A lung lesion recognition method based on endoscopic images

Abstract

The present invention discloses a method for identifying lung lesions based on endoscopic images, which comprises: acquiring a lung image through an endoscope in real time; converting the lung image into a grayscale image, and extracting texture features in the grayscale image; calculating the texture feature statistics of each segment using a grayscale co-occurrence matrix according to the texture features; inputting the texture feature statistics of each segment into a pre-trained deep convolutional neural network model, and outputting a lung lesion identification result.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a lung lesion recognition method based on endoscopic images. Background Art

[0002] With the development of medical technology, endoscopy has become a common means of inspection and treatment. Through endoscopy, doctors can examine the human cavity, such as the human respiratory tract, without surgery to determine whether the patient has a tumor, infection or other diseases. For example, bronchoscopy or gastroenteroscopy can be performed to check whether the patient has respiratory diseases or digestive diseases, respectively. Endoscopes usually include an image sensor, such as a CCD or CMOS, located at the distal end of the endoscope insertion part and connected to a display device to provide the doctor with an image of the inspected area.

[0003] However, due to limitations in image sensor performance or display device resolution, images captured by an endoscope cannot well display the texture features of the inspection site, especially the target lesion site, which makes it difficult to use the endoscope system for diagnosis.

[0004] The above contents are only used to assist in understanding the technical solution of the present invention and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of the present invention is to provide a lung lesion recognition method based on endoscopic images, aiming to solve the above-mentioned problems in the prior art.

[0006] To achieve the above object, the present invention provides a method for identifying lung lesions based on endoscopic images, characterized in that it comprises:

[0007] Real-time endoscopy to obtain lung images;

[0008] Converting the lung image into a grayscale image, and extracting texture features from the grayscale image;

[0009] According to the texture features, a gray level co-occurrence matrix is ​​used to calculate the texture feature statistics of each segment;

[0010] The texture feature statistics of each segment are input into a pre-trained deep convolutional neural network model to output the lung lesion recognition results.

[0011] Preferably, in the lung lesion recognition method based on endoscopic images, before inputting the texture feature statistics of each segment into a pre-trained deep convolutional neural network model and outputting the lung lesion recognition result, the method further includes:

[0012] Acquire characteristic information of a patient corresponding to the lung image, wherein the characteristic information includes living habits, age, and gender;

[0013] According to the feature information, inputting into a pre-established first convolutional neural network to determine the key index of each of the fragments;

[0014] Accordingly, the texture feature statistics of each segment are input into a pre-trained deep convolutional neural network model to output the lung lesion recognition results, including:

[0015] According to the key index of each segment and the texture feature statistics of each segment, a pre-trained deep convolutional neural network model is input to output the lung lesion recognition results.

[0016] Preferably, in the lung lesion recognition method based on endoscopic images, after inputting the pre-established first convolutional neural network according to the feature information and determining the key index of each segment, the method further includes:

[0017] The key index of each segment is input into a pre-established feed-forward sequence memory network model to modify the key index of each segment;

[0018] The hidden layer of the feedforward sequence memory network model is connected with a feedback connection block, and the feedback connection block stores historical information and future information of the patient corresponding to the lung image.

[0019] Preferably, in the pulmonary lesion recognition method based on endoscopic images, the key index of each segment is input into a pre-established feedforward sequence memory network model to modify the key index of each segment, including:

[0020] According to the lung image and the patient's historical information, the future information of the patient's lung image is predicted by the feedforward sequential memory network model;

[0021] The key index of each segment is modified based on the predicted future information of the patient's lung image.

[0022] Preferably, in the endoscopic image-based lung lesion recognition method, the texture feature statistics of each segment are calculated using a gray level co-occurrence matrix according to the texture features, including:

[0023] Generate a corresponding gray-level co-occurrence matrix according to the gray-level image and the parameters of the preset gray-level co-occurrence matrix, wherein the parameters include at least a direction, an offset, and an order;

[0024] Calculating the texture feature statistics of each segment according to the generated gray level co-occurrence matrix, wherein the texture feature statistics include energy, contrast, correlation and entropy;

[0025] in,

[0026] The energy is used to characterize the texture detail level of the texture of the lung image;

[0027] The contrast is used to characterize the contrast between different gray levels in the lung image texture;

[0028] The correlation is used to characterize the correlation of pixel gray levels in the lung image texture;

[0029] The entropy is used to characterize the uncertainty and complexity of the lung image texture.

