Method, device, equipment, medium and program product for fatigue recovery management of a blood donor

Abstract

This application discloses a method, device, medium, and program product for managing plasma donor fatigue recovery, addressing the problem in existing technologies where individual physiological responses of different plasma donors during plasma donation are not considered, leading to fatigue recovery predictions that deviate from individual realities and a lack of targeted intervention timing and content. The method includes: acquiring stressor time-series data and individual digital profiles of plasma donors; characterizing the donor's sensitivity to anticoagulant metabolism and fatigue recovery ability based on the stressor time-series data and individual digital profiles; predicting the plasma donor's fatigue recovery trajectory after plasma donation based on the stressor time-series data and individual digital profiles; determining at least one intervention time point based on the fatigue recovery trajectory, and pushing fatigue recovery guidance information to the plasma donor at the intervention time point.

Description

Technical Field

[0001] This application relates to the field of intelligent medical technology, and in particular to a method, device, equipment, medium, and program product for managing fatigue recovery of plasma donors. Background Technology

[0002] In the field of plasma donation health management, existing technologies typically rely on static models or fixed rules based on population average data to predict the fatigue recovery process of plasma donors, and provide uniform recovery guidance at preset fixed time points based on these predictions. For example, general hydration or rest recommendations are sent out at 2, 4, and 6 hours after plasma collection.

[0003] However, since the existing technologies do not take into account the individualized physiological response differences of different plasma donors to the plasma donation process, the predicted results of the fatigue recovery process may deviate significantly from the individual's actual situation, and the timing and content of interventions may lack specificity, which seriously affects the effectiveness of health management and the plasma donor's experience. Summary of the Invention

[0004] This application provides a method for managing fatigue recovery of plasma donors, which addresses the problem in the prior art that the fatigue recovery prediction results deviate from the individual reality and the timing and content of interventions lack specificity because the individual physiological response differences of different plasma donors during the plasma donation process are not considered.

[0005] This application also provides a plasma donor fatigue recovery management device, an electronic device, a computer-readable storage medium, and a computer program product.

[0006] The embodiments of this application adopt the following technical solutions: In a first aspect, embodiments of this application provide a method for managing fatigue recovery in plasma donors, including: Acquire time-series data on stressors and individual digital profiles of plasma donors; individual digital profiles characterize plasma donors' sensitivity to anticoagulant metabolism and fatigue recovery ability; Based on stressor time series data and individual digital profiles, predict the fatigue recovery trajectory of plasma donors after plasma donation; At least one intervention time point is determined based on the fatigue recovery trajectory, and fatigue recovery guidance information is sent to plasma donors at the intervention time point.

[0007] Optionally, obtain individual digital profiles of plasma donors, including: Acquire historical plasma donation data from plasma donors, including time-series data of historical stressors and corresponding historical recovery response data from at least one historical plasma donation process; Key stress features characterizing stress intensity were extracted from historical stress source time-series data. These key stress features included cumulative anticoagulant usage and standardized anticoagulant load. Based on historical recovery response data, key fatigue recovery features that characterize the change of fatigue degree over time are extracted. These key fatigue recovery features include fatigue half-life and fatigue decay rate within a preset recovery time window. Key stress features and key fatigue recovery features are input into the trained physiological parameter prediction model to obtain sensitivity and fatigue recovery ability; Individual digital profiles of plasma donors are constructed based on their sensitivity and fatigue recovery capabilities.

[0008] Optionally, key stress features characterizing stress intensity can be extracted from historical stressor time-series data, including: Based on historical stress source time series data, the start and end times of historical plasma collection were determined, and the collection duration was determined based on the start and end times of plasma collection. The cumulative amount of anticoagulant used within the time range from the start to the end of plasma collection was determined based on historical stress source time series data. The anticoagulant pumping rate time series was extracted from historical stress source time series data, and the pumping rate fluctuation characteristics were calculated based on the anticoagulant pumping rate time series. The pumping rate fluctuation characteristics include the sample standard deviation of the anticoagulant pumping rate time series. The standardized anticoagulant load is determined based on the cumulative amount of anticoagulant used, the duration of slurry collection, and the fluctuation characteristics of the pumping rate.

[0009] Optionally, key fatigue recovery features characterizing the change of fatigue degree over time can be extracted from historical recovery response data, including: Determine the recovery response sequence of fatigue level over time based on historical recovery response data; Historical fatigue recovery curves are constructed based on the recovery response sequence, and the initial fatigue level at the end of slurry harvesting is determined based on the historical fatigue recovery curves. The fatigue half-life is determined based on the historical fatigue recovery curve, where the fatigue half-life is the time length corresponding to when the historical fatigue recovery curve drops to half of the initial fatigue level; Based on the decrease in the historical fatigue recovery curve within the preset recovery time window and the duration of the preset recovery time window, the fatigue decay rate within the preset recovery time window is determined. The fatigue decay rate is used to characterize how quickly the fatigue level decreases over time within the preset recovery time window.

[0010] Optionally, before inputting key stress features and key fatigue recovery features into the trained physiological parameter prediction model to obtain the sensitivity coefficient and recovery rate coefficient, the method further includes: Multiple plasma donation data samples were acquired, including stressor time-series data samples and corresponding recovery response data samples. Determine the actual sensitivity and actual fatigue recovery ability corresponding to each plasma donor sample, and use them as label data; Key stress feature samples were extracted based on stressor time series data samples, and key fatigue recovery feature samples were extracted based on recovery response data samples. The key stress feature samples and key fatigue recovery feature samples are combined into an input feature vector; Using the input feature vector as the model input and the label data as the model output, the initial physiological parameter prediction model is trained to obtain the trained physiological parameter prediction model.

[0011] Optionally, the initial physiological parameter prediction model is trained using the input feature vector as the model input and the label data as the model output, resulting in a trained physiological parameter prediction model, including: The prediction sensitivity and fatigue recovery ability of plasma donation data samples were predicted using an initial physiological parameter prediction model. The prediction error of the initial physiological parameter prediction model is calculated based on the prediction sensitivity, predicted fatigue recovery ability, actual sensitivity, and actual fatigue recovery ability. The model parameters of the initial physiological parameter prediction model are iteratively updated based on the prediction error until the prediction error of the updated initial physiological parameter prediction model reaches the preset convergence condition, at which point the training is terminated and the trained physiological parameter prediction model is obtained.

[0012] Optionally, at least one intervention time point is determined based on the fatigue recovery trajectory, including: Based on fatigue recovery trajectory identification, at least one trajectory feature point is used as the intervention time point; The trajectory feature points include at least one of the following: The intervention start time point is the time point corresponding to the first preset proportion of the initial fatigue level when the fatigue level in the fatigue recovery trajectory decreases to the initial fatigue level at the end of slurry collection; The inflection point intervention time point is the dividing point between the first recovery stage and the second recovery stage in the fatigue recovery trajectory. The fatigue decay rate in the first recovery stage is greater than the fatigue decay rate in the second recovery stage. Consolidation intervention time point: The consolidation intervention time point represents the time point in the fatigue recovery trajectory that meets the baseline stability condition; The baseline stability conditions include: the fatigue level remains within a preset tolerance range centered on the initial fatigue level, and in a sequence consisting of a preset number of consecutive sampling points, the absolute value of the change in fatigue level between adjacent sampling points does not exceed a preset fluctuation threshold.

[0013] Optionally, fatigue recovery guidance information can be sent to plasma donors at the intervention time point, including: The guidance stage label is determined based on the predicted fatigue level corresponding to the intervention time point in the fatigue recovery trajectory. The guidance stage label is used to identify the current fatigue recovery stage. Based on the guidance stage label, a basic guidance template is selected from the preset guidance information database. The basic guidance template includes at least one suggestion for a fatigue recovery activity. Based on the predicted fatigue level, the parameters of the basic guidance template are adjusted to generate fatigue recovery guidance information corresponding to the fatigue recovery trajectory; By linking fatigue recovery guidance information to intervention time points, fatigue recovery guidance information can be pushed to plasma donors at the intervention time points.

[0014] Secondly, embodiments of this application provide a plasma donor fatigue recovery management device, including an acquisition module, a prediction module, and a processing module, wherein: The acquisition module is used to acquire time-series data on stressors and individual digital profiles of plasma donors; the individual digital profiles characterize the plasma donors' sensitivity to anticoagulant metabolism and fatigue recovery ability. The prediction module is used to predict the fatigue recovery trajectory of plasma donors after plasma donation based on stressor time series data and individual digital profiles. The processing module is used to determine at least one intervention time point based on the fatigue recovery trajectory and push fatigue recovery guidance information to plasma donors at the intervention time point.

[0015] Thirdly, embodiments of this application provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the plasma donor fatigue recovery management method as described above.

[0016] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the plasma donor fatigue recovery management method described above.

[0017] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the plasma donor fatigue recovery management method as described above.

[0018] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: The method provided in this application predicts the fatigue recovery trajectory after plasma donation by acquiring stressor time-series data of plasma donors and combining it with individual digital profiles characterizing their sensitivity to anticoagulant metabolism and fatigue recovery ability. This allows for the incorporation of individualized physiological response differences among different plasma donors into the modeling and prediction, making the predicted fatigue recovery process closer to the actual situation of the plasma donors. Based on this, intervention time points matching the plasma donor's state can be determined from the fatigue recovery trajectory, and targeted recovery guidance information can be pushed at the corresponding time points. This avoids the prediction bias and inaccurate timing / content of interventions caused by existing technologies based on group averages and fixed rules, thereby improving the effectiveness of fatigue recovery management and the plasma donor experience. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A schematic diagram illustrating the implementation process of a fatigue recovery management method for plasma donors provided in this application embodiment; Figure 2 A schematic diagram illustrating the implementation process of obtaining an individual digital profile of a plasma donor, provided in an embodiment of this application; Figure 3 A schematic diagram illustrating the implementation process of a physiological parameter prediction model trained according to an embodiment of this application; Figure 4 A schematic diagram illustrating an application process of the method provided in the embodiments of this application in practice; Figure 5 A schematic diagram of an electronic digital rating scale provided in an embodiment of this application; Figure 6 A schematic diagram of an interface for a personalized guidance push scheme provided in an embodiment of this application; Figure 7 This application provides a schematic diagram of the specific structure of a plasma donor fatigue recovery management device according to an embodiment of the present application; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] It should be understood that the training and prediction processes of the AI ​​models involved in the various embodiments of this specification all adhere to multiple legal and compliant principles, including legal data sources, compliant data content, compliant data governance, compliant training objectives and schemes, compliant training processes, compliant training environments and tools, and compliant ethical verification of training results, and comply with the requirements of Article 5 of the Patent Law. Among them: Data source legitimacy: All datasets used for AI model training were obtained through legal means, covering three categories: publicly authorized data, data authorized by partners, and self-collected compliant data. Publicly authorized data comes from compliant data sources following open-source licenses such as Apache 2.0, with complete copyright attribution and authorization scope clearly marked, and no unauthorized open-source code or data reuse. Data authorized by partners has been subject to formal data usage agreements, clearly defining the scope, duration, and confidentiality obligations, and possessing a complete authorization chain. For self-collected data involving personal information, strict informed consent procedures have been followed, and anonymization processes (including but not limited to field masking, feature anonymization, and differential privacy technology applications) have been implemented to remove personally identifiable information, fully complying with the requirements of relevant laws and regulations such as the "Interim Measures for the Administration of Generative Artificial Intelligence Services" and the "Personal Information Protection Law."

