Platelet-rich plasma collection amount evaluation method

Abstract

The invention discloses a platelet-rich plasma collection quantity evaluation method, and relates to the technical field of plasma collection quantity evaluation, and the method adopts real-time dynamic prediction and closed-loop control, and upgrades a traditional evaluation mode of statically detecting platelet concentration and leukocyte mixing quantity only after collection to real-time quality monitoring controllable in the whole process. The consistency and the accuracy of platelet rich plasma (PRP) collection are obviously improved; continuous collection data are smoothed and predicted through a Kalman filtering algorithm, the problem of over-collection or under-collection caused by manual judgment or a fixed threshold value in a traditional method is avoided, and the plasma quality and the stability of the curative effect are improved. A dynamic evaluation mechanism effectively reduces the synchronous enrichment risk of leukocytes in the PRP collection process; in an existing automatic collecting device, the leukocyte content and the platelet concentration are increased synchronously, so that pro-inflammatory factors in PRP are excessive, and tissue repair is inhibited.

Description

Technical Field

[0001] The invention relates to the technical field of plasma collection volume assessment, in particular to a method for assessing the platelet-rich plasma collection volume. Background Art

[0002] In clinical practice, platelet-rich plasma prepared by automated devices is often judged by platelet concentration and white blood cell content. This single quantitative method can easily mask the dynamic changes in the collection process and the differences in the biological activity of the product itself. The current consensus on the collection amount of each component of platelet-rich plasma is to recommend directly expressing the PRP concentration by platelet count and controlling the residual white blood cells at an extremely low level. However, in actual operations, there are often huge batch differences under the same device and the same parameters: some commercial kits can simultaneously enrich platelets and white blood cells by more than 4 times, resulting in abnormally increased levels of pro-inflammatory factors in PRP, ultimately inhibiting the repair metabolism of target tissues.

[0003] This evaluation method, which takes high concentration as good and low contamination as excellent, cannot reflect the actual output of key bioactive substances in PRP, such as PDGF-BB and TGF-β1. It also ignores the different requirements for residual leukocytes in different clinical scenarios. For example, the repair of intervertebral disc degeneration requires the control of leukocytes to reduce inflammation, while the orthopedics or cosmetology fields may require the participation of leukocytes in local immune regulation. Furthermore, single endpoint testing fails to provide real-time feedback during the collection process. Once the target threshold is exceeded, subsequent adjustments are difficult to remedy, and it is easy to reach the two extremes of insufficient or excessive collection.

[0004] Traditional remedial measures are mostly focused on post-collection filtration or secondary centrifugation, but they cannot solve the problem of component deviation that occurs at the initial stage of collection; more advanced centrifugation parameter optimization, gradient centrifugation or preset fixed collection volume solutions also face the dilemma of blind operation and difficulty in balancing individual differences; in other words, the existing evaluation standards provide the most basic quality threshold for PRP preparation, but in order to improve product consistency and adaptability while ensuring efficacy, it is necessary to shift from static concentration and purity to dynamic real-time quality evaluation and clinical scenario customization. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a method for assessing the amount of platelet-rich plasma collected to solve the problem that traditional remedial measures mostly focus on filtration or secondary centrifugation after collection, but cannot solve the problem of component deviation that occurs at the initial stage of collection.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] An embodiment of the present invention provides a method for evaluating the collection volume of platelet-rich plasma, which includes:

[0009] Step S1: Use an automated apheresis machine certified as a Class III medical device in China to collect blood in a fully enclosed state and measure the component concentrations in the blood in real time, including platelet concentration Plt and white blood cell concentration WBC.

[0010] Step S2: Use the Kalman filter algorithm to perform dynamic smoothing and short-term prediction on the continuously collected data in Step S1 to obtain the time series change curves of Plt concentration and WBC concentration during the collection process.

[0011] Step S3: Set the target quality range, where the Plt concentration is 500 - 1000×10^9 / L and the WBC concentration < 1×10^8 / L.

[0012] Step S4: Automatically trigger the stop collection instruction when the prediction curves first simultaneously meet the target quality range.

[0013] Step S5: Upload the data of the entire collection process and the patient's baseline Plt concentration to the in-hospital PRP quality management database, and automatically generate a quality report including collection volume, Plt concentration, WBC mixing rate, and dynamic prediction error information.

[0014] Step S6: Automatically update the prediction model parameters according to the report results, and continuously iterate and optimize for different patients and clinical application scenarios.

