Medical image processing method and device based on imaging mark

Abstract

The invention belongs to the technical field of image processing, and discloses a medical image processing method and device based on an imaging marker, and the method comprises the steps: obtaining a medical scanning image, and carrying out the filtering processing, and obtaining a medical gray image; performing depth reconstruction on the medical grayscale image to obtain a display depth image; performing decision segmentation on the display depth image to obtain a plurality of pathological feature region images; performing imaging dimension conversion on each pathological feature region image to obtain a medical modulo perception matrix; determining a medical perception factor according to the medical modulo perception matrix; calculating a feature accumulation operator of each pathological feature region image; performing accumulation operator truncation estimation on each feature accumulation operator to obtain a medical truncation value; and according to the medical truncation value and the medical perceiving factor, performing perceiving screening on each pathological feature region image to obtain a medical abnormal site information set, and generating a medical imaging mark image. According to the invention, the pathological discrimination degree of medical imaging can be improved, the speed of identifying a focus by a doctor can be improved, and the diagnosis efficiency can be improved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and particularly to a medical image processing method and device based on imaging markers. Background Art

[0002] In recent years, imaging methods have developed rapidly in various fields of life, such as in the fields of optics, computer science, medicine, etc. Especially in the medical field, by combining different image detection or image recognition technologies, it can be used for examining internal organs such as the viscera, monitoring the development of fetuses in pregnant women, etc. By taking medical imaging pictures such as B-ultrasound pictures and CT pictures, the cause of the disease can be identified.

[0003] However, the currently scanned medical imaging pictures contain a large amount of useless textures, and the pathological discrimination degree of the images is very low. If the clinical doctor lacks experience, the low pathological discrimination degree will make it difficult for the clinical doctor to accurately identify the location of the lesion in the image, and the diagnosis efficiency will be greatly reduced. Therefore, how to improve the pathological discrimination degree of medical imaging pictures is an urgent problem that the industry needs to solve. Summary of the Invention

[0004] This application provides a medical image processing method and device based on imaging markers, which can improve the pathological discrimination degree of medical imaging, help improve the speed of doctors' identification of lesions, and improve the diagnosis efficiency.

[0005] In a first aspect, an embodiment of this application provides a medical image processing method based on imaging markers, including:

[0006] Obtain a medical scan image and perform filtering processing to obtain a medical grayscale image;

[0007] Perform depth reconstruction on the medical grayscale image to obtain a display depth image;

[0008] Perform decision bisection on the display depth image to obtain multiple pathological feature region images;

[0009] Perform imaging dimension conversion on each pathological feature region image to obtain a medical recognition perception matrix;

[0010] Determine a medical perception factor according to the medical recognition perception matrix;

[0011] Calculate the feature accumulation operator of each pathological feature region image;

[0012] Perform accumulation operator truncation estimation on each feature accumulation operator to obtain a medical truncation value;

[0013] Perform perception screening on each pathological feature region image according to the medical truncation value and the medical perception factor to obtain a medical abnormal site information set, and generate a medical imaging marker image.

[0014] Further, the above-mentioned obtaining of the medical scan image and performing filtering processing to obtain the medical grayscale image includes:

[0015] Performing histogram equalization on the medical scan image to obtain an enhanced histogram image;

[0016] Performing filtering processing on the enhanced histogram image to obtain the medical grayscale image.

[0017] Further, the above-mentioned performing filtering processing on the enhanced histogram image to obtain the medical grayscale image includes:

[0018] Calculating the weighted average value of each pixel value in the enhanced histogram image to obtain a Gaussian kernel;

[0019] Performing a convolution operation on the enhanced histogram image and the Gaussian kernel to obtain the medical grayscale image.

[0020] Further, the above-mentioned performing depth reconstruction on the medical grayscale image to obtain the display depth image includes:

[0021] Performing pixel recombination on the medical grayscale image to obtain a medical pixel image;

[0022] Performing depth frequency decomposition on the medical pixel image to obtain the display depth image.

[0023] Further, the above-mentioned performing depth frequency decomposition on the medical pixel image to obtain the display depth image includes:

[0024] Using wavelet basis functions to decompose the medical pixel image to obtain multiple high-frequency pixel components and multiple low-frequency pixel components;

[0025] Performing inverse transformation on each low-frequency pixel component and each high-frequency pixel component to obtain the display depth image.

[0026] Further, the method further includes:

[0027] After obtaining each high-frequency pixel component, removing the high-frequency pixel components greater than a preset high-frequency threshold.

[0028] Further, the above-mentioned performing decision bisection on the display depth image to obtain multiple pathological feature region images includes:

[0029] Calculating the average value of each pixel value in the display depth image as the pixel attribute value;

[0030] Obtaining a pathological decision value according to the pixel attribute value, each low-frequency pixel component and each high-frequency pixel component;

[0031] Performing decision bisection on the display depth image according to the pathological decision value to obtain multiple pathological feature region images.

[0032] Further, the above-mentioned decision segmentation of the displayed depth image according to the pathological decision value obtains multiple pathological feature region images, including:

[0033] Eliminate the high-frequency pixel components smaller than the pathological decision value in the displayed depth image;

[0034] Perform inverse wavelet transform on the remaining high-frequency pixel components to obtain corresponding high-frequency pixel points;

[0035] Connect the high-frequency pixel points to obtain an edge contour;

[0036] Segment the displayed depth image according to the edge contour to obtain multiple medical edge images;

[0037] Perform binarization processing on each medical edge image to obtain multiple pathological feature region images.

