

[0047] Figure 2 This is a schematic diagram illustrating the process of determining a patient's current core symptoms using a recursive decision tree in an embodiment of the present invention;

[0048] Figure 3 This is a flowchart of the intelligent triage assessment method for FSD symptom clusters in breast cancer chemotherapy patients in an embodiment of the present invention. Detailed Implementation

[0049] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0050] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0051] Example 1

[0052] In one or more embodiments, a smart triage assessment system for FSD symptom clusters in breast cancer chemotherapy patients is disclosed, specifically including:

[0053] (1) The multidimensional data acquisition module is configured to acquire the patient’s instantaneous sensory intensity data for three dimensions of fatigue (F), sleep disorder (S) and depression (D) in real time through the mobile terminal interactive interface, and map the data into a standardized symptom intensity vector.

[0054] In this embodiment, the multidimensional data acquisition module has a built-in digital slider based on the logic of the Visual Analog Scale (VAS). By sliding the slider, the patient's values ​​for the three dimensions of fatigue (F), sleep disorder (S), and depression (D) can be obtained. Table 1 shows examples of the three dimensions of the Visual Analog Scale.

[0055] Table 1

[0056] ① Fatigue: Have you felt physically / mentally exhausted in the past two weeks, and even rest can't help you recover? 0 = Not tired at all, 2.5 = Slightly tired, 5 = Moderately tired, 7.5 = Very tired, 10 = Extremely tired ② Sleep disorders: In the past month, have you experienced difficulty falling asleep, poor sleep, or waking up early? 0 = Excellent sleep, 2.5 = Occasionally poor sleep, 5 = Frequently poor sleep, 7.5 = Severely poor sleep, 10 = Completely unable to sleep. ③ Depression: Have you felt down, uninterested in doing anything, or lacking energy in the past two weeks? 0 = Very good mood, 2.5 = Occasionally down, 5 = Often down, 7.5 = Severely down, 10 = Completely unmotivated

[0057] By scoring each of the three dimensions, the patient's symptom intensity vector V is obtained. VAS =[S f ,S s ,S d ].

[0058] Among them, S f S s S d The standardized scores for the three dimensions of fatigue, sleep disorders, and depression selected by the patients are respectively.

[0059] (2) The dynamic threshold determination module is configured to determine whether to enable a fixed threshold or a dynamic threshold based on the number of currently valid assessment data, and calculate the weights of the three dimensions when the dynamic threshold is enabled, and calculate the personalized dynamic threshold in combination with the corrected symptom intensity vector.

[0060] In this embodiment, the dynamic threshold determination module communicates with the multidimensional data acquisition module, and the built-in storage unit automatically stores the symptom intensity vector, chemotherapy time sequence information, and corresponding core scores of the special scale for each patient assessment.

[0061] The dynamic threshold determination module mainly includes:

[0062] (2-1) Evaluation data validity determination submodule, used to determine the number of valid evaluation data.

[0063] In this embodiment, combined with Figure 1 The criteria for valid assessment data are defined as follows: the symptom intensity vector (VAS score) is complete, the specific scale is filled in completely, and the time interval between two assessments is greater than a set threshold (e.g., 1 week). The cumulative number of valid assessment data is denoted as n.

[0064] (2-2) Entropy weight method dimension weighting submodule: used to calculate the dynamic weights of the three dimensions of fatigue (F), sleep disorder (S) and depression (D);

[0065] When n≥3, the historical symptom intensity vector is first standardized.

[0066] Since the raw scores of different dimensions may differ, the historical VAS scores first need to be mapped to the [0,1] interval to eliminate the influence of dimensions; the standardization process is as follows:

[0067] ;

[0068] in, This represents the original VAS score for the j-th dimension in the i-th evaluation. The standardized score, , These represent the minimum and maximum values ​​of the original VAS score for the j-th dimension in the historical data, respectively.

