Medical image enhancement technology based on artificial intelligence and application thereof

Abstract

The invention relates to the technical field of lung CT image enhancement, in particular to a medical image enhancement technology based on artificial intelligence and application thereof. The method comprises the following steps: obtaining a gray feature coefficient according to gray distribution around a pixel point; according to the gradient magnitude of the pixel points around the pixel points, gray-scale highlighted pixel points are obtained; obtaining a mutation characteristic coefficient by combining the position distribution between the pixel point and a reference pixel point, and further obtaining a distance measurement coefficient between any two pixel points; performing hierarchical clustering on the lung CT image to obtain a pixel point clustering frequency and a clustering threshold value; obtaining a local feature fusion degree according to the threshold difference of two adjacent clustering times of the pixel points and the number of clustering times, and further obtaining the clustering priority of each clustering; and performing adaptive enhancement on the image according to the clustering priority to obtain an enhanced lung CT image. According to the method, the problem of excessive enhancement is avoided, the influence of noise on the image is reduced to a certain extent, and finally the lung CT image with a good enhancement effect is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of lung CT image enhancement, and in particular to an artificial intelligence-based medical imaging enhancement technology and its application. Background Art

[0002] Artificial intelligence (AI) is widely used in medical image processing, for example in assisting medical image segmentation, intelligent disease diagnosis, and intelligent prognosis assessment. These technologies can significantly improve the speed and efficiency of image reading, enhancing the accuracy of doctors' diagnoses and treatments. Lung CT images, the most commonly used medical image for lung diagnosis, are crucial for accurate diagnosis and treatment selection. Therefore, enhancing lung CT images using AI-based image enhancement technology has become a crucial step in diagnosis and treatment.

[0003] When enhancing lung CT images, an equalization algorithm is often used to enhance the visual quality of the image. This method divides the image into many local windows of fixed size and equalizes the area within each window, improving the CT image performance and preserving the details of the lung structure. However, this method usually uses a local window-based approach when dividing the area. Although this method is easy to implement, it can easily lead to uneven image enhancement. In addition, lung CT images contain a large number of detailed structures and a certain amount of noise. Ultimately, this may result in insufficient enhancement of important lung details or excessive enhancement of noise in the image, making it impossible to achieve a good enhancement effect. Summary of the Invention

[0004] In order to solve the technical problems that equalization of lung CT images easily leads to uneven image enhancement, and lung CT images contain a large number of detailed structures and a certain amount of noise, which may ultimately result in insufficient enhancement of important lung details or excessive enhancement of noise in the image, and thus fail to achieve a good image enhancement effect, the purpose of the present invention is to provide an artificial intelligence-based medical image enhancement technology and its application. The technical solutions adopted are as follows:

[0005] An artificial intelligence-based medical image enhancement technology and its application method, the method comprising:

[0006] Obtaining a CT image of the patient's lungs;

[0007] Select any pixel point in the lung CT image as a reference pixel point; obtain the grayscale feature coefficient of the reference pixel point based on the grayscale distribution of the preset neighborhood of the reference pixel point; obtain all grayscale prominent pixels in the preset neighborhood of the reference pixel point based on the gradient amplitude of other pixels in the preset neighborhood of the reference pixel point; obtain the mutation feature coefficient of the reference pixel point based on the position distribution between the reference pixel point and all grayscale prominent pixels; obtain the distance measurement coefficient between any two pixels in the lung CT image based on the position distribution, grayscale features, grayscale feature coefficient and mutation feature coefficient of any two pixels in the lung CT image;

[0008] Based on the distance measurement coefficient between each two pixels, the pixels in the lung CT image are hierarchically clustered to obtain the number of clusters for each pixel and the clustering threshold for each clustering. The local feature fusion degree of the reference pixel is obtained based on the difference in clustering thresholds between two adjacent clusterings of the reference pixel and the number of clusterings of all pixels. The clustering priority of each clustering is obtained based on the distribution characteristics of the local feature fusion degree of all pixels in the preset first neighborhood of each pixel.

[0009] The lung CT image is adaptively enhanced according to the clustering priority to obtain an enhanced lung CT image.

[0010] Furthermore, the method for obtaining the grayscale characteristic coefficient includes:

[0011] The grayscale characteristic coefficient is obtained according to the grayscale characteristic coefficient calculation formula. The grayscale characteristic coefficient calculation formula is as follows:

[0012]

[0013] Where, represents the grayscale characteristic coefficient of the reference pixel; I represents the number of pixels in the preset neighborhood of the reference pixel; H i Represents the grayscale value of the i-th pixel in the preset neighborhood of the reference pixel; Represents the grayscale mean of the pixels in the preset neighborhood of the reference pixel; H max，i Represents the maximum grayscale value of the pixel within the preset neighborhood of the reference pixel; || represents the absolute value function.