[0038] Further, the above-mentioned imaging dimension conversion of each pathological feature region image to obtain a medical recognition perception matrix includes:

[0039] Construct a medical feature matrix according to each pixel value in each pathological feature region image;

[0040] Perform eigenvalue decomposition on the medical feature matrix to obtain multiple eigenvalues and corresponding eigenvectors;

[0041] Sort each eigenvalue according to the rule from large to small;

[0042] Starting from the largest eigenvalue, select the number of eigenvalues equal to the preset dimensionality reduction depth as the target eigenvalues;

[0043] Take the eigenvectors corresponding to each target eigenvalue as the target eigenvectors;

[0044] Perform dimensionality reduction on the medical feature matrix according to each target eigenvector to obtain a medical recognition perception matrix.

[0045] Further, the above-mentioned determination of the medical perception factor according to the medical recognition perception matrix includes:

[0046] Estimate the covariance matrix of the medical recognition perception matrix to obtain a medical covariance matrix;

[0047] Perform eigen decomposition on the medical covariance matrix to determine the largest eigenvalue and the corresponding eigenvector;

[0048] Take the first vector component of the eigenvector corresponding to the largest eigenvalue as the covariance calibration component;

[0049] Obtain the medical perception factor based on the maximum eigenvalue, covariance calibration component, and pixel attribute value.

[0050] Furthermore, the above-mentioned calculation of the feature aggregation operator for each pathological feature region image includes:

[0051] Calculate the pixel density value and pixel difference value of the pathological feature region image;

[0052] Calculate the feature aggregation operator of the pathological feature region image based on the pixel density value, pixel difference value, pixel attribute value, and preset weight coefficient.

[0053] Furthermore, the above-mentioned truncation estimation of each feature aggregation operator to obtain the medical truncation value includes:

[0054] Obtain the total number of feature pixel points of each pathological feature region image;

[0055] Determine the total sequence length of all feature aggregation operators;

[0056] Calculate the medical truncation value based on the total number of feature pixel points, pixel attribute value, total sequence length, the number of pixel points of each pathological feature region image, and the feature aggregation operator.

[0057] Furthermore, the above-mentioned truncation estimation of each feature aggregation operator to obtain the medical truncation value includes:

[0058] Take the sequence mean of each feature aggregation operator as the medical truncation value.

[0059] Furthermore, the above-mentioned perception screening of each pathological feature region image according to the medical truncation value and medical perception factor to obtain the medical abnormal site information set and generate a medical imaging marker image includes:

[0060] Traverse and identify the pathological feature region image according to the medical perception factor to obtain intermediate pixel points;

[0061] Judge whether the pixel value of the intermediate pixel point is greater than the medical truncation value;

[0062] If so, take the intermediate pixel point as an abnormal pixel point;

[0063] Take each abnormal pixel point as the medical abnormal site information set;

[0064] Generate a medical imaging marker image according to the medical abnormal site information set.

[0065] Furthermore, the above-mentioned traversal and identification of the pathological feature region image according to the medical perception factor to obtain intermediate pixel points includes:

[0066] Calculate the pixel average value of all pixel points in the image of the pathological feature region;

[0067] Use the ratio of the medical perception factor to the pixel average value as the perception ratio coefficient;

[0068] Calculate the ratio of the medical perception factor to the pixel value of each pixel point in the image of the pathological feature region;

[0069] Use the pixel points with a ratio greater than the perception ratio coefficient as intermediate pixel points.

[0070] Further, the method further includes:

[0071] After obtaining each abnormal pixel point, perform noise reduction processing on each abnormal pixel point;

[0072] Perform connected region analysis on each processed abnormal pixel point to obtain multiple connected regions;

[0073] Calculate the center coordinates or bounding boxes of the connected regions as abnormal feature points;

[0074] Use each abnormal feature point as a medical abnormal site information set.

[0075] Further, generating a medical imaging marker image according to the medical abnormal site information set includes:

[0076] Increase the pixel value of each abnormal pixel point in the medical abnormal site information set by a preset pixel threshold;

[0077] Overlay each abnormal pixel point on the medical scan image to obtain a medical imaging marker image.

[0078] In a second aspect, an embodiment of the present application provides a medical image processing device based on an imaging marker, including:

[0079] An acquisition module for acquiring a medical scan image and performing filtering processing to obtain a medical grayscale image;

[0080] A reconstruction module for performing depth reconstruction on the medical grayscale image to obtain a displayed depth image;

[0081] A segmentation module for performing decision segmentation on the displayed depth image to obtain multiple pathological feature region images;

[0082] A conversion module for performing imaging dimension conversion on each pathological feature region image to obtain a medical pattern recognition perception matrix;

[0083] A perception module for determining a medical perception factor according to the medical pattern recognition perception matrix;

[0084] A calculation module for calculating the feature aggregation operator of each pathological feature region image;

[0085] An estimation module for performing aggregation operator truncation estimation on each feature aggregation operator to obtain a medical truncation value;

[0086] A marking module for performing perception screening on each pathological feature region image according to the medical truncation value and the medical perception factor to obtain a set of medical abnormal site information and generate a medical imaging marking image.

[0087] In a third aspect, an embodiment of the present application provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it performs the steps of a medical image processing method based on imaging marking as described in any one of the above embodiments.

[0088] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, it implements the steps of a medical image processing method based on imaging marking as described in any one of the above embodiments.

