Patient portrait generation method and system based on artificial intelligence
By preprocessing and feature learning of the patient data set, building a prediction model and iterative training, combined with encryption processing, the shortcomings of patient portrait generation in the existing technology are solved, and more accurate and comprehensive patient portrait generation is achieved, making it personalized Medical care provides more reliable support.
Patent Information
- Application Number
- CN202411227686.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-03
- Publication Date
- 2024-11-15
AI Technical Summary
The existing patient portrait generation technology has problems such as insufficient data preprocessing, inaccurate feature extraction, and poor model training results, resulting in the generated patient portraits being inaccurate and comprehensive enough, and lacking stability.
Using artificial intelligence-based methods, preprocessing and feature learning of patient data sets, building prediction models, and stop training through iterative conditions, using feature extraction networks and convolutional layers and other technologies to generate patient portrait features, and combining encryption processing to ensure data security .
It improves the accuracy and comprehensiveness of patient portraits, provides more reliable support for personalized medical care and precise treatment, and enhances the adaptability and accuracy of data processing.
Smart Images

Figure CN118969311A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of personalized medicine, and in particular to a method and system for generating a patient portrait based on artificial intelligence. Background Art
[0002] With the rapid development of information technology, artificial intelligence (AI) is increasingly being used in the medical field. In particular, significant progress has been made in recent years in the generation of patient portraits. Traditional patient portrait generation techniques mainly rely on doctors' subjective judgment and basic data analysis, which are usually based on patients' basic health records and medical history information. However, with the accumulation of electronic health records (EHR), genomics data, and wearable device data, modern technology has gradually introduced more complex algorithms, such as machine learning (ML) and deep learning (DL), to extract valuable health information from massive data. These technologies have greatly improved the accuracy and practicality of patient portraits through automated data processing and pattern recognition capabilities. For example, deep learning-based algorithms can analyze complex information from multiple data sources, generate detailed patient health portraits, and predict disease risks, thereby supporting personalized medicine and precision treatment.
[0003] Although existing technologies have achieved initial results in the field of patient portrait generation, there are still some significant deficiencies. Existing methods often face problems such as insufficient data preprocessing, inaccurate feature extraction, and poor model training effects. For example, deficiencies in the data preprocessing stage may cause noise data to interfere with feature extraction, affecting the accuracy of the final portrait. In addition, existing feature learning technologies often find it difficult to effectively integrate information from different data sources, resulting in the generated patient portraits being incomplete or lacking in depth. There are also problems of overfitting or underfitting in the training process of the prediction model, which affects the generalization ability and stability of the model. Summary of the invention
[0004] In order to overcome the defects in the above-mentioned prior art, the present invention provides a patient portrait generation method based on artificial intelligence, including: preprocessing a patient data set; performing feature learning on the preprocessed patient data set to extract patient portrait features; constructing a prediction model based on the patient portrait features, and training the prediction model, and stopping the training when an iteration condition is met; and generating a patient portrait using the trained prediction model.
[0005] As a preferred solution of the patient portrait generation method based on artificial intelligence described in the present invention, the extracting of patient portrait features includes: inputting the preprocessed patient data set into the feature extraction network for feature learning, the patient data set includes the patient's age, gender, genetic characteristics, and electronic health records, and the feature extraction network includes an input layer, a bidirectional LSTM layer, a fully connected layer, an encoder and a decoder; the forward and reverse information in the patient data set is captured by the bidirectional LSTM layer, and then the output of the bidirectional LSTM layer is mapped to the latent space through the fully connected layer; the encoder is connected to the fully connected layer to calculate the mean and variance of the latent space and generate latent samples; the decoder is connected to the encoder to decode and reconstruct the latent samples and output patient portrait features.
[0006] As a preferred solution of the artificial intelligence-based patient portrait generation method described in the present invention, wherein: the prediction model includes an input layer, a convolution layer, a feature extraction layer, a pooling layer and an output layer; the convolution layer includes 2 DW convolution kernels and 2 PW convolution kernels, the DW convolution kernel size is 1*1, and the PW convolution kernel size is 3*3; the feature extraction layer includes a residual layer and a CSPDarknet53 network layer; the residual layer includes a BN layer and a convolution kernel of size 1*1, and the BN layer is arranged before the convolution kernel of size 1*1; the pooling layer is overlapping pooling, and the pooling window size is 5*5.
