HCT prediction method and system, device, and medium

Abstract

Disclosed in the present application are an HCT prediction method and system, a device, and a medium. Photons in a photon group are emitted into an extracorporeal circulation blood model so that the photons are reflected, scattered, or absorbed during movement. The extracorporeal circulation blood model is a model constructed on the basis of the thickness of an extracorporeal circulation tube, the thickness of the blood, and the distance from a light source generator to a detector. Light intensity data corresponding to photons reaching the detector are acquired. The acquired light intensity data are input into a prediction model to obtain a hematocrit (HCT) prediction value. The prediction model is a model constructed on the basis of a BP neural network and the relationship between light intensity and HCT. The prediction model combining simulation and neural network increases the data volume, achieves continuous non-invasive estimation of HCT by means of algorithm model optimization, and improves the accuracy of HCT prediction.

Description

A method, system, device and medium for predicting HCT

[0001] This application claims priority to Chinese Patent Application No. 202410554655.7, filed on May 7, 2024, entitled "An HCT Prediction Method, System, Device and Medium", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of extracorporeal membrane oxygenation (ECMO) system development, and in particular to an HCT prediction method, system, device, and medium. Background Technology

[0003] Non-invasive methods, due to their elimination of the need for physical penetration into tissues or blood collection for ex vivo analysis, exhibit significant advantages in reducing the risk of thrombosis and enabling continuous real-time measurement. Among these, the application of near-infrared spectroscopy provides an effective means for non-invasive hematocrit (HCT) monitoring. Near-infrared spectroscopy measures the intensity change of light passing through blood and uses the relationship between light intensity and HCT to design monitoring algorithms. After receiving light passing through the blood, the sensor detection module acquires the light intensity value and then estimates the HCT value based on mathematical models such as Twersky's theory. This method not only avoids the invasive procedures of traditional methods but also enables real-time monitoring, providing crucial reference information for clinical diagnosis and treatment.

[0004] However, non-invasive methods require precise sensors and a stable measurement environment for long-term continuous monitoring to obtain sufficient data for analysis. Since the amount of data available in actual calculations is relatively small, and HCT values ​​are typically calculated through mathematical formula fitting, optimization and improvement become difficult. Limited data may not adequately reflect individual differences and dynamic changes, limiting the accuracy and generalization ability of the predictive model. Previous studies on optical hematocrit measurement based on Twersky's theory also found an error of approximately 5%, indicating that the accuracy of HCT monitoring still needs improvement.

[0005] Summary of the Invention

[0006] This application provides an HCT prediction method, system, device, and medium to improve the accuracy of HCT prediction.

[0007] To address the above problems, the technical solutions provided in this application are as follows:

[0008] The first aspect of this application provides an HCT prediction method, including:

[0009] Photons from the photon array are emitted into an extracorporeal blood model so that the photons are reflected, scattered, or absorbed during their movement. The extracorporeal blood model is a model constructed based on the thickness of the extracorporeal circulation tubing, the thickness of the blood, and the distance between the light source generator and the detector.

[0010] Obtain the light intensity data corresponding to the photons arriving at the detector;

[0011] The acquired light intensity data is input into the prediction model to obtain the HCT prediction value. The prediction model is based on the BP neural network and the relationship between light intensity and HCT.

[0012] In one possible implementation, before acquiring the light intensity data corresponding to the photons arriving at the detector, the method further includes:

[0013] In response to the photon being reflected during its movement, the variables of the photon are adjusted;

[0014] The acquisition of light intensity data corresponding to photons arriving at the detector includes:

[0015] If the variable of the photon does not meet the end tracking condition, the photon is tracked until it reaches the detector, the light intensity data corresponding to the photon is acquired, and the tracking of the photon is terminated.

[0016] The process of emitting photons from the photon group into the extracorporeal blood model and subsequent steps continues until the number of photons tracked meets the first threshold.