[0014] Furthermore, the method for obtaining all grayscale prominent pixel points in a preset neighborhood includes:

[0015] Get the grayscale extreme value of the pixel in the preset neighborhood; get the preset extreme value weight coefficient;

[0016] The ratio between the extreme gray value of the pixel point and the preset extreme value weight coefficient is used as the mutation index;

[0017] The pixels in the preset neighborhood whose gradient amplitude is greater than the mutation index are regarded as grayscale prominent pixels.

[0018] Furthermore, the method for obtaining the mutation characteristic coefficient includes:

[0019] Calculate the distance between all grayscale prominent pixels in a preset neighborhood of the reference pixel and the reference pixel as a first distance; sort the pixels in ascending order according to the first distance to obtain a grayscale mutation point sequence;

[0020] The mutation characteristic coefficient is obtained according to the mutation characteristic coefficient calculation formula. The mutation characteristic coefficient calculation formula is as follows:

[0021]

[0022] Where δ represents the mutation characteristic coefficient of the reference pixel; J represents the number of grayscale prominent pixels in the preset neighborhood of the reference pixel; θ j，j+1 R represents the angle between the jth grayscale prominent pixel point and the j+1th grayscale prominent pixel point in the grayscale mutation point sequence and the reference pixel point; j，j+1 represents the distance between the jth grayscale prominent pixel point and the j+1th grayscale prominent pixel point in the grayscale mutation point sequence; R j R represents the distance between the jth grayscale prominent pixel point and the reference pixel point in the grayscale mutation point sequence; j+1 Represents the distance between the grayscale mutation point sequences in the grayscale mutation point sequence.

[0023] Furthermore, the method for obtaining the distance metric coefficient includes:

[0024] The distance metric coefficient is obtained according to the distance metric coefficient calculation formula. The distance metric coefficient calculation formula is as follows:

[0025]

[0026] Where μ A，B Represents the distance measurement coefficient between the Ath pixel and the Bth pixel; R A，B Indicates the distance between the Ath pixel and the Bth pixel; H A Indicates the gray value of the Ath pixel; H B Represents the grayscale value of the B-th pixel; Represents the grayscale characteristic coefficient of the A-th pixel; Represents the grayscale characteristic coefficient of the Bth pixel; δ A Indicates the mutation characteristic coefficient of the A-th pixel; δ B represents the mutation characteristic coefficient of the B-th pixel; || represents the absolute value function.

[0027] Furthermore, the method for obtaining the local feature fusion degree includes:

[0028] The local feature fusion degree is obtained according to the local feature fusion degree calculation formula. The local feature fusion degree calculation formula is as follows:

[0029]

[0030] Where z represents the local feature fusion degree of the reference pixel; M represents the number of clustering times of the reference pixel; M max represents the pixel with the most clustering times in the lung CT image; K m+1 K represents the clustering threshold of the reference pixel in the m+1th hierarchical clustering; m represents the clustering threshold of the reference pixel point in the mth level clustering; U represents the union; || represents the absolute value function; exp represents the exponential function with a natural constant as the base.

[0031] Furthermore, the method for obtaining cluster priority includes:

[0032] Obtaining the edge possibility of the reference pixel as an edge pixel based on the local feature fusion degree of all pixels in the first neighborhood preset by the reference pixel;

[0033] All edge pixel points in the lung CT image during each clustering are obtained; and the edge likelihood mean of all edge pixel points in the lung CT image during each clustering is used as the clustering priority of each clustering.

[0034] Furthermore, the method for obtaining the edge possibility includes:

[0035] The marginal probability is obtained according to the marginal probability calculation formula. The marginal probability calculation formula is as follows:

[0036]

[0037] Where f represents the edge possibility of the reference pixel as an edge pixel; z represents the local feature fusion degree of the reference pixel; L represents the number of pixels in the preset first neighborhood of the reference pixel; z represents the edge probability of the reference pixel as an edge pixel; l Indicates the local feature fusion degree of the lth pixel in the preset first neighborhood of the reference pixel.