[0007] As a preferred solution of the method for generating patient portraits based on artificial intelligence described in the present invention, the iteration conditions include: freezing the convolution layer and the feature extraction module, and adding a maximum pooling layer after the feature extraction module to reconstruct the prediction model; training the reconstructed prediction model, setting the learning rate to 0.001, and stopping the training when the number of training times reaches the maximum value and the loss function value converges, and the loss function value Loss is:
[0008]
[0009] Where N is the number of patient portrait features, y′ is the actual output value of the prediction model, y is the expected output value of the prediction model, W is the weight parameter of the recognition model, α is the regularization coefficient, and ‖w‖1 is the first-order norm of the regularization coefficient α.
[0010] As a preferred solution of the method for generating a patient portrait based on artificial intelligence described in the present invention, it further includes: encrypting the patient data set; dividing the patient data set into n data blocks of the same size, and encrypting the data blocks; placing the n data blocks in order from left to right or from top to bottom Matrix, and The key matrix is XORed; the XOR result is permuted, all the permuted results are cyclically shifted left by two bits, and the cyclic left shift result is XORed with the round constant to complete the encryption; wherein the key matrix includes n random keys, and the random keys are generated by the key derivation function.
[0011] As a preferred solution of the artificial intelligence-based patient portrait generation system described in the present invention, it includes: a data preprocessing module, configured to perform preprocessing of the patient data set; a feature extraction module, configured to perform feature learning on the preprocessed patient data set and extract patient portrait features; a model construction module, configured to execute construction of a prediction model based on the patient portrait features and train the prediction model, and stop training when the iteration condition is met; a patient portrait generation module, configured to generate a patient portrait using the trained prediction model.
[0012] As a preferred solution of the patient portrait generation method based on artificial intelligence described in the present invention, wherein: the feature extraction module is specifically configured to execute: inputting the preprocessed patient data set into the feature extraction network for feature learning, the patient data set includes the patient's age, gender, genetic characteristics, and electronic health records, and the feature extraction network includes an input layer, a bidirectional LSTM layer, a fully connected layer, an encoder and a decoder; the forward and reverse information in the patient data set is captured by the bidirectional LSTM layer, and then the output of the bidirectional LSTM layer is mapped to the latent space through the fully connected layer; the encoder is connected to the fully connected layer to calculate the mean and variance of the latent space and generate latent samples; the decoder is connected to the encoder to decode and reconstruct the latent samples and output patient portrait features.
[0013] As a preferred solution of the patient portrait generation method based on artificial intelligence described in the present invention, wherein: the model construction module is specifically configured to execute: constructing a prediction model including an input layer, a convolution layer, a feature extraction layer, a pooling layer and an output layer; the convolution layer includes 2 DW convolution kernels and 2 PW convolution kernels, the DW convolution kernel size is 1*1, and the PW convolution kernel size is 3*3; the feature extraction layer includes a residual layer and a CSPDarknet53 network layer; the residual layer includes a BN layer and a convolution kernel of size 1*1, and the BN layer is arranged before the convolution kernel of size 1*1; the pooling layer is overlapping pooling, and the pooling window size is 5*5.
[0014] As a preferred solution of the method for generating patient portraits based on artificial intelligence described in the present invention, the model building module is specifically configured to execute: freezing the convolution layer and the feature extraction module, and adding a maximum pooling layer after the feature extraction module to reconstruct the prediction model; training the reconstructed prediction model, setting the learning rate to 0.001, and stopping the training when the number of training times reaches the maximum value and the loss function value converges, and the loss function value Loss is:
[0015]
[0016] Where N is the number of patient portrait features, y′ is the actual output value of the prediction model, y is the expected output value of the prediction model, W is the weight parameter of the recognition model, α is the regularization coefficient, and ‖w‖1 is the first-order norm of the regularization coefficient α.
[0017] As a preferred solution of the method for generating a patient portrait based on artificial intelligence of the present invention, it further includes a data encryption module configured to execute: dividing the patient data set into n data blocks of the same size and encrypting the data blocks; placing the n data blocks in sequence from left to right or from top to bottom Matrix, and The key matrix is XORed; the XOR result is permuted, all the permuted results are cyclically shifted left by two bits, and the cyclic left shift result is XORed with the round constant to complete the encryption; wherein the key matrix includes n random keys, and the random keys are generated by the key derivation function.
