An intelligent hemostasis system and method
By using data acquisition and prediction models in the intelligent hemostasis system, the problem of inaccurate pressure monitoring in existing hemostasis devices has been solved, enabling precise adjustment of hemostasis pressure, avoiding limb necrosis and poor hemostasis, and improving the safety and effectiveness of hemostasis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- THE FIRST MEDICAL CENT CHINESE PLA GENERAL HOSPITAL
- Filing Date
- 2024-12-25
- Publication Date
- 2026-05-29
AI Technical Summary
Existing hemostatic devices lack precise pressure monitoring mechanisms, leading to inappropriate hemostatic pressure, which may cause excessive bleeding or thrombosis. Furthermore, they are difficult to personalize hemostatic pressure adjustment, especially when dealing with individual patient differences, which carries the risk of limb necrosis.
An intelligent hemostasis system is adopted, which collects patient data through a data acquisition module and uses differential evolution algorithm and WaveNet model to build a hemostasis time-pressure prediction model to achieve precise adjustment of pressure between the hemostasis device and the skin. This includes data processing, model building and integration of the hemostasis module.
It achieves precise adjustment of tourniquet pressure, avoiding limb necrosis and poor hemostasis caused by excessive or insufficient tourniquet pressure, and improves the safety and effectiveness of hemostasis.
Smart Images
Figure CN119606476B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of clinical medical technology, specifically to an intelligent hemostasis system and method. Background Technology
[0002] During the rescue of the wounded, there are instances of tourniquet misuse, including unnecessary use of tourniquets, tourniquets placed too high, and tourniquets left in place for too long. The longer the tourniquet is left to fully tighten, the greater the risk of compartment syndrome, vascular thrombosis, rhabdomyolysis, and irreversible myonecrosis. This can lead to severe tissue loss and limb amputation, threatening life or causing unnecessary disability.
[0003] Furthermore, the effectiveness of hemostasis relies on medical staff's experience in controlling the hemostatic pressure and time, such as adjusting the air pressure inside the hemostatic cuff to regulate the pressure. This subjective judgment may lead to inappropriate pressure, resulting in excessive bleeding or thrombosis. Existing hemostatic devices lack precise pressure monitoring mechanisms, making it difficult to achieve personalized hemostatic pressure adjustment, especially when dealing with individual patient differences. Summary of the Invention
[0004] To address the shortcomings of the existing technology, this invention aims to provide an intelligent hemostasis system and method, which addresses the problems of excessive tourniquet pressure and prolonged hemostasis time leading to necrosis of the limb or the pressure-affected area, as well as poor hemostasis due to insufficient pressure during combat wound treatment.
[0005] To solve the above problems, the present invention adopts the following technical solution:
[0006] On the one hand, the present invention provides an intelligent hemostasis system, including a data acquisition module, a data processing module, a model building module, a prediction module, and a hemostasis module.
[0007] The data acquisition module is used to collect the patient's blood pressure, heart rate, and pressure between the hemostasis device and the skin through the hemostasis device, as well as to collect the patient's body data and wound data.
[0008] The data processing module is used to preprocess the collected data, constructing a time series based on the patient's blood pressure, heart rate, and pressure between the hemostatic device and the skin, and using the patient's body data and wound data as labels for the time series to form time series data.
[0009] The model building module is used to construct a hemostasis time-pressure prediction model using time series data that shows successful hemostasis without excessive pressure or prolonged hemostasis time, through differential evolution algorithm and WaveNet model.
[0010] The prediction module is used to obtain the relationship curve between the pressure between the hemostatic device and the skin and the hemostatic time by inputting the patient's blood pressure, heart rate, patient body data and patient wound data through the hemostasis time-pressure prediction model.
[0011] The hemostasis module is used to control the hemostasis device and adjust the pressure between the hemostasis device and the skin based on the relationship curve between the pressure between the hemostasis device and the skin and the hemostasis time.
[0012] As one possible implementation, the patient's body data includes age, gender, BMI, and body fat; the patient's wound data includes part, area, depth, and amount of bleeding per unit time.