[0069] Then, based on the standardized data, the entropy value of each dimension is calculated to characterize the "disorder" or "dispersion" of the data in each dimension. The specific process is as follows:

[0070] Calculate the weight of the i-th sample in the j-th dimension. : ;

[0071] Calculate the entropy value of the j-th dimension. : ;

[0072] The larger the value (closer to 1), the smaller the data variation in that dimension (less information). The smaller the value, the greater the data fluctuation in that dimension (the greater the amount of information provided).

[0073] Finally, the difference coefficient (1-) is calculated based on the entropy value. ), and normalize to obtain the final weights. :

[0074] ;

[0075] Where j=1,2,3: correspond to the three dimensions of fatigue (F), sleep disorder (S), and depression (D), respectively.

[0076] Entropy The smaller the coefficient of variation, the higher the coefficient of variation. The larger the value, the higher the weight. The larger.

[0077] Calculated weights The higher the value, the greater the fluctuation of that dimension in historical assessments or the more information it contains, thus the greater the system determines its influence on the core symptoms.

[0078] (2-3) LSTM temporal correction submodule, used to correct the symptom intensity vector (VAS score) using LSTM model.

[0079] Specifically, using the patient's VAS score sequence (symptom intensity vector) from the last three effective assessments and chemotherapy time-series characteristics as input, and based on a pre-trained symptom time-series correction model for breast cancer chemotherapy patients, the output is a corrected true symptom score vector V. true =[S f ′ S s ′ S d ′ This eliminates momentary reporting bias, that is, removes scoring errors caused by patients' momentary emotions or environmental factors.

[0080] In this embodiment, the symptom timing correction model includes an input layer, a processing layer, and an output layer.

[0081] The data input to the input layer includes:

[0082] Historical scoring sequence: The VAS score sequence of the patient's three most recent effective assessments (i.e., the F, S, D scores of the past three assessments).

[0083] Chemotherapy timing characteristics: specifically including "current chemotherapy cycle" and "time since last chemotherapy".

[0084] The processing layer uses LSTM units to process the aforementioned time-series data, capturing the nonlinear patterns and long-term dependencies of symptom changes with chemotherapy cycles, thereby identifying which fluctuations are regular and which are transient disturbances. The output layer outputs the corrected true symptom intensity vector.

[0085] This embodiment utilizes LSTM to correct the symptom intensity vector (VAS score), which can eliminate scoring errors caused by patients' momentary emotional fluctuations or interference from external environmental factors during the reporting process, i.e., instantaneous reporting bias. At the same time, the LSTM model is good at processing time series data and can effectively capture the nonlinear patterns and long-term dependencies of breast cancer patients' symptoms as they change with chemotherapy cycles. The model can accurately distinguish which changes in patients' symptoms are regular and which are merely momentary disturbances, improving data reliability. Through the above processing, the system can finally output the corrected true symptom intensity vector, thus providing a highly reliable data foundation for subsequent core symptom identification and threshold calculation.

[0086] (2-4) Dynamic threshold calculation submodule, used to calculate personalized dynamic thresholds based on dimensional weights and the corrected symptom intensity vector.

[0087] The specific process is as follows: First, a baseline value is calculated based on the mean of the corrected scores:

[0088] ;

[0089] in, , is the corrected symptom intensity vector The mean value represents the patient's overall symptom burden level at the current assessment time; The preset scoring benchmark, such as 8 points, is used as the benchmark (the usual clinical high-risk warning line). It can be objectively fine-tuned according to the patient's current overall symptom level, providing a basic warning line that fits the current physical condition.

[0090] Based on benchmark value Introduce weighting factors for personalized amplification:

[0091] ;

[0092] in, For personalized dynamic thresholds, This represents the maximum value of the dynamic weights in the three dimensions. This is an adjustment coefficient used to control the impact of dynamic weights on the threshold; for example, a value of 0.2.