[0018] Beneficial effects of the present invention: Compared with traditional technologies, the present invention has stronger adaptability and higher accuracy in processing complex data, can generate more accurate and comprehensive patient portraits, and provides more reliable support for personalized medicine and precision treatment. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:
[0020] Figure 1 This is a flowchart of the method for generating a patient portrait based on artificial intelligence according to the first embodiment of the present invention. DETAILED DESCRIPTION
[0021] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.
[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.
[0024] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.
[0025] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.
[0026] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0027] Example 1
[0028] Reference Figure 1, which is the first embodiment of the present invention, provides a method for generating a patient portrait based on artificial intelligence, comprising:
[0029] S1: Preprocess the patient dataset.
[0030] Before building a patient portrait, the original data needs to be preprocessed to ensure the quality and consistency of the data in order to improve the accuracy and robustness of the model. The patient data set includes the patient's age, gender, genetic characteristics, and electronic health records (including patient medical records, treatment records, drug use, etc.). The data preprocessing process includes the following steps:
[0031] Data cleaning: including missing value processing, outlier detection and removal, which can be done by mean filling, median filling, interpolation, etc.
[0032] Data standardization and normalization: Since different medical data have different dimensions (such as blood pressure, body temperature, gene expression levels, etc.), it is necessary to map the data to the same dimension range through standardization (Z-score) or normalization (Min-Max Scaling). This can not only accelerate the convergence of the algorithm, but also prevent certain features from causing unnecessary deviations to the model in terms of distance measurement.
[0033] Data dimensionality reduction and sampling: In high-dimensional data, dimensionality reduction can reduce feature redundancy and improve model efficiency. In addition to PCA and LDA, more dimensionality reduction techniques can be used, such as t-SNE (suitable for visualization of high-dimensional data) and UMAP (non-linear dimensionality reduction, retaining local structure). In the case of unbalanced samples, sampling techniques (such as oversampling, undersampling, SMOTE) can be combined to balance the data distribution.
[0034] In the process of building patient portraits, in order to protect the privacy of patients, the patient data set needs to be encrypted to ensure the security and privacy of the data. Specifically, the patient data set is encrypted according to the following steps:
[0035] Divide the patient data set into n data blocks of equal size and encrypt the data blocks;
[0036] Place n data blocks in order from left to right or from top to bottom Matrix, and The key array is XORed;
[0037] The XOR operation result is permuted, all the permuted results are cyclically shifted left by two bits, and the cyclic left shift result is XORed with the round constant to complete the encryption;
[0038] The key matrix includes n random keys, and the random keys are generated by a key derivation function.
[0039] A key derivation function uses a pseudorandom function to derive one or more keys from a secret value. KDF can be used to extend a key to a longer key or to obtain a key in a desired format. Key derivation functions are often used with non-secret parameters to derive one or more keys from a common secret value; in this way, an attacker who obtains the derived key is prevented from learning useful information about the input secret value or any other derived key. KDF can also be used to ensure that the derived key has other desired properties, for example, avoiding "weak keys" in some specific encryption systems.
[0040] S2: Perform feature learning on the preprocessed patient dataset to extract patient portrait features.
[0041] The preprocessed patient data set is input into the feature extraction network for feature learning. The feature extraction network includes an input layer, a bidirectional LSTM (Long Short-Term Memory) layer, a fully connected layer, an encoder, and a decoder.
[0042] The bidirectional LSTM layer includes two directional LSTM layers, one for processing the forward sequence and the other for processing the reverse sequence, so that the past and future information in the data can be learned at the same time. The bidirectional LSTM layer captures the forward and reverse information in the patient data set to improve the feature extraction capability, and then the output of the bidirectional LSTM layer is mapped to the latent space through the fully connected layer, and the representation of the latent space is output through linear transformation;
[0043] The encoder is connected to the fully connected layer. The encoder includes one or more fully connected layers for calculating the mean and variance of the latent space and using the mean and variance to generate latent samples from the standard normal distribution.
[0044] The decoder is connected to the encoder. The decoder includes one or more fully connected layers, which are used to decode and reconstruct the potential samples and output the patient portrait features.