[0022] Data content compliance: The AI ​​model's dataset undergoes multiple screenings and cleaning processes to remove all content that may violate social morality or harm public interests. It contains no information that violates laws, regulations, or public interests, nor does it involve the illegal acquisition or use of genetic resources. For data in sensitive fields (such as healthcare and finance), an additional privacy-preserving computation module (including federated learning and secure multi-party computation technologies) ensures that the data is "usable but not visible," avoiding compliance risks during the original data transmission process and ensuring that the data application scenarios and uses comply with public order, good morals, and industry regulatory requirements.

[0023] Data governance norms: A complete data traceability system is established during the AI ​​model training process to automatically record the source, collection time, annotation process, cleaning rules, and permission allocation of training data, generating traceable compliance reports to ensure that the data is verifiable throughout its entire lifecycle. The dataset annotation process for AI models is completed by a professional human R&D team, clearly defining the proportion of human creative contributions and avoiding reliance on AI-generated data that has not undergone substantial human modification, thus meeting the examination requirements for "human main contributions" in AI patent applications.

[0024] Training objectives and plans are compliant: The AI ​​model training aims to provide individualized predictions of plasma donor fatigue recovery trajectories and optimize the timing and content of recovery interventions, thereby improving the scientific rigor and effectiveness of post-donation health management. The training protocol is strictly limited to legitimate applications in plasma donation health management. All data sources are legal and compliant, adhering to the principle of minimum necessity. During collection, storage, processing, and use, security measures such as informed consent, de-identification / anonymization, access control, encrypted transmission and storage, and audit trails are implemented to avoid undue infringement on donor privacy. The model output is solely for providing general fatigue recovery guidance and will not be used for illegal or criminal activities, infringing on the privacy of others, disrupting public safety, or other purposes that violate social morality. Furthermore, content constraints and risk control strategies prevent misuse. The training protocol and final output do not violate any mandatory provisions of laws or administrative regulations, do not harm public interests or the legitimate rights of others, and pose no potential risk of being used for illegal or criminal activities, infringing on privacy, or disrupting public safety. The protocol strictly adheres to the ethical principle of "intelligent for good."

[0025] Training process compliance: A closed-loop training framework is adopted to ensure compliance and controllability of the training process. The specific process is as follows: First, training samples are obtained through compliant data sources. After the aforementioned data cleaning and desensitization, they are input into the neural network model to generate preliminary training results. Second, an expert system is introduced to verify the preliminary results. Based on preset rules and human expert experience, the feasibility of the results is evaluated, and outputs that may pose ethical risks or compliance hazards are corrected (such as removing decision-making logic that violates public order and good morals, and adjusting model parameters that do not comply with safety regulations). Finally, the loss function weights are dynamically optimized based on expert system feedback to strengthen the model's learning of compliant results, avoid overfitting errors or non-compliant labels, and form a closed-loop control of "data input - model training - expert verification - parameter optimization - result feedback" to ensure that the entire training process complies with A5 ethical review requirements.

[0026] Training environment and tool compliance: AI model training is implemented using nationally licensed chips and a compliant training platform. All open-source frameworks and components used in the training process have obtained their corresponding licenses, and copyright statements and patent citation information are fully retained, with no instances of infringement or reuse. The training environment is built using virtual devices (containers / virtual machines) with fixed random seeds and initial parameter configurations to ensure the reproducibility of the training process. Furthermore, through access control and operation log recording, risks such as data leakage and parameter tampering during training are prevented, ensuring the security and compliance of the training process.

[0027] Training results ethical verification compliance: After the model is trained, it undergoes additional third-party ethical compliance assessment and algorithm filing review to verify that the model output does not violate social morality or harm public interests. For potentially sensitive scenarios (such as public services and intelligent decision-making), a special result verification mechanism is established to ensure that the model always complies with Article 5 of the Patent Law and relevant laws and regulations in practical applications.

[0028] In summary, the data and training process used in the AI ​​model of this specification strictly comply with the relevant provisions of Article 5 of the Patent Law and the Patent Examination Guidelines (2023 Edition), and there are no violations of laws, social ethics, public interests, or illegal use of genetic resources. It fully meets the compliance requirements for patent authorization.

[0029] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0030] To address the problem in existing technologies that fail to consider individual physiological differences among plasma donors during the donation process, resulting in fatigue recovery predictions deviating from individual realities and intervention timing and content lacking specificity, this application provides a method for managing plasma donor fatigue recovery.

[0031] The execution subject of this method can be various types of computing devices, or it can be an application or app installed on the computing device. The computing device can be a user terminal such as a mobile phone, tablet computer, or smart wearable device, or it can be a server.

[0032] For ease of description, this application uses a server as the execution subject of the method in its embodiments to illustrate the method. Those skilled in the art will understand that this embodiment uses a server as an example to describe the method, which is merely an illustrative example and does not limit the scope of protection of the corresponding claims.

[0033] Specifically, the implementation flow of the method provided in this application embodiment is as follows: Figure 1 As shown, it includes the following steps: Step 102: Obtain time-series data of stressors and individual digital profiles of plasma donors.

[0034] Among them, individual digital profiles are used to characterize plasma donors' sensitivity to coagulant metabolism and their ability to recover from fatigue.

[0035] Stressor time-series data refers to a time-stamped sequence of process data collected or obtained during plasma collection, used to characterize the external physiological load input to the donor's body during the plasma donation process. This stressor time-series data can originate from plasma collection equipment, controllers / servers communicating with the equipment, external sensors, information system records, or manually entered data, including but not limited to anticoagulant cumulative volume time-series data and cumulative plasma volume time-series data.

[0036] In some embodiments, the data acquisition unit can continuously collect and record key process quantities during plasma collection to obtain time-series data on plasma donor stressors. Specifically, the data acquisition unit can establish a communication connection with the plasma collection equipment to obtain, through subscription or query, at least a sequence showing the cumulative amount of anticoagulant used over time and a sequence showing the cumulative amount of plasma collected over time, and attach a uniform timestamp to each record to form a structured record arranged in chronological order.

[0037] Furthermore, considering that plasma collection data may contain unclear start and end points, data loss, or abnormal fluctuations, some optional implementations can identify key events based on stressor time-series data to determine process nodes such as the start and end of plasma collection, and generate stress input records corresponding to a single plasma collection process. Simultaneously, preprocessing such as data integrity verification, outlier labeling, and missing data compensation can be performed on the collected time-series data, and the processed stressor time-series data can be stored on a server or in a personal health record for later retrieval. This improves the availability and consistency of the collected data, thereby enhancing the reliability and stability of the overall modeling and prediction.

[0038] like Figure 2 As shown in some application embodiments, the individual digital profiles of plasma donors can be obtained using the following steps 202 to 210: Step 202: Obtain the historical plasma donation data of the plasma donor. The historical plasma donation data includes the time series data of historical stressors and the corresponding historical recovery response data during at least one historical plasma donation process.

[0039] Historical plasma donation data refers to a set of historical records formed by the same plasma donor during historical plasma donation activities, which can be used to characterize the relationship between stress input and recovery output.

[0040] Historical stressor time series data refers to time-stamped process data sequences that correspond to a specific historical plasma donation process and can reflect the external physiological load input during the plasma collection process. Examples include changes in anticoagulant usage over time, changes in collected plasma volume over time, and their derived process characteristics.

[0041] Historical recovery response data refers to post-donation recovery status data associated with a specific historical plasma donation process within the same plasma donation event range. It characterizes the fatigue changes and recovery process of plasma donors after donation. This data can at least include fatigue scores submitted by plasma donors at different time points after donation and their corresponding collection timestamps.

[0042] In some optional implementations, historical recovery response data may further include physiological signal data collected by wearable devices or terminals, as well as observational data reflecting sleep and activity levels, to more comprehensively characterize the recovery state. Thus, for the same plasma donation event, the system can obtain a set of historical stress input records corresponding to that event (i.e., historical stressor time-series data) and a set of historical recovery output records matching it in the time and event dimensions (i.e., historical recovery response data). These two sets of data, presented as paired samples, together constitute historical plasma donation data.

[0043] In some embodiments, when acquiring historical plasma donation data of plasma donors, the data management unit can perform the following steps after receiving a request to build or update an individual digital profile: First, based on the plasma donor's unique identifier, query their historical plasma donation event list in their personal health record, plasma station business system, or cloud data warehouse; then, using each historical plasma donation event as the primary key, read the process data collected or recorded during plasma collection for that event, as the historical stress source time-series data for that event; and further, search for fatigue feedback records reported by mobile devices or other terminals after plasma donation within the same event range, as the historical recovery response data corresponding to that event. To ensure a one-to-one correspondence between historical stress source time-series data and historical recovery response data, the system can use event identifier association, plasma collection end time and feedback time window association, or multi-factor matching, such as plasma donation date, plasma station number, equipment record number, etc., for binding; when there are duplicate records from multiple sources, the optimal record can be selected according to rules such as data integrity, time continuity, and source credibility level, or the records from multiple sources can be merged to obtain a unified historical plasma donation data sample.

[0044] In some optional implementations, considering that historical plasma donation data may contain missing data, terminal underreporting, inconsistent timestamps, or outliers, quality control and structured processing can be performed on the acquired historical plasma donation data. This includes, but is not limited to: removing or labeling samples with missing key fields; correcting or isolating obviously unreasonable time sequences; triggering a supplementary collection mechanism for samples with insufficient recovery feedback, such as increasing the frequency of feedback collection after subsequent plasma donations to supplement the sample; and uniformly encapsulating the stress input and recovery output of each plasma donation event into a standardized data object and writing it into the individual's health record. In this way, while ensuring the validity of the samples, the consistency and reusability of the samples can be improved, providing a more stable data foundation for the subsequent construction and iterative updates of individual digital profiles.

[0045] Step 204: Extract key stress features characterizing stress intensity based on historical stress source time series data. Key stress features include cumulative anticoagulant usage and standardized anticoagulant load.

[0046] In this embodiment, the start and end times of historical plasma collection can be determined based on historical stress source time-series data, and the collection duration can be determined based on these times. Then, the cumulative amount of anticoagulant used within the time range from the start to the end of plasma collection can be determined based on the historical stress source time-series data. After that, the anticoagulant pumping rate time series is extracted based on the historical stress source time-series data, and the pumping rate fluctuation characteristics are calculated based on the anticoagulant pumping rate time series, including the sample standard deviation of the anticoagulant pumping rate time series. Finally, the standardized anticoagulant load is determined based on the cumulative amount of anticoagulant used, the collection duration, and the pumping rate fluctuation characteristics.