[0093] This embodiment introduces a dynamic weight (taking the maximum value) calculated based on historical assessment data as a weighting factor for personalized amplification. The greater the weight of a certain symptom dimension (indicating that the symptom fluctuates more significantly or is more important in historical data), the higher the dynamic threshold will be, thereby avoiding frequent false alarms due to the normal fluctuation of the symptom.

[0094] In addition, this embodiment also sets boundary constraints for personalized dynamic thresholds, such as: This is to prevent the final calculated threshold from being too high, which could lead to missed diagnosis of high-risk symptoms, or from being too low, which could cause the system to be overly sensitive and increase the burden on doctors and patients.

[0095] In this embodiment, the system determines whether to enable a fixed threshold or a dynamic threshold based on the number of valid assessment data n. In the early stage of data accumulation (cold start period), when the amount of valid assessment data n accumulated by the patient is less than 3, the system defaults to enabling a fixed threshold (e.g., τ=8) because the data is insufficient to support the algorithm operation.

[0096] When n≥3, dynamic threshold calculation is automatically triggered, and the personalized dynamic threshold is applied. The consistency of the core symptom determination results is checked against the fixed threshold τ=8. If they are consistent, the dynamic threshold is switched directly. If they are inconsistent, the fixed threshold is retained once, and the dynamic threshold is switched again during the next evaluation. After switching, the fixed threshold is retained as a backup. If the accuracy of the dynamic threshold determination is lower than 80%, the fixed threshold is automatically switched back temporarily and the model optimization is triggered.

[0097] (3) The core symptom determination module is configured to determine the patient’s current core symptoms using a recursive decision tree based on a standardized symptom intensity vector and the currently enabled dynamic or fixed threshold.

[0098] In this embodiment, a recursive decision tree built based on Python is used to analyze the symptom intensity vector V. VAS Perform nonlinear path calculations to pinpoint the core symptoms of the current stage, combined with... Figure 2 The specific process is as follows:

[0099] First priority: Security-enforced blocking logic; the system first determines the currently enabled threshold type (fixed threshold or dynamic threshold), if the symptom intensity vector V VAS There exists a component S i If the value is greater than or equal to the current threshold, the system will skip extreme value comparison and forcibly lock the corresponding dimension of the component as the core symptom.

[0100] Second priority: Weight maximization mapping logic; if all components satisfy S i If the current threshold is enabled, the system calculates operator C. core =argmax(V VAS The item with the highest score was selected as the core symptom;

[0101] Third priority: Heuristic tie-breaking logic; if multiple dimensions have equal scores and all are the highest scores, the system triggers an interactive branch, revising the judgment by obtaining the user-defined weight of the most prominent subjective concern. For example, a pop-up window asks the patient, "These two scores are the same, which one do you think is the most important?" After the user selects, the system locks the selected score.

[0102] The first priority level triggers a high-risk assessment process when symptoms become severe enough to exceed a threshold, ensuring that no high-risk symptoms (such as severe depression) are missed. If none symptoms exceed the threshold, the system selects the most severe symptom based on the second priority level. If the system cannot determine the most severe symptom (tie-tied), a third priority level is used, allowing the user to intervene and select the most prominent subjective concern.

[0103] The system uses thresholds to determine whether a patient's symptoms have reached a critical level requiring unconditional priority treatment, thereby deciding whether to adopt the "forced lockout" mode or the "comparison mode".

[0104] When a score on a symptom dimension is greater than or equal to a threshold, it indicates that the symptom is a core symptom. The system considers this symptom to be of high risk and must be addressed first. The system will skip the comparison process with other symptoms (no longer comparing which score is the highest), directly and forcibly lock this dimension as a core symptom, and proceed to in-depth evaluation.

[0105] When the scores for all symptom dimensions are below the threshold, it indicates that the patient is not currently in an extremely critical condition, but it is still necessary to identify the most troubling issue for the patient. System behavior (weight maximization mapping logic): The system enters a "relative comparison" mode. The system compares the scores of the three triage questions and selects the one with the highest score as the dominant symptom for identification.

