Artificial intelligence-based patient care assistance system for hepatobiliary surgery

Abstract

The application discloses a liver and gall surgical patient nursing auxiliary system based on artificial intelligence, which comprises a data acquisition module, a data expansion model construction module, a nursing auxiliary model construction module and a patient nursing module. The application relates to the technical field of patient nursing assistance, in particular to a liver and gall surgical patient nursing auxiliary system based on artificial intelligence. The core architecture comprising a generator and a discriminator is constructed, a chaotic vector generation mechanism and a weight updating strategy are introduced, the fidelity and diversity of generated data are improved, the real data distribution is simulated more effectively, and thus the quality and scale of the data set are expanded. Through the innovative design of the model framework comprising a reset unit, an update unit, an integration unit, a candidate state unit and a state update unit, an improved loss function is introduced, the modeling capability of the model for time series data is improved, the long-term dependence relationship is better captured, and the prediction accuracy and stability of the nursing auxiliary model are improved.

Description

Technical Field

[0001] This invention relates to the field of patient care assistance technology, specifically to an artificial intelligence-based patient care assistance system for hepatobiliary surgery. Background Technology

[0002] The AI-based hepatobiliary surgery patient care assistance system is an innovative intelligent medical solution. It utilizes artificial intelligence algorithms and big data analysis technology to deeply mine and analyze multi-dimensional patient data, enabling accurate prediction and personalized intervention of patient care needs. The system can monitor patient status in real time, predict potential complications, and issue timely warning information, providing scientific nursing decision support for medical staff.

[0003] Traditional methods for generating nursing data for hepatobiliary surgery patients suffer from low data quality, insufficient diversity, and difficulty in effectively simulating the distribution of real data. Traditional auxiliary models for nursing care of hepatobiliary surgery patients have problems such as insufficient ability to model time series data, difficulty in capturing long-term dependencies, and low model prediction accuracy. Summary of the Invention

[0004] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides an artificial intelligence-based nursing assistance system for hepatobiliary surgery patients. Addressing the problems of low data quality, insufficient diversity, and difficulty in effectively simulating real data distribution in traditional hepatobiliary surgery patient nursing data generation methods, this solution constructs a core architecture including a generator and a discriminator, introduces a chaotic vector generation mechanism, and optimizes the loss function and weight update strategy. This improves the realism and diversity of the generated data, more effectively simulating real data distribution, thereby expanding the quality and scale of the dataset. Furthermore, addressing the problems of insufficient modeling ability for time-series data, difficulty in capturing long-term dependencies, and low model prediction accuracy in traditional hepatobiliary surgery patient nursing assistance models, this solution innovatively designs a model framework including a reset unit, update unit, integration unit, candidate state unit, and state update unit, and introduces an improved loss function. This enhances the model's ability to model time-series data and strengthens its ability to capture long-term dependencies, thereby improving the prediction accuracy and stability of the nursing assistance model.

[0005] The technical solution adopted by the present invention is as follows: The artificial intelligence-based patient care assistance system for hepatobiliary surgery provided by the present invention includes a data acquisition module, a data expansion model construction module, a nursing assistance model construction module, and a patient care module;

[0006] The data acquisition module collects historical data on the physical condition, disease status, vital signs, and complications of hepatobiliary surgery patients.

[0007] The data expansion model building module expands the dataset by generating a core architecture, generating chaotic vectors, evaluating generator loss, evaluating discriminator loss, and updating weights.

[0008] The nursing assistance model construction module constructs the nursing assistance model by designing a basic framework, a reset unit, an update unit, an integration unit, a candidate state unit, a state update unit, and a loss function.

[0009] The patient care module uses a nursing assistance model to predict patient complication data. When the model predicts that a patient will experience complications, it issues a prompt message.