[0015] As a preferred solution of the method for evaluating the collection volume of platelet-rich plasma according to the present invention, wherein: in Step S2, the step of using the Kalman filter algorithm to perform dynamic smoothing and short-term prediction on the continuously collected data in Step S1 is as follows:

[0016] Define the state vector x k and the state transition matrix F as:

[0017]

[0018] wherein, x k represents the state vector at time k, Plt k represents the platelet concentration at time k, WBC k represents the white blood cell concentration at time k, represents the instantaneous change rate of the platelet concentration at time k, represents the instantaneous change rate of the white blood cell concentration at time k, Δt represents the time interval between adjacent sampling times, and k is the sampling sequence number;

[0019] Define the observation matrix H as:

[0020]

[0021] Among them, H is used to extract the platelet and white blood cell concentrations from the state vector, and the 1s and 0s in the matrix are fixed constants;

[0022] Construct the process noise covariance matrix Q and the measurement noise covariance matrix R as follows:

[0023]

[0024] Among them, Q represents the process noise covariance matrix, and q is the process noise intensity scalar. represents the platelet measurement noise variance. represents the white blood cell measurement noise variance;

[0025] In the filtering prediction step, the prediction formula is:

[0026]

[0027] Among them, represents the predicted state estimate at time k. represents the updated state estimate at time k - 1, P k|k-1 represents the predicted error covariance at time k, P k-1|k-1 represents the updated error covariance at time k - 1. is the transpose of the state transition matrix.

[0028] The filtering update step is defined as:

[0029]

[0030] Among them, K k represents the Kalman gain matrix at time k. is the transpose of the observation matrix, z k represents the actual measurement vector at time k, including platelet and white blood cell concentrations, and I is the identity matrix with the same dimension as the state vector. represents the updated state estimate at time k, P k|k represents the updated error covariance matrix at time k.

[0031] As a preferred solution of the method for evaluating the collection volume of platelet-rich plasma described in the present invention, among them: in step S2, the step of using the Kalman filtering algorithm to perform dynamic smoothing and short-term prediction on the continuously collected data in step S1 further includes defining a prediction formula:

[0032]

[0033] Among them, represents the state estimate predicted m steps after time k, F mDenotes the m-th power of the state transition matrix, where m is the prediction step, and P k+m|k Denotes the error covariance after predicting m steps, and F i Denotes the i-th power of the state transition matrix, where i ranges from 0 to m - 1, and the summation symbol represents the contribution of the cumulative process noise to the error covariance;

[0034] The normalized mean square error NRMSE is used to evaluate the prediction performance, and the calculation formula of NRMSE is:

[0035]

[0036] Where, Denotes the predicted value at time k, and y k Denotes the actual measured value at time k, N represents the total number of sampled data, and y max And y min Represent the maximum and minimum values of the actual data during the acquisition process respectively.

[0037] As a preferred scheme of the platelet-rich plasma collection amount evaluation method described in the present invention, wherein: in step S4, based on the Kalman filter prediction result, it is judged whether the acquisition data first stably enters the target quality interval, so as to automatically trigger the stop acquisition instruction;

[0038] In step S4, the determination logic and process are:

[0039] Obtain the predicted value at time k from the Kalman filter output And Construct a binary index function: if And Then I k = 1, otherwise I k = 0, where I k Represents whether the predicted value at the sampling time k meets the target interval, Represents the platelet concentration predicted at time k, Represents the white blood cell concentration predicted at time k;

[0040] Design a continuous satisfaction condition counter:

[0041] If I k = 1, then C k = C k-1 + 1, if I k = 0, then C k = 0;

[0042] Where, C k Represents the number of sampling times that continuously meet the target interval since the last interruption, C k-1 Is the count value at the previous moment, and the initial condition is set to C0 = 0. When C kWhen the preset threshold n is reached for the first time, record this moment as k * , it is determined that the prediction curve has stably entered the target range, that is, the stop collection process is triggered.

[0043] As a preferred solution of the method for evaluating the collection volume of platelet-rich plasma according to the present invention, wherein: in step S4, the stop collection instruction is constructed and sent in the form of a data packet, and its format is defined as:

[0044] M = {ID, CMD, T},

[0045] wherein, ID represents the unique identifier of the collection device, CMD is the stop collection instruction code, preset as a constant S, T is the timestamp when the instruction is sent, and the data packet is transmitted through the encryption interface.

[0046] As a preferred solution of the method for evaluating the collection volume of platelet-rich plasma according to the present invention, wherein: in step S4, a fault tolerance mechanism and a failure fallback strategy are set:

[0047] Define a retry counter r and a maximum retry number R, and retry when the stop collection instruction fails to be sent:

[0048] If the sending fails, then r = r + 1, if the sending is successful, then r = 0,

[0049] At the same time, set a failure flag:

[0050] If the sending is successful, then FF = 0, if it still fails after r ≥ R, then FF = 1;

[0051] If FF = 1, automatically record the fault log, including the failure time, device status and error code, and start the manual intervention process as a fallback strategy.

