Lung CT image analysis method and system
By using grayscale processing and optimizing target region segmentation, combined with an improved imbalanced sample recognition model, the problems of radiation and noise in lung CT image analysis are solved, improving the accuracy and speed of image analysis, and making it suitable for clinical diagnosis of lung CT images.
Patent Information
- Application Number
- CN202411630383.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-15
- Publication Date
- 2025-02-21
AI Technical Summary
In existing technologies, lung CT image analysis methods present a contradiction between radiation exposure and increased image noise. Furthermore, the generalization ability and accuracy of convolutional neural networks are limited, and the scarcity of high-quality image samples affects model performance.
Lung CT images are processed by grayscale, the target region is divided and optimized, and the image is optimized using a grayscale restoration model. An improved imbalanced sample image recognition model is used for recognition, and an activation function of EfficientNetV2 is constructed to improve the model training efficiency and accuracy.
It improves the accuracy and speed of lung CT image resolution, reduces computational complexity, provides more reliable image information support, and enhances the model's recognition efficiency and stability.
Smart Images
Figure CN119494803A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of medical image processing, and in particular to a lung CT image analysis method and system. Background Art
[0002] The quality of lung CT images is directly affected by the X-ray flux. On the one hand, if you want to obtain high-quality images, you need to increase the X-ray flux, which will inevitably increase the patient's radiation exposure and cause further damage to the patient's body. On the other hand, if you reduce the radiation exposure of the patient by reducing the radiation dose, the number of photons received during the imaging process will decrease, resulting in a significant increase in noise in the image and reducing the accuracy of lung CT image interpretation.
[0003] With the continuous development of convolutional neural networks, using convolutional neural networks to analyze lung CT images has gradually become a feasible option, that is, first use a convolutional neural network to reduce the noise of lung CT images, and then use another convolutional neural network to recognize lung CT images. However, in order to achieve better image analysis results, convolutional neural networks are often designed to be very complex, and a large number of image samples are required to train the model. High-quality lung CT images are often very scarce, which not only limits the performance of convolutional neural networks, but also has a negative impact on the generalization ability of convolutional neural networks. Therefore, it is necessary to explore a lung CT image analysis method with stronger generalization ability and better accuracy. Summary of the invention
[0004] In view of the defects in the prior art, the present invention provides a lung CT image analysis method and system.
[0005] In order to achieve the above-mentioned purpose, in the first aspect, the present invention provides a lung CT image analysis method, the method comprising the following steps: obtaining an original lung CT image, and using the original lung CT image to obtain a lung CT image; selecting a first optimization target area, a second optimization target area, and a third optimization target area in the lung CT image; optimizing the lung CT image according to the grayscale values of the first optimization target area, the second optimization target area, and the third optimization target area to obtain a clear image; constructing an unbalanced sample image recognition model, and using the unbalanced sample image recognition model to perform image recognition on the clear image. The present invention can improve the accuracy of lung CT image analysis and provide more reliable and accurate image information support for clinical diagnosis.
[0006] Optionally, the obtaining of an original lung CT image and using the original lung CT image to obtain a lung CT image comprises the following steps: Obtain original lung CT images; The original lung CT image is grayed to obtain the lung CT image.
[0007] Furthermore, the present invention improves image contrast by graying the original lung CT image, reduces computational complexity, and improves the speed and accuracy of image analysis.
[0008] Optionally, selecting the first optimization target area, the second optimization target area, and the third optimization target area in the lung CT image comprises the following steps: Taking any pixel point in the lung CT image as a target pixel point, and determining a rectangular area with the target pixel point as the center as the first optimization target area; In the first optimization target area, a rectangular area is determined with the target pixel point as the center as the second optimization target area; Any pixel point in the first optimization target area is taken as a candidate pixel point, and a rectangular area is determined with the candidate pixel point as the center as the third optimization target area.
[0009] Furthermore, the present invention provides a data basis for calculating the true grayscale value of the target pixel by dividing the first optimization target area, the second optimization target area and the third optimization target area.
[0010] Optionally, the number of pixel points included in the first optimization target area is greater than the number of pixel points included in the second optimization target area.