[0047] The start of plasma collection can be defined as the moment when the anticoagulant pumping rate first becomes consistently greater than zero. Correspondingly, the end of plasma collection can be defined as the moment when pumping stops and the plasma collection weight stabilizes.

[0048] In this embodiment, a single plasma donation process can be divided into two core phases based on the start and end times of plasma collection: the in vivo stress period and the in vitro recovery period. The in vivo stress period can be understood as the phase from the start to the end of plasma collection, denoted as phase [stage name missing]. The in vitro recovery period can be understood as the phase from the end of plasma collection until the donor's physical functions (e.g., recovery from fatigue) recover, denoted as the phase. Furthermore, it can be All anticoagulant-related data during the phase, such as cumulative volume and pumping rate curves, will be marked as stress source inputs. The subjective fatigue scores collected during the phase are labeled as recovery response outputs.

[0049] In some embodiments, to obtain key stress features that can be used for modeling from historical stressor time-series data, it can be achieved in the following manner: First, key events in plasma collection are identified based on the changes in process quantities in historical stressor time-series data to determine the start and end times of historical plasma donations. Specifically, the start time of historical plasma donations can be determined by criteria such as the cumulative anticoagulant dosage changing from static to continuous growth over time, and the cumulative collected plasma volume simultaneously entering a continuous growth phase. The end time of historical plasma donations can be determined by criteria such as the cessation of the growth in cumulative anticoagulant dosage and the stabilization of the cumulative collected plasma volume, or the equipment status changing from a collection state to a termination state.

[0050] Optionally, to avoid misjudgments caused by short-term jitter, continuous verification and multi-signal consistency verification can be introduced into the above-mentioned start and end criteria. Furthermore, when there are process segments such as pauses or backflows, invalid plasma collection segments can be identified and eliminated to obtain start and end points consistent with the actual plasma collection process, and the plasma collection duration can be determined accordingly. This plasma collection duration can be understood as the total plasma collection duration or the effective plasma collection duration.

[0051] After determining the start and end times of slurry collection, the cumulative amount of anticoagulant used can be statistically analyzed within the time range between the start and end times of slurry collection to obtain the characteristics of the cumulative amount of anticoagulant used.

[0052] Meanwhile, the anticoagulant pumping rate time series can also be extracted from historical stress source time series data. The anticoagulant pumping rate time series refers to the temporal change in the amount of anticoagulant pumped per unit time. It can be obtained from the pumping rate records directly output by the equipment, or derived from the cumulative anticoagulant usage series through differencing, smoothing, and time alignment.

[0053] Subsequently, based on the anticoagulant pumping rate time series, the pumping rate fluctuation characteristics are calculated to characterize the pumping stability. These fluctuation characteristics may include the sample standard deviation of the pumping rate time series; the sample standard deviation measures the dispersion of the pumping rate around its mean, with greater dispersion indicating greater pumping instability. Optionally, to improve robustness, the pumping rate series can be denoised, outlier suppressed, and missing data imputed. Additional fluctuation indices, such as the coefficient of variation, quantile difference, or short-term fluctuation intensity, can be calculated for cross-validation.

[0054] After obtaining the cumulative anticoagulant usage, slurry collection time, and pumping rate fluctuation characteristics, the standardized anticoagulant load can be further determined. The standardized anticoagulant load is a comprehensive index that normalizes and makes comparable the intensity of anticoagulant use, used to characterize factors such as usage amount, process duration, and pumping stability as a single load within a unified framework.

[0055] In practical implementation, the cumulative amount of anticoagulant and the duration of plasma collection can be used as the basic terms characterizing the input intensity per unit time. The pumping rate fluctuation characteristics can be used as a modulation or weighting term reflecting the degree of input disturbance, constructing a load calculation function that satisfies monotonicity constraints: under the same conditions, the larger the cumulative amount of anticoagulant, the higher the load; the more unstable the pumping, the higher the load; and when the cumulative amount of anticoagulant is the same, the shorter the plasma collection time, the higher the input intensity and the higher the load per unit time. This load can serve as a key input for subsequent individual digital profiling and fatigue recovery trajectory prediction, thereby making the stress intensity comparable between different historical plasma donation processes, such as different plasma collection times and different pumping strategies, and reducing the bias caused by process differences in individual sensitivity estimation and recovery prediction, thus improving the stability and reliability of model training and inference.

[0056] In some embodiments, the standardized anticoagulant load can be defined as a function of the total anticoagulant dosage and the plasma collection time, used to compare the stress intensity of different plasma collection times. Its calculation method can be expressed as follows:

[0057] in, Indicates the standardized anticoagulant load; This is the total dose of anticoagulant, in mL; This is the total plasma collection time, in minutes. This is the sample standard deviation of the anticoagulant infusion rate during the stress period in vivo, in mL / min. This formula simultaneously considers dosage, time span, and fluctuations in the input process, enabling a more precise characterization of the quality and quantity of stressors.

[0058] For example, suppose that in a historical plasma donation event, the system retrieves the time-series data of historical stressors for that donation, including the cumulative anticoagulant dosage and the cumulative plasma volume collected. Key event identification reveals that before a certain point in time, both cumulative sequences remain unchanged; from that point onward, both the cumulative anticoagulant dosage and the cumulative plasma volume simultaneously enter a state of continuous increase, and this increase remains stable over a continuous period. Therefore, the system determines this point as the start time of plasma collection. Subsequently, when the cumulative anticoagulant dosage stops increasing and the cumulative plasma volume stabilizes and remains unchanged, the system determines this point as the end time of plasma collection. For example, suppose the start time of plasma collection is 10:05:00 and the end time is 10:53:00, therefore the plasma collection duration is 48 minutes.

[0059] Within the aforementioned plasma collection time range, the cumulative amount of anticoagulant used at the start of plasma collection can be used as the starting point and the cumulative amount of anticoagulant used at the end of plasma collection can be used as the ending point to calculate the cumulative amount of anticoagulant used in this historical plasma collection, which is assumed to be 62 mL.

[0060] Furthermore, an anticoagulant pumping rate time series can be generated based on the cumulative anticoagulant dosage sequence. This anticoagulant pumping rate time series can be obtained from the pumping rate record directly output by the device, or it can be obtained by the difference between adjacent sampling points of the cumulative sequence. Assuming that the pumping rate sequence (unit: mL / min) obtained in this embodiment is: 1.2, 1.3, 1.1, 1.4, 1.2, 1.3, 1.0, 1.5; Therefore, the sample standard deviation of this sequence is approximately 0.16 mL / min.

[0061] After obtaining the cumulative anticoagulant dosage (62 mL), slurry collection time (48 min), and pumping rate fluctuation characteristics (sample standard deviation approximately 0.16 mL / min), the standardized anticoagulant load can be calculated. That is: Standardized anticoagulant load = (cumulative anticoagulant usage / slurry collection time) × log(1 + sample standard deviation) = (62 mL / 48 min) × log(1 + 0.16 mL / min) ≈ 0.192 (dimensionless).

[0062] It should be noted that the above-described methods for extracting key stress features that characterize stress intensity are merely illustrative examples of the embodiments in this application and do not impose any limitations on the embodiments in this application.

[0063] Step 206: Extract key fatigue recovery features that characterize the change of fatigue degree over time based on historical recovery response data. Key fatigue recovery features include fatigue half-life and fatigue decay rate within a preset recovery time window.

[0064] The fatigue half-life is the time length corresponding to the point when the historical fatigue recovery curve drops to half of the initial fatigue level at the end of slurry production.

[0065] Fatigue decay rate is used to characterize how quickly the fatigue level decreases over time within a preset recovery time window.

[0066] In some embodiments, key fatigue recovery features characterizing the change of fatigue degree over time can be extracted according to the following procedure: First, historical recovery response data associated with the same historical plasma donation event are organized to form a recovery response sequence showing the change in fatigue level over time. This recovery response sequence refers to a data sequence arranged chronologically, consisting of multiple feedback time markers and fatigue scores (or equivalent fatigue indicators). Specifically, historical recovery response data can be aggregated based on the plasma donation event identifier and the plasma collection end time, using the plasma collection end time as a time reference point to perform relative time conversion and sorting of each feedback record. Simultaneously, consistency checks and quality control are performed to address issues such as duplicate submissions, omissions, obvious abrupt changes, or inconsistent time markers. For example, the higher-quality record within the same time window is retained, unreasonable values ​​are marked or removed, and missing points are optionally imputed or left blank, resulting in a clearly structured recovery response sequence suitable for fitting.

[0067] Secondly, a historical fatigue recovery curve is constructed based on the recovery response sequence, and the initial fatigue level at the end of slurry sampling is determined by this curve. The historical fatigue recovery curve is a fatigue change function obtained by continuously representing the discrete recovery response sequence, reflecting the overall trend of fatigue decay over time. Specifically, the curve can be generated using interpolation or regression fitting methods, such as piecewise linear interpolation, piecewise polynomial fitting, locally weighted regression, or spline fitting with smoothing constraints. To enhance robustness, monotonicity constraints or penalty terms can be introduced to suppress local rebounds caused by noise. The initial fatigue level can be defined as the curve value corresponding to the end of slurry sampling; if an actual feedback point exists at the end of slurry sampling, this feedback point value can be directly used; otherwise, the extrapolated / neighborhood estimate of the curve at the end of slurry sampling can be used as the initial fatigue level.

[0068] Then, the fatigue half-life is determined based on the historical fatigue recovery curve. The fatigue half-life is used to characterize the rate at which fatigue recovers from the initial level. In practice, the time point on the curve that satisfies the curve value being equal to half of the initial fatigue level can be found, and the time difference between this time point and the end of slurry collection can be calculated. When the curve has multiple intersections or local fluctuations, the time when the level is first reached can be used as the half-life.

[0069] Furthermore, the fatigue decay rate within the preset recovery time window is determined based on the decrease amplitude of the historical fatigue recovery curve within the preset recovery time window and the duration of the time window. The preset recovery time window refers to a fixed observation interval from the end of plasma collection, used to characterize the early or mid-stage recovery process after plasma donation; the fatigue decay rate characterizes how quickly the fatigue level decreases over time within this time window. Specifically, the difference between the curve values ​​corresponding to the start and end points of the time window can be taken as the decrease amplitude, and a ratio is formed with the time window length to obtain the decay rate. In other embodiments, the average slope of the curve within the time window or the fitted exponential decay parameter can be used as an equivalent expression of the decay rate to improve adaptability to noise and sampling sparsity.