[0052] As a preferred solution of the method for evaluating the collection volume of platelet-rich plasma according to the present invention, wherein: in step S5, the method for generating the quality report is:

[0053] Calculate the collection volume V through the data during the collection process: wherein, v k represents the blood volume collected during the k-th time period, N represents the total number of sampling time periods, and V represents the total collection volume;

[0054] During the collection process, the actually measured platelet concentration and white blood cell concentration are respectively recorded as Plt k and WBC k , at the same time, the predicted concentrations obtained by Kalman filtering are respectively and The dynamic prediction error is defined as:

[0055]

[0056] Among them, represents the predicted platelet concentration at time k, represents the predicted white blood cell concentration at time k, and are the prediction errors of platelets and white blood cells respectively;

[0057] The overall prediction performance is evaluated by the root mean square error RMSE, and the RMSE calculation formula is:

[0058]

[0059] Among them, N represents the total number of sampled data, RMSE Plt and RMSE WBC represent the root mean square errors of platelets and white blood cells respectively;

[0060] The white blood cell mixing rate η is defined as the ratio of the white blood cell concentration to the platelet concentration in the blood, and its calculation formula is:

[0061]

[0062] Among them, WBC avg represents the average value of the white blood cell concentration at all sampling times, Plt avg represents the average value of the platelet concentration at all sampling times, η represents the mixing rate, and the unit is percentage.

[0063] As a preferred solution of the platelet-rich plasma collection amount evaluation method described in the present invention, wherein: in step S5, the concentration data is standardized by the min-max normalization method, and its formula is:

[0064]

[0065] Among them, y k represents the original concentration data, taking Plt k or WBC k , y min and y max represent the minimum and maximum concentration values during the collection respectively, represents the standardized data, and the standardized platelet and white blood cell concentration curves are respectively denoted as and

[0066] Package each field of the report into a data packet and connect to the HIS system through a predetermined interface. The data packet format is defined as: Among them, ID represents the unique identifier of the collection device, T represents the timestamp at the end of the collection, and D represents the report data packet;

[0067] The fields in the data packet respectively correspond to the collection volume, the standardized concentration curve, the prediction error index, and the mixing rate. The interface adopts an encrypted transmission protocol and is docked with the hospital information system HIS.

[0068] As a preferred solution of the platelet-rich plasma collection volume evaluation method described in the present invention, wherein: in step S6, the method of automatically updating the prediction model parameters according to the report result is:

[0069] Construct a training set D within the update period, defined as: where y i represents the actual measurement value of the i-th sample, taken from the blood component data in the collection report, and φ i represents the feature vector of the i-th sample, including various input information related to the prediction. N0 represents the total number of sampled data in this training period;

[0070] The root mean square error RMSE and the mean absolute percentage error MAPE are used to evaluate the model performance. The formula for the root mean square error is:

[0071]

[0072] where θ represents the parameter vector of the current prediction model, represents the model prediction value of the i-th sample;

[0073] The formula for the mean absolute percentage error MAPE is:

[0074]

[0075] where the percentage represents the ratio of the error to the actual value;

[0076] To achieve incremental update, the online recursive least squares method is adopted, and the update formula is defined as:

[0077]

[0078] where θ k represents the model parameter vector before the k-th update, y k is the actual measurement value of the k-th sample, φ k is the feature vector of the k-th sample, and θ k+1 represents the updated model parameter vector;

[0079] The update gain K k The calculation formula is:

[0080]

[0081] where P θ,kis the parameter estimation error covariance matrix before the k-th update, λ is the forgetting factor, with a value range in (0,1], and P in the numerator θ,k φ k represents the contribution of the current sample to the covariance;

[0082] The error covariance matrix update formula is:

[0083]

[0084] where, P θ,k+1 represents the updated error covariance matrix;

[0085] After the update is completed, the final model parameter is denoted as θ final , and the updated data and evaluation indicators are transmitted to the HIS system in the form of data packets. The data packet is defined as: U = {ID, T, θ final , RMSE, MAPE}, where ID represents the unique identifier of the acquisition device, T represents the timestamp when the parameter update is completed, and U represents the update data packet.

[0086] As a preferred solution of the method for evaluating the collection amount of platelet-rich plasma described in the present invention, wherein: the target quality interval is set for different patients and clinical application scenarios.