[0106] If the scores for several dimensions are the same and all are the highest, the system will pop up a window asking the patient which one they subjectively feel is the most serious (subjective weighting adjustment).

[0107] Through the above path calculation, usually only one core symptom is identified, and the patient only needs to complete one scale, reducing the assessment workload by 2 / 3.

[0108] (4) The special assessment scale loading module is configured to respond to the core symptoms, call the special assessment scale of the core symptoms from the cloud for the patient to fill in, and generate a structured report with severity gradient based on the filling results.

[0109] In this embodiment, in response to the identified core symptoms, the deep assessment tools coupled with them are asynchronously invoked from the cloud knowledge base. The assessment tools include: Cancer Fatigue Scale (CFS), Pittsburgh Sleep Index Scale (PSQI), and Chinese version of 9-item health questionnaire (PHQ-9).

[0110] This module features item-level logical blocking to identify and remove duplicate measurements across dimensions. It also filters irrelevant items based on context awareness to optimize the assessment load. For example, if a patient's three symptoms all reach the threshold, there are three core symptoms. The three scales need to be pushed to the system in descending order of score. However, filling out three scales is too burdensome, and there are too many duplicate items among them. Therefore, duplicate items will not be pushed again; they only need to be filled out once.

[0111] Meanwhile, based on the user-completed specialized assessment scale, the scale options were structured into three sub-dimensions: physical fatigue, emotional fatigue, and cognitive fatigue; the impairment weight P for each sub-dimension was... i Based on the injury coefficient R, a structured report with severity gradients (none, mild, moderate, severe) is generated according to the clinical cutoff value.

[0112] Among them, the damaged weight P of each sub-dimension i The calculation method is as follows:

[0113] ;

[0114] For example, even with the same symptom of fatigue, some patients experience different symptoms (P). 躯体 Very high (physical fatigue), some patients P 情感 Very high (mentally exhausted). The system, based on P... i The focus of the intervention should be either "physical rehabilitation" or "psychological counseling".

[0115] The damage factor R is calculated as follows:

[0116] ;

[0117] For the Cancer Fatigue Scale (CFS), sub-dimension scores refer to the scores of the three dimensions of physical, emotional, and cognitive fatigue; the total scale score is the sum of the scores of all sub-dimensions; the measured total score for all items under a sub-dimension refers to the sum of the scores actually selected for all items within a specific sub-dimension. Taking the physical fatigue sub-dimension as an example, assuming physical fatigue includes 4 items, each scored from 0 to 4; and the patient actually selected scores of 3, 2, 4, and 1 respectively; then, the measured total score for all items under this sub-dimension = 3 + 2 + 4 + 1 = 10. The theoretical maximum score for this sub-dimension refers to the highest possible score achievable when all items under the corresponding sub-dimension have the "most severe" option selected.

[0118] As a specific example, the Cancer Fatigue Scale (CFS) includes 15 items across three dimensions: physical fatigue (7 items), emotional fatigue (4 items), and cognitive fatigue (4 items). It uses a Likert 5-point scale (0-4), with a total score of 0-60. 0 points indicates no fatigue; a higher total score indicates more severe fatigue, as shown in Table 2.

[0119] Table 2

[0120]

[0121] In the CFS scale, if the score for the physical fatigue dimension is >18, the system determines that the damage coefficient R is >60%. This means that the degree of damage in this dimension has exceeded 60%, which is considered "severe damage".

[0122] As a further implementation method, the intelligent triage assessment system for FSD symptom clusters in breast cancer chemotherapy patients also includes:

[0123] (5) The adaptive intervention strategy push and early warning module is configured to determine the corresponding damage coefficient based on the score of each sub-dimensional in the special assessment scale and automatically generate intervention suggestions; when self-injury risk indicators are identified, an encrypted high-risk early warning package is immediately generated and pushed to the medical terminal to enable remote monitoring.