[0013] As one possible implementation, the WaveNet model consists of a two-dimensional convolutional feature extraction layer, a one-dimensional sequence reconstruction layer, two residual modules, a fully connected layer, and an output layer. The residual module obtains output one by performing two causal dilated convolutional layers, a batch normalization layer, and a random deactivation layer, and obtains output two by performing a 1×1 convolution. The output one and output two are combined to obtain the output of the residual module.
[0014] The causal dilated convolutional layer uses dilated convolutions with a dilation coefficient to expand the receptive field:
[0015]
[0016] in, Given time-series input data, W represents the convolutional kernel weights, and b is the bias term. It is the output of the convolutional layer. It is an activation function.
[0017] The WaveNet model introduces a gated recurrent unit:
[0018]
[0019] in, These are the update door and the reset door. It is the Sigmoid function. For element-wise multiplication, It is in a hidden state.
[0020] The output layer performs linear regression:
[0021]
[0022] in, For predicted values, and These are the weights and biases of the output layer, respectively.
[0023] As one possible implementation, the construction of the hemostasis time-pressure prediction model using the differential evolution algorithm and the WaveNet model includes: optimizing the weights of the WaveNet model using the differential evolution algorithm, including:
[0024] Randomly initialize individuals in the population, with each individual representing a set of weights in the WaveNet model;
[0025] Perform mutation operations on each individual to generate experimental individuals;
[0026] The experimental individuals are cross-operated with the original individuals to generate candidate solutions;
[0027] A selection process is performed between the original individual and the candidate solutions, choosing the individual with higher fitness to enter the next generation; this continues until the loss function of the WaveNet model meets the requirements.
[0028] As one possible implementation, the hemostatic device includes a detection and control body and a tourniquet. The detection and control body is equipped with a blood pressure detection device, a heart rate monitoring device, a skin pressure detection device, a time module, a release and retraction device, and / or a patient body data and patient wound data input device. One end of the tourniquet is inserted into the detection and control body and detachably connected to it, while the other end is embedded in the release and retraction device. The release and retraction device adjusts the pressure between the hemostatic device and the skin by tightening or loosening.
[0029] As one possible implementation, the tourniquet has a serrated outer side, and the retraction device includes a motor and a rotor that is fixedly connected to the motor and matches the serrated shape of the tourniquet.
[0030] On the other hand, the present invention provides an intelligent hemostasis method, comprising:
[0031] The patient's blood pressure, heart rate, and pressure between the hemostatic device and the skin are collected using a hemostatic device, along with patient body data and wound data.
[0032] The collected data were preprocessed, and the patient's blood pressure, heart rate, and pressure between the hemostatic device and the skin were constructed into a time series based on the hemostasis time. The patient's body data and the patient's wound data were used as labels for the time series to form time series data.
[0033] Using time series data on successful hemostasis without excessive pressure or prolonged hemostasis time, a hemostasis time-pressure prediction model was constructed using differential evolution algorithm and WaveNet model.
[0034] By using a hemostasis time-pressure prediction model, the relationship curve between the pressure between the hemostasis device and the skin and the hemostasis time can be obtained by inputting the patient's blood pressure, heart rate, patient body data, and patient wound data.
[0035] Based on the relationship curve between the pressure between the hemostatic device and the skin and the hemostasis time, the pressure between the hemostatic device and the skin is adjusted by controlling the hemostatic device.
[0036] As one possible implementation, the patient's body data includes age, gender, BMI, and body fat; the patient's wound data includes part, area, depth, and amount of bleeding per unit time.
[0037] As one possible implementation, the WaveNet model consists of a two-dimensional convolutional feature extraction layer, a one-dimensional sequence reconstruction layer, two residual modules, a fully connected layer, and an output layer. The residual module obtains output one through two causal dilated convolutional layers, a batch normalization layer, and a random deactivation layer, and obtains output two through a 1×1 convolution. The output one and output two are combined to obtain the output of the residual module.