[0038] Furthermore, the lung CT image is adaptively enhanced according to the clustering priority to obtain an enhanced lung CT image, including:

[0039] The clustering result with the highest clustering priority is taken as the final clustering result of the lung CT image;

[0040] The area corresponding to the pixel points in each cluster in the final clustering result is used as each final division area of the lung CT image;

[0041] According to the grayscale characteristic coefficient and mutation characteristic coefficient of each pixel in each final divided area, adaptive histogram equalization is performed on each final divided area to obtain the equalization intensity in each final divided area. The calculation formula is as follows:

[0042]

[0043] Where p represents the equalization intensity in each final divided area; Represents the mean value of the grayscale characteristic coefficient of each pixel in the final divided area; It represents the mean value of the mutation characteristic coefficient of each pixel point in each final divided area; p0 represents the basic equalization strength, which can be obtained by the adaptive histogram equalization algorithm; norm() represents the normalization function.

[0044] A medical image enhancement technology based on artificial intelligence and its application system. The system includes a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the steps of the above-mentioned medical image enhancement technology based on artificial intelligence and its application method are implemented.

[0045] The present invention has the following beneficial effects:

[0046] The present invention uses lung CT images as a template for medical imaging for subsequent analysis; since different types of local areas have different grayscale features, in order to distinguish different types of local areas, the grayscale feature coefficient of the reference pixel in the preset neighborhood is obtained according to the grayscale distribution of the preset neighborhood of the reference pixel, and then the grayscale prominent pixel in the preset neighborhood of the reference pixel is selected; since the grayscale mutation may be that the pixel is at the edge of different types of local areas, or it may be a noise point, and the position distribution of the two types of pixels is different, the mutation feature coefficient of the reference pixel is obtained according to the position distribution between the reference pixel and all grayscale prominent pixels; since the lung CT image needs to be divided into regions before image enhancement to obtain local areas with the same features, the lung CT image needs to be hierarchically clustered, so according to the lung CT image The position distribution, grayscale features, grayscale feature coefficients and mutation feature coefficients of any two pixels in the lung CT image are used to obtain the distance measurement coefficient between any two pixels in the lung CT image; since the number of clustering of other pixels except boundary pixels and noise pixels is relatively large, the local feature fusion degree of the reference pixel is obtained according to the difference in clustering thresholds when the reference pixel is clustered twice adjacently and the number of clustering of all pixels; since the smaller the local feature fusion degree, the more likely it is an edge pixel, and the greater the edge possibility of the edge pixel of each cluster extracted during each clustering, the better the clustering effect, so the local feature fusion degree of all pixels in the first neighborhood of each pixel is preset to obtain the clustering priority of each cluster; the lung CT image is adaptively enhanced according to the clustering priority to obtain an enhanced lung CT image. The present invention performs equalization of different intensities on different areas of the lung CT image, increases the visibility of the image while avoiding the problem of excessive enhancement, and reduces the influence of noise on the image to a certain extent, and finally obtains a lung CT image with better enhancement effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 A flowchart of an artificial intelligence-based medical image enhancement technology and its application method provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0049] To further illustrate the technical means and effects employed by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail an artificial intelligence-based medical image enhancement technology and its applications, including its specific implementation, structure, features, and effects. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0050] 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.

[0051] The following describes in detail an artificial intelligence-based medical image enhancement technology and a specific application scheme thereof provided by the present invention in conjunction with the accompanying drawings.

[0052] See also Figure 1 , which illustrates an artificial intelligence-based medical image enhancement technology and its application method provided by an embodiment of the present invention, the method comprising:

[0053] Step S1: Acquire a lung CT image of the patient.

[0054] The embodiments of the present invention are mainly applied to image enhancement scenarios of medical images. Since the medical images of patients include CT images, magnetic resonance images, nuclear medicine PET images, etc., and lung CT images are the most commonly used medical images, the enhancement method of each medical image is the same. Therefore, in the embodiments of the present invention, the lung CT image of the patient is obtained and used as a template. The lung CT image is operated using the following steps.

[0055] In one embodiment of the present invention, semantic segmentation is performed on lung CT images to remove background areas unrelated to the lungs, thereby obtaining a lung CT image containing only the lungs. The semantic segmentation network is a well-known technique to those skilled in the art and will not be described in detail here.

[0056] Step S2: Select any pixel point in the lung CT image as a reference pixel point; obtain the grayscale feature coefficient of the reference pixel point based on the grayscale distribution of the preset neighborhood of the reference pixel point; obtain all grayscale prominent pixels in the preset neighborhood based on the gradient amplitude of other pixels in the preset neighborhood of the reference pixel point; obtain the mutation feature coefficient of the reference pixel point based on the position distribution between the reference pixel point and all grayscale prominent pixels; obtain the distance measurement coefficient between any two pixels in the lung CT image based on the position distribution, grayscale features, grayscale feature coefficient and mutation feature coefficient of any two pixels in the lung CT image.