[0030] Preferably, in the endoscopic image-based lung lesion recognition method, converting the lung image into a grayscale image and extracting texture features in the grayscale image includes:

[0031] Converting the lung image into a grayscale image;

[0032] Using a K-means clustering algorithm, segmenting the grayscale image into related segments;

[0033] According to the Gabor filter, the texture features of each segment are extracted.

[0034] Preferably, in the endoscopic image-based lung lesion identification method, the step of converting the lung image into a grayscale image comprises:

[0035] The lung image is converted into a grayscale image using the weighted average method, and the three components of R, G, and B pixels at the corresponding positions are weighted averaged with different weights. The specific formula is:

[0036] Gray(i,j)=0.299*R(i,j)+0.578*G(i,j)+0.114*B(i,j).

[0037] Preferably, in the lung lesion recognition method based on endoscopic images, in the step of extracting the texture features of each segment according to the Gabor filter, the Gabor filter function is:

[0038] ;

[0039] Among them, (x, y) is the pixel coordinate;

[0040] σx and σy are the standard deviations of the Gaussian envelope;

[0041] f is the frequency;

[0042] θ is the direction of the filter;

[0043] φ is the phase shift.

[0044] In order to achieve the above object, the present invention further provides a computer device, characterized in that it includes:

[0045] at least one processor; and,

[0046] a memory communicatively connected to the at least one processor; wherein,

[0047] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned endoscopic image-based lung lesion identification method.

[0048] In order to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned lung lesion identification method based on endoscopic images.

[0049] The present invention has at least the following beneficial effects:

[0050] The present invention acquires a lung image through an endoscope in real time; converts the lung image into a grayscale image, and extracts texture features in the grayscale image; calculates the texture feature statistics of each segment using a grayscale co-occurrence matrix based on the texture features; inputs the texture feature statistics of each segment into a pre-trained deep convolutional neural network model, and outputs a lung lesion recognition result, thereby accurately improving the accuracy of lesion recognition and reducing processing time;

[0051] Furthermore, the present invention incorporates the gray-level co-occurrence matrix into the deep learning framework, and extracts the texture information of the target part of the endoscopic image to assist in the identification and judgment of lung lesions, so that the judgment is more accurate.

[0052] Furthermore, the present invention addresses the limitations of existing methods by providing a more reliable and effective lung lesion detection method. By incorporating the gray-level co-occurrence matrix into the deep learning framework, the method not only improves the accuracy of lesion identification, but also reduces processing time.

[0053] Furthermore, artificial intelligence can be used to automatically detect lesions to reduce the number of lesions missed by doctors; artificial intelligence can be used to automatically identify lesions to reduce the number of biopsies.

[0054] Furthermore, the method provided by the present invention uses lung image lesion auxiliary identification and judgment obtained by endoscope, and is not used for processing the image displayed in real time on the monitor during endoscopic examination. The monitor still outputs the color image obtained by the endoscope for the doctor's intuitive observation and judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A schematic diagram of a first embodiment of a lung lesion recognition method based on endoscopic images provided by the present invention;

[0056] Figure 2 A schematic diagram of a second embodiment of the lung lesion recognition method based on endoscopic images provided by the present invention;

[0057] Figure 3 A schematic diagram of a third embodiment of the lung lesion recognition method based on endoscopic images provided by the present invention;

[0058] Figure 4 A schematic diagram of a computer device according to the present invention.

[0059] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0060] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0061] In the embodiments of the present invention, the term "and / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0062] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0063] In the embodiments of the present invention, the term "plurality" refers to two or more than two, and other quantifiers are similar.

[0064] In the present invention, unless otherwise specified, the directional words used, such as "up, down, top, bottom", usually refer to the directions shown in the drawings, or to the components themselves in the vertical, perpendicular or gravity directions; similarly, for ease of understanding and description, "inside and outside" refer to the inside and outside relative to the outline of each component itself, but the above-mentioned directional words are not used to limit the present invention.