[0011] Furthermore, the first optimized target area contains more pixels, which helps to capture image features in a wider range, provides a more comprehensive information basis for calculating the true grayscale value of the target pixel, and tries to maintain enough image details.
[0012] Optionally, the size of the second optimization target area is the same as the size of the third optimization target area.
[0013] Furthermore, the size of the second optimization target area and the size of the third optimization target area are set to be the same to facilitate the subsequent construction of a grayscale restoration model to calculate the true grayscale value of the target pixel.
[0014] Optionally, optimizing the lung CT image according to the grayscale values of the first optimization target area, the second optimization target area, and the third optimization target area to obtain a clear image comprises the following steps: Based on the grayscale values of all pixels in the second optimization target area and the third optimization target area, a grayscale restoration model is used to calculate the true grayscale value of the target pixel; Each pixel point in the lung CT image is taken as the target pixel point in turn and its true grayscale value is calculated, thereby completing the optimization of the lung CT image and obtaining the clear image.
[0015] Furthermore, by comprehensively considering the grayscale values of the second optimization target area and the third optimization target area and using the grayscale restoration model to calculate the true grayscale value of the target pixel, the impact of noise and artifacts on image quality can be effectively reduced and the image clarity can be improved.
[0016] Optionally, the grayscale restoration model satisfies the following relationship: Wherein, Gr is the real gray value of the target pixel, is the original grayscale value of the target pixel, M is the number of pixels in the first optimization target area, is the mean of all gray values in the first linear array A, , n is the number of pixels in the second optimization target area, is the gray value of the i-th position in the first linear array, and the first linear array is obtained by converting the gray value matrix corresponding to the second optimization target area, is the mean of all gray values in the j-th second linear array B, , is the gray value of the i-th position in the j-th second linear array, and the second linear array is obtained by converting the gray value matrix corresponding to the third optimization target area, For a about and The algebraic expression of .
[0017] Furthermore, the grayscale restoration model of the present invention finds information in the first optimization target area that has a similar structure to the second optimization target area, and performs weighted summation of the grayscale values of each pixel in the first optimization target area according to the degree of similarity to obtain the pixel value of the target pixel in the second optimization target area, thereby retaining the details and texture information of the lung CT image as much as possible when repairing the image, thereby improving the accuracy of lung CT image analysis.
[0018] Optionally, the constructing of an unbalanced sample image recognition model and using the unbalanced sample image recognition model to perform image recognition on the clear image comprises the following steps: Improve the activation function of EfficientNetV2 to obtain the original image recognition model; Using low-quality lung CT images to construct a low-level image dataset, and using the model initialization dataset to complete the initialization of the original image recognition model to obtain a backup image recognition model; Using high-quality lung CT images to construct an advanced image dataset to complete the training and verification of the backup image recognition model, thereby obtaining the unbalanced sample image recognition model; The unbalanced sample image recognition model is used to perform image recognition on the clear image.
[0019] Furthermore, the unbalanced sample image recognition model constructed by the present invention has higher recognition accuracy and faster recognition speed, and can improve the recognition efficiency of lung CT images.
[0020] Optionally, the activation function of the original image recognition model satisfies the following relationship: in, is the dependent variable of the activation function of the original image recognition model, and x is the independent variable of the activation function of the original image recognition model.
[0021] Furthermore, the activation function used in the present invention is relatively simple when x is greater than 0, which can accelerate the training process of the model, reduce the problem of gradient disappearance, and help improve the calculation efficiency. The activation function used in the present invention does not have the problem of gradient disappearance in the negative region when x is less than 0, which can improve the model's ability to handle negative values, enhance the model's nonlinear expression ability, improve the model's convergence speed and stability, and thus improve the recognition efficiency of lung CT images.
