Heart disease risk assessment method based on deep learning algorithm

Abstract

The invention discloses a heart disease risk assessment method based on a deep learning algorithm. The method comprises the following steps: carrying out standardization processing on a data set through a robust standardization method; performing feature preprocessing on the input physical sign data; the method comprises the following steps: acquiring feature information through a Swin Transform module; and performing parameter optimization on the model by using an optimizer. The method can more accurately predict the risk of the individual suffering from the heart disease, thereby being beneficial to early intervention and treatment by medical care professionals so as to improve the life quality of the patient.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data analysis, and in particular to a heart disease risk assessment method based on a deep learning algorithm. Background Art

[0002] Heart disease is a common and serious cardiovascular disease in life. Cardiovascular disease is one of the biggest threats to the health of people in my country and even the world. This disease has brought a serious burden to my country's medical system. Therefore, it is necessary to train a heart disease prediction model based on existing medical data to provide health guidance for patients.

[0003] However, the existing field of heart disease prediction faces an important challenge: it is difficult to accurately extract features related to heart disease risk from large-scale physical examination data and build an interpretable heart disease prediction model. Summary of the invention

[0004] The purpose of the present invention is to provide a heart disease risk assessment method based on a deep learning algorithm, which can more accurately predict an individual's risk of developing heart disease, thereby helping healthcare professionals to conduct early intervention and treatment to improve the patient's quality of life.

[0005] In order to achieve the above object, the present invention provides a method for assessing heart disease risk based on a deep learning algorithm, the method comprising:

[0006] The dataset was standardized using a robust normalization method;

[0007] Perform feature preprocessing on the input vital sign data;

[0008] Obtain feature information through the Swin Transformer module;

[0009] Use the optimizer to tune the model parameters.

[0010] Preferably, the standardization of the data set by the robust standardization method includes standardizing the data according to formula (1):

[0011]

[0012] Among them, x is the original data point, M is the median of the data, Q1 is the first quartile, and Q3 is the third quartile.

[0013] Preferably, feature preprocessing of the input vital sign data includes multi-branch multi-scale convolution to extract input features and channel attention calculation, feature fusion, feature map data filling and dimensionality enhancement.

[0014] Preferably, the multi-branch multi-scale convolution extracts input features and calculates channel attention, including: selecting a neural network with convolution kernel sizes of 1, 3, and 5, setting the padding of the convolution kernel and performing a convolution operation, calculating the weight of each channel, and adjusting the channel weight of the feature map to output three different feature maps.

[0015] Preferably, feature fusion includes input feature preparation, feature concatenation and outputting fused features.

[0016] Preferably, feature map data padding and dimensionality enhancement include analyzing size differences, determining padding locations and amounts, performing padding operations, understanding target dimensional structures, and performing dimensionality expansion operations.

[0017] Preferably, the internal components of the Swin Transformer module include a Patch merging layer, a window multi-head self-attention mechanism W-MSA, and a sliding window multi-head self-attention mechanism SW-MSA, wherein:

[0018] The patch merging layer is used to retain more feature information during downsampling;

[0019] The window multi-head self-attention mechanism W-MSA is used to limit the self-attention calculation to a window of a specific size;

[0020] The sliding window multi-head self-attention mechanism SW-MSA is used to move the window.

[0021] Preferably, using an optimizer to tune the parameters of the model includes using an Adam optimizer to update the parameters:

[0022] According to formula (2), the first-order moment estimate m is calculated t Updated value of:

[0023] m t =β1* m t-1 + (1- β1) *g t (2)

[0024] According to formula (3), the second-order moment estimate v is calculated t Updated value of:

[0025]

[0026] And according to formula (4), using the modified m ′ t and v t ′ Update model parameters θ t :

[0027]

[0028] Among them, g t is the gradient of the current time step, β1 and β2 are hyperparameters used to control the update speed, and m t-1 is the first-order moment estimate of the previous time step t-1, v t-1 is the second-order moment estimate of the previous time step t-1, α is the learning rate, and θ t-1 are the model parameters at the previous time step, and ε is a constant used to prevent the denominator from being zero.

[0029] Preferably, the important parameters updated include the number of iterations Epochs, the batch size Batch Size, the learning rate Learning_rate and the L1 regularization coefficient Reg_alpha.