[0038] The causal dilated convolutional layer uses dilated convolutions with a dilation coefficient to expand the receptive field:
[0039]
[0040] in, Given time-series input data, W represents the convolutional kernel weights, and b is the bias term. It is the output of the convolutional layer. It is an activation function;
[0041] The WaveNet model introduces a gated recurrent unit:
[0042]
[0043] in, These are updating the door and resetting the door. It is the Sigmoid function. For element-wise multiplication, It is in a hidden state;
[0044] The output layer performs linear regression:
[0045]
[0046] in, For predicted values, and These are the weights and biases of the output layer, respectively.
[0047] As one possible implementation, the construction of the hemostasis time-pressure prediction model using the differential evolution algorithm and the WaveNet model includes: optimizing the weights of the WaveNet model using the differential evolution algorithm, including:
[0048] Randomly initialize individuals in the population, with each individual representing a set of weights in the WaveNet model;
[0049] Perform mutation operations on each individual to generate experimental individuals;
[0050] The experimental individuals are cross-operated with the original individuals to generate candidate solutions;
[0051] A selection process is performed between the original individual and the candidate solutions, choosing the individual with higher fitness to enter the next generation; this continues until the loss function of the WaveNet model meets the requirements.
[0052] As one possible implementation, the hemostatic device includes a detection and control body and a tourniquet. The detection and control body is equipped with a blood pressure detection device, a heart rate monitoring device, a skin pressure detection device, a time module, a release and retraction device, and / or a patient body data and patient wound data input device. One end of the tourniquet is inserted into the detection and control body and detachably connected to it, while the other end is embedded in the release and retraction device. The release and retraction device adjusts the pressure between the hemostatic device and the skin by tightening or loosening.
[0053] As one possible implementation, the tourniquet has a serrated outer side, and the retraction device includes a motor and a rotor that is fixedly connected to the motor and matches the serrated shape of the tourniquet.
[0054] The beneficial effects of this invention are as follows: This invention collects blood pressure, heart rate, hemostasis pressure, patient body data, and patient wound data through a hemostasis device, and constructs a hemostasis time-pressure prediction model through differential evolution algorithm and WaveNet model to predict the relationship curve between hemostasis pressure and hemostasis time. Through the relationship curve, the pressure applied by the hemostasis device can be adjusted as needed, avoiding the problems of excessive tourniquet pressure and excessively long hemostasis time leading to necrosis of limbs or pressure-affected areas due to different human factors in the process of combat wound treatment, as well as poor hemostasis due to insufficient pressure. Attached Figure Description
[0055] Figure 1 This is a schematic diagram of an intelligent hemostasis system according to the present invention.
[0056] Figure 2 This is a schematic diagram of the hemostasis device of the present invention.
[0057] Figure 3 This is a flowchart of an intelligent hemostasis method according to the present invention. Detailed Implementation
[0058] The present invention will be further described in detail below with reference to specific embodiments.
[0059] It should be noted that these embodiments are only used to illustrate the present invention and are not intended to limit the present invention. Simple improvements to the method under the premise of the present invention are all within the scope of protection claimed by the present invention.
[0060] See Figure 1 This is an intelligent hemostasis system, comprising a data acquisition module 100, a data processing module 200, a model building module 300, a prediction module 400, and a hemostasis module 500.
[0061] The data acquisition module 100 is used to collect patient blood pressure, heart rate, and pressure between the hemostatic device and the skin via a hemostatic device, as well as patient body data and wound data. Patient body data includes age, gender, BMI, and body fat percentage; wound data includes location, area, depth, and bleeding volume per unit time. This data collection process can be performed through detection by the hemostatic device, or by data input, and can also incorporate relevant parameters such as patient blood flow rate and blood pressure at the wound site.
[0062] The data processing module 200 is used to preprocess the collected data, construct a time series based on the patient's blood pressure, heart rate, and pressure between the hemostasis device and the skin, and use the patient's body data and the patient's wound data as labels for the time series to form time series data.