[0022] In the second aspect, the present invention also provides a lung CT image analysis system, the lung CT image analysis system uses a lung CT image analysis method provided by the present invention, and the lung CT image analysis system includes: a data acquisition and preprocessing module, the data acquisition and preprocessing module is used to obtain the original lung CT image, and use the original lung CT image to obtain the lung CT image; a data optimization module, the data optimization module is used to select the first optimization target area, the second optimization target area and the third optimization target area in the lung CT image; according to the gray value of the first optimization target area, the second optimization target area and the third optimization target area, the lung CT image is optimized to obtain a clear image; an image analysis module, the image analysis module is used to construct an unbalanced sample image recognition model, and use the unbalanced sample image recognition model to perform image recognition on the clear image; a data output module, the data output module is used to output the image recognition result. The system provided by the present invention can stably run the method provided by the present invention, and improve the practicality of the method of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0024] Figure 1 A schematic diagram of a process of analyzing a lung CT image according to an embodiment of the present invention; Figure 2 Schematic diagram of the framework of a lung CT image analysis system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0025] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not need to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, software or methods are not specifically described.
[0026] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "in one embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or subcombination. In addition, it should be understood by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale.
[0027] It should be noted in advance that, in an optional embodiment, except for independent explanations, the same symbols or letters appearing in all formulas have the same meanings and values.
[0028] In an alternative embodiment, see Figure 1 The present invention provides a lung CT image analysis method, the method comprising the following steps: S1. Obtain an original lung CT image, and use the original lung CT image to obtain a lung CT image.
[0029] Wherein, step S1 specifically includes the following steps: S1. Obtain original lung CT images.
[0030] S2. Gray-scaling the original lung CT image to obtain the lung CT image.
[0031] Specifically, in this embodiment, the image contrast is improved by graying the original lung CT image, and the computational complexity is reduced, thereby improving the speed and accuracy of image analysis.
[0032] S2. Selecting a first optimization target region, a second optimization target region, and a third optimization target region in the lung CT image.
[0033] The present invention provides a data basis for calculating the true grayscale value of the target pixel by dividing the first optimization target area, the second optimization target area and the third optimization target area. Step S2 specifically includes the following steps: S21. Taking any pixel point in the lung CT image as a target pixel point, and determining a rectangular area with the target pixel point as the center as the first optimization target area.
[0034] Specifically, in this embodiment, the length and width of the first optimization target area both include 21 pixels.
[0035] S22: In the first optimization target area, determine a rectangular area with the target pixel point as the center as the second optimization target area.
[0036] Specifically, in this embodiment, the number of pixel points included in the first optimization target area is greater than the number of pixel points included in the second optimization target area.
[0037] More specifically, the length and width of the second optimized target area both contain 7 pixels. The first optimized target area contains more pixels, which helps to capture image features in a wider range in the future, provides a more comprehensive information basis for calculating the true grayscale value of the target pixel, and tries to keep enough image details.
[0038] S23: Take any pixel point in the first optimization target area as a candidate pixel point, and determine a rectangular area with the candidate pixel point as the center as the third optimization target area.
[0039] Specifically, in this embodiment, the size of the third optimization target area is the same as the size of the second optimization target area, that is, the length and width of the third optimization target area both contain 7 pixels.
[0040] Furthermore, the size of the second optimization target area and the size of the third optimization target area are set to be the same to facilitate the subsequent construction of a grayscale restoration model to calculate the true grayscale value of the target pixel.
[0041] S3. Optimize the lung CT image according to the grayscale values of the first optimization target area, the second optimization target area, and the third optimization target area to obtain a clear image.
[0042] Wherein, step S3 specifically includes the following steps: S31, calculating the true gray value of the target pixel using a gray restoration model based on the gray values of all pixels in the second optimization target area and the third optimization target area; Specifically, in this embodiment, this embodiment searches for information having a similar structure to the second optimization target area by continuously changing the position of the third optimization target area in the first optimization target area, and then obtains the true gray value of the target pixel point, so as to reduce the noise in the image and retain the details and texture information of the lung CT image as much as possible. This process can be represented by a gray restoration model, that is: Wherein, Gr is the real gray value of the target pixel, is the original grayscale value of the target pixel, M is the number of pixels in the first optimization target area, is the mean of all gray values in the first linear array A, , n is the number of pixels in the second optimization target area, is the gray value of the i-th position in the first linear array, and the first linear array is obtained by converting the gray value matrix corresponding to the second optimization target area, is the mean of all gray values in the j-th second linear array B, , is the gray value of the i-th position in the j-th second linear array, and the second linear array is obtained by converting the gray value matrix corresponding to the third optimization target area, For a about and The algebraic expression of . Specifically, the following relationship is satisfied: More specifically, when calculating the true grayscale value of the target pixel, the grayscale restoration model uses As the weight of the candidate pixel gray value. To express the similarity between the second optimization target area and the third optimization target area, use To reflect the correlation between the second optimization target area and the third optimization target area, The smaller the value, the greater the similarity. Larger values indicate greater correlation.