[0045] S3: Build a prediction model based on the patient portrait features and train the prediction model. When the iteration conditions are met, stop training.
[0046] The prediction model includes input layer, convolution layer, feature extraction layer, pooling layer and output layer;
[0047] The convolution layer includes 2 DW convolution kernels (Depthwise Convolution) and 2 PW (Pointwise Convolution) convolution kernels. The size of the DW convolution kernel is 1*1. The DW convolution kernel processes the input data channel by channel. Each convolution kernel only performs convolution operation on one channel of the input to reduce the computational complexity. The size of the PW convolution kernel is 3*3. A 3*3 convolution is applied at each position to combine the output of the deep convolution and increase the nonlinear expression ability of the network.
[0048] The feature extraction layer includes a residual layer and a CSPDarknet53 network layer; the residual layer includes a BN layer (BatchNormalization) and a 1*1 convolution kernel. The BN layer is set before the 1*1 convolution kernel to normalize the input and accelerate the training process. The CSPDarknet53 network is used for deep feature extraction to capture more advanced features.
[0049] The pooling layer is Overlapping Pooling, and the pooling window size is 5*5, which reduces the spatial dimension of the feature map, retains key information, and enhances the local information of the feature through overlapping pooling.
[0050] Iteration conditions include:
[0051] Freeze the convolutional layer and feature extraction module, and add a maximum pooling layer after the feature extraction module to reconstruct the prediction model;
[0052] The reconstructed prediction model is trained, and the learning rate is set to 0.001. The training is stopped when the number of training times reaches the maximum value and the loss function value converges. The loss function value Loss is:
[0053]
[0054] Where N is the number of patient portrait features, y′ is the actual output value of the prediction model, y is the expected output value of the prediction model, W is the weight parameter of the recognition model, α is the regularization coefficient, and ‖w‖1 is the first-order norm of the regularization coefficient α.
[0055] S4: Generate patient profiles using the trained prediction model.
[0056] A patient's age, gender, genetic characteristics, and electronic health records are input into the trained prediction model. The prediction model combines genomic data (genetic variants associated with cardiovascular disease) and her medical records (past heart attacks and treatment responses). Through comprehensive analysis by the prediction model, a health risk map is output, showing the risk distribution of cardiovascular disease, influencing factors and predicted health intervention effects.
[0057] Example 2
[0058] This embodiment is different from the first embodiment in that it provides a patient portrait generation system based on artificial intelligence, including:
[0059] A data preprocessing module is configured to perform preprocessing on the patient data set;
[0060] The data encryption module is configured to perform: dividing the patient data set into n data blocks of equal size and encrypting the data blocks; placing the n data blocks in sequence from left to right or from top to bottom Matrix, and The key matrix is XORed; the XOR result is permuted, all permuted results are circularly shifted left by two bits, and the circularly shifted left result is XORed with the round constant to complete the encryption; wherein, the key matrix includes n random keys, and the random keys are generated by the key derivation function. The key derivation function uses a pseudo-random function to derive one or more keys from a secret value. KDF can be used to extend a key to a longer key or obtain a key in a desired format. The key derivation function is usually used with non-secret parameters to derive one or more keys from a public secret value; in this way, an attacker who obtains a derived key can be prevented from learning useful information about the input secret value or any other derived key. KDF can also be used to ensure that the derived key has other desired properties, for example, avoiding "weak keys" in certain specific encryption systems.
[0061] The feature extraction module is configured to perform feature learning on the preprocessed patient data set and extract patient portrait features; the feature extraction module is specifically configured to perform: inputting the preprocessed patient data set into the feature extraction network for feature learning, the patient data set includes the patient's age, gender, genetic characteristics, and electronic health records, the feature extraction network includes an input layer, a bidirectional LSTM layer, a fully connected layer, an encoder and a decoder; capturing the forward and reverse information in the patient data set through the bidirectional LSTM layer, and then mapping the output of the bidirectional LSTM layer to the latent space through the fully connected layer; the encoder is connected to the fully connected layer to calculate the mean and variance of the latent space and generate latent samples; the decoder is connected to the encoder to decode and reconstruct the latent samples and output patient portrait features.