[0017] In one possible implementation, prior to adjusting the variable of the photon in response to its reflection during movement, the method further includes:

[0018] Based on the photon's movement step size and scattering direction, it is determined whether the photon collides with the blood boundary and is reflected. The movement step size is determined based on the absorption coefficient, scattering coefficient, and random variables. The scattering direction is determined based on the Henyey-Greenstein function and random variables, where the random variables are values ​​uniformly distributed between 0 and 1.

[0019] In one possible implementation, the photon variable includes a photon weight, and after adjusting the photon variable in response to the photon being reflected during movement, it further includes:

[0020] In response to the photon weight being less than the second threshold, the tracking of the photon ends, and the process of emitting photons from the photon group into the extracorporeal blood model and subsequent steps is executed until the number of photons whose tracking ends meets the first threshold.

[0021] In one possible implementation, the photon group comprises a set of photon packets, and the step of emitting photons from the photon group into the extracorporeal blood model and subsequent steps until the number of photons tracked meets a first threshold includes:

[0022] The process of emitting photons from the photon group into the extracorporeal blood model and subsequent steps is performed until each photon in the detected photon packet has finished tracking and the number of detected photon packets meets the third threshold.

[0023] In one possible implementation, the process of constructing the prediction model includes:

[0024] Based on the model parameters of the blood model, the light intensity data corresponding to the blood model at a preset wavelength is generated;

[0025] The generated light intensity data is normalized to obtain normalized light intensity data;

[0026] The normalized light intensity data is used as the input feature, and the HCT value corresponding to the light intensity data is used as the output target to construct the training database.

[0027] The preparatory model containing the inverse backpropagation neural network is trained based on the training database until a prediction model with a prediction accuracy greater than the fourth threshold is obtained.

[0028] In one possible implementation, the model parameters include a set of HCT concentration values ​​and a set of SO2 concentration values. The set of HCT concentration values ​​includes several HCT concentration values ​​increasing from a first HCT concentration to a second HCT concentration in steps of one step. The set of SO2 concentration values ​​includes several SO2 concentration values ​​increasing from a first SO2 concentration to a second SO2 concentration in steps of two steps. The step of generating light intensity data corresponding to the blood model at a preset wavelength based on the model parameters includes:

[0029] For each combination of HCT concentration and SO2 concentration, light intensity data corresponding to a preset wavelength is generated.

[0030] A second aspect of this application provides an HCT prediction system, comprising:

[0031] The transmitting unit is used to emit photons from the photon group to the extracorporeal blood model so that the photons are reflected, scattered or absorbed during the movement. The extracorporeal blood model is a model constructed based on the thickness of the extracorporeal circulation tubing, the thickness of the blood and the distance between the light source generator and the detector.

[0032] The acquisition unit is used to acquire the light intensity data corresponding to the photons arriving at the detector.

[0033] The prediction unit is used to input the acquired light intensity data into the prediction model to obtain the HCT prediction value. The prediction model is a model constructed based on the BP neural network and the relationship between light intensity and HCT.

[0034] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the HCT prediction method described in the first aspect above.

[0035] A fourth aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a terminal device, cause the terminal device to perform the HCT prediction method as described in the first aspect above.

[0036] Compared with the prior art, this application has the following beneficial effects:

[0037] This application proposes a method for acquiring sampling data through simulation and for estimating HCT using neural networks. Compared to traditional methods, simulation is not only simpler and faster to operate, but also generates a large number of sampling points, providing a rich dataset for subsequent HCT estimation. This data acquisition method not only improves efficiency but also provides a solid foundation for subsequent algorithm design and optimization. By training the prediction model, it can learn the complex relationship between light intensity and HCT from a large amount of simulation data and make accurate HCT predictions based on these learned relationships. Thus, this application combines simulation and neural network prediction models, increasing the amount of data available and achieving continuous, non-invasive HCT estimation through optimized algorithm models, thereby improving the accuracy of HCT prediction.