[0043] Further, the formula It has a larger value when the similarity between the second optimization target area and the third optimization target area is larger, and as the similarity decreases to a certain value, the value of the formula also decreases rapidly, thereby reducing the influence of the third optimization target area with smaller similarity to the second optimization target area on the calculation of the true grayscale value of the target pixel point, which can really reduce the blurriness of the lung CT image, improve the image optimization quality, and thus improve the accuracy of lung CT image analysis.
[0044] S32, taking each pixel point in the lung CT image as the target pixel point in turn and calculating its true grayscale value, thereby completing the optimization of the lung CT image and obtaining the clear image.
[0045] Specifically, in this embodiment, the original gray value of the target pixel is replaced with the real gray value of the target pixel, thereby optimizing the lung CT image and obtaining the clear image.
[0046] Furthermore, this embodiment uses a grayscale restoration model to calculate the true grayscale value of the target pixel, which can effectively reduce the impact of noise and artifacts on image quality, improve image clarity, and thereby improve the accuracy of lung CT image analysis.
[0047] S4. Construct an unbalanced sample image recognition model, and use the unbalanced sample image recognition model to perform image recognition on the clear image.
[0048] Wherein, step S4 specifically includes the following steps: S41. Improve the activation function of EfficientNetV2 to obtain the original image recognition model.
[0049] Specifically, in this embodiment, the original SiLU activation function in EfficientNetV2 is replaced with an improved activation function to obtain the original image recognition model. The activation function of the original image recognition model satisfies the following relationship: in, is the dependent variable of the activation function of the original image recognition model, and x is the independent variable of the activation function of the original image recognition model.
[0050] Furthermore, the activation function used in this embodiment is relatively simple when x is greater than 0, that is, , which can speed up the training process of the original image recognition model and reduce the problem of gradient disappearance, which helps to improve computing efficiency. The activation function used in this embodiment satisfies when x is not greater than 0 It has the characteristics of non-monotonicity, no upper bound, lower bound, continuous differentiability, and image smoothness. It will not have the problem of gradient disappearance in negative areas. It can improve the original image recognition model's ability to handle negative values, and can enhance the nonlinear expression ability of the original image recognition model, improve the convergence speed and stability of the original image recognition model, and thus improve the recognition efficiency of the unbalanced sample image recognition model for lung CT images.
[0051] S42. Use low-quality lung CT images to construct a low-level image dataset, and use the model initialization dataset to complete the initialization of the original image recognition model to obtain a backup image recognition model.
[0052] Specifically, in this embodiment, low-quality lung CT images refer to relatively blurred lung CT images, which usually contain various noises and have artifacts. These lung CT images are easy to obtain and are large in number, and can be used to construct a low-level image dataset.
[0053] Furthermore, the low-level image dataset is used to pre-train the original image recognition model, and then the original image recognition model is initialized to obtain a backup image recognition model.
[0054] Initializing the original image recognition model allows the model to learn some basic image features and structures and accumulate some prior knowledge. Since low-quality lung CT images contain more noise and interference, initializing the original image recognition model is actually training the model to extract useful information in complex and unfavorable environments, laying a solid foundation for improving the accuracy of lung CT image analysis.
[0055] S43. Use high-quality lung CT images to construct an advanced image data set to complete the training and verification of the backup image recognition model, and obtain the unbalanced sample image recognition model.
[0056] Specifically, in this embodiment, high-quality lung CT images refer to lung CT images with clear imaging, no noise and artifacts, or very little noise and artifacts, and these lung CT images are rare.