[0010] Furthermore, the data acquisition module collects historical data on the physical condition, disease status, vital signs, and complications of hepatobiliary surgery patients. The physical condition data includes the patient's age, gender, height, and weight. The disease status data refers to the type and stage of the hepatobiliary disease. The vital signs data includes the patient's body temperature, blood pressure, and heart rate. The complications data refers to whether the patient has experienced postoperative bleeding, infection, bile leakage, or liver failure.

[0011] Furthermore, the data expansion model construction module specifically includes the following:

[0012] The core architecture is generated, and the model includes a generator and a discriminator. The generator consists of an input layer, a hidden layer, and a generation layer. The input layer receives a chaotic vector, the hidden layer is activated by a linear rectified function, and skip connections are introduced in the hidden layer. The generation layer is activated by a hyperbolic tangent function. The generator learns the mapping from the chaotic vector to the real data distribution to generate data similar to the real data. The discriminator is a classifier that outputs a probability value, representing the probability that the input data is the real data.

[0013] Generate a chaotic vector, represented as follows:

[0014] ;

[0015] Where i represents the dimension index of the data vector. This represents the value of the chaotic vector in the i-th dimension. This represents the sigmoid function. This represents the value of the random noise vector in the i-th dimension;

[0016] The generator loss is evaluated as follows:

[0017] ;

[0018] in, This represents the loss value of the generator. Represents the chaotic vector Expectations Indicates that z follows the prior distribution Mid-sampling, Represents a generator. Indicates the discriminator, This indicates that the discriminator detects fake samples generated by the generator. The probability of identification, , and Indicates the loss weights of the generator. Indicates the real sample Expectations express From the distribution of real data Mid-sampling, Indicates encoder, This indicates that the encoder encodes real samples as chaotic vectors. This indicates that the generator obtains the chaotic vector from the encoder. The reconstructed sample, This represents taking the square of the L2 norm. This represents the gradient of the generator with respect to the input chaotic vector;

[0019] The discriminator loss is evaluated as follows:

[0020] ;

[0021] in, This represents the loss value of the discriminator. This represents the probability that the discriminator can distinguish a real sample x. The loss weights of the discriminator are represented. This represents the sample reconstructed by the generator based on the chaotic vector z;

[0022] Weight updates are represented as follows:

[0023] ;

[0024] Where it represents the number of weight updates. Indicates the first Model weights during the second weight update Indicates the first Model weights during the second weight update and This represents the learning rate of the model weights. This represents the increment of model weights. This represents the total number of data samples. Indicates the index of the data sample. Indicates the first The probability of the discriminator identifying the q-th data sample during the next weight update. This indicates that the discriminator operates on the q-th generated data sample. The probability of identification;

[0025] Expand the dataset by using a generator to generate simulation data and adding the simulation data to the original dataset.

[0026] Furthermore, the module for constructing the nursing assistance model specifically includes the following:

[0027] The basic framework of the design includes a reset unit, an update unit, an integration unit, a candidate state unit, and a state update unit, with complication data set as the model's label data.

[0028] The design includes a reset unit to control the influence of the previous hidden state on the current state, and introduces a weighted sum of historical hidden states to dynamically adjust the importance of historical information.

[0029] The design update unit controls the fusion of the current input and the previous hidden state. By adjusting the fusion ratio of the current input and historical information, a new intermediate state is generated.

[0030] The design integrates units that utilize the hyperbolic tangent function to capture nonlinear characteristics in the input data.

[0031] Design candidate state units, combine the output of the reset unit and the output of the integration unit to generate the final candidate state;

[0032] The design incorporates a state update unit that controls the fusion ratio between the previous hidden state and the candidate state through its output, and introduces a scale transformation matrix to adjust the magnitude of the state update.

[0033] The loss function is designed as follows:

[0034] ;

[0035] in, This represents the loss value of the nursing assistance model, where c represents the data index used for training the nursing assistance model. This represents the total number of data points used for training the nursing assistance model. This represents the true label value of the c-th data point. This represents the model prediction value for the c-th data point. Let T represent the logarithmic function, and T represent the total number of time steps. This represents the hidden state at the k-th time step. This represents the hidden state at time step (k-1). and Indicates the loss weight. Indicates the first The model prediction value for the c-th data point at time step c. This represents the model prediction value of the c-th data point at the k-th time step. Indicates the first The model prediction value for the c-th data point at time step c.