[0057] In reality, there are different structures such as lung parenchyma, trachea, and nodules in the lung area. The grayscale characteristics of local areas corresponding to different structures are different. When enhancing different types of local areas, it is necessary to consider the grayscale characteristics of the pixels in each local area. If the grayscale is more prominent, the local area will be more significant during image enhancement, while the local area with more uniform grayscale needs more enhancement. In addition, there may be noise areas in the lung CT image. In order to distinguish different types of local areas, in an embodiment of the present invention, the grayscale characteristic coefficient of the reference pixel in the preset neighborhood is obtained according to the grayscale distribution of the preset neighborhood of the reference pixel, so as to facilitate the subsequent clustering of all pixels in the lung CT image.

[0058] Preferably, in one embodiment of the present invention, the method for obtaining the grayscale characteristic coefficient includes:

[0059] The grayscale characteristic coefficient is obtained according to the grayscale characteristic coefficient calculation formula. The grayscale characteristic coefficient calculation formula is as follows:

[0060]

[0061] Where, represents the grayscale characteristic coefficient of the reference pixel; I represents the number of pixels in the preset neighborhood of the reference pixel; H i Represents the grayscale value of the i-th pixel in the preset neighborhood of the reference pixel; Represents the grayscale mean of the pixels in the preset neighborhood of the reference pixel; H max，i Represents the maximum grayscale value of the pixel within the preset neighborhood of the reference pixel; || represents the absolute value function.

[0062] In the grayscale characteristic coefficient calculation formula, It represents the average difference between the grayscale value of each pixel in the preset neighborhood of the reference pixel and the grayscale mean of the preset neighborhood. The larger the average difference, the greater the contrast of the reference pixel. At this time, the more obvious the grayscale feature of the reference pixel is, that is, the larger the grayscale feature coefficient;

[0063] It represents the mean difference between each pixel in the preset neighborhood and the pixel with the largest grayscale value in the preset neighborhood. The larger the mean difference, the less uniform the grayscale of the pixels in the preset neighborhood. At this time, the more prominent the grayscale features of the pixels in the preset neighborhood, that is, the larger the grayscale feature coefficient.

[0064] In one embodiment of the present invention, a circular area with a reference pixel as the center and a radius of 10 is used as the preset neighborhood. It should be noted that in other embodiments of the present invention, the preset neighborhood can be set arbitrarily, which is not limited here.

[0065] In addition to the grayscale characteristics of the pixel, whether there are mutation points around the pixel and the distribution of mutation points are also important characteristics. If there are many grayscale mutation pixels within the preset neighborhood of a pixel, and the grayscale mutation pixels are relatively continuous, it means that the pixel may be close to important tissues such as the trachea and blood vessels, or may be near lesions such as lung nodules and bullae. If these grayscale mutation pixels are relatively scattered, it is considered that there are many noise points around the pixel. Therefore, in this embodiment of the present invention, all grayscale prominent pixels within the preset neighborhood are obtained, and the mutation characteristic coefficient of the reference pixel is obtained based on the position distribution between the reference pixel and all grayscale prominent pixels.

[0066] Preferably, in one embodiment of the present invention, the method for obtaining all grayscale prominent pixel points within a preset neighborhood includes:

[0067] Obtain the grayscale extreme value of the pixel within the preset neighborhood; obtain the preset extreme value weight coefficient. In the embodiment of the present invention, the preset extreme value weight coefficient is set to 10. It should be noted that in other embodiments of the present invention, the preset extreme value weight coefficient can be set by itself and is not limited here.

[0068] The ratio between the grayscale extreme value of the pixel point and the preset extreme value weight coefficient is used as the mutation index; the pixels in the preset neighborhood whose gradient amplitude is greater than the mutation index are regarded as grayscale prominent pixels.

[0069] Preferably, in one embodiment of the present invention, the method for obtaining the mutation characteristic coefficient includes:

[0070] The distance between all grayscale prominent pixels in a preset neighborhood of the reference pixel and the reference pixel is calculated as the first distance; the pixels are sorted from small to large according to the first distance to obtain a grayscale mutation point sequence.