[0089] In summary, compared with the prior art, the beneficial effects brought by the technical solution provided by the embodiment of the present application at least include:

[0090] A medical image processing method based on imaging marking provided by an embodiment of the present application first performs filtering processing on a medical scan image to obtain a medical grayscale image, performs depth reconstruction on the medical grayscale image to obtain a display depth image, and performs decision segmentation on the display depth image to obtain multiple pathological feature region images. Here, the decision segmentation operation can retain the edge information of the objects in the image, improving the accuracy of abnormal point detection and positioning; secondly, the present application further performs dimension conversion, calculation of medical perception factors and feature aggregation operators based on each pathological feature region image, and finally performs perception screening according to the medical truncation value and the medical perception factor, and can obtain the abnormal points in the pathology, mark them in the medical scan image, improving the pathological discrimination of medical imaging, helping to improve the speed of doctors' identification of lesions, and improving the diagnosis efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 It is a flowchart of a medical image processing method based on imaging marking provided by an exemplary embodiment of the present application.

[0092] Figure 2 It is a flowchart of the decision segmentation step of the display depth image provided by an exemplary embodiment of the present application.

[0093] Figure 3Flowchart of the medical pattern recognition perception matrix determination step provided for an exemplary embodiment of the present application.

[0094] Figure 4 Structural diagram of a medical image processing device based on imaging markers provided for an exemplary embodiment of the present application. Detailed implementation manners

[0095] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments.

[0096] All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0097] Please refer to Figure 1 , an exemplary embodiment of the present application provides a medical image processing method based on imaging markers, including:

[0098] Step S1, obtain a medical scan image and perform filtering processing to obtain a medical grayscale image.

[0099] Among them, the medical scan image may include scan images of the patient's head, chest, abdomen, limbs, etc. The medical scan image can be collected by a medical detection device (such as an X-ray machine, a CT scanner, an MRI, etc.) and uploaded to the medical cloud service system. The medical display obtains the medical scan image from the medical system, and the medical scan image is transmitted to the medical display in digital form. Specifically, when implemented, after obtaining the patient's permission in the cloud service center, the medical scan image can be obtained from the medical system through data calling methods such as Application Programming Interface (API) data calling and web crawler.

[0100] Specifically, the above-mentioned obtaining a medical scan image and performing filtering processing to obtain a medical grayscale image may include:

[0101] Step S11, equalize the medical scan image using the histogram equalization method to obtain an enhanced histogram image.

[0102] In the specific implementation process, the medical scan image can be enhanced by using the histogram equalization method to improve the visualization effect of the medical scan image and obtain an enhanced histogram image.

[0103] Step S12, perform filtering processing on the enhanced histogram image to obtain a medical grayscale image.

[0104] Specifically, the above-mentioned filtering process for the enhanced histogram image to obtain a medical grayscale image includes:

[0105] Step S121, calculating the weighted average value of each pixel value in the enhanced histogram image to obtain a Gaussian kernel.

[0106] Step S122, performing a convolution operation on the enhanced histogram image and the Gaussian kernel to obtain a medical grayscale image.

[0107] In the specific implementation process, the filtering process for the enhanced histogram image can be implemented by using Gaussian filtering. For example, calculating the weighted average value of all pixel values in the enhanced histogram image, taking the weighted average value of all pixel values as the Gaussian kernel, and then performing a convolution operation on the obtained Gaussian kernel and the enhanced histogram image. Thus, the high-frequency noise in the medical scan image can be effectively removed through Gaussian filtering, while the edges and details of the medical scan image are retained, reducing the interference of noise on the detection of abnormal points, and thus obtaining a smooth medical grayscale image for subsequent further processing of the image.

[0108] It should be noted that in other embodiments, other filtering methods can also be used to process the enhanced histogram image, such as median filtering, mean filtering, etc. In addition, the convolution operation in this application is a process of placing the Gaussian kernel at each pixel position of the enhanced histogram image, multiplying each element of the Gaussian kernel by the corresponding image pixel value and then adding them to obtain a new pixel value, which can be implemented by existing technologies such as OpenCV and MATLAB, and will not be elaborated here.

[0109] Step S2, performing depth reconstruction on the medical grayscale image to obtain a displayed depth image.

[0110] Specifically, the above-mentioned depth reconstruction of the medical grayscale image to obtain a displayed depth image may include:

[0111] Step S21, performing pixel recombination on the medical grayscale image to obtain a medical pixel image.

[0112] In the specific implementation process, pixel recombination of medical grayscale images can be obtained by rearranging the pixels in the medical grayscale images, that is: extract all the pixel points in the medical grayscale image, and perform permutations and combinations on the extracted pixel points. For example, exchange the pixel points in the left half of the medical grayscale image with the pixel points in the symmetric positions of the right half of the medical grayscale image to achieve pixel recombination of the medical grayscale image and obtain a medical pixel image; in other embodiments, other permutation and combination rules for the rows or columns of pixel points can also be used for pixel recombination. It should be noted that pixel recombination is an image processing method that enhances the contrast information of an image by recombining the pixel arrangement in the original image. In this application, pixel recombination of medical grayscale images can obtain medical pixel images with more texture details.

[0113] Step S22, perform depth-frequency decomposition on the medical pixel image to obtain a display depth image.

[0114] Specifically, the above-mentioned depth-frequency decomposition of the medical pixel image to obtain a display depth image may include:

[0115] Step S221, decompose the medical pixel image using a wavelet basis function to obtain a plurality of high-frequency pixel components and a plurality of low-frequency pixel components. Among them, each low-frequency pixel component includes the structure and global features of the medical pixel image, and each high-frequency pixel component contains the details and texture information of the medical pixel image.