[0087] The beneficial effects of the present invention are as follows: The present invention adopts real-time dynamic prediction and closed-loop control, upgrades the traditional evaluation method of only statically detecting the platelet concentration and white blood cell mixing amount after the collection to a real-time quality monitoring with the whole process controllable, and significantly improves the consistency and accuracy of platelet-rich plasma PRP collection; smooths and predicts continuous collection data through the Kalman filter algorithm, avoids the over-collection or under-collection problems caused by manual judgment or fixed thresholds in the traditional method, and improves the stability of plasma quality and efficacy; the dynamic evaluation mechanism effectively reduces the risk of synchronous enrichment of white blood cells during the PRP collection process; in the existing automated collection devices, the white blood cell content often increases synchronously with the platelet concentration, resulting in excessive pro-inflammatory factors in PRP and inhibiting tissue repair; the present invention controls the white blood cell mixing amount within <1×10^8 / L through real-time prediction curve judgment, achieving double optimization of purity and efficacy; at the same time, it has the function of automatically recording and reporting the whole process data, and constructs a whole-process traceable system for PRP quality; key information such as collection amount, concentration time series, dynamic prediction error, and white blood cell mixing rate are structured and encapsulated and uploaded to the hospital management system, providing a basis for clinical decision-making and subsequent auditing. Description of the Drawings

[0088] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0089] Figure 1 It is a schematic flow chart of the method for evaluating the collection volume of platelet-rich plasma of the present invention. Specific Embodiments

[0090] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will provide a detailed description of the specific embodiments of the present invention with reference to the accompanying drawings of the specification.

[0091] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0092] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The phrase "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it an individual or alternative embodiment that mutually excludes other embodiments.

[0093] Embodiment 1, referring to Figure 1 , this embodiment provides a method for evaluating the collection volume of platelet-rich plasma, including the following steps:

[0094] Step S1, using an automated apheresis machine certified as a Class III medical device in China, collect blood in a fully enclosed state and measure the component concentrations in the blood in real time, including the platelet concentration Plt and the white blood cell concentration WBC;

[0095] Step S2, using the Kalman filter algorithm to perform dynamic smoothing and short-term prediction on the continuously collected data in Step S1, and obtain the time series change curves of the Plt concentration and the WBC concentration during the collection process;

[0096] In Step S2, the steps of using the Kalman filter algorithm to perform dynamic smoothing and short-term prediction on the continuously collected data in Step S1 are as follows:

[0097] Define the state vector x k and the state transition matrix F as:

[0098]

[0099] where x k represents the state vector at time k, Plt k represents the platelet concentration at time k, WBC k represents the white blood cell concentration at time k, represents the instantaneous change rate of the platelet concentration at time k, represents the instantaneous change rate of the white blood cell concentration at time k, Δt represents the time interval between adjacent sampling times, and k is the sampling sequence number;

[0100] Define the observation matrix H as:

[0101]

[0102] where H is used to extract the platelet and white blood cell concentrations from the state vector, and the 1s and 0s in the matrix are fixed constants;

[0103] Construct the process noise covariance matrix Q and the measurement noise covariance matrix R as:

[0104]

[0105] where Q represents the process noise covariance matrix, q is the process noise intensity scalar, represents the platelet measurement noise variance, represents the white blood cell measurement noise variance;

[0106] In the filtering prediction step, the prediction formula is:

[0107]

[0108] where, represents the predicted state estimate at time k, represents the updated state estimate at time k - 1, P k|k-1 represents the predicted error covariance at time k, P k-1|k-1 represents the updated error covariance at time k - 1, is the transpose of the state transition matrix,

[0109] The filtering update step is defined as:

[0110]

[0111] P k|k =(I - K k H)P k|k-1 ,

[0112] where K k represents the Kalman gain matrix at time k, is the transpose of the observation matrix, z kDenote the actual measurement vector at time k, which includes platelet and white blood cell concentrations. I is the identity matrix with the same dimension as the state vector. Denote the updated state estimate at time k, P k|k Denote the updated error covariance matrix at time k;

[0113] In step S2, the steps of using the Kalman filter algorithm to perform dynamic smoothing and short-term prediction on the continuously collected data in step S1 further include defining a prediction formula:

[0114]

[0115] where, Denote the state estimate predicted m steps after time k, F m Denote the m-th power of the state transition matrix, where m is the number of prediction steps, P k+m|k Denote the error covariance predicted m steps after, F i Denote the i-th power of the state transition matrix, where i ranges from 0 to m - 1, and the summation symbol represents the contribution of the cumulative process noise to the error covariance;