[0071] The mutation characteristic coefficient is obtained according to the mutation characteristic coefficient calculation formula. The mutation characteristic coefficient calculation formula is as follows:

[0072]

[0073] Where δ represents the mutation characteristic coefficient of the reference pixel; J represents the number of grayscale prominent pixels in the preset neighborhood of the reference pixel; θ j，j+1 R represents the angle between the jth grayscale prominent pixel point and the j+1th grayscale prominent pixel point in the grayscale mutation point sequence and the reference pixel point; j，j+1 represents the distance between the jth grayscale prominent pixel point and the j+1th grayscale prominent pixel point in the grayscale mutation point sequence; R j R represents the distance between the jth grayscale prominent pixel point and the reference pixel point in the grayscale mutation point sequence; j+1 Represents the distance between the grayscale mutation point sequences in the grayscale mutation point sequence.

[0074] In the calculation formula of the mutation characteristic coefficient, the angle θ between the line connecting two adjacent grayscale prominent pixels and the reference pixel in the grayscale mutation point sequence of the reference pixel is j，j+1 The larger the distance R between two adjacent grayscale prominent pixels is, the j，j+1 The larger the value is, the more discontinuous the grayscale prominent pixels in the preset neighborhood of the reference pixel are, that is, the higher the possibility of noise pixels around the reference pixel, and the larger the mutation characteristic coefficient of the reference pixel is. If the distance between the grayscale prominent pixel and the reference pixel is closer, the greater the influence of the grayscale prominent pixel on the reference pixel, so As θ j，j+1 ×R j，j+1 The weight of the grayscale mutation point sequence is the value between two adjacent grayscale prominent pixels. Calculate the average and obtain the mutation feature coefficient of the reference pixel.

[0075] Before image enhancement, the lung CT image needs to be divided into regions to obtain local regions with the same characteristics, and the pixels in each local region belong to the same body part, and the noise pixels are separated. Therefore, in an embodiment of the present invention, hierarchical clustering is performed on the lung CT image, and the distance measurement coefficient between the pixels is first analyzed.

[0076] Preferably, in one embodiment of the present invention, the method for obtaining the distance metric coefficient includes:

[0077] The distance metric coefficient is obtained according to the distance metric coefficient calculation formula. The distance metric coefficient calculation formula is as follows:

[0078]

[0079] Where μ A，B Represents the distance measurement coefficient between the Ath pixel and the Bth pixel; R A，B Indicates the distance between the Ath pixel and the Bth pixel; H A Indicates the gray value of the Ath pixel; H B Represents the grayscale value of the B-th pixel; Represents the grayscale characteristic coefficient of the A-th pixel; Represents the grayscale characteristic coefficient of the Bth pixel; δ A Indicates the mutation characteristic coefficient of the A-th pixel; δ B represents the mutation characteristic coefficient of the B-th pixel; ‖ represents the absolute value function.

[0080] In the distance metric coefficient calculation formula, the smaller the distance between pixels A and B, the smaller the grayscale value difference, the smaller the grayscale feature coefficient difference, and the smaller the mutation feature coefficient difference, the smaller the distance metric coefficient of pixels A and B when clustering.

[0081] Step S3: Based on the distance measurement coefficient between each two pixels, the pixels in the lung CT image are hierarchically clustered to obtain the number of clustering times for each pixel and the clustering threshold for each clustering; based on the difference in clustering thresholds between two adjacent clusterings of the reference pixel and the number of clustering times of all pixels, the local feature fusion degree of the reference pixel is obtained; based on the local feature fusion degree distribution characteristics of all pixels in the first neighborhood preset for each pixel, the clustering priority of each clustering is obtained.

[0082] In one embodiment of the present invention, all pixels in a lung CT image are hierarchically clustered from bottom to top to obtain the number of clusters for each pixel and the corresponding clustering threshold for each clustering. Hierarchical clustering is a well-known technique to those skilled in the art and will not be described in detail here.

[0083] After counting the number of clusterings for each pixel in the lung CT image, it can be found that the number of clusterings for boundary pixels and noise pixels of different types of local regions is relatively small, and the distribution of noise pixels is relatively isolated. In order to distinguish the boundaries and noise points of different local regions, in an embodiment of the present invention, the possibility of each pixel being a boundary pixel of a different local region is analyzed. Since the number of clusterings for pixels other than boundary pixels and noise pixels is relatively large, the local feature fusion degree of the reference pixel is obtained based on the difference in clustering thresholds when the reference pixel is clustered twice adjacently and the number of clusterings for all pixels.