[0124] In this embodiment, the damaged weight P of each sub-dimension can be... i Based on the characteristics of damage coefficient distribution, it retrieves and matches corresponding empowerment programs from the non-pharmacological intervention knowledge base, and pushes personalized intervention suggestions, including energy-saving strategies, sleep restrictions, and psychological adjustment, to patients through a real-time feedback mechanism; it has a self-learning optimization function, which can dynamically adjust the weight and priority of intervention strategies based on the symptom relief data after the patient's intervention is implemented.

[0125] At the same time, extreme deviation signals during the assessment process are monitored around the clock. For example, when self-harm risk indicators such as question 9 of PHQ-9 are identified, the routine process is forcibly interrupted and a remote expert consultation and monitoring procedure is immediately initiated. High-risk status encrypted packages are synchronously pushed to the medical management terminal and expert platform through a dual-end asynchronous transmission protocol.

[0126] Furthermore, the matched interventions, combined with the patient's current treatment timeline and chemotherapy cycle characteristics, can be converted into a visual rehabilitation prescription that includes charts, timeline reminders, and side effect annotations. This prescription can then be encrypted and stored in the case management portal for patients to access and follow up on at any time.

[0127] (6) The efficacy verification module is configured to compare the core symptom assessment results corresponding to different chemotherapy times with the results of assessing ordinary questionnaire data using network analysis. If the accuracy meets the predetermined requirements, the current assessment logic will continue to be used; otherwise, the parameters will be adjusted until the requirements are met.

[0128] In this embodiment, different chemotherapy times can be: before chemotherapy for early breast cancer (T0), after the first chemotherapy (T1), after the second chemotherapy (T2), after the third chemotherapy (T3), after the fourth chemotherapy (T4), after the fifth chemotherapy (T5), after the sixth chemotherapy (T6), after the seventh chemotherapy (T7), and after the eighth chemotherapy (T8).

[0129] Simultaneously, the study subjects underwent symptom measurement using a standard questionnaire and a mini-program assessment method. Network analysis was employed, identifying symptoms with high tight centrality as core symptoms. These were then compared with the core symptoms assessed by the system in this embodiment. An accuracy rate of ≥80% was considered satisfactory. If the cumulative accuracy rate falls below 80%, the system triggers recursive parameter calibration logic, such as adjusting VAS triage thresholds, weighting factors, and dynamic thresholds to achieve closed-loop optimization.

[0130] This embodiment uses a dynamic threshold that can be adaptively adjusted based on the patient's symptom sensitivity, chemotherapy cycle characteristics, and historical assessment data, solving the problem of missed or incorrect diagnoses caused by a fixed threshold. The accuracy of core symptom determination is improved by 15%-20%.

[0131] Through a mechanism of "fixed threshold fallback and dynamic threshold iteration," the system ensures the normal operation of the assessment process during the initial stage of system launch, when new patient data is insufficient, or when dynamic thresholds are abnormal. Once the data meets the standards, it automatically upgrades to personalized assessment, balancing usability and accuracy. The required historical assessment data and chemotherapy timeline information are all obtained from the system's original assessment process, eliminating the need for additional data collection items and reducing the operational burden on patients and medical staff, thus improving assessment compliance. A model performance validation unit combining network analysis is used to achieve recursive calibration of dynamic threshold parameters, continuously optimizing assessment accuracy as clinical data accumulates and adapting to the symptom characteristics of different patient groups.

[0132] This system also features functions such as item filtering, visualized prescriptions, and high-risk alerts, which improve accuracy while maintaining the efficiency of assessment and the targeted nature of intervention, further reducing the burden on doctors and patients.