[0038] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 is a flowchart of an HCT prediction method provided in an embodiment of this application;

[0041] Figure 2 is a schematic diagram of the extracorporeal blood model provided in an embodiment of this application;

[0042] Figure 3 is a schematic diagram of the MC simulation process provided in an embodiment of this application;

[0043] Figure 4 is a schematic diagram of the prediction model provided in the embodiment of this application;

[0044] Figure 5 is a schematic diagram of the HCT estimation results of the BP neural network algorithm provided in the embodiments of this application;

[0045] Figure 6 is a structural diagram of an HCT prediction system provided in an embodiment of this application. Detailed Implementation

[0046] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0047] To facilitate understanding of the technical solutions provided in the embodiments of this application, the terms involved in the embodiments of this application will be explained below.

[0048] Hematocrit (HCT), also known as erythrocyte volume, refers to the volume ratio of red blood cells that settle in a certain amount of anticoagulated whole blood after centrifugation. It is a simple method to indirectly reflect the number and size of red blood cells.

[0049] Twersky's theory has important applications in fields such as optical hematocrit measurement, especially in non-invasive blood monitoring. By utilizing Twersky's theory, hematocrit (HCT) in the blood can be estimated more accurately, providing strong support for clinical diagnosis and treatment.

[0050] The Henyey-Greenstein function, also known as the Henyey-Greenstein function, is a function used to describe the scattering of light. It was proposed in 1941 by American physicists Louis Henyey and John Greenstein. It is a three-dimensional function primarily used to describe the changes in light during the scattering process.

[0051] Backpropagation Neural Network (BP) is a commonly used artificial neural network model. Its basic principle is to transmit and process information through the connections and weights between multiple layers of neurons. The backpropagation algorithm continuously adjusts the connection weights between neurons to learn and predict input data. A BP neural network consists of an input layer, hidden layers, and an output layer, with each hidden and output layer containing multiple neurons. The hidden layer performs a weighted summation of the input data, applies a non-linear transformation using an activation function, and then passes the result to the output layer.

[0052] To facilitate understanding of the technical solutions provided in the embodiments of this application, the background technology involved in the embodiments of this application will be described below.

[0053] As mentioned earlier, near-infrared spectroscopy is a commonly used technique in non-invasive methods. It utilizes the propagation characteristics of near-infrared light in blood, measuring changes in light intensity to reflect changes in hemochromatographic coagulation (HCT). Specifically, a sensor module receives light passing through the blood and measures its intensity. Then, based on Twersky's theory, the HCT value can be estimated by analyzing the relationship between light intensity and HCT. This method has advantages such as high sensitivity and fast response speed, and therefore has been widely adopted in clinical applications.

[0054] However, existing non-invasive methods also have some drawbacks. First, the process of acquiring clinical sampling data is relatively complex and lengthy. Because non-invasive methods require specific instruments and equipment for measurement, the operation and maintenance of these devices typically require a certain level of professional knowledge and skills. Furthermore, due to the relatively complex measurement principles of non-invasive methods, a high level of technical expertise is also required for data processing and analysis. These factors all increase the difficulty and time cost of acquiring clinical sampling data.

[0055] Secondly, non-invasive methods rely on relatively limited data and typically require mathematical formula fitting to calculate HCT values. Due to the limited data, this fitting process may contain certain errors and uncertainties. Furthermore, optimizing and improving the mathematical model also faces challenges because sufficient experimental data and theoretical support are lacking to verify the model's accuracy and reliability. These factors all limit the accuracy and stability of non-invasive methods in HCT measurement.

[0056] Previous studies on optical hematocrit measurements based on Twersky's theory also identified an error of approximately 5%. This error primarily stems from limitations in the measurement principle itself and variations in experimental conditions. While this error range is somewhat acceptable, further research and improvements to non-invasive methods are needed to enhance measurement accuracy and stability for clinical applications requiring high precision.