[0116] Use the normalized mean squared error NRMSE to evaluate the prediction performance. The NRMSE calculation formula is:

[0117]

[0118] where, Denote the predicted value at time k, y k Denote the actual measured value at time k, N represents the total number of sampled data, y max and y min respectively represent the maximum and minimum values of the actual data during the acquisition process;

[0119] Specifically, construct a four-dimensional state vector including platelet and white blood cell concentrations and their rates of change here, and use the Kalman filter algorithm to achieve dynamic smoothing and short-term prediction of continuously sampled data;

[0120] The state transition matrix describes the evolution of the sampled data over time, and the observation matrix is used to extract the actual measured data. The process noise and measurement noise covariance matrices play a key role in the filtering effect and can adjust the sensitivity of the system to random fluctuations. The predicted state estimate and error covariance are calculated separately in the prediction stage and the update stage, and the dynamic correction of the observed data is achieved by combining the Kalman gain. The short-term prediction formula applies the state transition matrix multiple times to the updated state to obtain the estimate of the future state;

[0121] The normalized mean squared error is used as an evaluation index, which reflects the prediction performance of the model in different sampling intervals, takes into account the requirements of data smoothing and short-term prediction, and provides reliable timing information for subsequent automatic acquisition control;

[0122] Step S3, set the target quality range, where the Plt concentration is 500 - 1000×10^9 / L and the WBC concentration < 1×10^8 / L;

[0123] Step S4, automatically trigger the stop acquisition instruction when the prediction curve first satisfies the target quality range simultaneously;

[0124] In step S4, based on the Kalman filter prediction result, determine whether the acquired data first stably enters the target quality range, so as to automatically trigger the stop acquisition instruction;

[0125] In step S4, the determination logic and process are as follows:

[0126] Obtain the predicted value at time k from the Kalman filter output and Construct a binary index function: If and then I k = 1, otherwise I k = 0, where I k represents whether the predicted value at the sampling time k satisfies the target range, represents the predicted platelet concentration at time k, represents the predicted white blood cell concentration at time k;

[0127] Design a continuous satisfaction condition counter:

[0128] If I k = 1, then C k = C k-1 + 1, if I k = 0, then C k = 0;

[0129] where C k represents the number of sampling times that continuously satisfy the target range since the last interruption, C k-1 is the count value at the previous moment, and the initial condition is set to C0 = 0. When C k first reaches the preset threshold n, record this moment as k * , determine that the prediction curve stably enters the target range, that is, trigger the stop acquisition process;

[0130] In step S4, the stop acquisition instruction is constructed and sent in the form of a data packet, and its format is defined as:

[0131] M = {ID, CMD, T},

[0132] where ID represents the unique identifier of the acquisition device, CMD is the stop acquisition instruction code, preset as the constant S, and T is the timestamp when the instruction is sent. The data packet is transmitted through the encryption interface;

[0133] In step S4, set up a fault tolerance mechanism and a failure fallback strategy:

[0134] Define a retry counter r and a maximum retry count R, and retry when the stop collection instruction fails to be sent:

[0135] If the sending fails, then r = r + 1; if the sending is successful, then r = 0.

[0136] Meanwhile, set a failure flag:

[0137] If the sending is successful, then FF = 0; if it still fails after r ≥ R, then FF = 1.

[0138] If FF = 1, automatically record the fault log, including the failure time, device status, and error code, and start an artificial intervention process as the fallback strategy;

[0139] Specifically, here a stable determination logic is formed by a binary index function and a continuous counter. The stop collection is triggered only when the collected data meets the target quality range. A standardized stop collection instruction data packet is constructed to achieve seamless docking with the device and the in - hospital control system. The fault tolerance mechanism introduces retry counting and a failure flag to effectively handle network or device communication anomalies, and starts artificial intervention after the retry limit is exceeded;

[0140] In step S5, upload all the data during the entire collection process and the patient's basic Plt concentration to the in - hospital PRP quality management database, and automatically generate a quality report including the collection volume, Plt concentration, WBC mixing rate, and dynamic prediction error information;

[0141] In step S5, the way to generate the quality report is as follows:

[0142] Calculate the collection volume V through the data during the collection process: where, v k represents the blood volume collected in the k - th time period, N represents the total number of sampling time periods, and V represents the total collection volume;

[0143] During the collection process, the actually measured platelet concentration and white blood cell concentration are respectively denoted as Plt k and WBC k , and meanwhile, the predicted concentrations obtained by Kalman filtering are respectively and The dynamic prediction error is defined as:

[0144]