[0070] Optionally, considering that the feedback frequency and time distribution of different plasma donors may be inconsistent, in some optional implementations, confidence or quality labels can be attached to the above features. For example, feature confidence can be generated based on the number of valid feedback points, time coverage, and the size of the fitting residual. When the confidence is insufficient, a supplementary sampling strategy can be triggered, such as increasing the frequency of feedback collection after subsequent plasma donations or adopting a more conservative profile update strategy. In this way, the stability and comparability of key fatigue recovery features can be improved without increasing the burden on plasma donors too much, providing a more reliable input basis for subsequent individual digital profile construction and fatigue recovery trajectory prediction.

[0071] For example, assuming a plasma donation event by a plasma donor, the system records the end time of plasma collection as a time reference point and collects the fatigue score records submitted by the plasma donor after donation. The fatigue scores submitted by this plasma donor at multiple time points after plasma collection are as follows (score range 0-10, with higher scores indicating greater fatigue): Immediately after slurry collection: fatigue score 7.5; Two hours after the end of slurry collection: fatigue score 5.1; Four hours after the end of slurry collection: fatigue score 3.6; Eight hours after the end of slurry collection: fatigue score 2.4; 1) Determine the recovery response sequence; The above records are sorted chronologically, and the time elapsed since the end of plasma collection is used as a unified timeline to form a recovery response sequence: (0 hours, 7.5 points), (2 hours, 5.1 points), (4 hours, 3.6 points), (8 hours, 2.4 points), which, after sorting, are: (0, 7.5), (2, 5.1), (4, 3.6), (8, 2.4). If there are duplicate submissions or obvious outliers within the same time window, such as a sudden increase in a short period of time without any reason, they can be removed or labeled according to the consistency rules before proceeding to the fitting step.

[0072] 2) Construct historical fatigue recovery curves based on the recovery response sequence and determine the initial fatigue level; In some embodiments, piecewise linear interpolation can be used to make the discrete sequence continuous, constructing a historical fatigue recovery curve. Since there is an effective scoring point at the end of slurry collection, the value of the curve at the end of slurry collection is directly used as the initial fatigue level, that is, the initial fatigue level = 7.5.

[0073] 3) Determine the fatigue half-life; In this embodiment, the fatigue half-life is defined as the time length corresponding to the fatigue recovery curve dropping to half of the initial fatigue level. Therefore, half of the initial fatigue level can be initially determined as: 7.5 ÷ 2 = 3.75.

[0074] Secondly, since the fatigue score decreases from 5.1 after 2 hours to 3.6 after 4 hours, it can be determined that 3.75 will fall within this range. Based on this, linear interpolation can be performed within this range: Within the 2-4 hour timeframe, the total decrease was 5.1. 3.6 = 1.5; The decrease from 5.1 to 3.75 is equal to 5.1. 3.75 = 1.35; The corresponding time ratio = 1.35 ÷ 1.5 = 0.9; Corresponding time point = 2 hours + 0.9 × (4) (2 hours) = 2 hours + 1.8 hours = 3.8 hours, therefore we can determine that the fatigue half-life is 3.8 hours.

[0075] 4) Determine the fatigue decay rate within the preset recovery time window; For example, suppose the preset recovery time window is 0–3 hours after the end of slurry extraction. In this case, we can first calculate the fatigue value of the curve at 3 hours. Since 3 hours falls within the 2–4 hour range, linear interpolation can be performed. The slope over 2-4 hours = (3.6) 5.1) ÷ (4 2) = 0.75 minutes / hour; Fatigue value after 3 hours = 5.1 + (3) 2)×( 0.75) = 4.35; The decrease within that time window is: Decrease amount = 7.5 4.35 = 3.15; The time window duration is 3 hours; Therefore, the fatigue decay rate within the preset recovery time window is: Attenuation rate = 3.15 ÷ 3 = 1.05 minutes / hour; Therefore, the key fatigue recovery features output in this embodiment include at least the following: Recovery response sequence: (0, 7.5), (2, 5.1), (4, 3.6), (8, 2.4); Initial fatigue level: 7.5; Fatigue half-life: 3.8 hours; Fatigue decay rate within the preset time window (0-3 hours): 1.05 minutes / hour.

[0076] It should be noted that the above-described implementation methods for extracting key fatigue recovery features that characterize the change of fatigue degree over time are merely illustrative examples in this application and do not impose any limitations on the embodiments of this application.

[0077] Step 208: Input the key stress features and key fatigue recovery features into the trained physiological parameter prediction model to obtain sensitivity and fatigue recovery ability.

[0078] In some embodiments, the extracted key stress features and key fatigue recovery features can be used as model inputs to call the trained physiological parameter prediction model to perform inference calculations, thereby outputting the corresponding physiological parameters, namely sensitivity and fatigue recovery ability. Specifically, firstly, the features can be encapsulated in a unified data structure to form an input feature vector corresponding to a single plasma donation event; subsequently, the input feature vector is preprocessed, including but not limited to missing value handling, outlier truncation or labeling, unit unification and normalization, feature sorting and encoding, so that the input meets the interface and distribution requirements of the trained model.

[0079] After preprocessing, the input feature vector is input into the trained physiological parameter prediction model for forward inference. Optionally, the physiological parameter prediction model can be a regression model or an ensemble learning model, or a multi-task learning model with multiple output heads. Its output includes at least: a sensitivity parameter characterizing the strength of an individual's response to plasma collection stress input, and a fatigue recovery ability parameter characterizing an individual's ability to recover from a fatigued state to a baseline.

[0080] Optionally, to improve stability and interpretability, in some optional implementations, the model can adopt an integrated structure to fuse the outputs of multiple base learners and can perform calibration processing on the outputs, such as mapping the outputs to a preset interpretability range or level; at the same time, it can also output confidence or uncertainty metrics to indicate the credibility of the prediction result. When the confidence is insufficient, a more conservative intervention strategy is triggered or a prompt is made to collect supplementary recovery response data.

[0081] Finally, the obtained sensitivity and fatigue recovery capabilities are written into the individual's digital profile and stored in association with the metadata of the plasma donation event. This data can be used for subsequent fatigue recovery trajectory prediction, intervention time point determination, and guidance information generation, thereby realizing automatic mapping and individualized modeling from stress-recovery observation characteristics to computable physiological parameters.

[0082] In some embodiments, before performing step 208, it is necessary to first train a physiological parameter prediction model. For example... Figure 3 As shown in some embodiments, the trained physiological parameter prediction model can be obtained using the following steps 302 to 310: Step 302: Obtain multiple plasma donation data samples, including stress source time series data samples and corresponding recovery response data samples.

[0083] Among them, the stress source time series data samples are used to characterize the process data sequence related to stress input during a specific plasma donation process; the recovery response data samples are used to characterize the post-donation recovery state observation data corresponding to that plasma donation process. The two are paired samples based on the same plasma donation event to ensure that the mapping relationship learned in subsequent training is the correspondence between stress input and recovery output.

[0084] In some alternative implementations, plasma donation data samples can be retrieved in batches from personal health records, plasma station business systems, or training data warehouses; and data quality screening and consistency verification can be performed before the samples are stored or trained, such as removing, labeling, or downweighting samples with missing key fields, conflicting time stamps, severely insufficient or obviously abnormal recovery response data, in order to reduce the impact of noisy samples on the training results.

[0085] Step 304: Determine the actual sensitivity and actual fatigue recovery ability corresponding to each plasma donor sample, and use them as label data.

[0086] Actual sensitivity is used to characterize the strength of a plasma donor's response to plasma donation stress input; actual fatigue recovery capacity is used to characterize the strength of a plasma donor's ability to return from a fatigued state to a baseline state.

[0087] In some implementations, the label data can be obtained through statistical inference based on historical data or constructed based on expert rules. For example, the actual sensitivity can be derived from the relationship between changes in stress intensity and changes in early fatigue response, and the actual fatigue recovery capacity can be derived from the time required to recover to a steady state or the recovery speed relative to the group baseline. Optionally, the label data can be further normalized, segmented, or assessed for confidence to improve the stability and interpretability of the label data.

[0088] Step 306: Extract key stress feature samples based on stress source time series data samples, and extract key fatigue recovery feature samples based on recovery response data samples.

[0089] In some implementations, key stress feature samples are used to characterize the stress intensity and process morphology of the plasma donation process, such as the intensity of anticoagulant use, plasma collection continuity characteristics, and pumping stability; key fatigue recovery feature samples are used to characterize the recovery kinetics of fatigue changes over time after plasma donation, such as the initial fatigue level, recovery rate characteristics, and stage transition characteristics.

[0090] In some alternative implementations, fatigue recovery curves can be constructed first on the recovery response data samples, for example by interpolation or regression fitting, and then features that can stably characterize the recovery process can be extracted from the curves; at the same time, the time series data samples can be denoised, outlier suppressed and missing data imputed to improve the robustness of feature extraction.

[0091] Step 308: Combine the key stress feature samples and the key fatigue recovery feature samples into an input feature vector.

[0092] In this embodiment, key stress feature samples and key fatigue recovery feature samples are combined into an input feature vector. This combination can include splicing, encoding, and structured encapsulation, ensuring that each plasma donation data sample corresponds to a unique input feature vector.

[0093] In some alternative implementations, preprocessing can be performed on the input feature vectors to meet the model interface and distribution requirements. This may include, but is not limited to, missing value handling, outlier truncation or labeling, unit unification, normalization, discrete feature encoding, and fixed feature order, thereby ensuring input consistency between the training and inference phases.

[0094] Step 310: Using the input feature vector as the model input and the label data as the model output, train the initial physiological parameter prediction model to obtain the trained physiological parameter prediction model.

[0095] In this embodiment, the initial physiological parameter prediction model can be trained using the input feature vector as the model input and the label data as the model output to obtain the trained initial physiological parameter prediction model. Then, the trained initial physiological parameter prediction model is used to predict the prediction sensitivity and predicted fatigue recovery ability of plasma donation data samples. Next, the prediction error of the initial physiological parameter prediction model is calculated based on the prediction sensitivity, predicted fatigue recovery ability, actual sensitivity, and actual fatigue recovery ability. Finally, the model parameters of the initial physiological parameter prediction model are iteratively updated based on the prediction error until the prediction error of the updated initial physiological parameter prediction model reaches a preset convergence condition, at which point training is terminated to obtain the trained physiological parameter prediction model.

[0096] In some implementations, the prediction error can be in the form of a joint error, that is, the error in the sensitivity dimension and the error in the fatigue recovery capability dimension are calculated separately and then weighted and summed. Optionally, the prediction error can also be superimposed with a complexity penalty term to constrain the model size or parameter range, thereby improving the generalization ability.

[0097] In some embodiments, the prediction error of the model can be expressed as:

[0098] in, This represents the model's prediction error; N represents the number of plasma donation data samples. This represents a loss function, such as the mean squared error loss function; It is a hyperparameter that balances the predictive importance of the two coefficients. It is a regularization term for model complexity. Indicates plasma donation data sample i Predictive sensitivity; Indicates plasma donation data sample i The actual sensitivity; Indicates plasma donation data sample i Predictable fatigue recovery ability; Indicates plasma donation data sample i The actual fatigue recovery ability.