[0133] As a specific implementation method, a specific working example of the system in this embodiment is given below:

[0134] (1) System initialization and data accumulation phase (n<3):

[0135] After logging into the case management portal for the first time and entering basic health data (age, chemotherapy regimen, underlying diseases, etc.), the patient fills in the instantaneous scores of three dimensions: fatigue, sleep disorders, and depression through the mobile VAS collection module. The system then constructs an initial perception vector VVAS = [Sf, Ss, Sd].

[0136] Since the cumulative effective assessment data volume n < 3, the dynamic threshold determination module triggers the fixed threshold mode, and the system uses τ=8 for core symptom assessment: if V VAS There exists a component S i If the score is ≥8, the corresponding dimension will be locked as the core symptom, triggering the asynchronous loading of the corresponding special scale (CFS / PSQI / PHQ-9). When the scale is loaded, the system will automatically filter out redundant items that are not related to the current VAS score (e.g., if the sleep dimension score is 2, the "sleep quality satisfaction" item in the PSQI will be filtered out).

[0137] If all components S i <8, via C core =argmax(V VAS The item with the highest score is selected as the core symptom; if there are ties, an interactive interface will pop up to obtain the weight correction result of the patient's "most prominent subjective distress".

[0138] After the specialized assessment is completed, the multi-dimensional damage feature in-depth analysis module calculates the damage weight and damage coefficient of each sub-dimension, the adaptive intervention strategy coupled with the push engine matches the personalized intervention plan, and the visualization conversion module generates a rehabilitation prescription and pushes it to the patient's terminal; at the same time, the remote monitoring backend monitors the risk signals throughout the process, the model performance verification unit records the judgment results, and the system automatically stores the assessment data and chemotherapy time sequence information.

[0139] (2) Dynamic threshold switching stage (n=3):

[0140] When a patient completes a total of 3 valid assessments (meeting the requirements of no missing VAS scores, complete completion of specific scales, and an interval of ≥1 week between assessments), the system automatically triggers the dynamic threshold adjustment process:

[0141] Data retrieval: The dynamic threshold determination module retrieves three historical VAS score sequences, chemotherapy timeline information (such as chemotherapy cycle 2 → cycle 3 → cycle 4), and core scores of the special scale from the case management port.

[0142] Entropy weighting: Standardize the three historical VAS scores to calculate the dimension weights (example: w=[0.45,0.25,0.30], the fatigue dimension has the highest weight).

[0143] LSTM temporal correction: Load a pre-trained symptom temporal prediction model, input three VAS score sequences and chemotherapy cycle information, and output the corrected true score vector V. true =[7.2,4.5,6.8].

[0144] Perform dynamic threshold calculation: =(7.2+4.5+6.8) / 3=6.17;

[0145] =8+(6.17 5) / 10=8.12; max(w j ) = 0.45;

[0146] =8.12×(1+0.45×0.2)=8.85; Because it exceeds the interval [6.5,8.5], the final result is... =8.5.

[0147] Consistency check: using When the current VAS score vector V=[7.5,5.0,7.0] is judged using a fixed threshold τ=8 and a fixed threshold τ=8 respectively, fatigue is identified as the core symptom in both cases, and the judgment results are consistent. The system then automatically switches to the dynamic threshold.

[0148] (3) Dynamic threshold iterative optimization stage (n>3):

[0149] Each time a patient receives a new valid assessment, the system automatically updates the historical data, and the dynamic threshold determination module recalculates the dimension weights and dynamic thresholds (Example: After one new assessment, w=[0.48,0.22,0.30], V...). true =[7.5,4.3,6.5] =8.4).

[0150] The performance verification module periodically compares the core symptoms of dynamic threshold locking with the gold standard extracted by network analysis. If the cumulative accuracy is ≥85%, the current dynamic threshold parameters are continued to be used; if the accuracy is below 80%, the entropy weight standardization coefficient, the LSTM model time window, and the weight factors in the dynamic threshold calculation formula are automatically adjusted to achieve closed-loop optimization of the model.