[0057] More specifically, we first construct an initial advanced image dataset using high-quality lung CT images, and then expand the initial advanced image dataset through data enhancement to obtain an advanced image dataset. We then use the advanced image dataset to train and verify the backup image recognition model, optimize and adjust the backup image recognition model, and obtain an unbalanced sample image recognition model.
[0058] Furthermore, the unbalanced sample image recognition model constructed by the present invention has higher recognition accuracy and faster recognition speed, and can improve the recognition efficiency of lung CT images.
[0059] S44: Use the unbalanced sample image recognition model to perform image recognition on the clear image.
[0060] Specifically, in this embodiment, an unbalanced sample image recognition model is used to identify various types of lesions in clear images, such as tumors, inflammations, fractures, vascular lesions, lung nodules, calcifications, etc.
[0061] It should be noted that, in some cases, the actions described in the specification can be performed in a different order and still achieve the desired results. In this embodiment, the order of steps given is only to make the embodiment appear clearer and easier to explain, rather than to limit it.
[0062] In an alternative embodiment, see Figure 2 The present invention also provides a lung CT image analysis system, which uses a lung CT image analysis method provided by the present invention. The lung CT image analysis system includes a data acquisition and preprocessing module 1, a data optimization module 2, an image analysis module 3 and a data output module 4.
[0063] The data acquisition and preprocessing module 1 is used to obtain original lung CT images, and use the original lung CT images to obtain lung CT images.
[0064] Specifically, in this embodiment, the data acquisition and preprocessing module 1 is electrically connected to the data optimization module 2. The data acquisition and preprocessing module 1 can receive the original lung CT image acquired using X-ray imaging technology and grayscale it to obtain the lung CT image. The data acquisition and preprocessing module 1 specifically performs the content described in step S1.
[0065] The data optimization module 2 is used to select a first optimization target area, a second optimization target area and a third optimization target area in the lung CT image; optimize the lung CT image according to the grayscale values of the first optimization target area, the second optimization target area and the third optimization target area to obtain a clear image.
[0066] Specifically, in this embodiment, the data optimization module 2 is electrically connected to the data acquisition and preprocessing module 1 and the image analysis module 3. The data optimization module 2 specifically executes the contents described in step S2 and step S3.
[0067] The image analysis module 3 is used to construct an unbalanced sample image recognition model, and use the unbalanced sample image recognition model to perform image recognition on the clear image.
[0068] Specifically, in this embodiment, the image analysis module 3 is electrically connected to the data optimization module 2 and the data output module 4. The image analysis module 3 specifically executes the content described in step S4.
[0069] The data output module 4 is used to output the image recognition result.
[0070] The system provided by the present invention can stably run the method provided by the present invention, thereby improving the practicability of the method provided by the present invention.
[0071] In summary, firstly, the present invention selects a first optimization target area, a second optimization target area and a third optimization target area in the lung CT image, and then uses a grayscale restoration model to optimize the lung CT image to obtain a clear image, and the obtained clear image can retain the detail features and texture features of the original image. Secondly, in view of the scarcity of high-quality lung CT images, the present invention improves the activation function in EfficientNetV2 to improve the image recognition efficiency and accuracy of the convolutional neural network, obtains the original image recognition model, and initializes the original image recognition model by pre-training, thereby having an unbalanced sample image recognition model with higher recognition accuracy and faster recognition speed. Finally, the unbalanced sample image recognition model is used to identify the clear image, thereby improving the accuracy of the analysis of the lung CT image, and providing more reliable and accurate image information support for clinical diagnosis. In addition, the system provided by the present invention can stably run the method provided by the present invention, thereby improving the practicality of the method of the present invention.
[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.
Claims
1. A lung CT image analysis method, characterized in that: The steps include: Acquire an original lung CT image, and use the original lung CT image to acquire a lung CT image; Selecting a first optimization target area, a second optimization target area, and a third optimization target area in the lung CT image; Optimizing the lung CT image according to the grayscale values of the first optimization target area, the second optimization target area, and the third optimization target area to obtain a clear image; An unbalanced sample image recognition model is constructed, and the unbalanced sample image recognition model is used to perform image recognition on the clear image.