[0116] Further, after obtaining each high-frequency pixel component, high-frequency pixel components greater than a preset high-frequency threshold can also be removed.

[0117] In the specific implementation process, perform frequency-domain selection on the high-frequency components and remove the high-frequency pixel components greater than the preset high-frequency threshold, which can reduce the interference of the noise generated in the processing process on the image.

[0118] Step S222, perform inverse transformation on each low-frequency pixel component and each high-frequency pixel component to obtain a display depth image.

[0119] It should be noted that the wavelet basis function and inverse transformation in this application both use wavelet transform in the prior art for processing. Wavelet transform can decompose an image into components of different frequencies, and different components correspond to details and structures at different frequencies. In other embodiments, other depth-frequency decomposition methods can also be used, which are not limited here.

[0120] Step S3, perform decision bisection on the display depth image to obtain a plurality of pathological feature region images.

[0121] Specifically, the above-mentioned decision bisection of the displayed depth image to obtain multiple pathological feature region images may include:

[0122] Step S31: Calculate the average value of each pixel value in the displayed depth image as the pixel attribute value.

[0123] It can be considered that the pixel attribute value is the average value of the pixel values of the entire displayed depth image. For example, add up each pixel value in the displayed depth image to obtain the total pixel value, and divide the total pixel value by the total number of pixels to determine the pixel attribute value of the displayed depth image. The pixel attribute value reflects the average level of the acoustic echo intensity within the displayed depth image, and the pixel attribute value can be calculated through image processing libraries such as OpenCV and Pillow in the prior art.

[0124] Step S32: Obtain the pathological decision value based on the pixel attribute value, each low-frequency pixel component, and each high-frequency pixel component.

[0125] Among them, the pathological decision value is a decision parameter used to divide the pixels in the image into two or more different regions, and is used to decide whether a pixel point in the image is a target (feature) region or a background region. The pathological decision value can be determined according to the following formula:

[0126]

[0127] Among them, y represents the weighted average value of each high-frequency pixel component, x represents the weighted average value of each low-frequency pixel component, μ represents the pixel attribute value, and exp is the xth power of the base e of the natural logarithm.

[0128] In addition, it should be noted that calculating the weighted average values of each high-frequency pixel component and each low-frequency pixel component aims to simplify the calculation. Reasonably, the Python programming language in the prior art can be used for calculation, and other calculation methods can also be used in other embodiments, which will not be elaborated here.

[0129] Step S33: Perform decision bisection on the displayed depth image according to the pathological decision value to obtain multiple pathological feature region images.

[0130] Specifically, please refer to Figure 2 The above-mentioned decision bisection of the displayed depth image according to the pathological decision value to obtain multiple pathological feature region images can be implemented by the following steps:

[0131] Step S331: Eliminate the high-frequency pixel components smaller than the pathological decision value in the displayed depth image.

[0132] It can be considered that by using the pathological decision value as the segmentation threshold to screen the high-frequency pixel components in the displayed depth image and retaining the larger high-frequency pixel components, the decision analysis of the displayed depth image can be realized, that is, analyzing the pixel points with various texture information, shape information and other features existing in the displayed depth image.

[0133] Step S332, perform inverse wavelet transform on the remaining high-frequency pixel components respectively to obtain the corresponding high-frequency pixel points.

[0134] Step S333, connect the high-frequency pixel points respectively to obtain the edge contour.

[0135] Step S334, segment the displayed depth image according to the edge contour to obtain multiple medical edge images.

[0136] It can be considered that performing inverse wavelet transform on the screened high-frequency pixel components, connecting all the image pixel points after the inverse transform to obtain the edge contour and the connected region, and segmenting the edge contour in the displayed depth image realizes the decision cut of the displayed depth image, thereby obtaining multiple medical edge images.

[0137] Step S335, perform binarization processing on each medical edge image to obtain multiple pathological feature region images.

[0138] In the specific implementation process, when performing binarization processing on all medical edge images, according to the preset medical binarization threshold, the pixel points in all medical edge images that are greater than the medical binarization threshold can be marked as feature pixel points, and other pixel points are marked as the background. The largest connected region composed of the feature pixel points in each medical edge image is used as the pathological feature region image corresponding to the medical edge image, and then the pathological feature region images corresponding to all medical edge images are determined respectively.

[0139] In addition, it should be noted that by using the depth reconstruction method for the image, the displayed depth image is obtained. This displayed depth image provides more comprehensive and detailed image information than the planar medical image, and retains the structural information of the medical image, can better identify the image edge information, and improves the recognition accuracy of the abnormal points in the image.

[0140] The pathological feature region image is obtained by performing decision screening on the displayed depth image, so as to segment the image with more local details in the displayed depth image from the image background. Targeted detection and monitoring of the lesions in the pathological feature region image can avoid the interference of the background region on the imaging result during the medical image imaging process.

[0141] Step S4, perform imaging dimension conversion on each pathological feature region image to obtain the medical pattern recognition perception matrix.

[0142] Specifically, please refer to Figure 3 , the above-mentioned imaging dimension conversion of each pathological feature region image to obtain a medical pattern recognition perception matrix may specifically include the following steps:

[0143] Step S41, construct a medical feature matrix according to each pixel value in each pathological feature region image.