[0145] where, represents the predicted platelet concentration at time k, represents the predicted white blood cell concentration at time k, and They are the prediction errors of platelets and white blood cells respectively;

[0146] The overall prediction performance is evaluated by the root mean square error RMSE, and the calculation formula of RMSE is:

[0147]

[0148] Among them, N represents the total number of sampling data, RMSE Plt and RMSE WBC represent the root mean square errors of platelets and white blood cells respectively;

[0149] The white blood cell mixing rate η is defined as the ratio of the white blood cell concentration to the platelet concentration in the blood, and its calculation formula is:

[0150]

[0151] Among them, WBC avg represents the average value of the white blood cell concentration at all sampling times, Plt avg represents the average value of the platelet concentration at all sampling times, η represents the mixing rate, and the unit is percentage;

[0152] In step S5, the minimum-maximum normalization method is used to standardize the concentration data, and its formula is:

[0153]

[0154] Among them, y k represents the original concentration data, taking Plt k or WBC k , y min and y max represent the minimum and maximum concentration values during the collection respectively, represents the data after standardization, and the standardized platelet and white blood cell concentration curves are respectively recorded as and

[0155] Package each field of the report into a data packet and connect to the HIS system through a predetermined interface. The data packet format is defined as: Among them, ID represents the unique identifier of the collection device, T represents the timestamp at the end of the collection, and D represents the report data packet;

[0156] The fields in the data packet respectively correspond to the collection volume, the standardized concentration curve, the prediction error index, and the mixing rate. The interface uses an encrypted transmission protocol and is connected to the hospital information system HIS;

[0157] Specifically, a quality report is generated here by comprehensively integrating the collected data, recording key information such as the collection volume, concentration time series, dynamic prediction error, and white blood cell mixing rate;

[0158] Step S6, automatically update the prediction model parameters according to the report results, and continuously iterate and optimize for different patients and clinical application scenarios;

[0159] In step S6, the method of automatically updating the prediction model parameters according to the report results is as follows:

[0160] Construct a training set D within the update period, defined as: where y i represents the actual measured value of the i-th sample, taken from the blood component data in the collection report, and φ i represents the feature vector of the i-th sample, containing various input information related to the prediction, and N0 represents the total number of sampled data in this training period;

[0161] Use the root mean square error RMSE and the mean absolute percentage error MAPE to evaluate the model performance. The formula for the root mean square error is:

[0162]

[0163] where θ represents the parameter vector of the current prediction model, represents the model prediction value of the i-th sample;

[0164] The formula for the mean absolute percentage error MAPE is:

[0165]

[0166] where the percentage represents the ratio of the error to the actual value;

[0167] To achieve incremental update, use the online recursive least squares method, and the update formula is defined as:

[0168]

[0169] where θ k represents the model parameter vector before the k-th update, y k is the actual measured value of the k-th sample, φ k is the feature vector of the k-th sample, and θ k+1 represents the updated model parameter vector;

[0170] Update gain K k The calculation formula is:

[0171]

[0172] where Pθ,k is the parameter estimation error covariance matrix before the k-th update, λ is the forgetting factor, and its value range is in (0, 1]. In the numerator, P θ,k φ k represents the contribution of the current sample to the covariance;

[0173] The error covariance matrix update formula is:

[0174]

[0175] where P θ,k+1 represents the updated error covariance matrix;

[0176] After the update is completed, the final model parameter is denoted as θ final , and the updated data and evaluation indicators are transmitted to the HIS system in the form of a data packet. The data packet is defined as: U = {ID, T, θ final , RMSE, MAPE}, where ID represents the unique identifier of the acquisition device, T represents the timestamp when the parameter update is completed, and U represents the update data packet;

[0177] Specifically, for the data after the acquisition cycle ends, the automatic update of the model parameters is realized by constructing the training set D;

[0178] Each sample collected during the training cycle consists of the actual measurement value and the feature vector. The performance of the model is evaluated by the root mean square error and the mean absolute percentage error. The online recursive least squares method is used to gradually update the parameters. During the update process, the incremental update formula is used to enable each new data to be immediately incorporated into the model. At the same time, the error covariance matrix reflects the uncertainty of the update process, and the forgetting factor is introduced to gradually attenuate the influence of historical data, thereby enhancing the response ability of the model to the latest data;

[0179] The target quality interval is set for different patients and clinical application scenarios.