[0036] Furthermore, the patient care module collects patients' physical condition data, disease data, and vital sign data in real time, and uses a nursing assistance model to predict patients' complication data. When the model predicts that a patient will have complications, it issues a prompt message to remind medical staff to prepare for complication response.

[0037] The beneficial effects achieved by the present invention using the above solution are as follows:

[0038] (1) In view of the problems of low quality, insufficient diversity and difficulty in effectively simulating the distribution of real data in traditional methods of generating nursing data for hepatobiliary surgery patients, this solution improves the realism and diversity of the generated data by constructing a core architecture that includes a generator and a discriminator, introducing a chaotic vector generation mechanism and optimizing the loss function and weight update strategy, thereby more effectively simulating the distribution of real data and expanding the quality and scale of the dataset.

[0039] (2) In view of the problems that traditional nursing auxiliary models for hepatobiliary surgery patients have insufficient ability to model time series data, difficulty in capturing long-term dependencies and low model prediction accuracy, this solution improves the model's ability to model time series data and enhances its ability to capture long-term dependencies by innovatively designing a model framework that includes reset unit, update unit, integration unit, candidate state unit and state update unit, and introduces an improved loss function, thereby improving the prediction accuracy and stability of the nursing auxiliary model. Attached Figure Description

[0040] Figure 1 A schematic diagram of the artificial intelligence-based patient care assistance system for hepatobiliary surgery provided by the present invention;

[0041] Figure 2 A schematic diagram for constructing the data extension model module;

[0042] Figure 3 A schematic diagram for constructing a nursing support model module.

[0043] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0044] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0045] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention 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 invention.

[0046] Example 1, see Figure 1 The artificial intelligence-based patient care assistance system for hepatobiliary surgery provided by this invention includes a data acquisition module, a data expansion model construction module, a nursing assistance model construction module, and a patient care module.

[0047] The data acquisition module collects historical data on the physical condition, disease status, vital signs, and complications of hepatobiliary surgery patients and sends the data to the data expansion model building module.

[0048] The data expansion model building module expands the dataset by generating a core architecture, generating chaotic vectors, evaluating generator loss, evaluating discriminator loss, and updating weights, and then sends the data to the nursing assistance model building module.

[0049] The nursing assistance model construction module constructs the nursing assistance model by designing a basic framework, a reset unit, an update unit, an integration unit, a candidate state unit, a state update unit, and a loss function, and sends the data to the patient care module.

[0050] The patient care module uses a nursing assistance model to predict patient complication data. When the model predicts that a patient will experience complications, it issues a prompt message.

[0051] Example 2, see Figure 1 This embodiment is based on the above embodiment. The data acquisition module collects historical physical data, disease data, vital sign data, and complication data of hepatobiliary surgery patients. The physical data includes the patient's age, gender, height, and weight. The disease data refers to the type and stage of the patient's hepatobiliary disease. The vital sign data includes the patient's body temperature, blood pressure, and heart rate. The complication data refers to whether the patient has experienced postoperative bleeding, infection, bile leakage, or liver failure.

[0052] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment, and the data expansion model construction module specifically includes the following:

[0053] The core architecture is generated, and the model includes a generator and a discriminator. The generator consists of an input layer, a hidden layer, and a generation layer. The input layer receives a chaotic vector, the hidden layer is activated by a linear rectified function, and skip connections are introduced in the hidden layer. The generation layer is activated by a hyperbolic tangent function. The generator learns the mapping from the chaotic vector to the real data distribution to generate data similar to the real data. The discriminator is a classifier that outputs a probability value, representing the probability that the input data is the real data.