[0084] Preferably, in one embodiment of the present invention, the method for obtaining the local feature fusion degree includes:

[0085] The local feature fusion degree is obtained according to the local feature fusion degree calculation formula. The local feature fusion degree calculation formula is as follows:

[0086]

[0087] Where z represents the local feature fusion degree of the reference pixel; M represents the number of clustering times of the reference pixel; M max represents the pixel with the most clustering times in the lung CT image; K m+1 K represents the clustering threshold of the reference pixel in the m+1th hierarchical clustering; m represents the clustering threshold of the reference pixel point in the mth level clustering; U represents the union; ‖ represents the absolute value function; exp represents the exponential function with a natural constant as the base.

[0088] In the local feature fusion calculation formula, the ratio of the number of clustering of the reference pixel to the maximum number of clustering The larger it is, the less likely the reference pixel is to be an edge pixel or a noise pixel, and the higher the degree of fusion of the local features of the reference pixel is. A set of clustering threshold difference values of two adjacent clusterings in all clustering times of the reference pixel is obtained, where the smaller the maximum value of the clustering threshold difference between two adjacent clusterings is, the easier the reference pixel is to be clustered, and the higher the degree of fusion of the local features of the reference pixel is.

[0089] According to the local feature fusion degree, the smaller the local feature fusion degree, the more likely it is an edge pixel point. Among them, the greater the edge possibility of the edge pixel points of each cluster extracted during each clustering, the better the clustering effect. Therefore, in the embodiment of the present invention, the local feature fusion degree of all pixels in the first neighborhood preset for each pixel point is used to obtain the clustering priority of each clustering.

[0090] Preferably, in one embodiment of the present invention, the method for obtaining cluster priority includes:

[0091] According to the local feature fusion degree of all pixels in the first neighborhood of the reference pixel, the edge possibility of the reference pixel as an edge pixel is obtained. In the embodiment of the present invention, the edge possibility calculation formula is as follows:

[0092]

[0093] Where f represents the edge possibility of the reference pixel as an edge pixel; z represents the local feature fusion degree of the reference pixel; L represents the number of pixels in the preset first neighborhood of the reference pixel; z represents the edge probability of the reference pixel as an edge pixel; l Indicates the local feature fusion degree of the lth pixel in the preset first neighborhood of the reference pixel.

[0094] In the edge possibility calculation formula, the lower the square of the local feature fusion degree of the reference pixel point and the average of the local feature fusion degrees of other pixels in the preset first neighborhood, the higher the possibility that the reference pixel point is located at the edge of different types of local areas, that is, the greater the edge possibility of the reference pixel point as an edge pixel point.

[0095] In one embodiment of the present invention, the reference pixel is taken as the center and the eight surrounding pixels are used as the preset first neighborhood. It should be noted that in other embodiments of the present invention, the preset first neighborhood can be set arbitrarily and is not limited here.

[0096] All edge pixel points in the lung CT image during each clustering are obtained according to the edge detection algorithm; the edge likelihood mean of all edge pixel points in the lung CT image during each clustering is used as the clustering priority of each clustering.

[0097] It should be noted that the edge detection algorithm is a technical means well known to those skilled in the art and will not be described in detail here.

[0098] Step S4: Adaptively enhance the lung CT image according to the clustering priority to obtain an enhanced lung CT image.

[0099] Preferably, in one embodiment of the present invention, adaptively enhancing the lung CT image according to the clustering priority to obtain the enhanced lung CT image includes:

[0100] The clustering result with the highest clustering priority is taken as the final clustering result of the lung CT image;

[0101] The area corresponding to the pixel points in each cluster in the final clustering result is used as each final division area of the lung CT image;

[0102] According to the grayscale characteristic coefficient and mutation characteristic coefficient of each pixel in each final divided area, adaptive histogram equalization is performed on each final divided area to obtain the equalization intensity in each final divided area. The calculation formula is as follows:

[0103]

[0104] Where p represents the equalization intensity in each final divided area; Represents the mean value of the grayscale characteristic coefficient of each pixel in the final divided area; It represents the mean value of the mutation characteristic coefficient of each pixel point in each final divided area; p0 represents the basic equalization strength, which can be obtained by the adaptive histogram equalization algorithm; norm() represents the normalization function.

[0105] In the equalization strength calculation formula, the larger the mean value of the grayscale characteristic coefficient and the mean value of the mutation characteristic coefficient in each final divided area, the lower the equalization strength required for the final divided area. At this time, the lower the equalization strength in each final divided area, the lower the equalization strength. As the weight coefficient of the basic equalization strength, the equalization strength in each final divided area is obtained.

[0106] According to the equalization intensity in each final divided area, a histogram equalization algorithm is applied to all final divided areas to obtain an enhanced lung CT image. It should be noted that the histogram equalization algorithm is a technical means well known to those skilled in the art and will not be described in detail here.