[0144] In the specific implementation process, each column of the medical feature matrix corresponds to a pathological feature region image, and the matrix element value in this column represents the pixel value of the feature pixel point in this pathological feature region image, thereby determining the medical feature matrix.

[0145] Step S42, perform eigenvalue decomposition on the medical feature matrix to obtain multiple eigenvalues and corresponding eigenvectors.

[0146] Step S43, sort each eigenvalue in descending order.

[0147] Step S44, starting from the largest eigenvalue, select the number of eigenvalues equal to the preset dimensionality reduction depth as the target eigenvalues.

[0148] Step S45, use the eigenvectors corresponding to each target eigenvalue as the target eigenvectors.

[0149] Step S46, perform dimensionality reduction on the medical feature matrix according to each target eigenvector to obtain a medical pattern recognition perception matrix.

[0150] It can be understood that according to the size of the eigenvalues, the first w largest eigenvalues and the corresponding eigenvectors are selected to perform dimensionality reduction processing on the medical feature matrix, thereby determining the medical pattern recognition perception matrix; where w is the preset dimensionality reduction depth, usually calibrated as a positive integer according to the dimensionality reduction requirement; the medical pattern recognition perception matrix is mainly used to characterize the distribution of feature pixel points in medical images.

[0151] In some embodiments, the number of rows of the medical feature matrix can be determined according to the size of the largest pathological feature region image. To facilitate subsequent dimensionality reduction processing of the medical feature matrix, when there are empty elements in the medical feature matrix, the empty elements are regarded as zero elements and matrix dimensionality reduction is performed; the above-mentioned feature pixel points represent the abnormal distribution of image pixels, so there may be pixel points of pathological features such as abnormal dot-like structures, cystic structures, and abnormal accumulation of liquids.

[0152] In some embodiments, the dimensionality reduction process can be performed using the PCA method in the prior art. Specifically, when implementing, the eigenvectors corresponding to the first w largest eigenvalues can also be used to form a medical sample matrix, where w is the preset dimensionality reduction depth. Multiply the transposed matrix of the medical sample matrix by the medical feature matrix to complete the dimensionality reduction process of the medical feature matrix and obtain a medical pattern recognition matrix, which can be specifically determined according to the following formula:

[0153] M = (U w ) T ·Q

[0154] where, (U w ) T represents the transposed matrix of the medical sample matrix, Q represents the medical feature matrix, and M represents the medical pattern recognition matrix; in other embodiments, other dimensionality reduction methods can also be used, which are not limited here.

[0155] Step S5, determine the medical perception factor according to the medical pattern recognition matrix.

[0156] Specifically, the determination of the medical perception factor according to the medical pattern recognition matrix may include:

[0157] Step S51, perform covariance matrix estimation on the medical pattern recognition matrix to obtain a medical covariance matrix.

[0158] In the specific implementation process, the covariance matrix estimation can be performed on the medical pattern recognition matrix to reduce the dispersion degree between different feature pixel values in the medical pattern recognition matrix and obtain a medical covariance matrix.

[0159] It should be noted that covariance matrix estimation is a data processing method that performs statistical analysis on a given data matrix to calculate the covariance relationship between various variables and thus generates a covariance matrix.

[0160] Step S52, perform eigen decomposition on the medical covariance matrix to determine the largest eigenvalue and the corresponding eigenvector.

[0161] In the specific implementation process, for the largest eigenvalue of the medical covariance matrix, the multiple eigenvalues of the medical covariance matrix can be determined by performing eigen decomposition on the medical covariance matrix, and the largest eigenvalue can be selected from the multiple eigenvalues. In some embodiments, the largest eigenvalue of the medical covariance matrix can be calculated according to NumPy in the Python library.

[0162] Step S53, use the first vector component of the eigenvector corresponding to the largest eigenvalue as the covariance calibration component.

[0163] The covariance calibration component is the first vector component of the eigenvector corresponding to the largest eigenvalue of the medical covariance matrix, that is, the value of the first dimension element of the eigenvector in the vector space. The first vector component in the eigenvector can be obtained through indexing in the Python library and is used as a numerical value for convenient calculation and processing.

[0164] Step S54, obtain the medical perception factor based on the largest eigenvalue, the covariance calibration component, and the pixel attribute value.

[0165] The medical perception factor α can be determined by the following formula:

[0166]

[0167] where λ max represents the largest eigenvalue of the medical covariance matrix, β represents the covariance calibration component, and μ is the pixel attribute value.

[0168] In addition, it should be noted that the medical perception factor is a feature recognition coefficient determined based on the largest eigenvalue of the medical covariance matrix, the covariance calibration component, and the pixel attribute value, and is used to identify abnormal features of pixel points in the medical imaging diagram according to the discrimination degree of each pathological feature region image.

[0169] Step S6, calculate the feature accumulation operator of each pathological feature region image.

[0170] Specifically, the above calculation of the feature accumulation operator of each pathological feature region image may include:

[0171] Step S61, calculate the pixel density value and the pixel difference value of the pathological feature region image.

[0172] Among them, the pixel density value represents the density quantization value of the feature pixel points in the pathological feature region image, that is, the ratio of the number of feature pixel points to ordinary pixel points in the pathological feature region image; the pixel difference value is the average of the number difference between abnormal feature points and normal pixel points in the medical scan image obtained through machine learning.

[0173] Step S62, calculate the feature accumulation operator of the pathological feature region image based on the pixel density value, the pixel difference value, the pixel attribute value, and the preset weight coefficient.