[0054] The chaotic vector is generated by introducing a random noise vector and the golden ratio, as shown below:

[0055] ;

[0056] Where i represents the dimension index of the data vector. This represents the value of the chaotic vector in the i-th dimension. This represents the sigmoid function. This represents the value of the random noise vector in the i-th dimension;

[0057] To evaluate the generator loss, we innovatively introduce reconstruction loss and gradient regularization to encourage the generator to reconstruct real samples from the chaotic vectors obtained from the encoder, as shown below:

[0058] ;

[0059] in, This represents the loss value of the generator. Represents the chaotic vector Expectations Indicates that z follows the prior distribution Mid-sampling, Represents a generator. Indicates the discriminator, This indicates that the discriminator detects fake samples generated by the generator. The probability of identification, , and Indicates the loss weights of the generator. Indicates the real sample Expectations express From the distribution of real data Mid-sampling, Indicates encoder, This indicates that the encoder encodes real samples as chaotic vectors. This indicates that the generator obtains the chaotic vector from the encoder. The reconstructed sample, This represents taking the square of the L2 norm. This represents the gradient of the generator with respect to the input chaotic vector;

[0060] The discriminator loss is evaluated by combining the real sample loss, generated sample loss, and gradient regularization, as shown below:

[0061] ;

[0062] in, This represents the loss value of the discriminator. This represents the probability that the discriminator can distinguish a real sample x. The loss weights of the discriminator are represented. This represents the sample reconstructed by the generator based on the chaotic vector z;

[0063] The weight update incorporates the discriminator's probability of distinguishing between real and generated samples to design the weight increment, thereby updating the model's weights, as shown below:

[0064] ;

[0065] Where it represents the number of weight updates. Indicates the first Model weights during the second weight update Indicates the first Model weights during the second weight update and This represents the learning rate of the model weights. This represents the increment of model weights. This represents the total number of data samples. Indicates the index of the data sample. Indicates the first The probability of the discriminator identifying the q-th data sample during the next weight update. This indicates that the discriminator operates on the q-th generated data sample. The probability of identification;

[0066] Expand the dataset by using a generator to generate simulation data and adding the simulation data to the original dataset.

[0067] By performing the above operations, this solution addresses the problems of low data quality, insufficient diversity, and difficulty in effectively simulating the distribution of real data in traditional methods for generating nursing data for hepatobiliary surgery patients. It improves the realism and diversity of the generated data by constructing a core architecture that includes a generator and a discriminator, introducing a chaotic vector generation mechanism, and optimizing the loss function and weight update strategy. This allows the solution to more effectively simulate the distribution of real data and thus expand the quality and scale of the dataset.

[0068] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the module for constructing the nursing assistance model specifically includes the following:

[0069] The basic framework of the design includes a reset unit, an update unit, an integration unit, a candidate state unit, and a state update unit, with complication data set as the model's label data.

[0070] The design incorporates a reset unit to control the impact of the previous hidden state on the current state. A weighted sum of historical hidden states is introduced to dynamically adjust the importance of historical information, as shown below:

[0071] ;

[0072] Where t represents the index of the time step, This represents the output value of the reset unit at time step t. Indicates the weight of the reset unit. This indicates the input value for the reset unit. This represents the weight from the previous hidden state to the reset unit. Let k represent the hidden state at time step (t-1), and k represent the index of the time step. This represents the state weight of the reset unit at time step k. This represents the hidden state at the k-th time step;

[0073] The design update unit controls the fusion of the current input and the previous hidden state. By adjusting the fusion ratio of the current input and historical information, a new intermediate state is generated, as shown below:

[0074] ;

[0075] in, This represents the output value of the update unit at time step t. This indicates the weight of the updated unit. This represents the input value of the update unit. This represents the weight from the previous hidden state to the update unit. This represents the state weight of the updating unit at time step k. This indicates the state adjustment bias of the update unit;

[0076] The design integrates a unit that uses the hyperbolic tangent function to capture the nonlinear characteristics in the input data, as shown below:

[0077] ;

[0078] in, This represents the output value of the integrated unit. Represents the hyperbolic tangent function. Indicates the weight of the integration unit. Indicates the input of the integrated unit, This represents the weight from the previous hidden state to the ensemble unit. This represents the state weight of the integrated unit at the k-th time step;

[0079] Design a candidate state unit, combine the output of the reset unit and the output of the integration unit to generate the final candidate state, as shown below:

[0080] ;

[0081] in, This represents the output value of the candidate state unit at time step t. The symbol for multiplication by one variable is used. The weights of the candidate state units are represented. This represents the input value of the candidate state unit. This represents the state weight of the integrated unit at time step t;

[0082] The design incorporates a state update unit. This unit controls the fusion ratio between the previous hidden state and the candidate state through its output. A scaling transformation matrix is ​​introduced to adjust the magnitude of the state update, as shown below:

[0083] ;

[0084] in, This represents the hidden state at time step t. Let represent the scaling transformation matrix at time step t;

[0085] The loss function is designed, and a smoothness constraint is introduced for the hidden state and the predicted value, as follows:

[0086] ;

[0087] in, This represents the loss value of the nursing assistance model, where c represents the data index used for training the nursing assistance model. This represents the total number of data points used for training the nursing assistance model. This represents the true label value of the c-th data point. This represents the model prediction value for the c-th data point. Let T represent the logarithmic function, and T represent the total number of time steps. and Indicates the loss weight. Indicates the first The model prediction value for the c-th data point at time step c. This represents the model prediction value of the c-th data point at the k-th time step. Indicates the first The model prediction value for the c-th data point at time step c.

[0088] By performing the above operations, this solution addresses the problems of insufficient modeling ability for time series data, difficulty in capturing long-term dependencies, and low prediction accuracy of traditional nursing auxiliary models for hepatobiliary surgery patients. It innovatively designs a model framework including reset, update, integration, candidate state, and state update units, and introduces an improved loss function. This enhances the model's ability to model time series data and capture long-term dependencies, thereby improving the prediction accuracy and stability of the nursing auxiliary model.

[0089] Example 5, see Figure 1 This embodiment is based on the above embodiment. The patient care module collects the patient's physical condition data, disease data and vital sign data in real time, and uses a nursing assistance model to predict the patient's complication data. When the model predicts that the patient will have complications, it issues a prompt message to remind medical staff to prepare for the complication response.

[0090] It should 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 process, method, article, or apparatus.