[0180] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for evaluating the collection volume of platelet-rich plasma, characterized in that: including Step S1: Collect blood in a fully enclosed state and measure the concentration of components in the blood in real time, including platelet concentration Plt and white blood cell concentration WBC; Step S2: Use the Kalman filter algorithm to perform dynamic smoothing and short-term prediction on the continuously collected data in Step S1 to obtain the time series change curves of Plt concentration and WBC concentration during the collection process; Step S3: Set the target quality range, where the Plt concentration is 500 - 1000×10^9 / L and the WBC concentration < 1×10^8 / L; Step S4: Automatically trigger the stop collection instruction when the prediction curve first simultaneously meets the target quality range; Step S5: Upload the whole process data of the collection and the patient's basic Plt concentration to the in-hospital PRP quality management database, and automatically generate a quality report containing information such as the collection volume, Plt concentration, WBC mixing rate, and dynamic prediction error; Step S6: Automatically update the prediction model parameters according to the report results, and continuously iterate and optimize for different patients and clinical application scenarios.

2. The platelet-rich plasma collection volume evaluation method according to claim 1, characterized in that: In Step S2, the steps of using the Kalman filter algorithm to perform dynamic smoothing and short-term prediction on the continuously collected data in Step S1 are as follows: Define the state vector x k and the state transition matrix F as: where, x k represents the state vector at time k, Plt k represents the platelet concentration at time k, WBC k represents the white blood cell concentration at time k, represents the instantaneous change rate of the platelet concentration at time k, represents the instantaneous change rate of the white blood cell concentration at time k, Δt represents the time interval between adjacent sampling times, and k is the sampling sequence number; Define the observation matrix H as: where H is used to extract the platelet and white blood cell concentrations from the state vector, and the 1s and 0s in the matrix are fixed constants; Construct the process noise covariance matrix Q and the measurement noise covariance matrix R as: where Q represents the process noise covariance matrix, and q is the process noise intensity scalar, represents the variance of platelet measurement noise, represents the variance of white blood cell measurement noise; In the filtering prediction step, the prediction formula is: Among them, represents the predicted state estimate at time k, represents the updated state estimate at time k-1, P k|k-1 represents the predicted error covariance at time k, P k-1|k-1 represents the updated error covariance at time k-1, is the transpose of the state transition matrix, The filtering update step is defined as: P k|k = (I - K k H)P k|k-1 , Among them, K k represents the Kalman gain matrix at time k, is the transpose of the observation matrix, z k represents the actual measurement vector at time k, including platelet and white blood cell concentrations, I is the identity matrix with the same dimension as the state vector, represents the updated state estimate at time k, P k|k represents the updated error covariance matrix at time k.

3. The method for evaluating the collection volume of platelet-rich plasma according to claim 2, wherein: In Step S2, the steps of using the Kalman filter algorithm to perform dynamic smoothing and short-term prediction on the continuously collected data in Step S1 also include defining the prediction formula: Among them, represents the state estimate m steps after prediction from time k, and F m represents the m-th power of the state transition matrix, where m is the number of prediction steps, and P k+m|k represents the error covariance m steps after prediction, and F i represents the i-th power of the state transition matrix, where i ranges from 0 to m - 1, and the summation symbol represents the contribution of the cumulative process noise to the error covariance; Use the normalized root mean square error NRMSE to evaluate the prediction performance, and the NRMSE calculation formula is: Among them, represents the predicted value at time k, y k represents the actual measured value at time k, N represents the total number of sampled data, y max and y min respectively represent the maximum and minimum values of the actual data during the acquisition process.

4. The platelet-rich plasma collection volume evaluation method according to claim 1, wherein: In Step S4, based on the Kalman filter prediction results, determine whether the collected data first stably enters the target quality range, so as to automatically trigger the stop collection instruction; In Step S4, the determination logic and process are: Obtain the predicted value at time k from the Kalman filter output and Construct a binary index function: If and then I k = 1, otherwise I k = 0, where I k indicates whether the predicted value at the sampling time k satisfies the target interval, represents the platelet concentration predicted at time k, represents the white blood cell concentration predicted at time k; Design a continuous condition satisfaction counter: If I k = 1, then C k = C k-1 + 1, if I k = 0, then C k = 0; Among them, C k represents the number of sampling times that continuously meet the target interval since the last interruption. C k-1 is the count value at the previous moment, and the initial condition is set as C0 = 0. When C k first reaches the preset threshold n, record this moment as k * , it is determined that the prediction curve stably enters the target interval, that is, the stop-sampling process is triggered.

5. The method for evaluating the collection volume of platelet-rich plasma according to claim 4, characterized in that: In Step S4, the stop collection instruction is constructed and sent in the form of a data packet, and its format is defined as: M = {ID, CMD, T}, where ID represents the unique identifier of the collection device, CMD is the stop collection instruction code, preset as the constant S, T is the timestamp when the instruction is sent, and the data packet is transmitted through an encrypted interface.