[0174] In the specific implementation process, the feature accumulation operator corresponding to the i-th pathological feature region image can be determined by the following formula:

[0175]

[0176] where c is the image identification coefficient, β is the covariance calibration component, μ is the pixel attribute value, ε is the image dissimilarity value, d iis the pixel mean of the image of the i-th pathological feature region.

[0177] The image identification coefficient c is a quantitative value used to measure attributes such as the quantity, size, and distribution of specific abnormal features in a medical image. The larger the image identification coefficient, the more obvious the abnormal features; the smaller the image identification coefficient, the less obvious the abnormal features. Usually, the image identification coefficient can be obtained by calculation through a machine model. The ratio between the correct identification rate and the wrong identification rate of historical image identification determined by the machine model is used as the image identification coefficient.

[0178] In some embodiments, the image identification coefficient can be calibrated as a constant 2.

[0179] The image dissimilarity value ε is a quantitative value used to determine whether the feature points in the image of the pathological feature region are abnormal points. Specifically, in implementation, the image dissimilarity value can be determined by the pixel density value, pixel difference value, pixel attribute value, and a preset weight coefficient:

[0180]

[0181] where ρ represents the pixel density value, γ represents the pixel difference value, represents the preset weight coefficient, and μ is the pixel attribute value.

[0182] Preset weight coefficient represents the influence degree of different abnormal features on the medical imaging marker map. By calculating the proportional coefficient between the weighted average value of all abnormal pixel points in the historical medical imaging marker map and the average value of all pixel points in the historical medical imaging marker map, this proportional coefficient is used as the preset weight coefficient; the pixel density value, pixel difference value, and preset weight coefficient can all be calculated and determined using Python programming software in the prior art, which will not be elaborated here.

[0183] Step S7, perform accumulation operator truncation estimation on each feature accumulation operator to obtain a medical truncation value.

[0184] Specifically, the above-mentioned performing accumulation operator truncation estimation on each feature accumulation operator to obtain a medical truncation value includes:

[0185] Step S71, obtain the total number of feature pixel points of each pathological feature region image.

[0186] Step S72, determine the total sequence length of all feature accumulation operators.

[0187] After obtaining the feature accumulation operators corresponding to each pathological feature region image respectively, arrange all the feature accumulation operators in ascending order to determine the feature accumulation operator sequence; the total sequence length is the sequence length of the feature accumulation operator sequence.

[0188] Step S73: Calculate the medical cut-off value based on the total number of characteristic pixel points, pixel attribute values, total sequence length, number of pixel points in each pathological feature region image, and characteristic accumulation operator.

[0189] In a specific implementation process, the medical cut-off value η can be determined by the following formula:

[0190]

[0191] where N represents the total number of characteristic pixel points in all pathological feature region images, μ represents the pixel attribute value, A is the sequence length of the characteristic accumulation operator sequence, ξ i is the characteristic accumulation operator corresponding to the i-th pathological feature region image, and O i is the number of pixel points in the i-th pathological feature region image.

[0192] In some embodiments, the sequence mean value of each characteristic accumulation operator can also be used as the medical cut-off value.

[0193] Step S8: Perform perception screening on each pathological feature region image according to the medical cut-off value and medical perception factor to obtain a set of medical abnormal site information, and generate a medical imaging marked image.

[0194] Specifically, the above-mentioned perception screening of each pathological feature region image according to the medical cut-off value and medical perception factor to obtain a set of medical abnormal site information and generate a medical imaging marked image includes:

[0195] Step S81: Traverse and identify the pathological feature region image according to the medical perception factor to obtain intermediate pixel points.

[0196] In a specific implementation process, the above-mentioned traversing and identifying the pathological feature region image according to the medical perception factor to obtain intermediate pixel points can specifically include the following steps:

[0197] Step S811: Calculate the pixel average value of all pixel points in the pathological feature region image.

[0198] Step S812: Take the ratio of the medical perception factor to the pixel average value as the perception ratio coefficient.

[0199] Step S813: Calculate the ratio of the medical perception factor to the pixel value of each pixel point in the pathological feature region image.

[0200] Step S814: Take the pixel points whose ratio is greater than the perception ratio coefficient as intermediate pixel points.

[0201] It can be considered that the intermediate pixel points are the pixel points of the image with drastic change feature points in the pathological feature region image, and thus contain most of the pathological features and texture features in the pathological feature region image.

[0202] Step S82: Determine whether the pixel value of the intermediate pixel point is greater than the medical cut-off value.

[0203] Step S83: If so, regard the intermediate pixel point as an abnormal pixel point.

[0204] Step S84: Regard each abnormal pixel point as a set of medical abnormal site information.

[0205] Specifically, regard the intermediate pixel points greater than the medical cut-off value as abnormal pixel points, and form a set of medical abnormal site information from all abnormal pixel points. The abnormal pixel points are the pixel points in the pathological feature region image that are most likely to contain pathological features. Highlighting the abnormal pixel points can improve the doctor's diagnosis efficiency of the patient's lesion.

[0206] In another embodiment, the method may further include:

[0207] Step S841: After obtaining each abnormal pixel point, perform noise reduction processing on each abnormal pixel point.

[0208] Performing noise reduction processing on the abnormal pixel points here can reduce possible noises or artifacts.

[0209] Step S842: Perform connected component analysis on each processed abnormal pixel point to obtain multiple connected components;

[0210] Step S843: Calculate the center coordinates or bounding boxes of the connected components as abnormal feature points;

[0211] Step S844: Regard each abnormal feature point as a set of medical abnormal site information.