[0065] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. However, it can be understood by those skilled in the art that in the embodiments of the present invention, many technical details are proposed in order to enable the reader to better understand the present invention. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed in the present invention can also be implemented. The division of the following embodiments is for the convenience of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined and referenced with each other under the premise of no contradiction.

[0066] In order to solve the above problems, the present embodiment relates to a lung lesion identification method based on endoscopic images, which can be applied to computer devices. The computer devices can be electronic devices with data processing capabilities such as desktop computers, tablet computers, notebooks, etc. In other embodiments, they can also be other electronic devices with data processing capabilities, and no specific restrictions are made here.

[0067] The following is an explanation of the implementation details of the lung lesion recognition method based on endoscopic images according to the first embodiment of the present invention. The following content is only the implementation details provided for the convenience of understanding and is not necessary for the implementation of this solution.

[0068] The specific process of this implementation is as follows: Figure 1 As shown, specifically including:

[0069] Step S100, acquiring a lung image through an endoscope in real time;

[0070] It should be understood that the lung image obtained by endoscope in real time can be a lung image obtained by medical staff through endoscope in real time. The lung image can be one or more images, for example, multiple lung images obtained by medical staff within 2 seconds.

[0071] Step S200, converting the lung image into a grayscale image, and extracting texture features in the grayscale image;

[0072] It should be understood that converting the lung image into a grayscale image, that is, converting the lung image from color to grayscale, reduces the color space and simplifies the image through conversion, so that it is easier to apply to subsequent processing.

[0073] Specifically, step S200 includes: converting the lung image into a grayscale image; using a K-means clustering algorithm to segment the grayscale image into related segments; and extracting texture features of each segment according to a Gabor filter.

[0074] The lung image is converted into a grayscale image, and the image is converted from color to grayscale. This conversion simplifies the image by reducing the color space, making it easier to apply subsequent processing techniques. Image grayscale processing uses a weighted average method to convert the color RGB color image into a grayscale image.

[0075] In this embodiment, the lung image is converted into a grayscale image using a weighted average method, and the three components of R, G, and B pixels at corresponding positions are weighted averaged with different weights. The specific formula is:

[0076] Gray(i,j)=0.299*R(i,j)+0.578*G(i,j)+0.114*B(i,j).

[0077] In other embodiments, the grayscale image needs to be filtered using a median filter.

[0078] The K-means clustering algorithm is used to segment the image into highly correlated segments to facilitate image recognition.

[0079] The steps to implement the k-means algorithm are as follows:

[0080] (1) Determine the K value, that is, cluster the data set into K clusters or groups;

[0081] (2) Randomly select K data points as center points;

[0082] (3) Calculate the distance between each value in the pixel array and each center point respectively, and determine which class the value is closer to;

[0083] (4) Calculate the average value of each category as the new center;

[0084] (5) Compare the new center with the old center. If the distance between the new center and the old center is less than a certain threshold, it means that the recalculated center position has not changed much and the convergence is stable. It is considered that the clustering has achieved the desired result and the iteration is terminated.

[0085] (6) If the new center and the old center change greatly, that is, the distance is greater than the threshold, continue to iterate the third to fifth steps until the iteration terminates.

[0086] Texture features of each segment are extracted based on Gabor filters. Gabor filters are very effective in texture representation and recognition. Gabor filters are Gaussian kernel functions modulated by sinusoidal plane waves. Multichannel filtering is achieved by a bank of Gabor filters of different scales and orientations. The Gabor filter bank performs a robust multiresolution decomposition, allowing the calculation of frequency and orientation information. Gabor texture features are calculated based on the statistical distribution of Gabor amplitude responses. Gabor features are robust to photometric disturbances such as illumination changes and noise.

[0087] The Gabor filter is a linear filter that can be viewed as a Gaussian envelope of a sinusoidal wave with a specific frequency and direction. The Gabor filter function is:

[0088] ;

[0089] Among them, (x, y) is the pixel coordinate;

[0090] σ x and σ y is the standard deviation of the Gaussian envelope;

[0091] f is the frequency;

[0092] θ is the direction of the filter;

[0093] φ is the phase shift.