[0099] In some optional implementations, when the initial physiological parameter prediction model is a gradient boosting decision tree ensemble structure, the iterative update can be manifested as follows: In each iteration, a new base learner is constructed based on the residuals or loss function gradients of the current model on the training samples to fit the error direction, and the new base learner is incorporated into the model set according to a preset update rule, thereby gradually reducing the overall prediction error; at the same time, learning rate, tree structure constraints, or early stopping strategies can be set to further suppress overfitting. Training is terminated when the prediction error of the updated initial physiological parameter prediction model reaches the preset convergence condition, resulting in the trained physiological parameter prediction model.

[0100] In some implementations, the preset convergence conditions may include: the training error or validation error being lower than a threshold, the decrease in error over multiple consecutive iterations being lower than a threshold, or early stopping being triggered when the maximum number of iterations is reached; and the model version with the best validation performance may be selected as the model to be trained.

[0101] In some alternative implementations, after training is terminated, the trained physiological parameter prediction model and its feature configuration and version information can be persistently stored so that they can be directly reused when key stress features and key fatigue recovery features are subsequently input into the trained physiological parameter prediction model. At the same time, prediction confidence or uncertainty measure can be output to help determine the credibility of the output results during the inference stage, thereby supporting more robust individualized decision-making.

[0102] Step 210: Construct individual digital profiles of plasma donors based on their sensitivity and fatigue recovery capabilities.

[0103] In some implementations, after obtaining the plasma donor's sensitivity and fatigue recovery ability, these parameters can be written into a structured profile object as core physiological parameters to construct an individual digital profile of the plasma donor. Specifically, the profile construction unit can first standardize the definitions of sensitivity and fatigue recovery ability to ensure consistency with the parameter meanings used by subsequent prediction and decision-making modules. For example, sensitivity can be converted into a sensitivity coefficient characterizing the strength of an individual's response to plasma donation stress input, and fatigue recovery ability can be converted into a recovery rate coefficient characterizing the speed of an individual's fatigue recovery. Necessary range constraints, outlier corrections, or rank mappings can then be applied to these coefficients to ensure comparability and interpretability across different plasma donation events.

[0104] Subsequently, using the donor's unique identifier as an index, a corresponding individual digital profile data structure is generated. Each individual digital profile includes at least two core fields: sensitivity coefficient and recovery rate coefficient. Optionally, in addition to these two core fields, metadata fields related to the profile may also be included, such as profile generation time, applicable data range, corresponding physiological parameter prediction model version number, historical plasma donation sample coverage, data quality identifier, and prediction confidence level, to support subsequent model invocation, version tracking, and reliability assessment.

[0105] Optionally, to facilitate the sharing and reuse of individual digital profiles across different business systems, in some embodiments, individual digital profiles can be encapsulated in a scalable structured storage format and stored in personal health records or server databases. Simultaneously, update strategies can be configured for individual digital profiles; for example, when subsequent plasma donation recovery response data deviates continuously from the prediction results based on this profile, profile reassessment or retraining can be triggered, thereby maintaining the profile's adaptability to long-term individual changes. Through the above methods, sensitivity and fatigue recovery capabilities can be solidified from single-inference results into sustainably reusable individual digital profiles, providing a unified and traceable individualized parameter foundation for subsequent fatigue recovery trajectory prediction, intervention time point determination, and personalized guidance information generation.

[0106] Step 104: Based on the time series data of stressors and individual digital profiles, predict the fatigue recovery trajectory of plasma donors after plasma donation.

[0107] In some implementations, after acquiring the stressor time-series data of the plasma donor and obtaining the individual digital profile of the plasma donor, fatigue recovery trajectory prediction can be performed to obtain the predicted results of the change in the fatigue level of the plasma donor over time after the end of plasma donation.

[0108] Specifically, the time-series data of the stress source can first be processed to extract stress characterization information that can represent the intensity and process characteristics of the stress input during the plasma donation. This processing may include: determining the cumulative amount of anticoagulant and the duration of plasma collection within the time range from the start to the end of plasma collection; extracting the pumping rate fluctuation characteristics based on the pumping characteristics formed by the change of the cumulative amount of anticoagulant over time; and then determining the standardized anticoagulant load based on the cumulative amount of anticoagulant, the duration of plasma collection, and the pumping rate fluctuation characteristics, as a comprehensive quantitative indicator of the stress intensity of this plasma donation. Subsequently, the sensitivity coefficient and recovery rate coefficient corresponding to the plasma donor are read from the individual digital profile, and the standardized anticoagulant load is correlated and combined with the sensitivity coefficient and recovery rate coefficient to form a set of input parameters for fatigue recovery trajectory prediction.

[0109] Based on this, a fatigue recovery trajectory prediction model is invoked to generate the fatigue recovery trajectory after plasma donation. The fatigue recovery trajectory prediction model is used to map stress input characterization information and individualized physiological parameters into the process of fatigue degree changing over time. Its output can be a fatigue prediction sequence at discrete time points or a continuous fatigue recovery curve.

[0110] In generating the fatigue recovery trajectory, the sensitivity coefficient can be used to individualize the initial fatigue level under the same stress intensity, reflecting the differences in individual responses to plasma donation stress input; the recovery rate coefficient can be used to individualize the rate of fatigue decay over time, reflecting the differences in individual recovery capabilities; and the standardized anticoagulant load can be used to apply a stress intensity-related influence to the recovery process, so that, under the same conditions, the higher the stress intensity, the slower the predicted recovery process, thus allowing the output trajectory to simultaneously reflect the combined effects of individual and process differences. Therefore, a fatigue recovery trajectory corresponding to this plasma donation event can be obtained, serving as the basis for subsequently determining intervention time points and generating fatigue recovery guidance information.

[0111] Step 106: Determine at least one intervention time point based on the fatigue recovery trajectory, and push fatigue recovery guidance information to plasma donors at the intervention time point.

[0112] In this embodiment of the application, at least one trajectory feature point can be identified as an intervention time point based on the fatigue recovery trajectory; wherein, the trajectory feature point includes at least one of the following: The intervention start time point is the time point corresponding to the first preset proportion of the initial fatigue level when the fatigue level in the fatigue recovery trajectory decreases to the initial fatigue level at the end of slurry collection; The inflection point intervention time point is the dividing point between the first recovery stage and the second recovery stage in the fatigue recovery trajectory. The fatigue decay rate in the first recovery stage is greater than the fatigue decay rate in the second recovery stage. Consolidation intervention time point: The consolidation intervention time point represents the time point in the fatigue recovery trajectory that meets the baseline stability condition; The baseline stability conditions include: the fatigue level remains within a preset tolerance range centered on the initial fatigue level, and in a sequence consisting of a preset number of consecutive sampling points, the absolute value of the change in fatigue level between adjacent sampling points does not exceed a preset fluctuation threshold.

[0113] In some embodiments, the initial fatigue level at the end of slurry sampling can be determined, and a target fatigue level corresponding to a first preset ratio can be obtained accordingly. Then, the fatigue recovery trajectory is retrieved along the time axis to determine the time point at which the fatigue level first reaches or exceeds the target fatigue level, and this time point is determined as the intervention initiation time point. In some optional embodiments, considering that the trajectory may have local fluctuations, a continuous satisfaction strategy can also be adopted, that is, requiring the fatigue level to meet the threshold condition at several consecutive sampling points before confirming the intervention initiation time point, in order to improve the stability of identification.

[0114] In some embodiments, the changing trend of the fatigue decay rate can be calculated based on the fatigue recovery trajectory, and the first recovery stage and the second recovery stage can be identified accordingly. When the fatigue decay rate is detected to change from a larger value range to a smaller value range, the boundary position corresponding to this change is determined as the stage boundary point, and the corresponding time point is determined as the inflection point intervention time point. The fatigue decay rate in the first recovery stage is greater than the fatigue decay rate in the second recovery stage. In some optional embodiments, the boundary point can be constrained or verified in conjunction with changes in curve shape to ensure that the inflection point intervention time point reflects the stage transition characteristics of the recovery process from fast to slow.

[0115] In some embodiments, sliding detection can be performed on the continuous sampling point sequence on the time axis: when, starting from a certain moment, several subsequent consecutive sampling points all fall within a preset tolerance range and adjacent changes all meet a preset fluctuation threshold, the starting time point that meets the conditions is determined as the consolidation intervention time point. In some optional embodiments, the preset tolerance range and preset fluctuation threshold can be adaptively set according to the overall amplitude of the fatigue recovery trajectory or the differences in individual digital profiles, so as to improve the adaptability to different individuals and different plasma donation processes.

[0116] In some alternative implementations, considering that the identified multiple intervention time points may be too dense in time or not conducive to push scheduling, sequence constraints can also be imposed on the intervention time points, such as setting the minimum interval between adjacent intervention time points and performing uniform granular processing on the intervention time points, thereby outputting an ordered and easy-to-execute sequence of intervention time points.

[0117] In some embodiments, when pushing fatigue recovery guidance information to plasma donors at the intervention time point, the predicted fatigue level corresponding to the intervention time point can be read and combined with its relative position or stage attribute in the fatigue recovery trajectory to map it to a preset stage set, thereby obtaining a guidance stage label. Then, a basic guidance template is selected from a preset guidance information database based on the guidance stage label. The basic guidance template includes at least one suggestion for a fatigue recovery activity.

[0118] In some implementations, a pre-defined guidance information database is indexed and organized according to guidance stage labels, so that different stages correspond to different types of recovery activity suggestions, thereby ensuring that the basic guidance template matches the current fatigue recovery stage.

[0119] Subsequently, the predicted fatigue level can be used as the basis for template modulation, and the variable fields in the template can be personalized or graded to make the generated fatigue recovery guidance information consistent with the fatigue recovery trajectory of the plasma donor.

[0120] In some preferred embodiments, the parameter adjustments can be further refined by combining the type of intervention time point with the individual digital profile: on the one hand, the guidance objectives are determined based on whether the intervention time point is the start-up intervention time point, the inflection point intervention time point, or the consolidation intervention time point; on the other hand, the guidance intensity and focus are individually adjusted by combining the sensitivity coefficient and recovery rate coefficient in the individual digital profile, thereby improving the pertinence and feasibility of the guidance information.

[0121] Finally, the intervention time point and fatigue recovery guidance information can be formed into a task to be executed and written into the scheduled task queue. When the corresponding intervention time point is reached, the task execution is triggered, and the fatigue recovery guidance information is delivered to the plasma donor via mobile push. The corresponding content display entry is provided in the application to support plasma donors to view and execute it.

[0122] The method provided in this application predicts the fatigue recovery trajectory after plasma donation by acquiring stressor time-series data of plasma donors and combining it with individual digital profiles characterizing their sensitivity to anticoagulant metabolism and fatigue recovery ability. This allows for the incorporation of individualized physiological response differences among different plasma donors into the modeling and prediction, making the predicted fatigue recovery process closer to the actual situation of the plasma donors. Based on this, intervention time points matching the plasma donor's state can be determined from the fatigue recovery trajectory, and targeted recovery guidance information can be pushed at the corresponding time points. This avoids the prediction bias and inaccurate intervention timing / content problems caused by existing technologies based on group averages and fixed rules, thereby improving the effectiveness of fatigue recovery management and the plasma donor experience.