[0057] To address this issue, this application provides an HCT prediction method that acquires sampling data through simulation. This innovative method significantly overcomes the complexity and time-consuming nature of traditional non-invasive methods in acquiring clinical sampling data. Compared to traditional methods, simulation is not only simpler and faster to operate, but also generates a large number of sampling points, thus providing a rich dataset for subsequent HCT estimation. This data acquisition method not only improves efficiency but also provides a solid foundation for subsequent algorithm design and optimization.

[0058] Furthermore, this application utilizes neural networks to estimate HCT. Neural networks, with their powerful learning and nonlinear mapping capabilities, excel in handling complex relationships and pattern recognition. By training a neural network model to form a prediction model, it can learn the complex relationship between light intensity and HCT from a large amount of simulation data, and make accurate HCT predictions based on these learned relationships. This method not only improves the accuracy of HCT prediction but also adapts to changes in different individuals and physiological states, making HCT estimation more accurate and reliable.

[0059] Therefore, by combining simulation and neural network technologies, this application not only simplifies the data acquisition process and increases the amount of data, but also achieves continuous non-invasive estimation of HCT by optimizing the algorithm model.

[0060] It should be noted that the HCT prediction method, system, device, and medium provided in this application can be applied to the field of computer technology. The above are merely examples and do not limit the application field of the HCT prediction method, system, device, and medium provided in this application. Furthermore, the embodiments of this application may not limit the executing entity of HCT prediction. For example, the HCT prediction method of this application embodiment can be applied to data processing devices such as terminal devices or servers. The terminal device can be an electronic device such as a computer or a personal digital assistant (PDA). The server can be a standalone server, a cloud server, or a cluster server composed of multiple servers.

[0061] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0062] The following embodiment illustrates an HCT prediction method provided in this application. Referring to Figure 1, which is a flowchart of an HCT prediction method provided in this embodiment, the method includes:

[0063] S101. Photons from the photon group are emitted into an extracorporeal circulating blood model so that the photons are reflected, scattered, or absorbed during the movement.

[0064] S102. Obtain the light intensity data corresponding to the photons arriving at the detector.

[0065] The Monte Carlo (MC) photon simulation process involves emitting photon packets and simulating their transmission, reflection, absorption, and scattering behavior. Tracking of a photon ends when its weight falls below a certain set value or when it escapes from the upper or lower surface of biological tissue. The process then returns to the first step to record another photon, continuing until all the set photons have been tracked. Finally, the light intensity data on the detector is obtained.

[0066] In one possible implementation, based on bio-optical characteristics, MC optical transmission simulation can be used to obtain light intensity data at specific wavelengths (e.g., 660nm, 940nm, and 1300nm) and under specific optical path settings (distance d3 between the light source and the detector, e.g., 4mm). Once photons enter the blood model, they are randomly scattered or absorbed. By collecting the remaining photons that reach the detector and analyzing their intensity, HCT can be determined.

[0067] The extracorporeal blood model was constructed based on the thickness of the extracorporeal circulation tubing, the thickness of the blood, and the distance between the light source and the detector.

[0068] The extracorporeal circulation (ECC) blood model needs to simulate the flow of blood in the tubing as realistically as possible under actual conditions. This includes the spatial distribution of blood varying with depth, as well as the constancy of the spatial distribution of the tubing surrounding the blood. Such a model design ensures that the data obtained in the simulation experiment is closer to reality, thereby improving the accuracy of HCT monitoring. Secondly, in terms of the choice of light transmission mode, the reflection mode is chosen because of its advantage of not being limited by the placement location. In the reflection mode, the light source emitter and detector are placed on the same side of the object being measured, which means that in practical applications, measurements can be easily performed regardless of the location in the ECC tubing, without worrying about the installation and position adjustment of the equipment. The ECC blood model in the reflection mode proposed in this application is shown in Figure 2. Figure 2 is a schematic diagram of the ECC blood model provided in the embodiment of this application, in which a detector and a light source emitter are set on the tubing, d1 is the thickness of the ECC tubing, set to 2 mm; d2 is the blood thickness, which is considered semi-infinite since the light reaching the detector is collected by the reflection method; d3 is the distance between the light source and the detector, which can be set to 4 mm.