[0030] According to the above technical scheme, the present invention first obtains the original data set, and then uses a robust normalization method to standardize the data set so that the original data has the same range; then, the input vital sign data is preliminarily processed through the feature preprocessing module to meet the input requirements of the subsequent backbone network; then, the feature preprocessing is sent to the Swin Transformer module, and different components inside the module are used to obtain rich feature information while reducing the amount of calculation; finally, the optimizer is used to tune the parameters during the model training process to further improve the performance of the model on this data set. In this way, the method can more accurately predict the risk of an individual suffering from heart disease, thereby helping healthcare professionals to conduct early intervention and treatment and improve the quality of life of patients. In addition, the method also provides a powerful tool for medical research to be able to analyze and predict the risk of heart disease on large-scale data sets to promote progress in the field of disease prevention and management.

[0031] Other features and advantages of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention but do not constitute a limitation of the present invention. In the accompanying drawings:

[0033] Figure 1 is a flow chart of a heart disease risk assessment method based on a deep learning algorithm provided according to the present invention. DETAILED DESCRIPTION

[0034] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0035] See also Figure 1 The present invention provides a method for assessing heart disease risk based on a deep learning algorithm, the method comprising:

[0036] The dataset was standardized using a robust normalization method;

[0037] Perform feature preprocessing on the input vital sign data;

[0038] Obtain feature information through the Swin Transformer module;

[0039] Use the optimizer to tune the model parameters.

[0040] Specifically, in this embodiment, in order to make the original data have the same range, it is preferred to standardize the data set by a robust standardization method, including standardizing the data according to formula (1):

[0041]

[0042] Among them, x is the original data point, M is the median of the data, Q1 is the first quartile, and Q3 is the third quartile.

[0043] In this embodiment, in order to make the input vital sign data meet the input requirements of the subsequent backbone network, preferably, feature preprocessing of the input vital sign data includes multi-branch multi-scale convolution to extract input features and channel attention calculation, feature fusion, feature map data filling and dimensionality enhancement.

[0044] Among them, the above-mentioned multi-branch multi-scale convolution extracts input features and calculates channel attention, including: selecting neural networks with convolution kernel sizes of 1, 3, and 5, setting the padding of the convolution kernel and performing convolution operations, calculating the weight of each channel, adjusting the channel weight of the feature map, and outputting three different feature maps.

[0045] In this embodiment, the feature fusion includes input feature preparation, feature concatenation, and output fused features.

[0046] In this embodiment, the above-mentioned feature map data filling and dimensionality enhancement include analyzing size differences, determining filling positions and quantities, performing filling operations, understanding target dimensional structures, and performing dimensionality expansion operations.

[0047] In this embodiment, in order to utilize different internal components of the Swin Transformer module to obtain rich feature information while reducing the amount of calculation, the internal components of the Swin Transformer module preferably include a patch merging layer, a window multi-head self-attention mechanism W-MSA, and a sliding window multi-head self-attention mechanism SW-MSA, wherein:

[0048] The patch merging layer is used to retain more feature information during downsampling;

[0049] The window multi-head self-attention mechanism W-MSA is used to limit the self-attention calculation to a window of a specific size;

[0050] The sliding window multi-head self-attention mechanism SW-MSA is used to move the window;

[0051] In this way, the above components can reduce a certain amount of calculation, realize feature interaction between global features and local features, enrich feature information, and thus improve model efficiency.

[0052] During the model training process, in order to further improve the performance of the model on this data set, preferably, the optimizer is used to tune the model parameters, including using the Adam optimizer to update the parameters:

[0053] According to formula (2), the first-order moment estimate m is calculated t Updated value of:

[0054] m t =β1* m t-1 + (1- β1) *g t (2)

[0055] According to formula (3), the second-order moment estimate v is calculated t Updated value of:

[0056]

[0057] And according to formula (4), using the modified m ′ t and v t ′ Update model parameters θ t :

[0058]

[0059] Among them, g t is the gradient of the current time step, β1 and β2 are hyperparameters used to control the update speed, and m t-1 is the first-order moment estimate of the previous time step t-1, v t-1 is the second-order moment estimate of the previous time step t-1, α is the learning rate, and θ t-1 are the model parameters at the previous time step, and ε is a constant used to prevent the denominator from being zero.

[0060] A specific embodiment is provided below to illustrate the present invention:

[0061] S1: The original data set was collected, including 13 indicator characteristic variables: age, gender, chest pain type, resting blood pressure, serum cholesterol, resting electrocardiogram, fasting blood glucose, maximum heart rate, exercise-induced angina, exercise-induced ST value, exercise peak ST slope, number of large vessels, and thalassemia;

[0062] S2: The data set is standardized using a robust normalization method so that the original data have the same range;

[0063] S3: Preliminary processing of the input vital sign data through the feature preprocessing module to make it meet the input requirements of the subsequent backbone network;

[0064] S4: After feature preprocessing, the features are sent to the Swin Transformer module, and different components inside the module are used to obtain rich feature information while reducing the amount of calculation;

[0065] S5: During the model training process, use the optimizer to tune parameters to further improve the performance of the model on this dataset.