[0123] The following describes how the methods provided in the embodiments of this application are applied in practice, taking into account real-world scenarios.

[0124] Please see Figure 4 This is a schematic diagram illustrating an application process of the method provided in this application embodiment. The process specifically includes the following steps: Step 1: During the first plasma collection, the core stressor data of this plasma collection process can be collected and recorded in real time through the plasma collection machine data interface. After the plasma collection is completed, the recovery period data can be triggered and collected through the plasma donor's mobile APP.

[0125] For first-time plasma donors, during the first collection, the system sends a request command to the plasma collection machine to collect two core stressor time-series data streams: one is the cumulative volume of the anticoagulant pumped in real time. Secondly, the cumulative volume of collected plasma. These two data streams are provided by high-precision metering pumps and weighing sensors inside the slurry sampling machine. The acquisition module adds a high-precision timestamp to each data point, forming two strictly aligned time series. At the end of the slurry sampling process, the total amount of anticoagulant used and the total sampling time are calculated and stored. The complete time-series data stream is then uploaded to the server.

[0126] After the plasma donor completes plasma collection and enters the rest area, the system pushes a message to the app linked to the plasma donor, such as... Figure 5 The electronic digital rating scale shown is mandatory.

[0127] like Figure 5 As shown, the electronic digital rating scale is a visual analog slider, with 0 points on the left indicating no feeling and 10 points on the right indicating extreme fatigue. Plasma donors report their immediate overall fatigue level by sliding the slider. At preset key time points after the donor leaves the plasma station, such as 3 and 6 hours after plasma collection, the app sends notifications again, guiding the donor to report their fatigue level at that time using the same scale. All scores and their corresponding collection time points are encrypted and synchronized to the server. This process not only captures static fatigue scores but, more importantly, dynamically tracks and obtains a discrete trajectory describing the changes in fatigue over time, especially during the critical recovery period after plasma collection.

[0128] Step 2: The collected multi-dimensional heterogeneous data is time-stamped and cleaned. The anticoagulant dosage is used as the stress input, and the time-related fatigue data is used as the recovery response output. The data are then fused to generate complete time-series data characterizing the stress-recovery process of individuals who are donating plasma for the first time.

[0129] The aforementioned heterogeneous data streams include: slurry sampler time-series data and subjective fatigue assessment time-point data. Starting from the slurry sampling start time... A unified timeline with absolute zero. For each raw data point in each data stream, based on its built-in timestamp and... The difference is remapped onto a unified time axis. For data with different sampling frequencies, such as high-frequency slurry sampler data and low-frequency fatigue scores, a time window alignment strategy is adopted to transform all heterogeneous data into feature sequences with consistent dimensions at a unified time point.

[0130] (1) In the plasma collection machine data sequence with a unified time axis, two landmark time points were identified: the start time of plasma collection and the end time of plasma collection. The start time of plasma collection was defined as the moment when the pumping rate of anticoagulant first remained above zero. The end time of plasma collection was defined as the moment when the pumping stopped and the plasma collection weight stabilized. These two time points divided the single plasma donation process into two core phases: the in vivo stress period, from the start to the end of plasma collection, denoted as phase. The in vitro recovery period, starting from the end of plasma collection, is denoted as stage 1. .Will All anticoagulant-related data during the phase, such as cumulative volume and pumping rate curves, will be marked as stress source inputs. The subjective fatigue scores collected during the phase are labeled as recovery response outputs.

[0131] (2) To The stressors of each stage are quantified. This data not only records the total anticoagulant dosage but also extracts key features describing the dynamic process of anticoagulant input, such as the average injection rate and the variance of the injection rate. The standardized anticoagulant load L is calculated, defined as a function of the total dosage and the plasma collection time, to compare the stress intensity of different plasma collection times. Its calculation method can be expressed as follows:

[0132] in, This is the total dose of anticoagulant, in mL; This is the total plasma collection time, in minutes. yes The sample standard deviation of the anticoagulant infusion rate within a given period, in mL / min. This formula takes into account dosage, time span, and variability in the input process, providing a more precise characterization of the quality and quantity of stressors.

[0133] (3) Processing The module uses piecewise linear interpolation or nonparametric regression, such as locally weighted scatter smoothing, to construct a continuous or quasi-continuous subjective fatigue recovery curve S(t) starting from the end of plasma collection, directly depicting the trajectory of the donor's psychological indicators recovering over time.

[0134] (4) Extracted stressor features, such as The stressor characteristics (L, L) are correlated with the dynamic characteristics of the recovery curve S(t). Specifically, features characterizing the recovery rate, such as the time required for the fatigue score to decrease to half of its initial value and the initial slope of the recovery curve, are extracted from the recovery curve S(t). These stressor characteristics and recovery dynamic characteristics are combined into a multidimensional, individualized stress-recovery feature vector. , can be represented as:

[0135] in, It represents the fatigue half-life. The fatigue decay rate within a preset recovery time window (e.g., within 30 minutes after slurry extraction), in minutes. 1. The slope of the curve is obtained by calculating the instantaneous slope at the starting point, i.e., within 30 minutes after the end of slurry collection; it reflects the initiation speed of recovery. Together with the corresponding original data fragments along a unified timeline, they are encapsulated into a structured time-series record of an individual's stress-recovery process and stored in the database. This record serves as the direct data foundation for subsequent personalized modeling and profile generation.

[0136] Step 3: Call the pre-trained prediction model, input the generated time-series data, and the model quantitatively analyzes the mapping relationship between anticoagulant dosage and recovery index curve, outputting the anticoagulant metabolism sensitivity coefficient of the first plasma donor. With recovery rate coefficient This constitutes a digital health profile of him after his first plasma donation.

[0137] The prediction model is a pre-trained gradient boosting decision tree ensemble learning model. The model is an ensemble of numerous regression decision trees using gradient boosting. Each tree... The algorithm learns to calculate the residual between the predicted true label and the sum of all previous tree predictions. For input X, each tree starts from the root node and, based on the pre-defined "feature index j" and "threshold τ" at that node, guides X to a leaf node. This leaf node stores a residual for the predicted true label. and Predicted value vector Among them, the anticoagulant metabolism sensitivity coefficient A continuous value between 0 and 2, quantifying the intensity of an individual's response to physiological stress from anticoagulants; a higher value indicates greater sensitivity. Recovery rate coefficient. This is a continuous value between 0.5 and 3, quantifying an individual's intrinsic ability to recover from stress; a higher value indicates a stronger recovery ability. The final prediction result is a weighted sum of the prediction vectors from all trees.

[0138] The labeled data required for model training is not obtained through direct measurement, but rather indirectly constructed from historical data using expert rules and statistical methods: Anticoagulant metabolic sensitivity coefficient Construction: Based on extensive historical data, this study analyzed the individual slope between the standardized anticoagulant load L and the corresponding early psychological response intensity, such as the initial increase in fatigue within 30 minutes after plasma collection, across multiple plasma donations for each donor. The higher the initial label value.

[0139] Recovery rate coefficient Construction: Based on historical data, this study analyzes the ratio of the time required for each plasma donor to recover their fatigue index to baseline under similar control conditions (L), to the average recovery time of the entire population. A smaller ratio indicates faster recovery, and thus a higher contribution rate. The higher the initial label value.

[0140] The model's training objective function aims to simultaneously minimize and The prediction error is in the form of:

[0141] in, This represents the model's prediction error; N represents the number of training samples. This represents a loss function, such as the mean squared error loss function; It is a hyperparameter that balances the predictive importance of the two coefficients. It is a regularization term for model complexity. Indicates plasma donation data sample i Predictive sensitivity; Indicates plasma donation data sample i The actual sensitivity; Indicates plasma donation data sample i Predictable fatigue recovery ability; Indicates plasma donation data sample i The actual fatigue recovery capacity. The specific technical implementation process of this step is described below: A vector is obtained from the generated individualized stress-recovery time-series records. Specifically, it includes: ,as well as Four-dimensional features. Input a pre-trained prediction model. The final prediction result is a vector of predicted values ​​for all trees. Weighted sum:

[0142] in, The learning rate is used to control the contribution weight of each tree and prevent overfitting.

[0143] In addition, a confidence score for each predicted value is calculated. and The confidence level is calculated by considering the input features. The confidence level is evaluated by the average Mahalanobis distance of the sample features in the training dataset. The smaller the distance, the closer the sample is to the feature space region that the model is familiar with, and the higher the confidence level.

[0144] Will and The data, along with its confidence level and key metadata of this plasma collection, such as date and total duration, are combined into a structured JSON-LD format digital profile. This profile not only contains numerical values ​​but also machine-readable semantic tags, such as... and The data is categorized as high-sensitivity, low-recovery type, providing a direct and clear classification basis for subsequent personalized guidance. This digital profile is stored in the individual's health record and serves as input for subsequent steps.

[0145] Step 4: When the plasma donor donates plasma for the next time, the theoretical fatigue recovery trajectory of the plasma donor is predicted based on the stored health profile. Based on the theoretical fatigue recovery trajectory, multiple reminder time points are determined, and personalized guidance plans corresponding to each reminder time point are determined and pushed to the plasma donor's APP interface for display.

[0146] This step aims to address the technical challenge of applying stored personalized health profiles to subsequent plasma donation activities and providing proactive, forward-looking health guidance. Unlike traditional passive responses or fixed-time reminders, this innovative solution is designed to adaptively plan the timing and content of interventions based on predicted individual recovery curves, ensuring the right guidance is provided at the right time.

[0147] (1) When a plasma donor begins a new plasma donation, the latest health profile is retrieved from the user's health record, the core of which is the personalization coefficient: anticoagulant metabolism sensitivity coefficient. With recovery rate coefficient Simultaneously, key stress parameters were obtained from the real-time data stream of this slurry sampling: the total amount of anticoagulant used. Together with the standardized anticoagulant load L, this constitutes the input context for this prediction.

[0148] (2) Call the built-in recovery trajectory prediction model. This model is a parameterized nonlinear function with the following form:

[0149] in, The predicted theoretical fatigue score at time t; The predicted initial fatigue value at the end of slurry harvesting at t=0 is given by the formula. calculate, The calibration coefficient represents the initial fatigue impact of a unit load on a sensitive individual; The predicted long-term fatigue baseline, i.e. the level after full recovery, is usually set as the user's resting fatigue constant.

[0150] The personalized recovery rate constant is a function determined by both the health profile and the current workload.

[0151] in, It is the baseline recovery rate. This is the damping coefficient. The meaning of this formula is: the rate of recovery coefficient. Positively promotes recovery, while having a high anticoagulant metabolic sensitivity coefficient. Combining it with a high load (L) will slow down the recovery process. By adjusting... , The model can dynamically simulate various theoretical curves from rapid recovery to slow recovery.

