Method for constructing lung dysfunction recognition model by combining wavelet-based greedy algorithm and neural network
Through a combination of wavelet-based greedy algorithm and neural network, a lung dysfunction recognition model is constructed, which solves the flexibility and computational complexity problems in the existing technology, and realizes efficient identification and diagnosis of abnormal lung function signals.
Patent Information
- Application Number
- CN202510372082.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-27
- Publication Date
- 2025-08-12
AI Technical Summary
The existing lung function recognition methods lack flexibility and accuracy, have high computational complexity, and are difficult to conduct effective multi-scale analysis and information extraction. The limitations of feature extraction affect the timeliness and accuracy of diagnosis.
The greedy algorithm based on wavelet is used to analyze the lung function detection image data, construct the basis function and obtain feature coefficients, and train it in combination with neural network models to build a lung dysfunction recognition model.
It improves the ability to identify abnormal lung function signals, enhances the adaptability and flexibility of the model, simplifies the calculation process, and realizes real-time recognition.
Smart Images
Figure CN120471825A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image recognition, and in particular to a method for constructing a lung dysfunction recognition model by combining a greedy algorithm based on wavelet with a neural network. Background Art
[0002] With the intensification of industrialization and environmental pollution, the incidence of respiratory diseases such as chronic obstructive pulmonary disease (COPD), asthma and interstitial lung disease has increased significantly. These diseases lead to impaired lung ventilation function, affecting the daily life and quality of life of patients. In the identification of pulmonary dysfunction, the application of mathematical methods has received more and more attention. As an effective signal processing tool, wavelet analysis can analyze signals in both time and frequency dimensions, thereby extracting information with practical significance. By introducing parameterized The systematic construction of unconditional bases can provide new tools for analyzing lung function report data. This method can not only process basic physiological signals, but can also be extended to the analysis of more complex reproducing kernel Hilbert spaces and reproducing kernel Banach spaces, thus supporting the diagnosis of lung dysfunction.
[0003] Current lung function recognition methods generally have the following defects:
[0004] Lack of adaptive methods: Most existing technologies rely on specific wavelet bases or selective algorithms, which often lack flexibility and accuracy when faced with diverse lung function data.
[0005] High computational complexity: The use of complex wavelet algorithms and models makes real-time data analysis extremely difficult and cannot meet the needs of clinical applications.
[0006] Insufficient benchmark basis: Existing research lacks the construction of unconditional basis, which makes it difficult to conduct effective multi-scale analysis and information extraction.
[0007] Limitations of feature extraction: Many methods cannot fully identify potential abnormal data in feature extraction, affecting the timeliness and accuracy of diagnosis. Summary of the Invention
[0008] In order to solve the problems of the existing technology and the defects, the purpose of the present invention is to provide a method for constructing a lung dysfunction recognition model based on a greedy algorithm combined with a neural network based on wavelet. The technical solution adopted is as follows:
[0009] A method for constructing a pulmonary dysfunction recognition model based on a wavelet-based greedy algorithm combined with a neural network, the method comprising the following steps:
[0010] Acquire pulmonary function test image data; the pulmonary function test image data includes index data and image parts;
[0011] Preprocess the index data and image parts respectively;
[0012] The pre-processed image is analyzed using a greedy algorithm based on wavelet and a basis function is constructed. The characteristic coefficients of each curve are obtained according to the basis function.
[0013] The characteristic coefficients and indicator data are used as training data and input into the neural network model for training to obtain a trained neural network model.
[0014] Preferably, the preprocessing of the indicator data is one or more of processing missing values, removing outliers, normalizing the indicator data, and unifying the numerical range.
[0015] Preferably, the preprocessing of the image portion is one or more of data cropping, normalization, and data enhancement.
[0016] The present invention adopts a greedy algorithm based on wavelet to analyze the pre-processed image part and construct a basis function; and obtains the characteristic coefficient of each curve according to the basis function.