[0063] Preprocessing includes:
[0064] Each parameter is normalized to map the original data to the interval [0, 1], as shown in the following formula:
[0065]
[0066] in:
[0067] Xnorm represents the normalized data;
[0068] X represents the original data;
[0069] Xmin is the minimum value in the original dataset;
[0070] Xmax is the maximum value in the original dataset.
[0071] The purpose of normalization is to eliminate the influence of different dimensions between parameters, so that the contribution of each parameter to model training is at the same level.
[0072] Outliers are detected and handled. If the original value of a parameter exceeds the known reasonable range, it is corrected using linear interpolation or a neighborhood-based method.
[0073] The model building module 300 is used to construct a hemostasis time-pressure prediction model using time series data that shows successful hemostasis without excessive pressure or excessive hemostasis time, through differential evolution algorithm and WaveNet model.
[0074] The WaveNet model consists of a two-dimensional convolutional feature extraction layer, a one-dimensional sequence reconstruction layer, two residual modules, a fully connected layer, and an output layer. The residual module obtains output one through two causal dilated convolutional layers, a batch normalization layer, and a random deactivation layer, and obtains output two through a 1×1 convolution. The output one and output two are combined to obtain the output of the residual module.
[0075] The causal dilated convolutional layer uses dilated convolutions with a dilation coefficient to expand the receptive field:
[0076]
[0077] in, Given time-series input data, W represents the convolutional kernel weights, and b is the bias term. It is the output of the convolutional layer. It is an activation function.
[0078] The WaveNet model introduces a gated recurrent unit:
[0079]
[0080] in, These are updating the door and resetting the door. It is the Sigmoid function. For element-wise multiplication, It is in a hidden state.
[0081] The output layer performs linear regression:
[0082]
[0083] in, For predicted values, and These are the weights and biases of the output layer, respectively.
[0084] The WaveNet model can be used to explore the relationship between time series under different labels, that is, the appropriate hemostasis time and pressure under different patient conditions and wound conditions.
[0085] The construction of a hemostasis time-pressure prediction model using differential evolution algorithm and WaveNet model includes: optimizing the weights of the WaveNet model using differential evolution algorithm, including:
[0086] Randomly initialize individuals in the population, with each individual representing a set of weights for the WaveNet model:
[0087]
[0088] in, Population size, i.e., the number of individuals; Let be the i-th weight parameter of the i-th individual.
[0089] Perform mutation operations on each individual to generate experimental individuals, and then perform crossover operations between the experimental individuals and the original individuals to generate candidate solutions.
[0090]
[0091] in, Let j be the j-th parameter of the i-th candidate solution in the (t+1)-th generation; Three distinct individuals are randomly selected from the current population; F is a scaling factor that controls the difference weights.
[0092] The crossover operation is used to combine the mutated candidate solutions with the current solution:
[0093]
[0094] in, is the j-th parameter of the i-th candidate solution after crossover; J is the randomly selected crossover index; CR is the crossover probability; To generate random numbers between [0, 1].
[0095] A selection process is performed between the original individual and the candidate solutions, choosing the individual with higher fitness to enter the next generation; this continues until the loss function of the WaveNet model reaches the required level or the maximum number of iterations is reached.
[0096] The prediction module 400 is used to obtain the relationship curve between the pressure between the hemostatic device and the skin and the hemostasis time by inputting the patient's blood pressure, heart rate, physical data, and wound data through a hemostasis time-pressure prediction model. Through the trained hemostasis time-pressure prediction model, by inputting the patient's blood pressure, heart rate, physical data, and wound data, the appropriate hemostasis pressure for a future period can be obtained, i.e., the relationship curve between the pressure between the hemostatic device and the skin and the hemostasis time.
[0097] The hemostasis module 500 is used to control and adjust the pressure between the hemostasis device and the skin based on the relationship curve between the pressure between the hemostasis device and the skin and the hemostasis time. Through the above process, the pressure applied by the hemostasis device can be adjusted as needed based on the hemostasis time, avoiding problems such as excessive tourniquet pressure or excessively long hemostasis time leading to necrosis of the limb or pressure-affected area, and insufficient pressure leading to poor hemostasis during combat wound treatment.