[0069] When light enters the bloodstream through the circulatory system, photons are both absorbed and scattered by the tissues. Therefore, MC photon simulation and determination of photon intensity require the bio-optical properties of the proposed blood model: the absorption coefficient μ. a scattering coefficient μ s And anisotropy g.

[0070] In practical applications, the system determines whether a photon collides with a blood boundary and is reflected based on its movement step size and scattering direction. The movement step size is determined by the absorption coefficient, scattering coefficient, and a random variable. The scattering direction is determined by the Henyey-Greenstein function and a random variable that is uniformly distributed between 0 and 1. The photon's variables are adjusted in response to reflection during its movement.

[0071] A photon group can include several photon packets. These photon packets are emitted into the proposed blood model. The size of the emitted photon packet is called the photon weight (ω), and the initial weight can be considered as 1. After entering the blood model, the photon moves by a random step size (Δs). Δs is calculated by randomly sampling the photon scattering probability, using the following formula:

[0072] In equation (1), ξ is a random variable uniformly distributed between 0 and 1. Equation (1) represents the distance a photon travels before interacting with blood (absorption or scattering).

[0073] Then, it is determined whether the photon has come into contact with the blood surface. Once the photon hits the blood surface, it will be reflected, and the variables corresponding to that photon will be updated and adjusted. The variables include the photon's position vector, direction vector, and weight.

[0074] The direction of photon propagation is determined by the azimuth angle ψ∈[0,2π] and the scattering deflection angle θ∈[0,π]. The cosine of the deflection angle follows the Henyey-Greenstein function:

[0075] After the photon variable is updated, the system checks whether the photon is alive. If the current photon is no longer alive, the tracking of the current photon ends, and the tracking of the next photon begins. If the current photon is alive, the tracking continues until the detector detects the photon's light intensity data, at which point the tracking of the current photon ends, and the tracking of the next photon begins.

[0076] One method for detecting whether a photon is alive is to determine the relationship between the photon weight and a threshold. For example, if the photon weight is less than a certain set value, or if the photon escapes from the upper or lower surface of the biological tissue, the tracking of that photon ends, and the process returns to record another photon, until all the set photons have been tracked.

[0077] In one possible implementation, the simulation process is repeated until the number of photons tracked meets a first threshold. For example, if the simulation requires acquiring the light intensity data of 1,000,000 photons, the simulation ends when the light intensity data of the current photon is detected and it is determined that the current photon is the 1,000,000th photon (i.e., the last photon). Alternatively, the tracking can be terminated based on the number of photon packets until every photon in the currently detected photon packet has been tracked and the number of detected photon packets meets a third threshold. The MC simulation process for light propagation in a blood model is shown in Figure 3, which is a schematic diagram of the MC simulation process provided in an embodiment of this application.

[0078] S103. Input the acquired light intensity data into the prediction model to obtain the HCT prediction value.

[0079] A prediction model constructed using a backpropagation (BP) neural network is used to obtain an estimate of the heliotropic light intensity (HCT). The light intensity data obtained from MC simulation is used as input, and the corresponding HCT is used as output to obtain the predicted HCT value.

[0080] As shown in Figure 4, which is a schematic diagram of the prediction model provided in the embodiment of this application, the training of a neural network can be used to predict the HCT value when the light intensity value is known.

[0081] The prediction model is based on a backpropagation (BP) neural network and the relationship between light intensity and hematocrit (HCT). During model construction, a BP neural network training database can be built first. The model parameters of the blood model can be adjusted to obtain light intensity data corresponding to different model parameters, which serves as the model's training database.