[0091] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0092] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An artificial intelligence-based patient care support system for hepatobiliary surgery, characterized by: The application relates to a liver and gallbladder surgery patient nursing system, which comprises a data collection module, a data expansion model construction module, a nursing auxiliary model construction module and a patient nursing module. The data collection module collects historical physical data, disease data, vital sign data and complication data of liver and gallbladder surgery patients. The data expansion model construction module expands the data set by generating a core architecture, generating a chaotic vector, evaluating generator loss, evaluating discriminator loss and weight updating. The core architecture is generated, and the model comprises a generator and a discriminator; the generator comprises an input layer, a hidden layer and a generation layer; the input layer receives the chaotic vector; the hidden layer is activated by a linear rectification function, and a skip connection is introduced in the hidden layer; the generation layer is activated by a hyperbolic tangent function; the generator learns the mapping from the chaotic vector to the real data distribution to generate data similar to the real data; the discriminator is a classifier, which outputs a probability value representing the probability that the input data is real data. The chaotic vector is generated and is represented as follows: ； where i denotes the dimension index of the data vector, denotes the value of the chaotic vector in the i-th dimension, denotes the sigmoid function, denotes the value of the random noise vector in the i-th dimension; The nursing auxiliary model construction module constructs the nursing auxiliary model by designing a basic framework, a reset unit, an update unit, an integration unit, a candidate state unit, a state update unit and a loss function. The basic framework is designed, and the model comprises a reset unit, an update unit, an integration unit, a candidate state unit and a state update unit; the complication data is set as the label data of the model. The reset unit is designed to control the influence of the previous hidden state on the current state, and the weighted sum of the historical hidden state is introduced to dynamically adjust the importance of the historical information. The update unit is designed to control the fusion of the current input and the previous hidden state, and a new intermediate state is generated by adjusting the fusion ratio of the current input and the historical information. The integration unit is designed to capture the nonlinear characteristics in the input data by using a hyperbolic tangent function. The candidate state unit is designed to combine the output of the reset unit and the output of the integration unit to generate the final candidate state. The state update unit is designed to control the fusion ratio of the previous hidden state and the candidate state through the output of the update unit, and a scale transformation matrix is introduced to adjust the amplitude of the state update. The loss function is designed and is represented as follows: ； wherein, represents a loss value of the care assistance model, c represents a data index for training of the care assistance model, represents a total number of data for training of the care assistance model, represents a true label value of the cth data, represents a model prediction value of the cth data, represents a logarithm function, T represents a total number of time steps, represents a hidden state of the kth time step, represents a hidden state of the k-1th time step, and represents a loss weight, represents a model prediction value of the cth data at the kth time step, represents a model prediction value of the cth data at the kth time step, represents a model prediction value of the cth data at the kth time step, represents a model prediction value of the cth data at the kth time step, represents a model prediction value of the cth data at the kth time step. The patient nursing module predicts the complication data of the patient by using the nursing auxiliary model, and sends a prompt information when the model predicts that the patient will have complications.

2. The patient care assistance system for hepatobiliary surgery based on artificial intelligence according to claim 1, characterized in that: The data expansion model construction module specifically comprises the following contents: Generating a core architecture; Generating a chaotic vector; Evaluating the generator loss, which is represented as follows: ； wherein, represents a loss value of the generator, represents an expectation of a chaotic vector , represents z sampled from a prior distribution , represents the generator, represents the discriminator, represents a discrimination probability of the discriminator on a fake sample generated by the generator, , and represent loss weights of the generator, represents an expectation on a real sample , represents z sampled from a real data distribution , represents the encoder, represents that the encoder encodes a real sample into a chaotic vector, represents a sample reconstructed by the generator according to the chaotic vector obtained by the encoder , represents taking a square of an L2 norm, represents a gradient of the generator on an input chaotic vector; Evaluating the discriminator loss, which is represented as follows: ； wherein, denotes a loss value of the discriminator, denotes a discrimination probability of the discriminator for a real sample x, denotes a loss weight of the discriminator, denotes a sample reconstructed by the generator according to the chaotic vector z; Weight updating, which is represented as follows: ； wherein it represents the number of weight updates, represents the model weight at the th weight update, represents the model weight at the th weight update, and represents the model weight learning rate, represents the model weight increment, represents the total number of data samples, represents the index of the data sample, represents the discrimination probability of the discriminator on the qth data sample at the th weight update, represents the discrimination probability of the discriminator on the qth generated data sample ; Expanding the data set by using the generator to generate simulation data and adding the simulation data to the original data set to expand the original data set. 3.The patient care assistance system for hepatobiliary surgery based on artificial intelligence according to claim 1, wherein: The data collection module collects historical physical data, disease data, vital sign data and complication data of liver and gallbladder surgery patients; the physical data comprises the age, gender, height and weight of the patient; the disease data refers to the type and stage of the liver and gallbladder diseases of the patient; the vital sign data comprises the body temperature, blood pressure and heart rate of the patient; and the complication data refers to whether the patient has postoperative bleeding, infection, biliary fistula and liver failure. 4.The patient care assistance system for hepatobiliary surgery based on artificial intelligence according to claim 1, wherein: The patient care module collects patients' physical condition data, disease data, and vital sign data in real time, and uses a nursing assistance model to predict patients' complication data. When the model predicts that a patient will have a complication, it issues a prompt message to remind medical staff to prepare for the complication response.

Images