[0098] A monitoring module can also be set up to monitor the hemostasis time and pressure, and issue an alarm when the duration and pressure reach a level that may cause wounds or bodily damage.
[0099] See Figure 2 The hemostasis device can be a detection and control body and a tourniquet. The detection and control body is equipped with a blood pressure detection device, a heart rate monitoring device, a skin pressure detection device, a time module (for synchronizing data time), a retraction device, and / or a patient body data and patient wound data input device. One end of the tourniquet is inserted into the detection and control body to achieve a detachable connection with the detection and control body, and the other end is embedded in the retraction device. The retraction device adjusts the pressure between the hemostasis device and the skin by tightening or loosening.
[0100] The tourniquet has a serrated outer edge. The tightening and loosening device includes a motor and a rotor fixedly connected to the motor and conforming to the serrated edge of the tourniquet. The motor drives the rotor to rotate, thus tightening or loosening the tourniquet.
[0101] See Figure 3 This is an intelligent hemostasis method, comprising:
[0102] S100: Collects patient blood pressure, heart rate, and pressure between the hemostatic device and the skin through the hemostatic device, and collects patient body data and patient wound data.
[0103] S200. The collected data is preprocessed. Data on patient blood flow rate, bleeding volume, blood pressure, heart rate, and pressure between the hemostatic device and the skin are constructed into a time series based on the hemostasis time. Patient body data and patient wound data are used as labels for the time series to form time series data.
[0104] S300 uses time series data of successful hemostasis without excessive pressure or excessive hemostasis time, and constructs a hemostasis time-pressure prediction model using differential evolution algorithm and WaveNet model.
[0105] S400: By using the hemostasis time-pressure prediction model, input the patient's blood pressure, heart rate, patient's physical data, and patient's wound data to obtain the relationship curve between the pressure between the hemostasis device and the skin and the hemostasis time.
[0106] S500: Based on the relationship curve between the pressure between the hemostatic device and the skin and the hemostasis time, control the hemostatic device to adjust the pressure between the hemostatic device and the skin.
[0107] The patient's physical data includes age, gender, BMI, and body fat percentage; the patient's wound data includes part, area, depth, and amount of bleeding per unit time.
[0108] The WaveNet model consists of a two-dimensional convolutional feature extraction layer, a one-dimensional sequence reconstruction layer, two residual modules, a fully connected layer, and an output layer. The residual module obtains output one through two causal dilated convolutional layers, a batch normalization layer, and a random deactivation layer, and obtains output two through a 1×1 convolution. The output one and output two are combined to obtain the output of the residual module.
[0109] Causal dilated convolutional layers use dilated convolutions with dilation coefficients to expand the receptive field:
[0110]
[0111] in, Given time-series input data, W represents the convolutional kernel weights, and b is the bias term. It is the output of the convolutional layer. It is an activation function.
[0112] The WaveNet model introduces gated recurrent units:
[0113]
[0114] in, These are updating the door and resetting the door. It is the Sigmoid function. For element-wise multiplication, It is in a hidden state.
[0115] Linear regression is performed on the output layer:
[0116]
[0117] in, For predicted values, and These are the weights and biases of the output layer, respectively.
[0118] As one possible implementation, the construction of the hemostasis time-pressure prediction model using the differential evolution algorithm and the WaveNet model includes: optimizing the weights of the WaveNet model using the differential evolution algorithm, including:
[0119] Randomly initialize individuals in the population, with each individual representing a set of weights in the WaveNet model;
[0120] Perform mutation operations on each individual to generate experimental individuals;
[0121] The experimental individuals are cross-operated with the original individuals to generate candidate solutions;
[0122] A selection process is performed between the original individual and the candidate solutions, choosing the individual with higher fitness to enter the next generation; this continues until the loss function of the WaveNet model meets the requirements.
[0123] The hemostatic device includes a detection and control body and a tourniquet. The detection and control body is equipped with a blood pressure detection device, a heart rate monitoring device, a skin pressure detection device, a time module, a release and retraction device, and / or a patient body data and patient wound data input device. One end of the tourniquet is inserted into the detection and control body and detachably connected to it, while the other end is embedded in the release and retraction device. The release and retraction device adjusts the pressure between the hemostatic device and the skin by tightening or loosening.