[0107] In summary, the patient's lung CT image is obtained; any pixel point in the lung CT image is selected as a reference pixel point; the grayscale characteristic coefficient of the reference pixel point is obtained according to the grayscale distribution of the preset neighborhood of the reference pixel point; all grayscale prominent pixels in the preset neighborhood are obtained according to the gradient amplitude of other pixels in the preset neighborhood of the reference pixel point; the mutation characteristic coefficient of the reference pixel point is obtained according to the position distribution between the reference pixel point and all grayscale prominent pixels; the position distribution, grayscale characteristics, grayscale characteristic coefficient and mutation characteristic coefficient of any two pixels in the lung CT image are obtained. The distance measurement coefficient between two pixels is used; based on the distance measurement coefficient between each two pixels, the pixels in the lung CT image are hierarchically clustered to obtain the number of clustering of each pixel and the clustering threshold of each clustering; based on the difference in clustering thresholds between two adjacent clustering of the reference pixel and the number of clustering of all pixels, the local feature fusion degree of the reference pixel is obtained; based on the local feature fusion degree distribution characteristics of all pixels in the first neighborhood preset for each pixel, the clustering priority of each clustering is obtained; the lung CT image is adaptively enhanced according to the clustering priority to obtain an enhanced lung CT image.

[0108] One embodiment of the present invention provides an artificial intelligence-based medical image enhancement technology and its application system, which includes a memory, a processor and a computer program, wherein the memory is used to store the corresponding computer program, and the processor is used to run the corresponding computer program. When the computer program runs in the processor, it can implement the method described in steps S1-S4.

[0109] 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.

[0110] 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 medical image enhancement technology based on artificial intelligence and its application method, characterized in that: The method comprises: Obtaining a CT image of the patient's lungs; Select any pixel point in the lung CT image as a reference pixel point; obtain the grayscale feature coefficient of the reference pixel point based on the grayscale distribution of the preset neighborhood of the reference pixel point; obtain all grayscale prominent pixels in the preset neighborhood based on the gradient amplitude of other pixels in the preset neighborhood of the reference pixel point; obtain the mutation feature coefficient of the reference pixel point based on the position distribution between the reference pixel point and all grayscale prominent pixels; obtain the distance measurement coefficient between any two pixels in the lung CT image based on the position distribution, grayscale features, grayscale feature coefficient and mutation feature coefficient of any two pixels in the lung CT image; Based on the distance measurement coefficient between each two pixels, the pixels in the lung CT image are hierarchically clustered to obtain the number of clusterings for each pixel and the clustering threshold for each clustering; based on the difference in clustering thresholds between two adjacent clusterings of the reference pixel and the number of clusterings of all pixels, the local feature fusion degree of the reference pixel is obtained; based on the distribution characteristics of the local feature fusion degree of all pixels in a preset first neighborhood of each pixel, the clustering priority of each clustering is obtained; The lung CT image is adaptively enhanced according to the clustering priority to obtain an enhanced lung CT image.

2. The medical image enhancement technology based on artificial intelligence and its application method according to claim 1, characterized in that: The method for obtaining the grayscale characteristic coefficient includes: The grayscale characteristic coefficient is obtained according to the grayscale characteristic coefficient calculation formula, and the grayscale characteristic coefficient calculation formula is as follows: Where, represents the grayscale characteristic coefficient of the reference pixel; I represents the number of pixels in the preset neighborhood of the reference pixel; H i Represents the grayscale value of the i-th pixel in the preset neighborhood of the reference pixel; Represents the grayscale mean of the pixels in the preset neighborhood of the reference pixel; H max，i Represents the maximum grayscale value of the pixel within the preset neighborhood of the reference pixel; || represents the absolute value function.

3. The artificial intelligence-based medical image enhancement technology and its application method according to claim 1, characterized in that: The method for obtaining all grayscale prominent pixel points in a preset neighborhood includes: Get the grayscale extreme value of the pixel in the preset neighborhood; get the preset extreme value weight coefficient; The ratio between the grayscale extreme value of the pixel point and the preset extreme value weight coefficient is used as a mutation index; Pixels within a preset neighborhood whose gradient amplitude is greater than the mutation index are regarded as grayscale prominent pixels.