[0062] The model building module is configured to execute the construction of a prediction model based on the patient portrait features and train the prediction model. When the iteration condition is met, the training is stopped. The model building module is specifically configured to execute: the prediction model includes an input layer, a convolution layer, a feature extraction layer, a pooling layer, and an output layer; the convolution layer includes 2 DW convolution kernels and 2 PW convolution kernels, the DW convolution kernel size is 1*1, and the PW convolution kernel size is 3*3; the feature extraction layer includes a residual layer and a CSPDarknet53 network layer; the residual layer includes a BN layer and a convolution kernel size of 1*1, and the BN layer is set before the convolution kernel size of 1*1. The pooling layer is an overlapping pooling, and the pooling window size is 5*5.
[0063] The model building module is specifically configured to perform: freezing the convolutional layer and feature extraction module, and adding a maximum pooling layer after the feature extraction module to reconstruct the prediction model; training the reconstructed prediction model, setting the learning rate to 0.001, and stopping the training when the number of training times reaches the maximum value and the loss function value converges. The loss function value Loss is:
[0064]
[0065] Where N is the number of patient portrait features, y′ is the actual output value of the prediction model, y is the expected output value of the prediction model, W is the weight parameter of the recognition model, α is the regularization coefficient, and ‖w‖1 is the first-order norm of the regularization coefficient α.
[0066] The patient portrait generation module is configured to generate a patient portrait using the trained prediction model.
[0067] It should be appreciated that embodiments of the present invention may be implemented or enforced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program may be implemented in an assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed ASIC for this purpose.
[0068] Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions that may be executed by one or more processors.
[0069] Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, a RAM, a ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or part thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques described in the present invention, the present invention also includes the computer itself. The computer program can be applied to input data to perform the functions described herein, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents a physical and tangible object, including a specific visual depiction of the physical and tangible object produced on a display.
[0070] As used in this application, the terms "component", "module", "system", etc. are intended to refer to a computer-related entity, which can be hardware, firmware, a combination of hardware and software, software, or software in operation. For example, a component can be, but is not limited to: a process running on a processor, a processor, an object, an executable file, a thread in execution, a program, and / or a computer. As an example, an application running on a computing device and the computing device can both be components. One or more components can exist in a process and / or thread in execution, and a component can be located in a computer and / or distributed between two or more computers. In addition, these components can be executed from various computer-readable media having various data structures thereon. These components can communicate in a local and / or remote process manner, such as according to a signal having one or more data packets (e.g., data from a component that interacts with another component in a local system, a distributed system, and / or interacts with other systems in a signal manner through a network such as the Internet).
[0071] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. A method for generating a patient portrait based on artificial intelligence, characterized in that: include: Preprocess the patient dataset; Perform feature learning on the preprocessed patient data set to extract patient portrait features; Building a prediction model based on the patient portrait features, and training the prediction model, and stopping the training when an iteration condition is met; Generate patient portraits using the trained predictive model.
2. The method for generating a patient portrait based on artificial intelligence according to claim 1, characterized in that: The extracting of patient portrait features includes: Inputting the preprocessed patient data set into a feature extraction network for feature learning, wherein the patient data set includes patient age, gender, genetic characteristics, and electronic health records, and the feature extraction network includes an input layer, a bidirectional LSTM layer, a fully connected layer, an encoder, and a decoder; The forward and reverse information in the patient dataset is captured by the bidirectional LSTM layer, and then the output of the bidirectional LSTM layer is mapped to the latent space by a fully connected layer; The encoder is connected to the fully connected layer to calculate the mean and variance of the latent space and generate latent samples; The decoder is connected to the encoder and is used to decode and reconstruct the potential samples and output patient portrait features.
3. The method for generating a patient portrait based on artificial intelligence according to claim 2, characterized in that: The prediction model includes an input layer, a convolution layer, a feature extraction layer, a pooling layer and an output layer; The convolution layer includes 2 DW convolution kernels and 2 PW convolution kernels, the size of the DW convolution kernel is 1*1, and the size of the PW convolution kernel is 3*3; The feature extraction layer includes a residual layer and a CSPDarknet53 network layer; the residual layer includes a BN layer and a convolution kernel of size 1*1, and the BN layer is arranged before the convolution kernel of size 1*1; The pooling layer is overlapping pooling, and the pooling window size is 5*5.