[0017] Furthermore, the step of constructing a basis function includes:
[0018] The two initial characteristic coefficients f0 and f1 are set as<f,H0> and At the same time, the coefficients of the basis functions are initialized and Among them, <> represents the inner product operation;
[0019] When m=2, write m=2 j-1 +1+k; where j=1 and k=0<2 j-1 At this time, we get and will Initialized to 0;
[0020] When 3≤m≤4, let j=2 and 2 j-1 +1≤m≤2 j ; written as m = 2 j-1 +1+k, where j≥2 and 0≤k<2 j-1 ;
[0021] Select f3 through the greedy algorithm and The maximum value in Then we get k0, so that
[0022] Define f3 selected by the greedy algorithm as And set Afterwards, define As the amount of the remainder;
[0023] When m≥5, then define
[0024] Find the general term and get it by generalization: Make and The coefficient distribution of the basis function is
[0025] The basis function is thus obtained according to the coefficients of the basis function.
[0026] Optionally, the neural network model is one of a Restnet neural network model constructed using a deep learning framework, a Restnet neural network model constructed using an RNN unit, a long short-term memory network and a gated recurrent unit, or a Transformer neural network model.
[0027] Furthermore, when the neural network model uses a deep learning framework to construct a Restnet neural network model, a greedy algorithm based on wavelet is used to analyze the preprocessed image portion and construct a basis function; and characteristic coefficients of each curve are obtained according to the basis function;
[0028] The characteristic coefficients and indicator data are used as training data and input into the Restnet neural network model built using the deep learning framework for training.
[0029] Furthermore, when the neural network model uses RNN units, long short-term memory networks, and gated recurrent units to construct a Restnet neural network model, it includes:
[0030] The pre-processed image is analyzed using a greedy algorithm based on wavelet and a basis function is constructed. The characteristic coefficients of each curve are obtained according to the basis function.
[0031] Extract exponential data from the greedy algorithm based on wavelet and convert the characteristic coefficients into format;
[0032] The converted exponential data and feature coefficients are then input into the Restnet neural network model as training data for training.
[0033] Furthermore, when the neural network model uses a Transformer neural network model, it includes:
[0034] The pre-processed image is analyzed using a greedy algorithm based on wavelet and a basis function is constructed. The characteristic coefficients of each curve are obtained according to the basis function.
[0035] The exponential data and feature coefficients are input as training data into the Transformer neural network model for training.
[0036] The present invention also provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method for constructing a pulmonary dysfunction recognition model by combining a wavelet-based greedy algorithm with a neural network.
[0037] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to execute the method for constructing a lung dysfunction recognition model by combining a greedy algorithm based on wavelets with a neural network.
[0038] The present invention has the following beneficial effects:
[0039] The present invention uses a wavelet-based greedy algorithm to analyze the preprocessed image portion and construct basis functions. Based on the basis functions, characteristic coefficients for each curve are obtained. These characteristic coefficients and indicator data are then used as training data to input into a neural network model for training, resulting in a trained neural network model. The constructed neural network model can be used for pulmonary function test image recognition, significantly improving the ability to identify abnormal pulmonary function signals.
[0040] The neural network model constructed in the present invention can process different types of lung function data and has greater adaptability and flexibility.
[0041] The neural network model constructed in the present invention uses a greedy algorithm based on wavelets to find the basis with the largest energy in the interval through interval division, which simplifies the calculation process, optimizes the calculation efficiency, and makes real-time recognition possible. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 The present invention provides a flowchart of a method for constructing a lung dysfunction recognition model by combining a greedy algorithm based on wavelets with a neural network.
[0043] Figure 2 It is a principle block diagram of the computer device described in the present invention. DETAILED DESCRIPTION
[0044] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.
[0045] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0046] The following describes in detail a method for constructing a lung dysfunction recognition model by combining a greedy algorithm based on wavelet and a neural network, provided by the present invention, with reference to the accompanying drawings.
[0047] like Figure 1 As shown, a method for constructing a lung dysfunction recognition model by combining a wavelet-based greedy algorithm with a neural network includes the following steps:
[0048] Acquire pulmonary function test image data; the pulmonary function test image data includes index data and image parts;
[0049] Preprocess the index data and image parts respectively;
[0050] The greedy algorithm based on wavelet is used to analyze the preprocessed image part and construct the basis function;
[0051] Obtain the characteristic coefficient of each curve according to the basis function;
[0052] The characteristic coefficients and indicator data are input into the neural network model as training data for training to obtain a trained neural network model.