4. The artificial intelligence-based medical image enhancement technology and its application method according to claim 1, characterized in that: The method for obtaining the mutation characteristic coefficient includes: Calculating the distance between all grayscale prominent pixels in a preset neighborhood of the reference pixel and the reference pixel as a first distance; sorting the pixels in ascending order according to the first distance to obtain a grayscale mutation point sequence; The mutation characteristic coefficient is obtained according to the mutation characteristic coefficient calculation formula, and the mutation characteristic coefficient calculation formula is as follows: Where δ represents the mutation characteristic coefficient of the reference pixel; J represents the number of grayscale prominent pixels in the preset neighborhood of the reference pixel; θ j，j+1 R represents the angle between the jth grayscale prominent pixel point and the j+1th grayscale prominent pixel point in the grayscale mutation point sequence and the reference pixel point; j，j+1 represents the distance between the jth grayscale prominent pixel point and the j+1th grayscale prominent pixel point in the grayscale mutation point sequence; R j R represents the distance between the jth grayscale prominent pixel point and the reference pixel point in the grayscale mutation point sequence; j+1 Represents the distance between the grayscale mutation point sequences in the grayscale mutation point sequence.

5. The artificial intelligence-based medical image enhancement technology and its application method according to claim 1, characterized in that: The method for obtaining the distance metric coefficient includes: The distance metric coefficient is obtained according to the distance metric coefficient calculation formula, and the distance metric coefficient calculation formula is as follows: Where μ A，B Represents the distance measurement coefficient between the Ath pixel and the Bth pixel; R A，B Indicates the distance between the Ath pixel and the Bth pixel; H A Indicates the gray value of the Ath pixel; H B Represents the grayscale value of the B-th pixel; Represents the grayscale characteristic coefficient of the A-th pixel; Represents the grayscale characteristic coefficient of the Bth pixel; δ A Indicates the mutation characteristic coefficient of the A-th pixel; δ B represents the mutation characteristic coefficient of the B-th pixel; || represents the absolute value function.

6. The artificial intelligence-based medical image enhancement technology and its application method according to claim 1, characterized in that: The method for obtaining the local feature fusion degree includes: The local feature fusion degree is obtained according to the local feature fusion degree calculation formula, and the local feature fusion degree calculation formula is as follows: Where z represents the local feature fusion degree of the reference pixel; M represents the number of clustering times of the reference pixel; M max represents the pixel with the most clustering times in the lung CT image; K m+1 K represents the clustering threshold of the reference pixel in the m+1th hierarchical clustering; m represents the clustering threshold of the reference pixel point in the mth level clustering; U represents the union; || represents the absolute value function; exp represents the exponential function with a natural constant as the base.

7. The artificial intelligence-based medical image enhancement technology and its application method according to claim 1, characterized in that: The method for obtaining the cluster priority includes: Obtaining the edge possibility of the reference pixel being an edge pixel according to the local feature fusion degree of all pixels in a first neighborhood preset by the reference pixel; All edge pixel points in the lung CT image during each clustering are obtained; and the edge likelihood mean of all edge pixel points in the lung CT image during each clustering is used as the clustering priority of each clustering.

8. The artificial intelligence-based medical image enhancement technology and its application method according to claim 7, characterized in that: The method for obtaining the edge possibility includes: The edge probability is obtained according to the edge probability calculation formula, which is as follows: Where f represents the edge possibility of the reference pixel as an edge pixel; z represents the local feature fusion degree of the reference pixel; L represents the number of pixels in the preset first neighborhood of the reference pixel; z represents the edge probability of the reference pixel as an edge pixel; l Indicates the local feature fusion degree of the lth pixel in the preset first neighborhood of the reference pixel.

9. The artificial intelligence-based medical image enhancement technology and its application method according to claim 1, characterized in that: Adaptively enhancing the lung CT image according to the clustering priority to obtain an enhanced lung CT image, comprising: taking the clustering result with the highest clustering priority as the final clustering result of the lung CT image; Using the area corresponding to the pixel points in each cluster in the final clustering result as each final divided area of the lung CT image; According to the grayscale characteristic coefficient and mutation characteristic coefficient of each pixel in each final divided area, adaptive histogram equalization is performed on each final divided area to obtain the equalization intensity in each final divided area. The calculation formula is as follows: Where p represents the equalization intensity in each final divided area; Represents the mean value of the grayscale characteristic coefficient of each pixel in the final divided area; It represents the mean value of the mutation characteristic coefficient of each pixel point in each final divided area; p0 represents the basic equalization strength, which can be obtained by the adaptive histogram equalization algorithm; norm() represents the normalization function.

10. An artificial intelligence-based medical image enhancement technology and its application system, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the artificial intelligence-based medical image enhancement technology and its application method as described in any one of claims 1 to 9 are implemented.

Images