2. A lung CT image analysis method according to claim 1, characterized in that: The step of obtaining an original lung CT image and using the original lung CT image to obtain a lung CT image comprises the following steps: Obtain original lung CT images; The original lung CT image is grayed to obtain the lung CT image.
3. A lung CT image analysis method according to claim 1, characterized in that: The selecting of the first optimization target area, the second optimization target area and the third optimization target area in the lung CT image comprises the following steps: Taking any pixel point in the lung CT image as a target pixel point, and determining a rectangular area with the target pixel point as the center as the first optimization target area; In the first optimization target area, a rectangular area is determined with the target pixel point as the center as the second optimization target area; Any pixel point in the first optimization target area is taken as a candidate pixel point, and a rectangular area is determined with the candidate pixel point as the center as the third optimization target area.
4. The lung CT image analysis method according to claim 1, characterized in that: The number of pixels included in the first optimization target area is greater than the number of pixels included in the second optimization target area.
5. The lung CT image analysis method according to claim 1, characterized in that: The size of the second optimization target area is the same as the size of the third optimization target area.
6. A lung CT image analysis method according to claim 1, characterized in that: The step of optimizing the lung CT image according to the grayscale values of the first optimization target area, the second optimization target area, and the third optimization target area to obtain a clear image comprises the following steps: Based on the grayscale values of all pixels in the second optimization target area and the third optimization target area, a grayscale restoration model is used to calculate the true grayscale value of the target pixel; Each pixel point in the lung CT image is taken as the target pixel point in turn and its true grayscale value is calculated, thereby completing the optimization of the lung CT image and obtaining the clear image.
7. A lung CT image analysis method according to claim 6, characterized in that: The grayscale restoration model satisfies the following relationship: Wherein, Gr is the real gray value of the target pixel, is the original grayscale value of the target pixel, M is the number of pixels in the first optimization target area, is the mean of all gray values in the first linear array A, , n is the number of pixels in the second optimization target area, is the gray value of the i-th position in the first linear array, and the first linear array is obtained by converting the gray value matrix corresponding to the second optimization target area, is the mean of all gray values in the j-th second linear array B, , is the gray value of the i-th position in the j-th second linear array, and the second linear array is obtained by converting the gray value matrix corresponding to the third optimization target area, For a about and The algebraic expression of .
8. The lung CT image analysis method according to claim 1, characterized in that: The step of constructing an unbalanced sample image recognition model and using the unbalanced sample image recognition model to perform image recognition on the clear image comprises the following steps: Improve the activation function of EfficientNetV2 to obtain the original image recognition model; Using low-quality lung CT images to construct a low-level image dataset, and using the model initialization dataset to complete the initialization of the original image recognition model to obtain a backup image recognition model; Using high-quality lung CT images to construct an advanced image dataset to complete the training and verification of the backup image recognition model, thereby obtaining the unbalanced sample image recognition model; The unbalanced sample image recognition model is used to perform image recognition on the clear image.
9. A lung CT image analysis method according to claim 8, characterized in that: The activation function of the original image recognition model satisfies the following relationship: in, is the dependent variable of the activation function of the original image recognition model, and x is the independent variable of the activation function of the original image recognition model.
10. A lung CT image analysis system, wherein the lung CT image analysis system uses a lung CT image analysis method according to any one of claims 1 to 9, characterized in that: The lung CT image analysis system comprises: A data acquisition and preprocessing module, the data acquisition and preprocessing module is used to acquire an original lung CT image, and use the original lung CT image to acquire a lung CT image; A data optimization module, wherein the data optimization module is used to select a first optimization target area, a second optimization target area, and a third optimization target area in the lung CT image; optimize the lung CT image according to the grayscale values of the first optimization target area, the second optimization target area, and the third optimization target area to obtain a clear image; An image analysis module, the image analysis module is used to construct an unbalanced sample image recognition model, and use the unbalanced sample image recognition model to perform image recognition on the clear image; A data output module, wherein the data output module is used to output image recognition results.