[0053] The preprocessing of the indicator data and the image part respectively includes:
[0054] Perform one or more of the following operations on the indicator data: processing missing values, removing outliers, normalizing, and unifying the numerical range;
[0055] One or more of data cropping, normalization, and data enhancement processing is performed on the image portion.
[0056] In a specific embodiment, a greedy algorithm based on wavelet is used to analyze the pre-processed image portion and construct a basis function, including:
[0057] The two initial characteristic coefficients f0 and f1 are set as<f,H0> and At the same time, the coefficients of the basis functions are initialized and Among them, 〈〉 represents the inner product operation;
[0058] When m=2, write m=2 j-1 +1+k; where j=1 and k=0<2 j-1 At this time, we get and will Initialized to 0;
[0059] When 3≤m≤4, let j=2 and 2 j-1 +1≤m≤2 j ; written as m = 2 j-1 +1+k, where j≥2 and 0≤k<2 j-1 ;
[0060] Select f3 through the greedy algorithm and The maximum value in Then we get k0, so that
[0061] Define f3 selected by the greedy algorithm as And set Afterwards, define As the amount of the remainder;
[0062] When m≥5, then define
[0063] Find the general term and get it by generalization: Make and The coefficient distribution of the basis function is
[0064] The basis function is thus obtained according to the coefficients of the basis function.
[0065] In this embodiment, it is assumed that a certain curve F can be decomposed into the following form:<F,H0> *H0+<F,H1> *H1+<F,H2> *H2+......+ <F,H n >*H n ;
[0066] Where n is infinite.
[0067] [H0,H1,H2,...,H n ] constitutes a whole basis function space, each H i It's all based on the same foundation.
[0068] The present invention selects a finite number of basis functions from the entire basis function space through a greedy algorithm so that F can be approximately represented.
[0069] and <F,H i > is the inner product of the original curve and the basis, also known as the characteristic coefficient; for convenience, f i express <F,H i >.
[0070] In the above formula, H can be understood as a formal definition. When j is 0 or 1, there is no screening process, so it is written directly.
[0071] or Indicates that the filter items are selected, such as and Selected It's big.
[0072] In the formula, the mark f in the upper right corner of H represents the final filtered value.
[0073] in addition, Corresponding to multiple This is because the relationship between m and j, k can be understood as:
[0074] m=[1]+[2]+[4]+.....+[2 j-1 +1+k]
[0075] If m is 7, then j is 3 and k is 2.
[0076] This method first gives the basis functions (i.e., initial characteristic coefficients) for the cases where j = 0 and 1, which cannot be expressed using a recursive formula. Then, j is incremented from 2 to a predetermined value, t. For example, in the example above, m is 7 and t is 3. The best possible value for k is selected. For example, when m = 7, k can be 0, 1, or 2. The best value is then selected from these three.
[0077] The present invention selects t bases through a greedy algorithm, and the formula expression method of the curve F is as follows:
[0078]
[0079] in, are called the coefficients of the basis functions.
[0080] For the selected basis, The value of is 1; for the unselected basis, The value of is 0.
[0081] In this embodiment, according to the definition of the characteristic coefficient, after the basis function is obtained, the characteristic coefficient of each curve can be directly obtained according to the basis function.
[0082] Preferably, the neural network model includes building a Restnet neural network model using a deep learning framework, building a Restnet neural network model using an RNN unit, a long short-term memory network and a gated recurrent unit, or a Transformer neural network model.
[0083] Preferably, when the neural network model uses a deep learning framework to construct a Restnet neural network model, a greedy algorithm based on wavelet is used to analyze the preprocessed image portion and construct a basis function; and characteristic coefficients of each curve are obtained according to the basis function;
[0084] The feature coefficients and indicator data are used as training data input and the Restnet neural network model is built using the deep learning framework for training.
[0085] The above method uses a deep learning framework (such as TensorFlow or PyTorch) to build a ResNet neural network model (using a 101-layer model) and determine its structural parameters (such as the number of layers, convolution kernel size, stride, etc.). Define the loss function (Cross Entropy Loss), the formula of the loss function is as follows:
[0086] H(p,q)=-∑(p(x)*log(q(x)))
[0087] Here, p(x) is the probability distribution of the true label, and q(x) is the probability distribution of the predicted label.