[0151] When a patient scores >0 on question 9 of PHQ-9, the remote monitoring and risk dynamic early warning system immediately generates a high-risk encrypted package and pushes it to the medical management terminal and the expert remote consultation platform, forcibly starting the remote monitoring program without interfering with the patient's current operation interface.

[0152] (4) Intervention implementation and feedback stage:

[0153] After receiving a visualized rehabilitation prescription, patients implement interventions chronologically (e.g., when physical fatigue is severe, a combination of physical energy conservation and aerobic exercise is implemented), and feedback on implementation status and symptom changes is provided through the terminal. The adaptive intervention strategy, coupled with a push engine, collects feedback data and dynamically adjusts the weight and priority of intervention strategies (e.g., if symptoms are significantly relieved after aerobic exercise, the matching priority of this type of program is increased).

[0154] Example 2

[0155] In one or more embodiments, a smart triage assessment method for FSD symptom clusters in breast cancer chemotherapy patients is disclosed, combining... Figure 3 Specifically, it includes the following process:

[0156] S101: Real-time sensory intensity data of patients in three dimensions—fatigue, sleep disorders, and depression—are obtained through a mobile interactive interface, and the data is mapped into a standardized symptom intensity vector.

[0157] S102: Based on the amount of currently valid assessment data, determine whether to enable a fixed threshold or a dynamic threshold, and when enabling a dynamic threshold, calculate the weights of the three dimensions, and combine them with the corrected symptom intensity vector to calculate a personalized dynamic threshold.

[0158] S103: Based on the standardized symptom intensity vector and the currently enabled dynamic or fixed threshold, use a recursive decision tree to determine the patient's current core symptoms.

[0159] S104: In response to the core symptoms, retrieve the specific assessment scale for the core symptoms from the cloud for the patient to complete, and generate a structured report with severity gradients based on the completion results.

[0160] It should be noted that the specific implementation methods of the above steps are the same as those in Embodiment 1, and will not be described in detail again.

[0161] Example 3

[0162] In one or more embodiments, a terminal device is disclosed, comprising a processor and a memory, wherein the processor is used to implement instructions; and the memory is used to store multiple instructions adapted to be loaded by the processor and executed by the intelligent triage assessment method for FSD symptom clusters of breast cancer chemotherapy patients as described in Embodiment 2.

[0163] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A smart triage assessment system for FSD symptom clusters in breast cancer chemotherapy patients, characterized in that, include: The multidimensional data acquisition module is configured to acquire instantaneous sensory intensity data of patients in three dimensions—fatigue, sleep disorders, and depression—in real time through a mobile interactive interface, and map the data into a standardized symptom intensity vector. Specifically, a digital slider based on the logic of a visual analog scale maps the patient's subjective feelings into a standardized symptom intensity vector. The dynamic threshold determination module is configured to determine whether to enable a fixed threshold or a dynamic threshold based on the amount of currently valid assessment data. When enabling a dynamic threshold, it calculates the weights of the three dimensions and combines them with the corrected symptom intensity vector to calculate a personalized dynamic threshold. Specifically: The historical scores of the three dimensions of fatigue, sleep disorder, and depression in the symptom intensity vector are standardized, and the information entropy of each dimension is calculated; the dynamic weight of each dimension is calculated based on the information entropy. Using the patient's historical symptom intensity vector and chemotherapy time sequence information as input, a pre-trained LSTM model is used to output a corrected symptom intensity vector. Based on the dynamic weights and the corrected symptom intensity vector, a personalized dynamic threshold is calculated, specifically as follows: ； ； in, For personalized dynamic thresholds, This represents the maximum value of the dynamic weights in the three dimensions. This is an adjustment coefficient used to control the magnitude of the impact of dynamic weights on the threshold. As the baseline value, As a preset scoring benchmark, The corrected symptom intensity vector The mean; The core symptom determination module is configured to determine the patient's current core symptoms using a recursive decision tree based on the standardized symptom intensity vector and the currently enabled dynamic or fixed threshold. The specialized assessment scale loading module is configured to, in response to the core symptoms, retrieve the specialized assessment scale for the core symptoms from the cloud for the patient to complete, and generate a structured report with severity gradients based on the completion results.