[0082] Taking HCT concentration and SO2 concentration as examples, the model parameters include a set of HCT concentration values ​​and a set of SO2 concentration values. The HCT concentration value set includes several HCT concentration values ​​increasing from a first HCT concentration to a second HCT concentration in step sizes. The SO2 concentration value set includes several SO2 concentration values ​​increasing from a first SO2 concentration to a second SO2 concentration in step sizes. For example, the HCT concentration of the blood model is adjusted, starting from 15% and increasing to 55% in step sizes of 5%, and the SO2 concentration is increased from 40% and increasing to 100% in step sizes of 2%. For each combination of HCT and SO2, simulated MC light intensity data at 1300, 940, and 660 nm are generated as input, and the corresponding HCT values ​​are used as output. A training database for the BP neural network is established, and these light intensity data will serve as the input features of the neural network. These parameter settings are intended to simulate blood characteristics under different physiological conditions.

[0083] Since light intensity data obtained from MC simulations may have different dimensions and ranges, normalization is necessary to improve the training efficiency and accuracy of the neural network. Normalization typically maps the data to the range [0, 1] or [-1, 1]. The normalized light intensity data is used as input, and the corresponding HCT value is used as output to construct the training database for the BP neural network. This training database will be used to train the prediction model.

[0084] Then, based on the aforementioned training database, the preparatory model containing the inverse backpropagation neural network is trained, and the input and output of each layer, as well as the mean square error, are calculated. When the error is not less than the preset error value (i.e., e in the figure), the network parameters are adjusted, and the process of calculating the input and output of each layer and the mean square error is repeated until the error is less than the preset error value.

[0085] A test database is constructed based on the light intensity data of each photon obtained through simulation in the above process. Similar to the method used to establish the training database, the three types of light intensity data after MC simulation are used as input, and the prediction result is the predicted value of HCT. The prediction process involves inverse normalization of the predicted data to finally obtain the estimated value of HCT. In practical applications, the error between the true value and the predicted value can be compared to verify the accuracy of the neural network. The result is shown in Figure 5, which is a schematic diagram of the HCT estimation result of the BP neural network algorithm provided in this embodiment. The horizontal axis represents the true HCT value, and the vertical axis represents the predicted HCT value. The model corresponding to the neural network algorithm is continuously adjusted until a prediction model with a prediction accuracy greater than the fourth threshold is obtained.

[0086] In summary, the method provided in this application acquires sampling data through simulation. This method is not only simple and fast to operate, but also capable of acquiring a large number of sampling points, thus obtaining rich data resources. These data provide a solid foundation for subsequent HCT estimation. Furthermore, this application estimates HCT using a neural network. Neural networks have powerful learning and fitting capabilities, enabling them to extract useful information from large amounts of data and make accurate predictions accordingly. In this application, a BP neural network is used to predict HCT. The BP neural network is a commonly used multilayer feedforward neural network that continuously adjusts network parameters through the backpropagation algorithm to minimize prediction errors.

[0087] The effectiveness of the HCT predicted in this application was verified through simulation and prediction in MATLAB. The results show that the BP neural network can predict HCT with high accuracy, with an error of only 0.61%. This result fully demonstrates the accuracy of the continuous non-invasive HCT estimation method proposed in this application.

[0088] The above are some specific implementations of the HCT prediction method provided in the embodiments of this application. Based on this, this application also provides a corresponding system for HCT prediction. The system provided in the embodiments of this application will be described below from the perspective of functional modularity. Figure 6 is a structural diagram of an HCT prediction system provided in the embodiments of this application.

[0089] The system includes:

[0090] The transmitting unit 110 is used to transmit photons from the photon group to the extracorporeal blood model so that the photons are reflected, scattered or absorbed during the movement. The extracorporeal blood model is a model constructed based on the thickness of the extracorporeal circulation tubing, the thickness of the blood and the distance between the light source generator and the detector.