[0094] When applying a Gabor filter, you need to choose appropriate parameters: The frequency f of the filter determines the scale of texture details that it can capture. The direction θ of the filter determines its sensitivity to textures in a specific direction. Multiple directions are usually selected to capture textures in different directions, and the valid value is a real number from 0° to 360°. The standard deviation (σ x and σ y ): controls the spatial extent of the filter. The phase shift φ is usually set to 0 or π / 2.

[0095] The steps for performing Gabor filtering on an image are as follows:

[0096] Create Gabor Filter: Generates a Gabor filter based on selected parameters.

[0097] Filtering: Use Gabor filters to convolve the image and get multiple filtered images. These filtered images represent the texture information of the image at different frequencies and directions.

[0098] After performing a Gabor filtering operation on an image, various texture features can be extracted from the Gabor filtered image.

[0099] Step S300, calculating the texture feature statistics of each segment using a gray level co-occurrence matrix according to the texture features;

[0100] Specifically, a corresponding gray-level co-occurrence matrix is ​​generated according to the gray-level image and the parameters of a preset gray-level co-occurrence matrix, wherein the parameters include at least a direction, an offset, and an order;

[0101] Calculating the texture feature statistics of each segment according to the generated gray level co-occurrence matrix, wherein the texture feature statistics include energy, contrast, correlation and entropy;

[0102] in,

[0103] The energy is used to characterize the texture detail level of the texture of the lung image;

[0104] The contrast is used to characterize the contrast between different gray levels in the lung image texture;

[0105] The correlation is used to characterize the correlation of pixel gray levels in the lung image texture;

[0106] The entropy is used to characterize the uncertainty and complexity of the lung image texture.

[0107] Step S400, inputting the texture feature statistics of each segment into a pre-trained deep convolutional neural network model, and outputting the lung lesion recognition result.

[0108] It should be noted that the results of lung lesion identification can usually include but are not limited to cancer, nodules, inflammation, and normal.

[0109] The training of a deep convolutional neural network model can include:

[0110] (1) Obtaining sample images, preparing multiple endoscopic lung images, labeling the lung images, and determining a label for each lung image. For example, the label may include but is not limited to cancer, nodule, inflammation, and normal.

[0111] (2) The sample image is then converted into a first grayscale image, and a first texture feature is extracted from the first grayscale image. For details, please refer to step 200, which will not be described in detail here.

[0112] (3) According to the first texture feature, the gray level co-occurrence matrix is ​​used to calculate the first texture feature statistic of each segment. For details, please refer to step 300, which will not be described in detail here.

[0113] (4) The first texture feature statistics of each fragment are input into the deep convolutional neural network model for training.

[0114] A deep convolutional neural network (DCNN model) was trained to classify and recognize different endoscopic lung images.

[0115] Specifically, the deep learning model may include an input layer IN, multiple hidden layers HID and an output layer OUT, and the hidden layer HID may include a feature extraction network FEN, a global pooling layer GP and a fully connected layer FC, and the feature extraction network FEN may include multiple layers selected from the group consisting of the above-mentioned convolutional layers, pooling layers and normalization layers.

[0116] The present invention acquires a lung image through an endoscope in real time; converts the lung image into a grayscale image, and extracts texture features in the grayscale image; calculates the texture feature statistics of each segment using a grayscale co-occurrence matrix based on the texture features; inputs the texture feature statistics of each segment into a pre-trained deep convolutional neural network model, and outputs a lung lesion recognition result, thereby accurately improving the accuracy of lesion recognition and reducing processing time;

[0117] Furthermore, the present invention addresses the limitations of existing methods by providing a more reliable and effective lung lesion detection method. By incorporating the gray-level co-occurrence matrix into the deep learning framework, the method not only improves the accuracy of lesion identification, but also reduces processing time.

[0118] Furthermore, artificial intelligence can be used to automatically detect lesions to reduce the number of lesions missed by doctors; artificial intelligence can be used to automatically identify lesions to reduce the number of biopsies.