6. The method for evaluating the collection volume of platelet-rich plasma according to claim 5, wherein: In Step S4, set the fault tolerance mechanism and the failure fallback strategy: Define the retry counter r and the maximum retry times R, and retry when the stop collection instruction fails to be sent: If the sending fails, then r = r + 1, if the sending is successful, then r = 0, At the same time, set the failure flag: If the sending is successful, then FF = 0, if it still fails after r ≥ R, then FF = 1; If FF = 1, automatically record the fault log, including the failure time, device status, and error code, and start the manual intervention process as the fallback strategy.

7. The method for evaluating the collection volume of platelet-rich plasma according to claim 1, wherein: In Step S5, the method of generating the quality report is: Calculate the collection volume V from the data during the collection process: where v k represents the blood volume collected during the k-th time period, N represents the total number of sampling time periods, and V represents the overall collection volume; During the acquisition process, the actually measured platelet concentration and white blood cell concentration are respectively denoted as Plt k and WBC k . Meanwhile, the predicted concentrations obtained by Kalman filtering are respectively and The dynamic prediction error is defined as: Among them, represents the predicted platelet concentration at time k, represents the predicted white blood cell concentration at time k, and are the prediction errors of platelets and white blood cells respectively; The overall prediction performance is evaluated using the root mean square error RMSE, and the RMSE calculation formula is: where N represents the total number of sampled data, and RMSE Plt and RMSE WBC respectively represent the root mean square errors of platelets and white blood cells; The white blood cell mixing rate η is defined as the ratio of white blood cells to platelet concentration in the blood, and its calculation formula is: Among them, WBC avg represents the average white blood cell concentration at all sampling times, and Plt avg represents the average platelet concentration at all sampling times. η represents the mixing rate, and the unit is percentage.

8. The method for evaluating the collection volume of platelet-rich plasma according to claim 7, characterized in that: In step S5, the concentration data is standardized using the min-max normalization method, and its formula is: Among them, y k represents the original concentration data, taking Plt k or WBC k , y min and y max respectively represent the minimum and maximum concentration values during collection, represents the data after standardization. The standardized platelet and white blood cell concentration curves are respectively denoted as and Encapsulate each field of the report into a data packet and interface with the HIS system through a predetermined interface. The data packet format is defined as: Among them, ID represents the unique identifier of the acquisition device, T represents the timestamp at the end of the acquisition, and D represents the report data packet; The fields in the data packet respectively correspond to the acquisition quantity, the standardized concentration curve, the prediction error index, and the mixing ratio. The interface uses an encrypted transmission protocol and is docked with the hospital information system HIS.

9. The method for evaluating the collection volume of platelet-rich plasma according to claim 8, wherein: In step S6, the method of automatically updating the prediction model parameters according to the report results is: Construct the training set D within the update period, defined as: where y i represents the actual measured value of the i-th sample, taken from the blood component data in the acquisition report, and φ i represents the feature vector of the i-th sample, containing various input information related to the prediction, and N0 represents the total number of sampled data in this training period; The root mean square error RMSE and the mean absolute percentage error MAPE are used to evaluate the model performance. The formula for the root mean square error is: where θ represents the parameter vector of the current prediction model, represents the model prediction value of the i-th sample; The formula for the mean absolute percentage error MAPE is: Among them, the percentage represents the ratio of the error to the actual value; To achieve incremental update, the online recursive least squares method is used, and the update formula is defined as: Among them, θ k represents the model parameter vector before the k-th update, y k is the actual measured value of the k-th sample, φ k is the feature vector of the k-th sample, and θ k+1 represents the updated model parameter vector; Updated gain K k The calculation formula is as follows: Among them, P θ,k is the parameter estimation error covariance matrix before the k-th update, λ is the forgetting factor, and its value range is in (0, 1]. In the numerator, P θ,k φ k represents the contribution of the current sample to the covariance; The error covariance matrix update formula is: where, P θ,k+1 represents the updated error covariance matrix; After the update is completed, the final model parameters are denoted as θ final , the updated data and evaluation metrics are transmitted to the HIS system in the form of data packets, and the data packet is defined as: U = {ID, T, θ final , RMSE, MAPE}, where D represents the unique identifier of the acquisition device, T represents the timestamp when the parameter update is completed, and U represents the update data packet.

10. The platelet-rich plasma collection volume assessment method according to claim 1, wherein: The target quality interval is set for different patients and clinical application scenarios.

Images