[0088] This formula actually performs a weighted summation of the predicted probabilities for each category, where the weights are the probability distribution of the true labels. And select an optimizer, such as Adam, SGD, etc., and set the learning rate and other hyperparameters.
[0089] The training data is fed into the ResNet neural network model for training, which is repeated over multiple rounds to update the parameters of the ResNet neural network model. In each round of training, the gradient is calculated using the backpropagation algorithm, and the optimizer is used to update the weights of the ResNet neural network model. Training of the ResNet neural network model is terminated (until the change in the loss function falls below a pre-set minimum change value in two consecutive rounds, or the pre-set maximum number of iterations is reached).
[0090] In a specific embodiment, when the neural network model uses an RNN unit, a long short-term memory network, and a gated recurrent unit to construct a Restnet neural network model, it includes:
[0091] The pre-processed image is analyzed using a greedy algorithm based on wavelet and a basis function is constructed. The characteristic coefficients of each curve are obtained according to the basis function.
[0092] Extract key features (exponential data, characteristic coefficients) from the wavelet coefficients and convert them into a format suitable for input into the gated recurrent unit. Then, the converted exponential data and characteristic coefficients are input into the Restnet neural network model as training data.
[0093] In this embodiment, the structural design of the Restnet neural network model: RNN unit selection: basic RNN units, long short-term memory networks (LSTMs), and gated recurrent units (GRUs) can be selected to build the Restnet neural network model, because LSTMs and GRUs perform better in long-term sequence learning.
[0094] Layer and node settings: Set the appropriate number of hidden layers and the number of nodes per layer. Cross-validation is usually required to optimize these hyperparameters.
[0095] Training process: Select an appropriate loss function, such as cross-entropy loss. Use optimization algorithms such as Adam and RMSprop to accelerate convergence and improve the training efficiency of the RestNet neural network model. Also, use a mini-batch training method to improve training speed and the generalization ability of the RestNet neural network model.
[0096] Restnet neural network model adjustment: Further improve training results by adjusting hyperparameters such as learning rate and batch size.
[0097] In a specific embodiment, when the neural network model uses a Transformer neural network model, it includes:
[0098] The pre-processed image is analyzed using a greedy algorithm based on wavelet and a basis function is constructed. The characteristic coefficients of each curve are obtained according to the basis function.
[0099] Characteristic coefficients and index data were analyzed to extract characteristic points, such as maximum points and change points, which are crucial for the identification of lung dysfunction.
[0100] A Transformer-based deep learning model was constructed, using the characteristic coefficients and exponential data extracted by a greedy wavelet-based algorithm as training input. The Transformer-based deep learning model can include an encoder-decoder architecture, using self-attention and multi-head attention modules to capture the temporal dependencies and complex characteristics of the data. The Transformer-based deep learning model was trained using labeled pulmonary function data, and cross-validation was used to evaluate model performance. The accuracy and stability of the model were further improved by adjusting parameters such as the learning rate, batch size, and loss function.
[0101] See also Figure 2 , Figure 2 Schematic block diagram of a computer device provided by an embodiment of the present invention. The computer device 500 is a server, which can be an independent server or a server cluster composed of multiple servers.
[0102] The computer device 500 includes a processor 502 , a memory, and a network interface 505 connected via a system bus 501 , wherein the memory may include a non-volatile storage medium 503 and an internal memory 504 .
[0103] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, the processor 502 can execute a method for constructing a lung dysfunction recognition model by combining a greedy algorithm based on wavelets with a neural network.
[0104] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500.
[0105] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a method for constructing a pulmonary dysfunction recognition model by combining a greedy algorithm based on wavelets with a neural network.
[0106] The network interface 505 is used for network communication, such as providing data information transmission. Those skilled in the art will understand that Figure 2 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device 500 to which the solution of the present invention is applied. The specific computer device 500 may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.
[0107] The processor 502 is configured to run a computer program 5032 stored in the memory to implement the method disclosed in an embodiment of the present invention for constructing a lung dysfunction recognition model by combining a greedy algorithm based on wavelets with a neural network.