[0124] The tourniquet has a serrated outer side, and the retraction device includes a motor and a rotor that is fixedly connected to the motor and matches the serrated shape of the tourniquet.
[0125] Finally, 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 it. Although the present invention has been described with reference to preferred embodiments, those skilled in the art should understand that various changes in form and detail can be made without departing from the spirit and scope of the invention as defined in the appended claims.
Claims
1. An intelligent hemostasis system, characterized in that, It includes a data acquisition module, a data processing module, a model building module, a prediction module, and a hemostasis module; The data acquisition module is used to collect the patient's blood pressure, heart rate, and pressure between the hemostasis device and the skin through the hemostasis device, as well as to collect the patient's body data and wound data. The data processing module is used to preprocess the collected data, construct a time series based on the patient's blood pressure, heart rate, and pressure between the hemostatic device and the skin, and use the patient's body data and wound data as labels for the time series to form time series data. The model building module is used to construct a hemostasis time-pressure prediction model using time series data of successful hemostasis without excessive pressure or excessive hemostasis time, through differential evolution algorithm and WaveNet model. The prediction module is used to obtain the relationship curve between the pressure between the hemostatic device and the skin and the hemostatic time by inputting the patient's blood pressure, heart rate, patient body data and patient wound data through the hemostasis time-pressure prediction model. The hemostasis module is used to control the hemostasis device and adjust the pressure between the hemostasis device and the skin according to the relationship curve between the pressure between the hemostasis device and the skin and the hemostasis time. The WaveNet model consists of a two-dimensional convolutional feature extraction layer, a one-dimensional sequence reconstruction layer, two residual modules, a fully connected layer, and an output layer. The residual module obtains output one through two causal dilated convolutional layers, a batch normalization layer, and a random deactivation layer, and obtains output two through a 1×1 convolution. The output one and output two are combined to obtain the output of the residual module. The causal dilated convolutional layer uses dilated convolutions with a dilation coefficient to expand the receptive field: in, Given time-series input data, W represents the convolutional kernel weights, and b is the bias term. It is the output of the convolutional layer. It is an activation function; The WaveNet model introduces a gated recurrent unit: in, These are the update door and the reset door. It is the Sigmoid function. For element-wise multiplication, It is in a hidden state; The output layer performs linear regression: in, For predicted values, and These are the weights and biases of the output layer, respectively. The construction of the hemostasis time-pressure prediction model using the differential evolution algorithm and the WaveNet model includes: optimizing the weights of the WaveNet model using the differential evolution algorithm, including: Randomly initialize individuals in the population, with each individual representing a set of weights in the WaveNet model; Perform mutation operations on each individual to generate experimental individuals; The experimental individuals are cross-operated with the original individuals to generate candidate solutions; A selection process is performed between the original individual and the candidate solutions, choosing the individual with higher fitness to enter the next generation; this continues until the loss function of the WaveNet model meets the requirements.
2. The intelligent hemostasis system according to claim 1, characterized in that, The patient's physical data includes age, gender, BMI, and body fat percentage; the patient's wound data includes location, area, depth, and amount of bleeding per unit time.
3. The intelligent hemostasis system according to claim 1, characterized in that, The hemostatic device includes a detection and control body and a tourniquet. The detection and control body is equipped with a blood pressure detection device, a heart rate monitoring device, a skin pressure detection device, a time module, a release and retraction device, and / or a patient body data and patient wound data input device. One end of the tourniquet is inserted into the detection and control body and detachably connected to it, and the other end is embedded in the release and retraction device. The release and retraction device adjusts the pressure between the hemostatic device and the skin by tightening or loosening.
4. The intelligent hemostasis system according to claim 3, characterized in that, The tourniquet has a serrated outer side, and the retraction device includes a motor and a rotor that is fixedly connected to the motor and matches the serrated shape of the tourniquet.