4. The method for generating a patient portrait based on artificial intelligence according to claim 3, characterized in that: The iteration conditions include: Freezing the convolutional layer and the feature extraction module, and adding a maximum pooling layer after the feature extraction module to reconstruct the prediction model; The reconstructed prediction model is trained, and the learning rate is set to 0.
001. The training is stopped when the number of training times reaches the maximum value and the loss function value converges. The loss function value Loss is: Where N is the number of patient portrait features, y′ is the actual output value of the prediction model, y is the expected output value of the prediction model, W is the weight parameter of the recognition model, α is the regularization coefficient, and ‖w‖1 is the first-order norm of the regularization coefficient α.
5. The method for generating a patient portrait based on artificial intelligence according to any one of claims 1 to 4, characterized in that: Also included is encrypting said patient data set; Dividing the patient data set into n data blocks of equal size, and encrypting the data blocks; Place n data blocks in order from left to right or from top to bottom Matrix, and The key array is XORed; The XOR operation result is permuted, all the permuted results are cyclically shifted left by two bits, and the cyclic left shift result is XORed with the round constant to complete the encryption; The key matrix includes n random keys, and the random keys are generated by a key derivation function.
6. A patient portrait generation system based on artificial intelligence, characterized in that: include: A data preprocessing module is configured to perform preprocessing on the patient data set; A feature extraction module is configured to perform feature learning on the preprocessed patient data set to extract patient portrait features; A model building module is configured to build a prediction model based on the patient portrait features, train the prediction model, and stop training when an iteration condition is met; The patient portrait generation module is configured to generate a patient portrait using the trained prediction model.
7. The patient portrait generation system based on artificial intelligence according to claim 6, characterized in that: The feature extraction module is specifically configured to perform: Inputting the preprocessed patient data set into a feature extraction network for feature learning, wherein the patient data set includes patient age, gender, genetic characteristics, and electronic health records, and the feature extraction network includes an input layer, a bidirectional LSTM layer, a fully connected layer, an encoder, and a decoder; The forward and reverse information in the patient dataset is captured by the bidirectional LSTM layer, and then the output of the bidirectional LSTM layer is mapped to the latent space by a fully connected layer; The encoder is connected to the fully connected layer to calculate the mean and variance of the latent space and generate latent samples; The decoder is connected to the encoder and is used to decode and reconstruct the potential samples and output patient portrait features.
8. The patient portrait generation system based on artificial intelligence according to claim 7, characterized in that: The model building module is specifically configured to perform: Constructing a prediction model includes input layer, convolution layer, feature extraction layer, pooling layer and output layer; The convolution layer includes 2 DW convolution kernels and 2 PW convolution kernels, the size of the DW convolution kernel is 1*1, and the size of the PW convolution kernel is 3*3; The feature extraction layer includes a residual layer and a CSPDarknet53 network layer; the residual layer includes a BN layer and a convolution kernel of size 1*1, and the BN layer is arranged before the convolution kernel of size 1*1; The pooling layer is overlapping pooling, and the pooling window size is 5*5.
9. The patient portrait generation system based on artificial intelligence according to claim 8, characterized in that: The model building module is specifically configured to perform: Freezing the convolutional layer and the feature extraction module, and adding a maximum pooling layer after the feature extraction module to reconstruct the prediction model; The reconstructed prediction model is trained, and the learning rate is set to 0.
001. The training is stopped when the number of training times reaches the maximum value and the loss function value converges. The loss function value Loss is: Where N is the number of patient portrait features, y′ is the actual output value of the prediction model, y is the expected output value of the prediction model, W is the weight parameter of the recognition model, α is the regularization coefficient, and ‖w‖1 is the first-order norm of the regularization coefficient α.
10. The artificial intelligence-based patient portrait generation system according to any one of claims 6 to 9, characterized in that: Also included is a data encryption module configured to perform: Dividing the patient data set into n data blocks of equal size, and encrypting the data blocks; Place n data blocks in order from left to right or from top to bottom Matrix, and The key array is XORed; The XOR operation result is permuted, all the permuted results are cyclically shifted left by two bits, and the cyclic left shift result is XORed with the round constant to complete the encryption; The key matrix includes n random keys, and the random keys are generated by a key derivation function.
Citation Information
Cited By
Chronic disease patient portrait generation method and system based on big data
CN121478843A