[0066] Specifically, in S2, the data is standardized:

[0067]

[0068] Where x is the original data point, M is the median of the data, Q1 is the first quartile (25% quantile), and Q3 is the third quartile (75% quantile).

[0069] In S3, the process of standardizing vital sign data includes multi-branch multi-scale convolution to extract input features, channel attention calculation, feature fusion, feature map data padding and dimension enhancement operations. Among them, a neural network with different convolution kernels is used for preliminary feature extraction. The input features are extracted through multi-branch multi-scale convolution, and the padding of the convolution kernel is set to make the output feature map size the same, and the receptive fields are different to extract feature data of different granularity to avoid information loss.

[0070] In S5, training uses the Adam optimizer for parameter update. Several important parameters involved include Epochs (number of iterations), Batch Size (batch size), Learning_rate (learning rate), and Reg_alpha (L1 regularization coefficient).

[0071] In summary, this heart disease risk assessment method based on deep learning algorithm combines multiple technologies to screen out relevant factors related to heart disease, and uses this method to explore new risk factors, treatment methods and prevention strategies for heart disease, thereby promoting the progress of medical research. At the same time, this method optimizes the parameters of the model, which further improves the model's predictive ability on this data set.

[0072] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0073] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0074] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0075] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0076] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0077] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0078] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0079] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0080] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A heart disease risk assessment method based on deep learning algorithm, characterized in that: The method comprises: The dataset was standardized using a robust normalization method; Perform feature preprocessing on the input vital sign data; Obtain feature information through the Swin Transformer module; Use the optimizer to tune the model parameters.

2. The heart disease risk assessment method based on deep learning algorithm according to claim 1, characterized in that: Standardizing the data set by the robust normalization method includes standardizing the data according to formula (1): Among them, x is the original data point, M is the median of the data, Q1 is the first quartile, and Q3 is the third quartile.

3. The heart disease risk assessment method based on deep learning algorithm according to claim 1, characterized in that: Feature preprocessing of the input vital sign data includes multi-branch and multi-scale convolution to extract input features and channel attention calculation, feature fusion, feature map data filling and dimensionality enhancement.

4. The heart disease risk assessment method based on deep learning algorithm according to claim 3, characterized in that: Multi-branch and multi-scale convolution extracts input features and calculates channel attention, including: selecting neural networks with convolution kernel sizes of 1, 3, and 5, setting the padding of the convolution kernel and performing convolution operations, calculating the weight of each channel, adjusting the channel weights of the feature map, and outputting three different feature maps.

5. The heart disease risk assessment method based on deep learning algorithm according to claim 3, characterized in that: Feature fusion includes input feature preparation, feature concatenation, and output fused features.

6. The heart disease risk assessment method based on deep learning algorithm according to claim 3, characterized in that: Feature map data padding and dimensionality enhancement include analyzing the size difference, determining the padding location and amount, performing the padding operation, understanding the target dimensional structure, and performing the dimensionality expansion operation.

7. The heart disease risk assessment method based on deep learning algorithm according to claim 1, characterized in that: The internal components of the SwinTransformer module include the Patch merging layer, the window multi-head self-attention mechanism W-MSA and the sliding window multi-head self-attention mechanism SW-MSA, among which, The patch merging layer is used to retain more feature information during downsampling; The window multi-head self-attention mechanism W-MSA is used to limit the self-attention calculation to a window of a specific size; The sliding window multi-head self-attention mechanism SW-MSA is used to move the window.

8. The heart disease risk assessment method based on deep learning algorithm according to claim 1, characterized in that: Using the optimizer to tune the model parameters includes using the Adam optimizer to update the parameters: According to formula (2), the first-order moment estimate m is calculated t Updated value of: m t =β1* m t-1 + (1- β1) *g t (2) According to formula (3), the second-order moment estimate v is calculated t Updated value of: And according to formula (4), using the modified m ′ t and v t ′ Update model parameters θ t : Among them, g t is the gradient of the current time step, β1 and β2 are hyperparameters used to control the update speed, and m t-1 is the first-order moment estimate of the previous time step t-1, v t-1 is the second-order moment estimate of the previous time step t-1, α is the learning rate, and θ t-1 are the model parameters at the previous time step, and ε is a constant used to prevent the denominator from being zero.

9. The method for assessing heart disease risk based on deep learning algorithm according to claim 8, characterized in that: The important parameters updated include the number of iterations Epochs, batch size Batch Size, learning rate Learning_rate and L1 regularization coefficient Reg_alpha.

Images