[0152] This time Substituting the values ​​into the formula above, the engine can calculate a continuous theoretical fatigue recovery curve starting from t=0. The curve represents the system's best prior prediction of the donor's recovery process in this instance.

[0153] (3) Divide the recovery period into three stages in advance and set reminder goals for each stage: Rapid decline phase: Fatigue subsides rapidly; the goal is to prevent initial discomfort and ensure a smooth start to the recovery process.

[0154] Platform transition period: The decline in fatigue slows down, and the goal is to overcome the recovery bottleneck and maintain the momentum for recovery.

[0155] Gradual stabilization period: Fatigue levels are close to baseline; the goal is to consolidate the recovery effect and form a closed loop of healthy behaviors.

[0156] For the recovery curve Perform mathematical analysis to identify characteristic time points that align with the goals of each stage: (1) Starting point Defined as the time corresponding to the fatigue score decreasing to 20% of its total decrease, i.e., satisfying... of This point indicates the early stage of recovery, during which it is necessary to replenish fluids and electrolytes promptly.

[0157] (2) Inflection point Defined as the point where the second derivative of the curve is zero, i.e., the moment when the direction of acceleration changes. The recovery velocity may slow down near this point, making it a crucial intervention window for preventing a recovery plateau. This is achieved by solving... get .

[0158] (3) Consolidation points : Defined as fatigue score entering a stable zone, for example, with The difference is less than 10% of the start time, which satisfies the condition. minimum time t This point is suitable for summarizing the recovery process and encouraging long-term healthy behaviors.

[0159] To prevent the calculated time points from being too densely or sparsely spaced, the module sets a minimum interval, such as 90 minutes. Simultaneously, the intervals between all time points and the end time of plasma collection are rounded up to the nearest multiple of 15 minutes for easier user understanding and system scheduling. Finally, an ordered, personalized sequence of prompt time points is output.

[0160] (4) For each planned reminder time point Based on the simulation context, the block executes a three-layer decision logic to generate a guidance scheme: First-level, timing-level strategy: based on Determine the core guidance objective by identifying the specific characteristic point, such as initiation, inflection point, or consolidation. For example, The goal is to replenish quickly. It is about breaking through the bottleneck. It reinforces habits.

[0161] Second-layer, profiling strategy: combining user... and Adjust the intensity and focus of guidance based on the value. For highly sensitive individuals... For high-performing users, the guidance includes a reassuring explanation that "mild discomfort is normal"; for users with low recovery efficiency... For low-level users, the guidance emphasizes specific steps and quantitative requirements, such as "be sure to drink 500ml" rather than "it is recommended to drink more water".

[0162] Third-layer, state-layer strategy: Combining The simulated state at any given time, such as whether the recovery rate is below a threshold, should be incorporated into the scheme as a dynamic variable. For example, in If the simulated recovery rate is very low, the program will add an active intervention suggestion such as "get up and walk slowly for 5 minutes to promote circulation".

[0163] (5) The results of the above three levels of decision-making are filled into a preset, parameterized natural language template to generate the final user-readable solution. Each solution includes: situational resonance, such as "You may now feel that your fatigue is decreasing more slowly, which is a common stage in the recovery process"; specific actions, such as "Try to eat a potassium-rich fruit, such as a banana"; scientific principles, such as "Potassium helps maintain cellular electrolyte balance and counteracts the mild effects of anticoagulants"; and expected goals, such as "Help you get through this recovery stage more smoothly".

[0164] (6) Register the generated, time-bound guidance plan to the system's scheduled task queue. Each task will be assigned to a specific timeframe. It is automatically triggered at any time. When a task is triggered, the module delivers the guidance plan to the current plasma donor through channels such as APP push notifications.

[0165] It should be noted that the above-listed guidance schemes are merely illustrative examples of embodiments of this application and do not impose any limitations on the embodiments of this application.

[0166] like Figure 6 The image shown is a schematic diagram of an interface for a personalized guidance push scheme provided in an embodiment of this application. Specifically, Figure 6 The left image in the image is a schematic diagram of "Lock Screen / Notification Bar Push Preview", which is in the form of an APP push notification. Figure 6 The image on the right is a schematic diagram of the "Details Page in the App that the user enters after clicking the notification". This page fully displays the personalized solution generated by the system's three-layer decision-making logic.

[0167] This approach transforms health profiles from static labels into predictive models capable of dynamically simulating an individual's recovery process and proactively planning interventions. By solving for the characteristic points of the theoretical curve, the optimal intervention timing is determined. Furthermore, by integrating information from the profile, timing, and predicted state, a highly customized guidance plan is generated. This enables a shift from generic, timed broadcasts to truly personalized health support tailored to each individual. The entire process runs automatically after plasma donation, requiring no manual intervention, demonstrating a high level of automation and intelligence.

[0168] After a plasma donor completes a plasma donation and the subsequent recovery feedback process, the system can further automatically detect whether the following two conditions are met simultaneously: (1) Condition A - Data validity: The number of valid fatigue feedback collected in this plasma donation reaches or exceeds the preset minimum requirement, for example, at least two of the planned three feedback points are successfully collected.

[0169] (2) Condition B - Persistent bias: The current health profile shows significant bias in predicting the recovery process after the two most recent consecutive plasma donations. Specifically, the mean absolute bias of each prediction must exceed the personalized threshold.

[0170] Calculate the mean absolute deviation: For a single plasma donation, calculate the absolute value of the difference between the donor's actual fatigue score and the theoretical fatigue value predicted by the profile at all valid feedback time points, and then take the average of these absolute values.

[0171] Significant deviation is defined as follows: the average absolute deviation of the plasma donation is greater than a dynamic threshold determined by the donor's historical fluctuation level, for example, exceeding 1.5 times its historical average deviation.

[0172] The system automatically triggers the health profile update process only when both conditions A and B are met. All historical valid data of the plasma donor are collected, the model is retrained, and a new profile is generated.

[0173] The method provided in this application predicts the fatigue recovery trajectory after plasma donation by acquiring stressor time-series data of plasma donors and combining it with individual digital profiles characterizing their sensitivity to anticoagulant metabolism and fatigue recovery ability. This allows for the incorporation of individualized physiological response differences among different plasma donors into the modeling and prediction, making the predicted fatigue recovery process closer to the actual situation of the plasma donors. Based on this, intervention time points matching the plasma donor's state can be determined from the fatigue recovery trajectory, and targeted recovery guidance information can be pushed at the corresponding time points. This avoids the prediction bias and inaccurate intervention timing / content problems caused by existing technologies based on group averages and fixed rules, thereby improving the effectiveness of fatigue recovery management and the plasma donor experience.

[0174] To address the problem in existing technologies that fail to consider individual physiological differences among plasma donors during the donation process, leading to deviations in fatigue recovery predictions from individual realities and a lack of targeted intervention timing and content, this application provides a plasma donor fatigue recovery management device. A schematic diagram of the device's specific structure is shown below. Figure 7 As shown, it includes an acquisition module 71, a prediction module 72, and a processing module 73. The functions of each module are as follows: The acquisition module 71 is used to acquire time-series data of stressors and individual digital profiles of plasma donors; the individual digital profiles characterize the plasma donors' sensitivity to anticoagulant metabolism and fatigue recovery ability. Prediction module 72 is used to predict the fatigue recovery trajectory of plasma donors after plasma donation based on stress source time series data and individual digital profiles; The processing module 73 is used to determine at least one intervention time point based on the fatigue recovery trajectory and push fatigue recovery guidance information to the plasma donor at the intervention time point.

[0175] Optionally, module 71 includes: The plasma donation data acquisition unit is used to acquire the historical plasma donation data of plasma donors. The historical plasma donation data includes the time series data of historical stressors and the corresponding historical recovery response data of at least one historical plasma donation process. The stress feature extraction unit is used to extract key stress features characterizing stress intensity based on historical stress source time series data. Key stress features include cumulative anticoagulant usage and standardized anticoagulant load. The fatigue recovery feature extraction unit is used to extract key fatigue recovery features that characterize the change of fatigue degree over time based on historical recovery response data. Key fatigue recovery features include fatigue half-life and fatigue decay rate within a preset recovery time window. The processing unit is used to input key stress features and key fatigue recovery features into the trained physiological parameter prediction model to obtain sensitivity and fatigue recovery ability. The digital profile building unit is used to create individual digital profiles of plasma donors based on their sensitivity and fatigue recovery capabilities.

[0176] Optional, a stress feature extraction unit, used for: Based on historical stress source time series data, the start and end times of historical plasma collection were determined, and the collection duration was determined based on the start and end times of plasma collection. The cumulative amount of anticoagulant used within the time range from the start to the end of plasma collection was determined based on historical stress source time series data. The anticoagulant pumping rate time series was extracted from historical stress source time series data, and the pumping rate fluctuation characteristics were calculated based on the anticoagulant pumping rate time series. The pumping rate fluctuation characteristics include the sample standard deviation of the anticoagulant pumping rate time series. The standardized anticoagulant load is determined based on the cumulative amount of anticoagulant used, the duration of slurry collection, and the fluctuation characteristics of the pumping rate.

[0177] Optional, fatigue recovery feature extraction unit, used for: Determine the recovery response sequence of fatigue level over time based on historical recovery response data; Historical fatigue recovery curves are constructed based on the recovery response sequence, and the initial fatigue level at the end of slurry harvesting is determined based on the historical fatigue recovery curves. The fatigue half-life is determined based on the historical fatigue recovery curve, where the fatigue half-life is the time length corresponding to when the historical fatigue recovery curve drops to half of the initial fatigue level; Based on the decrease in the historical fatigue recovery curve within the preset recovery time window and the duration of the preset recovery time window, the fatigue decay rate within the preset recovery time window is determined. The fatigue decay rate is used to characterize how quickly the fatigue level decreases over time within the preset recovery time window.

[0178] Optionally, the plasma donor fatigue recovery management device also includes a training module, specifically including: The sample acquisition unit is used to acquire multiple plasma donation data samples, which include stress source time-series data samples and corresponding recovery response data samples. The label determination unit is used to determine the actual sensitivity and actual fatigue recovery ability of each plasma donor sample, which are used as label data. The feature extraction unit is used to extract key stress feature samples based on stress source time series data samples and key fatigue recovery feature samples based on recovery response data samples. The feature combination unit is used to combine key stress feature samples and key fatigue recovery feature samples into an input feature vector. The model training unit is used to train the initial physiological parameter prediction model by taking the input feature vector as the model input and the label data as the model output, so as to obtain the trained physiological parameter prediction model.

[0179] Optional, model training unit, used for: The prediction sensitivity and fatigue recovery ability of plasma donation data samples were predicted using an initial physiological parameter prediction model. The prediction error of the initial physiological parameter prediction model is calculated based on the prediction sensitivity, predicted fatigue recovery ability, actual sensitivity, and actual fatigue recovery ability. The model parameters of the initial physiological parameter prediction model are iteratively updated based on the prediction error until the prediction error of the updated initial physiological parameter prediction model reaches the preset convergence condition, at which point the training is terminated and the trained physiological parameter prediction model is obtained.