[0108] Those skilled in the art will understand that Figure 2 The embodiment of the computer device shown in the figure does not constitute a limitation on the specific composition of the computer device. In other embodiments, the computer device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. For example, in some embodiments, the computer device may only include a memory and a processor. In such an embodiment, the structure and function of the memory and processor are the same as those in the figure. Figure 2 The embodiments shown are consistent and will not be described again here.
[0109] It should be understood that in the embodiment of the present invention, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.
[0110] In another embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium may be a non-volatile computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method disclosed in an embodiment of the present invention for combining a wavelet-based greedy algorithm with a neural network to construct a lung dysfunction recognition model.
[0111] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0112] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, or units with the same function may be combined into one unit. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices or units, or may be an electrical, mechanical or other form of connection.
[0113] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.
[0114] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk.
[0115] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0116] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.
Claims
1. A method for constructing a lung dysfunction recognition model based on a greedy algorithm based on wavelet and a neural network, characterized in that: The method comprises the following steps: Acquire pulmonary function test image data; the pulmonary function test image data includes index data and image parts; Preprocess the index data and image parts respectively; The pre-processed image is analyzed using a greedy algorithm based on wavelet and a basis function is constructed. The characteristic coefficients of each curve are obtained according to the basis function. The characteristic coefficients and indicator data are used as training data and input into the neural network model for training to obtain a trained neural network model.
2. The method for constructing a pulmonary dysfunction recognition model based on a greedy algorithm based on wavelet and a neural network according to claim 1, characterized in that: The preprocessing of the indicator data is one or more of processing missing values, removing outliers, normalizing the indicator data, and unifying the numerical range.
3. The method for constructing a pulmonary dysfunction recognition model based on a wavelet-based greedy algorithm combined with a neural network according to claim 1, characterized in that: The preprocessing of the image portion is one or more of data cropping, normalization, and data enhancement.
4. The method for constructing a pulmonary dysfunction recognition model by combining a wavelet-based greedy algorithm with a neural network according to claim 1, characterized in that: The neural network model is one of a Restnet neural network model constructed using a deep learning framework, a Restnet neural network model constructed using an RNN unit, a long short-term memory network and a gated recurrent unit, or a Transformer neural network model.
5. The method for constructing a pulmonary dysfunction recognition model by combining a wavelet-based greedy algorithm with a neural network according to claim 4, characterized in that: When the neural network model uses a deep learning framework to construct a Restnet neural network model, a greedy algorithm based on wavelet is used to analyze the preprocessed image portion and construct a basis function; and characteristic coefficients of each curve are obtained according to the basis function; The characteristic coefficients and indicator data are used as training data and input into the Restnet neural network model constructed using the deep learning framework for training.
6. The method for constructing a pulmonary dysfunction recognition model by combining a wavelet-based greedy algorithm with a neural network according to claim 4, characterized in that: When the neural network model uses an RNN unit, a long short-term memory network, and a gated recurrent unit to construct a Restnet neural network model, it includes: The pre-processed image is analyzed using a greedy algorithm based on wavelet and a basis function is constructed. The characteristic coefficients of each curve are obtained according to the basis function. Extract exponential data from the greedy algorithm based on wavelet and convert the characteristic coefficients into format; The converted exponential data and feature coefficients are then input into the Restnet neural network model as training data for training.
7. The method for constructing a pulmonary dysfunction recognition model by combining a wavelet-based greedy algorithm with a neural network according to claim 4, characterized in that: When the neural network model uses a Transformer neural network model, it includes: The pre-processed image is analyzed using a greedy algorithm based on wavelet and a basis function is constructed. The characteristic coefficients of each curve are obtained according to the basis function. The exponential data and feature coefficients are input as training data into the Transformer neural network model for training.
8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for constructing a pulmonary dysfunction recognition model by combining a greedy algorithm based on wavelet with a neural network as described in any one of claims 1 to 7 is implemented.
9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to execute the method for constructing a pulmonary dysfunction recognition model by combining a greedy algorithm based on wavelet with a neural network as described in any one of claims 1 to 7.