[0091] Acquisition unit 111 is used to acquire light intensity data corresponding to photons arriving at the detector;

[0092] The prediction unit 112 is used to input the acquired light intensity data into the prediction model to obtain the HCT prediction value. The prediction model is a model constructed based on the BP neural network and the relationship between light intensity and HCT.

[0093] In one possible implementation, the system further includes:

[0094] An adjustment unit is used to adjust the variables of a photon in response to its reflection during movement.

[0095] The acquisition unit includes:

[0096] The first termination unit is used to respond to the fact that the variable of the photon does not meet the termination tracking condition, track the photon until the photon reaches the detector, acquire the light intensity data corresponding to the photon, and terminate the tracking of the photon.

[0097] An execution unit is used to perform the step of emitting photons from the photon group into the extracorporeal blood model and subsequent steps until the number of photons tracked meets the first threshold.

[0098] In one possible implementation, the system further includes:

[0099] The judgment unit is used to determine whether a photon collides with the blood boundary and is reflected based on the photon's movement step size and scattering direction. The movement step size is determined based on the absorption coefficient, scattering coefficient, and random variables. The scattering direction is determined based on the Henyey-Greenstein function and random variables, where the random variables are values ​​uniformly distributed between 0 and 1.

[0100] In one possible implementation, the system further includes:

[0101] The second termination unit is used to terminate the tracking of a photon in response to the photon weight being less than a second threshold, and to execute the step of emitting photons from the photon group into the extracorporeal blood model and subsequent steps until the number of photons whose tracking has ended meets the first threshold.

[0102] In one possible implementation, the execution unit is used to perform the emission of photons from the photon group into the extracorporeal blood model and subsequent steps until the tracking of each photon in the detected photon packet ends and the number of detected photon packets meets a third threshold.

[0103] In one possible implementation, the process of constructing the prediction model includes:

[0104] Based on the model parameters of the blood model, the light intensity data corresponding to the blood model at a preset wavelength is generated;

[0105] The generated light intensity data is normalized to obtain normalized light intensity data;

[0106] The normalized light intensity data is used as the input feature, and the HCT value corresponding to the light intensity data is used as the output target to construct the training database.

[0107] The preparatory model containing the inverse backpropagation neural network is trained based on the training database until a prediction model with a prediction accuracy greater than the fourth threshold is obtained.

[0108] In one possible implementation, the model parameters include a set of HCT concentration values ​​and a set of SO2 concentration values. The set of HCT concentration values ​​includes several HCT concentration values ​​increasing from a first HCT concentration to a second HCT concentration in steps of one step. The set of SO2 concentration values ​​includes several SO2 concentration values ​​increasing from a first SO2 concentration to a second SO2 concentration in steps of two steps. The step of generating light intensity data corresponding to the blood model at a preset wavelength based on the model parameters includes:

[0109] For each combination of HCT concentration and SO2 concentration, light intensity data corresponding to a preset wavelength is generated.

[0110] This application also provides corresponding devices and computer storage media for implementing the HCT prediction method provided in this application.

[0111] The device includes a memory and a processor. The memory stores instructions or code, and the processor executes the instructions or code to enable the device to perform the HCT prediction method according to any embodiment of this application.

[0112] The computer storage medium stores code, and when the code is run, the device running the code implements the HCT prediction method described in any embodiment of this application.

[0113] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems or apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.

[0114] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0115] It should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0116] It should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0117] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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 limitations, 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.

[0118] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0119] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting HCT, characterized in that, include: Photons from the photon array are emitted into an extracorporeal blood model so that the photons are reflected, scattered, or absorbed during their movement. The extracorporeal blood model is constructed based on the thickness of the extracorporeal circulation tubing, the thickness of the blood, and the distance between the light source generator and the detector. Obtain the light intensity data corresponding to the photons arriving at the detector; The acquired light intensity data is input into the prediction model to obtain the predicted value of hematocrit (HCT). The prediction model is based on the backpropagation neural network (BP neural network) and the relationship between light intensity and HCT.