[0212] The above noise reduction processing can adopt deep learning methods in the prior art. By using deep learning models such as convolutional neural network (CNN) or generative adversarial network (GAN) to denoise the image, it has good processing capabilities for various types of noises and complex image structures. In other embodiments, other noise reduction methods can also be adopted, which are not limited here.

[0213] Step S85: Generate a medical imaging marker image according to the set of medical abnormal site information.

[0214] Specifically, generating a medical imaging marker image according to the set of medical abnormal site information includes:

[0215] Step S851: Increase the pixel values of each abnormal pixel in the medical abnormal site information set by a preset pixel threshold.

[0216] Step S852: Superimpose each abnormal pixel on the medical scan image to obtain a medical imaging marked image.

[0217] In specific implementation, an image processing software (such as the Matplotlib library in the Python library) can be used to locate and mark the abnormal pixels. The abnormal pixels refer to the pixels that may have pathological features such as abnormal dot structures, cystic structures, and abnormal accumulation of liquids. The abnormal pixels can be marked by increasing their pixel values, and then superimposed on the medical scan image to ensure that the abnormal pixels match the medical scan image. Furthermore, the superimposed image is visually presented to ensure that the abnormal pixels are clearly visible in the image, and finally a medical imaging marked image is generated. In the generated medical imaging marked image, the pathological discrimination degree of the medical imaging is improved, and the efficiency of doctors in judging the lesions of patients is increased.

[0218] It should be noted that the pathological discrimination degree = the sum of the pixel values of the pixels containing lesion features / (the sum of the pixel values of the pixels containing lesion features + the sum of the pixel values of the useless pixels). The higher the pathological discrimination degree, the more prominent the lesion area is presented in the medical imaging. In this application, the abnormal pixels in the medical abnormal site information set are highlighted, thereby improving the pathological discrimination degree of the medical imaging and increasing the efficiency of doctors in judging the lesions of patients.

[0219] The above-mentioned visual presentation is to adjust parameters such as color mapping, brightness, and contrast of the image, which can be realized by existing technologies such as image processing software, programming languages, or professional medical image processing, and will not be elaborated here.

[0220] In the above embodiment, a medical image processing method based on imaging marking is provided. First, the medical scan image is filtered to obtain a medical grayscale image, the medical grayscale image is depth-reconstructed to obtain a display depth image, and the display depth image is decision-segmented to obtain multiple pathological feature region images. This decision-segmentation operation can retain the edge information of the objects in the image and improve the accuracy of detecting and locating abnormal points. Secondly, this application further calculates the dimensional transformation, medical perception factor, and feature accumulation operator based on each pathological feature region image. Finally, perception screening is performed according to the medical truncation value and the medical perception factor, and the abnormal points in pathology can be obtained and marked on the medical scan image, which improves the pathological discrimination degree of medical imaging, helps to improve the speed of doctors in identifying lesions, and improves the diagnosis efficiency.

[0221] Please refer to Figure 4, another embodiment of the present application provides a medical image processing device based on imaging markers, including:

[0222] An acquisition module 101, configured to acquire a medical scan image and perform filtering processing to obtain a medical grayscale image.

[0223] A reconstruction module 102, configured to perform depth reconstruction on the medical grayscale image to obtain a displayed depth image.

[0224] A segmentation module 103, configured to perform decision segmentation on the displayed depth image to obtain multiple pathological feature region images.

[0225] A conversion module 104, configured to perform imaging dimension conversion on each pathological feature region image to obtain a medical recognition perception matrix.

[0226] A perception module 105, configured to determine a medical perception factor according to the medical recognition perception matrix.

[0227] A calculation module 106, configured to calculate the feature aggregation operator of each pathological feature region image.

[0228] An estimation module 107, configured to perform aggregation operator truncation estimation on each feature aggregation operator to obtain a medical truncation value.

[0229] A marking module 108, configured to perform perception screening on each pathological feature region image according to the medical truncation value and the medical perception factor to obtain a set of medical abnormal site information, and generate a medical imaging marker image.

[0230] The specific limitations provided in this embodiment regarding a medical image processing device based on imaging markers can be referred to the embodiment of a medical image processing method based on imaging markers in the above text, which will not be elaborated here. Each module in the above medical image processing device based on imaging markers can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above respective modules.

[0231] An embodiment of the present application provides a computer device, which may include a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the processor is caused to execute the steps of a medical image processing method based on imaging markers as described in any of the above embodiments.

[0232] For the working process, working details, and technical effects of the computer device provided in this embodiment, reference may be made to the embodiments of a medical image processing method based on imaging markers in the foregoing text, and details are not described herein again.

[0233] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of a medical image processing method based on imaging markers as described in any of the above embodiments. Among them, the computer-readable storage medium refers to a carrier for storing data, and may include, but is not limited to, a floppy disk, an optical disc, a hard disk, a flash memory, a USB flash drive, and / or a Memory Stick, etc. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. For the working process, working details, and technical effects of the computer-readable storage medium provided in this embodiment, reference may be made to the embodiments of a medical image processing method based on imaging markers in the foregoing text, and details are not described herein again.

[0234] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM).