[0119] like Figure 2 As shown, Figure 2 The second embodiment of the lung lesion recognition method based on endoscopic images of the present invention is illustrated. Based on the first embodiment, before step S400, the method further includes:

[0120] Step S510, acquiring characteristic information of the patient corresponding to the lung image, wherein the characteristic information includes living habits, age, and gender.

[0121] It should be understood that the probability of lung disease may be different for each patient of different ages, different living habits, and different genders. For example, men who smoke frequently are more likely to develop lung disease than those who do not smoke frequently; for example, elderly people who smoke frequently are more likely to develop lung disease than young people who smoke frequently. Of course, this is just a probability problem, and the above situation does not necessarily exist, so it is necessary to further determine it through the patient's historical lung image information. According to the patient's historical lung image situation, the amount of key index of each segment is further determined.

[0122] Step S520, inputting a pre-established first convolutional neural network according to the feature information to determine a key index of each segment;

[0123] It should be understood that the patient's characteristic information, input into the first convolutional neural network, will determine the initial value of the key index of each segment; it can be further modified according to the historical lung image conditions.

[0124] Accordingly, the step S400 includes:

[0125] Step S410, according to the key index of each segment and the texture feature statistics of each segment, a pre-trained deep convolutional neural network model is input to output the lung lesion recognition result.

[0126] like Figure 3 As shown, Figure 3 The third embodiment of the lung lesion recognition method based on endoscopic images of the present invention is illustrated. Based on the first embodiment, after step S520, the method further includes:

[0127] Step S530, inputting the key index of each segment into a pre-established feedforward sequence memory network model to modify the key index of each segment;

[0128] The hidden layer of the feedforward sequence memory network model is connected with a feedback connection block, and the feedback connection block stores historical information and future information of the patient corresponding to the lung image.

[0129] Specifically, step S530 further includes:

[0130] Step S531, predicting future information of the patient's lung image through the feedforward sequence memory network model according to the lung image and the patient's historical information;

[0131] It should be understood that by combining the historical information of the patient's lung image and the current lung image, the future information of the patient's lung image can be predicted. For example, if the patient's historical lung image shows small lung nodules, and the results of multiple reexaminations show that the size has not changed, then it may be inflammatory. At this time, it can be predicted that the patient's future information of the lung image may be relatively inflammatory. However, if the patient's multiple historical lung images have changed or slightly changed, then it is necessary to focus on the corresponding segment, and the patient's lung image at this point is likely to change in the future. Then the key index for this segment will be significantly higher than others.

[0132] Step S532, modifying the key index of each segment according to the predicted future information of the patient's lung image.

[0133] Specifically, assuming that the patient's lung image determines the criticality of each segment as a1, a2, ... based on the patient's characteristic information, then the critical index of each segment is corrected based on the patient's historical lung images to predict future information. For example, if there are historical small lung nodules and the historical images have changed very little, the criticality of the segment may be reduced; for another example, if there are historical lung nodules and the historical images have changed, the criticality of the segment may be increased.

[0134] In other embodiments, the initial probability of each segment having a lung disease in the future can be predicted based on the historical information of the patient's lung image, and the initial probability of each segment can be corrected based on the key index of each segment determined by the feature information. The specific method can also be determined according to different situations, such as age.

[0135] In order to achieve the above object, the present invention also provides a computer device, such as Figure 4 As shown, the cleaning base station includes at least one processor 601; and a memory 602 that is communicatively connected to the at least one processor 601; wherein the memory 602 stores instructions that can be executed by the at least one processor 601, and the instructions are executed by the at least one processor 601 so that the at least one processor 601 can execute the above-mentioned endoscopic image-based lung lesion recognition method.

[0136] The memory 602 and the processor 601 are connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors 601 and the memory 602 together. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. The data processed by the processor 601 is transmitted on a wireless medium via an antenna, and further, the antenna also receives data and transmits the data to the processor 601.

[0137] The processor 601 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management and other control functions. The memory 602 can be used to store data used by the processor 601 when performing operations.

[0138] In order to achieve the above-mentioned purpose, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned lung lesion identification method based on endoscopic images.

[0139] That is, those skilled in the art can understand that all or part of the steps in the above-mentioned implementation method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including several instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the method described in each implementation of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0140] Obviously, the embodiments described above are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, ordinary technicians in this field can make other different forms of changes or modifications without creative work, which should fall within the scope of protection of the present invention.