2. The method according to claim 1, characterized in that, Before acquiring the light intensity data corresponding to the photons arriving at the detector, the process also includes: In response to the photon being reflected during its movement, the variables of the photon are adjusted; The acquisition of light intensity data corresponding to photons arriving at the detector includes: If the variable of the photon does not meet the end tracking condition, the photon is tracked until it reaches the detector, the light intensity data corresponding to the photon is obtained, and the tracking of the photon ends. The process of emitting photons from the photon group into the extracorporeal blood model and subsequent steps continues until the number of photons tracked meets the first threshold.

3. The method according to claim 2, characterized in that, Before adjusting the variable of the photon in response to its reflection during movement, the method further includes: Based on the photon's movement step size and scattering direction, it is determined whether the photon collides with the blood boundary and is reflected. The movement step size is determined based on the absorption coefficient, scattering coefficient, and a random variable. The scattering direction is determined based on the Henyey-Greenstein function and a random variable, where the random variable is... Values ​​that are evenly distributed between 0 and 1.

4. The method according to claim 2, characterized in that, The variables of the photon include photon weights, and the adjustment of the photon variables in response to the reflection of the photon during its movement further includes: If the photon weight of a photon is less than a second threshold, the tracking of the photon ends, and the process of emitting photons from the photon group into the extracorporeal blood model and subsequent steps is performed until the number of photons whose tracking ends meets the first threshold.

5. The method according to claim 2, characterized in that, The photon group comprises a set of photon packets, and the step of emitting photons from the photon group into the extracorporeal blood model and subsequent steps until the number of photons tracked meets a first threshold includes: The process of emitting photons from the photon group into the extracorporeal blood model and subsequent steps is performed until each photon in the detected photon packet has finished tracking and the number of detected photon packets meets the third threshold.

6. The method according to claim 1, characterized in that, The process of constructing the prediction model includes: Based on the model parameters of the blood model, the light intensity data corresponding to the blood model at a preset wavelength is generated; The generated light intensity data is normalized to obtain normalized light intensity data; The normalized light intensity data is used as the input feature, and the HCT value corresponding to the light intensity data is used as the output target to construct the training database. The preparatory model containing the inverse backpropagation neural network is trained based on the training database until a prediction model with a prediction accuracy greater than the fourth threshold is obtained.

7. The method according to claim 6, characterized in that, The model parameters include a set of HCT concentration values ​​and a set of sulfur dioxide (SO2) concentration values. The HCT concentration value set includes several HCT concentration values ​​increasing from a first HCT concentration to a second HCT concentration in steps of one step. The SO2 concentration value set includes several SO2 concentration values ​​increasing from a first SO2 concentration to a second SO2 concentration in steps of two steps. The model parameters based on the blood model generate light intensity data corresponding to the blood model at a preset wavelength, including: For each combination of HCT concentration and SO2 concentration, light intensity data corresponding to a preset wavelength is generated.

8. An HCT prediction system, characterized in that, The system includes: The transmitting unit is used to emit photons from the photon group to the extracorporeal blood model so that the photons are reflected, scattered or absorbed during the movement. The extracorporeal blood model is a model constructed based on the thickness of the extracorporeal circulation tubing, the thickness of the blood and the distance between the light source generator and the detector. The acquisition unit is used to acquire the light intensity data corresponding to the photons arriving at the detector. The prediction unit is used to input the acquired light intensity data into the prediction model to obtain the HCT prediction value. The prediction model is a model constructed based on the BP neural network and the relationship between light intensity and HCT.

9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the HCT prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a terminal device, cause the terminal device to perform the HCT prediction method as described in any one of claims 1-7.

Images