[0235] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0236] The above-described embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A medical image processing method based on imaging markers, characterized in that: include: Acquire a medical scan image and perform filtering processing to obtain a medical grayscale image; Performing depth reconstruction on the medical grayscale image to obtain a display depth image; Performing decision segmentation on the displayed depth image to obtain a plurality of pathological characteristic area images; Performing imaging dimension conversion on each of the pathological characteristic region images to obtain a medical model recognition perception matrix; Determining a medical awareness factor according to the medical model awareness matrix; Calculating the feature accumulation operator of each of the pathological feature region images; Performing accumulation operator truncation estimation on each of the characteristic accumulation operators to obtain a medical truncation value; The images of the pathological characteristic regions are screened according to the medical cutoff value and the medical perception factor to obtain a medical abnormal site information set and generate a medical imaging marker image.

2. The medical image processing method based on imaging markers according to claim 1, characterized in that: The step of performing depth reconstruction on the medical grayscale image to obtain a display depth image includes: Performing pixel reorganization on the medical grayscale image to obtain a medical pixel image; Depth frequency decomposition is performed on the medical pixel image to obtain the display depth image.

3. The medical image processing method based on imaging markers according to claim 2, characterized in that: The performing depth frequency decomposition on the medical pixel image to obtain the display depth image comprises: The medical pixel image is decomposed by using wavelet basis function to obtain multiple high-frequency pixel components and multiple low-frequency pixel components; Inverse transformation is performed on each of the low-frequency pixel components and each of the high-frequency pixel components to obtain the display depth image.

4. The medical image processing method based on imaging markers according to claim 3, characterized in that: The decision segmentation of the displayed depth image to obtain a plurality of pathological characteristic area images includes: Calculate the average value of each pixel value in the display depth image as the pixel attribute value; Obtaining a pathology decision value according to the pixel attribute value, each of the low-frequency pixel components and each of the high-frequency pixel components; The displayed depth image is segmented and decided according to the pathological decision value to obtain a plurality of pathological characteristic area images.

5. The medical image processing method based on imaging markers according to claim 4, characterized in that: The decision segmentation of the displayed depth image according to the pathological decision value to obtain a plurality of pathological characteristic area images includes: Eliminating the high-frequency pixel component whose value is smaller than the pathological decision value in the displayed depth image; Performing inverse wavelet transform on the remaining high-frequency pixel components to obtain corresponding high-frequency pixel points; Connecting each of the high-frequency pixel points to obtain an edge contour; Segmenting the displayed depth image according to the edge contour to obtain a plurality of medical edge images; Binarization is performed on each of the medical edge images to obtain a plurality of pathological characteristic region images.

6. The medical image processing method based on imaging markers according to claim 1, characterized in that: The step of performing imaging dimension conversion on each of the pathological characteristic region images to obtain a medical model recognition perception matrix includes: Constructing a medical feature matrix according to each pixel value in each of the pathological feature region images; Performing eigenvalue decomposition on the medical feature matrix to obtain multiple eigenvalues ​​and corresponding eigenvectors; Sort the eigenvalues ​​in descending order; Starting from the maximum eigenvalue, select eigenvalues ​​equal to the preset dimensionality reduction depth as the target eigenvalues; Taking the feature vectors corresponding to the target feature values ​​as the target feature vectors; The medical feature matrix is ​​reduced in dimension according to each of the target feature vectors to obtain the medical model recognition matrix.

7. The medical image processing method based on imaging markers according to claim 4, characterized in that: The step of determining the medical perception factor according to the medical model perception matrix comprises: Performing covariance matrix estimation on the medical model recognition matrix to obtain a medical covariance matrix; Performing eigendecomposition on the medical covariance matrix to determine the maximum eigenvalue and the corresponding eigenvector; Taking the first vector component of the eigenvector corresponding to the maximum eigenvalue as the co-square calibration component; The medical perception factor is obtained according to the maximum eigenvalue, the covariance calibration component and the pixel attribute value.

8. The medical image processing method based on imaging markers according to claim 7, characterized in that: The step of calculating the feature accumulation operator of each of the pathological feature region images comprises: Calculating the pixel density value and the pixel difference value of the pathological characteristic region image; The feature accumulation operator of the pathological feature region image is calculated according to the pixel density value, the pixel difference value, the pixel attribute value and a preset weight coefficient.

9. The medical image processing method based on imaging markers according to claim 8, characterized in that: The step of performing an accumulation operator truncation estimation on each of the feature accumulation operators to obtain a medical truncation value comprises: Obtaining the total number of characteristic pixel points of each pathological characteristic region image; Determining the total sequence length of all the feature accumulation operators; The medical cutoff value is calculated based on the total number of characteristic pixel points, the pixel attribute value, the total sequence length, the number of pixels of each pathological characteristic region image and a characteristic accumulation operator.

10. A medical image processing device based on imaging markers, characterized in that: include: An acquisition module is used to acquire a medical scan image and perform filtering processing to obtain a medical grayscale image; A reconstruction module, used for performing depth reconstruction on the medical grayscale image to obtain a display depth image; A segmentation module, used for performing decision segmentation on the displayed depth image to obtain multiple pathological characteristic area images; A conversion module, used for performing imaging dimension conversion on each of the pathological characteristic region images to obtain a medical model recognition perception matrix; A perception module, used for determining a medical perception factor according to the medical model perception matrix; A calculation module, used for calculating the feature accumulation operator of each of the pathological feature region images; An estimation module, used for performing accumulation operator truncation estimation on each of the characteristic accumulation operators to obtain a medical truncation value; The marking module is used to perform perception screening on each of the pathological characteristic region images according to the medical cutoff value and the medical perception factor, obtain a medical abnormal site information set, and generate a medical imaging marking image.

Images