Claims

1. A method for identifying lung lesions based on endoscopic images, characterized in that: include: Real-time endoscopy to obtain lung images; Converting the lung image into a grayscale image, and extracting texture features from the grayscale image; Converting the lung image into a grayscale image, and extracting texture features from the grayscale image, comprising: converting the lung image into a grayscale image; using a K-means clustering algorithm to segment the grayscale image into associated segments; and extracting texture features of each segment according to a Gabor filter; According to the texture features, a gray level co-occurrence matrix is ​​used to calculate the texture feature statistics of each segment; The texture feature statistics of each segment are input into a pre-trained deep convolutional neural network model to output the lung lesion recognition results; Before inputting the texture feature statistics of each segment into a pre-trained deep convolutional neural network model and outputting the lung lesion recognition result, the method further includes: Acquire characteristic information of a patient corresponding to the lung image, wherein the characteristic information includes living habits, age, and gender; According to the feature information, inputting into a pre-established first convolutional neural network to determine the key index of each of the fragments; Accordingly, the texture feature statistics of each segment are input into a pre-trained deep convolutional neural network model to output the lung lesion recognition results, including: According to the key index of each segment and the texture feature statistics of each segment, a pre-trained deep convolutional neural network model is input to output the lung lesion recognition results; After inputting the pre-established first convolutional neural network according to the feature information to determine the key index of each segment, the method further includes: The key index of each segment is input into a pre-established feed-forward sequence memory network model to modify the key index of each segment; Wherein, the hidden layer of the feedforward sequence memory network model is connected with a feedback connection block, and the feedback connection block stores historical information and future information of the patient corresponding to the lung image; The key index of each segment is input into a pre-established feedforward sequence memory network model to modify the key index of each segment, including: According to the lung image and the patient's historical information, the future information of the patient's lung image is predicted by the feedforward sequential memory network model; The key index of each segment is modified based on the predicted future information of the patient's lung image.

2. The method for identifying lung lesions based on endoscopic images according to claim 1, characterized in that: The step of calculating the texture feature statistics of each segment using a gray level co-occurrence matrix according to the texture feature comprises: Generate a corresponding gray-level co-occurrence matrix according to the gray-level image and the parameters of the preset gray-level co-occurrence matrix, wherein the parameters include at least a direction, an offset, and an order; Calculating the texture feature statistics of each segment according to the generated gray level co-occurrence matrix, wherein the texture feature statistics include energy, contrast, correlation and entropy; in, The energy is used to characterize the texture detail level of the texture of the lung image; The contrast is used to characterize the contrast between different gray levels in the lung image texture; The correlation is used to characterize the correlation of pixel gray levels in the lung image texture; The entropy is used to characterize the uncertainty and complexity of the lung image texture.

3. The method for identifying lung lesions based on endoscopic images according to claim 1, characterized in that: The step of converting the lung image into a grayscale image comprises: The lung image is converted into a grayscale image using the weighted average method, and the three components of R, G, and B pixels at the corresponding positions are weighted averaged with different weights. The specific formula is: Gray(i,j)=0.299*R(i,j)+0.578*G(i,j)+0.114*B(i,j); Among them, Gray (i, j) is the gray value of the pixel with coordinates (i, j) in the lung image; R(i, j) is the red channel value of the pixel with coordinates (i, j) in the lung image; G(i, j) is the green channel value of the pixel with coordinates (i, j) in the lung image; B(i, j) is the blue channel value of the pixel with coordinates (i, j) in the lung image.

4. The method for identifying lung lesions based on endoscopic images according to claim 1, characterized in that: In the step of extracting the texture features of each segment according to the Gabor filter, the Gabor filter function is: ； Among them, (x, y) is the pixel coordinate; σx and σy are the standard deviations of the Gaussian envelope; f is the frequency; θ is the direction of the filter; φ is the phase shift.

5. A computer device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the lung lesion identification method based on endoscopic images as described in any one of claims 1 to 4.

6. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for identifying lung lesions based on endoscopic images as described in any one of claims 1 to 4 is implemented.

Images