2. The intelligent triage assessment system for FSD symptom clusters in breast cancer chemotherapy patients as described in claim 1, characterized in that, The dynamic threshold determination module determines whether to use a fixed threshold or a dynamic threshold based on the amount of currently valid evaluation data. Specifically: Automatically store the symptom intensity vector, chemotherapy timeline information, and corresponding core scores of specific scales for each patient assessment; If the following conditions are met: the symptom intensity vector is complete, the specific scale is filled in completely, and the time interval between two assessments is greater than the set threshold, then the assessment data is considered valid. If the number of valid assessment data for the patient is greater than the preset value n, then a dynamic threshold is enabled; otherwise, a fixed threshold is enabled; n is an integer greater than 2.

3. The intelligent triage assessment system for FSD symptom clusters in breast cancer chemotherapy patients as described in claim 1, characterized in that, The recursive decision tree is used to determine the patient's current core symptoms, specifically: If one or more dimensions of the symptom intensity vector have a score component that is greater than or equal to the current enabled threshold, then the one or more dimensions will be used as the patient's current core symptom. If the score components of all dimensions in the symptom intensity vector are less than the current activation threshold, then the dimension corresponding to the highest score component is selected as the patient's current core symptom. If two or three dimensions have equal scores and are the highest-scoring items, an interactive branch is triggered, allowing the patient to select the current core symptom.

4. The intelligent triage assessment system for FSD symptom clusters in breast cancer chemotherapy patients as described in claim 1, characterized in that, The specific assessment scale loading module calls the specific assessment scale corresponding to the core symptoms from the cloud and automatically filters out duplicate assessment items between multiple scales.

5. The intelligent triage assessment system for FSD symptom clusters in breast cancer chemotherapy patients as described in claim 1, characterized in that, Also includes: The adaptive intervention strategy push and early warning module is configured to determine the corresponding damage coefficient based on the scores of each sub-dimensional of the special assessment scale and automatically generate intervention suggestions. When signs of self-harm risk are detected, an encrypted high-risk warning package is immediately generated and pushed to the medical terminal to enable remote monitoring.

6. The intelligent triage assessment system for FSD symptom clusters in breast cancer chemotherapy patients as described in claim 1, characterized in that, Also includes: The efficacy verification module is configured to compare the core symptom assessment results corresponding to different chemotherapy times with the results of assessing ordinary questionnaire data using network analysis. If the accuracy meets the predetermined requirements, the current assessment logic will continue to be used; otherwise, the parameters will be adjusted until the requirements are met.

7. A method for intelligent triage assessment of FSD symptom clusters in breast cancer chemotherapy patients as described in any one of claims 1-6, characterized in that, include: The system acquires real-time sensory intensity data of patients in three dimensions: fatigue, sleep disorders, and depression through a mobile interactive interface, and maps the data into a standardized symptom intensity vector. Based on the amount of currently valid assessment data, determine whether to enable a fixed threshold or a dynamic threshold, and when enabling a dynamic threshold, calculate the weights of the three dimensions, and combine them with the corrected symptom intensity vector to calculate a personalized dynamic threshold. Based on the standardized symptom intensity vector and the currently enabled dynamic or fixed threshold, a recursive decision tree is used to determine the patient's current core symptoms. In response to the core symptoms, a specific assessment scale for the core symptoms is retrieved from the cloud for the patient to complete, and a structured report with severity gradients is generated based on the completion results.

8. A terminal device comprising a processor and a memory, the processor for implementing instructions; the memory for storing multiple instructions, characterized in that, The instructions are adapted to be loaded by a processor and executed by the method of intelligent triage assessment system for FSD symptom clusters in breast cancer chemotherapy patients as described in claim 7.

Images