[0180] Optionally, processing module 73 is used for: Based on fatigue recovery trajectory identification, at least one trajectory feature point is used as the intervention time point; The trajectory feature points include at least one of the following: The intervention start time point is the time point corresponding to the first preset proportion of the initial fatigue level when the fatigue level in the fatigue recovery trajectory decreases to the initial fatigue level at the end of slurry collection; The inflection point intervention time point is the dividing point between the first recovery stage and the second recovery stage in the fatigue recovery trajectory. The fatigue decay rate in the first recovery stage is greater than the fatigue decay rate in the second recovery stage. Consolidation intervention time point: The consolidation intervention time point represents the time point in the fatigue recovery trajectory that meets the baseline stability condition; The baseline stability conditions include: the fatigue level remains within a preset tolerance range centered on the initial fatigue level, and in a sequence consisting of a preset number of consecutive sampling points, the absolute value of the change in fatigue level between adjacent sampling points does not exceed a preset fluctuation threshold.

[0181] Optionally, processing module 73 is used for: The guidance stage label is determined based on the predicted fatigue level corresponding to the intervention time point in the fatigue recovery trajectory. The guidance stage label is used to identify the current fatigue recovery stage. Based on the guidance stage label, a basic guidance template is selected from the preset guidance information database. The basic guidance template includes at least one suggestion for a fatigue recovery activity. Based on the predicted fatigue level, the parameters of the basic guidance template are adjusted to generate fatigue recovery guidance information corresponding to the fatigue recovery trajectory; By linking fatigue recovery guidance information to intervention time points, fatigue recovery guidance information can be pushed to plasma donors at the intervention time points.

[0182] The device provided in this application predicts the fatigue recovery trajectory after plasma donation by acquiring the stressor time series data of plasma donors and combining it with individual digital profiles characterizing their sensitivity to anticoagulant metabolism and fatigue recovery ability. This allows for the incorporation of individualized physiological response differences among different plasma donors into the modeling and prediction, making the predicted fatigue recovery process closer to the actual situation of the plasma donors. Based on this, intervention time points matching the plasma donor's state can be determined from the fatigue recovery trajectory, and targeted recovery guidance information can be pushed at the corresponding time points. This avoids the prediction bias and inaccurate intervention timing / content caused by existing technologies based on group averages and fixed rules, thereby improving the effectiveness of fatigue recovery management and the plasma donor experience.

[0183] Figure 8 To illustrate the hardware structure of an electronic device according to various embodiments of this application, the electronic device may include a processor 801 and a memory 802 storing computer program instructions. Specifically, the processor 801 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of this application.

[0184] Memory 802 may include mass storage for data or instructions. For example, and not limitingly, memory 802 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 802 may include removable or non-removable (or fixed) media. Where appropriate, memory 802 may be internal or external to an electronic device. In a particular embodiment, memory 802 may be a non-volatile solid-state memory.

[0185] In one embodiment, memory 802 may be read-only memory (ROM). In one embodiment, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0186] The processor 801 reads and executes computer program instructions stored in the memory 802 to implement any of the plasma donor fatigue recovery management methods in the above embodiments.

[0187] In one example, the electronic device may also include a communication interface 803 and a bus 810. For example, Figure 8 As shown, the processor 801, memory 802, and communication interface 803 are connected through bus 810 and complete communication with each other.

[0188] The communication interface 803 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0189] Bus 810 includes hardware, software, or both, that couples components of an electronic device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 810 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0190] Furthermore, in conjunction with the plasma donor fatigue recovery management method in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any of the plasma donor fatigue recovery management methods in the above embodiments.

[0191] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0192] The above description is merely a specific implementation example of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0193] Secondly, those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0194] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0195] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0196] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0197] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0198] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0199] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0200] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0201] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for managing fatigue recovery of plasma donors, characterized in that, include: Obtain time-series data on stressors and individual digital profiles of plasma donors; The individual digital profile represents the plasma donor's sensitivity to anticoagulant metabolism and fatigue recovery ability; Based on the stressor time series data and the individual digital profile, predict the fatigue recovery trajectory of the plasma donor after plasma donation; Based on the fatigue recovery trajectory, at least one intervention time point is determined, and fatigue recovery guidance information is pushed to the plasma donor at the intervention time point.

2. The method as described in claim 1, characterized in that, The process of obtaining the individual digital profile of the plasma donor includes: Acquire the historical plasma donation data of the plasma donor, which includes historical stress source time-series data and corresponding historical recovery response data from at least one historical plasma donation process; Based on the historical stress source time series data, key stress features characterizing stress intensity are extracted, including cumulative anticoagulant usage and standardized anticoagulant load. Based on the historical recovery response data, key fatigue recovery features that characterize the change of fatigue degree over time are extracted. These key fatigue recovery features include fatigue half-life and fatigue decay rate within a preset recovery time window. The key stress features and key fatigue recovery features are input into the trained physiological parameter prediction model to obtain the sensitivity and the fatigue recovery ability. Based on the sensitivity and fatigue recovery ability, an individual digital profile of the plasma donor is constructed.

3. The method as described in claim 2, characterized in that, The extraction of key stress features characterizing stress intensity based on the historical stressor time-series data includes: Based on the historical stress source time series data, the start and end times of historical plasma collection are determined, and the collection duration is determined based on the start and end times of plasma collection. The cumulative amount of anticoagulant used within the time range from the start time of the slurry collection to the end time of the slurry collection is determined based on the historical stress source time series data. The anticoagulant pumping rate time series is extracted from the historical stress source time series data, and the pumping rate fluctuation characteristics are calculated based on the anticoagulant pumping rate time series, wherein the pumping rate fluctuation characteristics include the sample standard deviation of the anticoagulant pumping rate time series. The standardized anticoagulant load is determined based on the cumulative amount of anticoagulant used, the slurry collection time, and the fluctuation characteristics of the pumping rate.

4. The method as described in claim 2, characterized in that, The extraction of key fatigue recovery features characterizing the change of fatigue degree over time based on the historical recovery response data includes: Based on the historical recovery response data, a recovery response sequence in which fatigue level changes over time is determined; Based on the recovery response sequence, a historical fatigue recovery curve is constructed, and the initial fatigue level at the end of slurry harvesting is determined according to the historical fatigue recovery curve. The fatigue half-life is determined based on the historical fatigue recovery curve, wherein the fatigue half-life is the time length corresponding to when the historical fatigue recovery curve drops to half of the initial fatigue level; Based on the decrease of the historical fatigue recovery curve within the preset recovery time window and the duration of the preset recovery time window, the fatigue decay rate within the preset recovery time window is determined, wherein the fatigue decay rate is used to characterize how quickly the fatigue level decreases over time within the preset recovery time window.

5. The method as described in claim 2, characterized in that, Before inputting the key stress features and key fatigue recovery features into the trained physiological parameter prediction model to obtain the sensitivity coefficient and the recovery rate coefficient, the method further includes: Multiple plasma donation data samples were acquired, including stress source time-series data samples and corresponding recovery response data samples. The actual sensitivity and actual fatigue recovery ability corresponding to each plasma donor sample were determined and used as label data. Key stress feature samples are extracted based on the stress source time series data samples, and key fatigue recovery feature samples are extracted based on the recovery response data samples; The key stress feature samples and the key fatigue recovery feature samples are combined into an input feature vector; Using the input feature vector as the model input and the label data as the model output, the initial physiological parameter prediction model is trained to obtain the trained physiological parameter prediction model.

6. The method as described in claim 5, characterized in that, The process of training an initial physiological parameter prediction model using the input feature vector as model input and the label data as model output to obtain the trained physiological parameter prediction model includes: The prediction sensitivity and fatigue recovery ability of the plasma donation data samples are predicted using the initial physiological parameter prediction model. The prediction error of the initial physiological parameter prediction model is calculated based on the predicted sensitivity, the predicted fatigue recovery ability, the actual sensitivity, and the actual fatigue recovery ability. The model parameters of the initial physiological parameter prediction model are iteratively updated based on the prediction error until the prediction error of the updated initial physiological parameter prediction model reaches the preset convergence condition, at which point the training is terminated and the trained physiological parameter prediction model is obtained.

7. The method as described in claim 1, characterized in that, Determining at least one intervention time point based on the fatigue recovery trajectory includes: Based on the fatigue recovery trajectory, at least one trajectory feature point is identified as the intervention time point; The trajectory feature points include at least one of the following: The intervention start time point is the time point corresponding to the first preset proportion of the fatigue level in the fatigue recovery trajectory when the fatigue level drops to the initial fatigue level at the end of the slurry collection. The inflection point intervention time point is the dividing point between the first recovery stage and the second recovery stage in the fatigue recovery trajectory, where the fatigue decay rate of the first recovery stage is greater than the fatigue decay rate of the second recovery stage. The consolidation intervention time point refers to the time point in the fatigue recovery trajectory that meets the baseline stability condition; The baseline stability conditions include: the fatigue level remains within a preset tolerance range centered on the initial fatigue level, and in a sequence of a preset number of consecutive sampling points, the absolute value of the change in fatigue level between adjacent sampling points does not exceed a preset fluctuation threshold.

8. The method as described in claim 1 or 7, characterized in that, The step of sending fatigue recovery guidance information to the plasma donor at the intervention time point includes: Based on the predicted fatigue level corresponding to the intervention time point in the fatigue recovery trajectory, a guidance stage label is determined, and the guidance stage label is used to identify the current fatigue recovery stage. A basic guidance template is selected from a preset guidance information database based on the guidance stage label. The basic guidance template includes at least one suggestion for a fatigue recovery activity. Based on the predicted fatigue level, the parameters of the basic guidance template are adjusted to generate fatigue recovery guidance information corresponding to the fatigue recovery trajectory; The fatigue recovery guidance information is linked to the intervention time point so that the fatigue recovery guidance information is pushed to the plasma donor at the intervention time point.

9. A fatigue recovery management device for plasma donors, characterized in that, It includes an acquisition module, a prediction module, and a processing module, wherein: The acquisition module is used to acquire time-series data of stressors and individual digital profiles of plasma donors; the individual digital profiles characterize the plasma donors' sensitivity to anticoagulant metabolism and fatigue recovery ability. The prediction module is used to predict the fatigue recovery trajectory of the plasma donor after plasma donation based on the stress source time series data and the individual digital profile. The processing module is used to determine at least one intervention time point based on the fatigue recovery trajectory, and push fatigue recovery guidance information to the plasma donor at the intervention time point.

10. An electronic device, characterized in that, include: The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the plasma donor fatigue recovery management method as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the plasma donor fatigue recovery management method as described in any one of claims 1 to 8.

12. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the plasma donor fatigue recovery management method according to any one of claims 1 to 8.

Images