An intelligent analysis method for medical images of liver fibrosis

Abstract

The present invention relates to the technical field of medical image analysis, and discloses a method for intelligent analysis of medical images of liver fibrosis, the method comprising: extracting micro-liver fibrosis regions from liver tissue images based on similarities and differences of pixel regions in liver tissue images to obtain micro-liver fibrosis regions; quantitatively rating the degree of fibrosis in the micro-liver fibrosis regions; and comprehensively evaluating liver fibrosis based on the degree of fibrosis in the micro-liver fibrosis regions to obtain a result of liver fibrosis degree evaluation. The present invention generates an initial contour of a liver region, iterates based on the smoothness, curvature and gradient energy of the contour to obtain a smooth and continuous contour curve that approximates the liver boundary, accurately segments the liver tissue image of the medical image data, extracts multi-scale fusion fibrosis features from the pixel region, and realizes micro-image region extraction of liver fibrosis lesion regions in liver tissue images and early liver fibrosis degree evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image analysis, and in particular to an intelligent analysis method for medical images of liver fibrosis. Background Art

[0002] The liver is the largest metabolic organ in the human body and undertakes many important functions, such as detoxification, protein synthesis, participation in blood sugar metabolism, and fat storage. When the liver suffers from long-term inflammation or toxin damage, liver cells will be damaged and fibrosis will be initiated. The occurrence of liver fibrosis is usually caused by excessive accumulation of collagen and other extracellular matrix (ECM) components, which will destroy the normal structure and function of the liver. As the degree of fibrosis increases, the elasticity of the liver gradually loses, which in turn causes complications such as liver failure, portal hypertension, ascites, esophageal varices, and even liver cancer. Therefore, early diagnosis and timely treatment of liver fibrosis are crucial to the patient's prognosis. At present, the evaluation of liver fibrosis mainly relies on liver biopsy, serum marker detection, imaging examination and other methods. Although traditional liver biopsy can provide accurate histological diagnosis, its invasiveness and the risk of sample error and complications limit its widespread application in clinical practice. Although serum markers can provide a certain basis for diagnosis, their accuracy and sensitivity are relatively low. Therefore, imaging examinations, especially liver imaging analysis based on ultrasound, CT, MRI and other technologies, have become an important means of non-invasive detection of liver fibrosis. Traditional imaging examinations have certain limitations in the quantitative assessment of liver fibrosis. They are not sensitive enough to identify mild liver fibrosis and cannot perceive subtle structural changes in the liver. Therefore, how to achieve early liver fibrosis analysis has become an important issue that needs to be solved urgently. Summary of the invention

[0003] In view of this, the present invention provides an intelligent analysis method for medical images of liver fibrosis, which obtains the early liver fiber change structure through comparative learning, and performs intelligent identification of liver fibrosis based on this, thereby realizing automated detection of liver fibrosis.

[0004] To achieve the above object, the present invention provides a method for intelligent analysis of liver fibrosis medical images, comprising the following steps:

[0005] S1: Collect liver medical image data and perform preprocessing to obtain liver tissue images, wherein the image preprocessing adopts liver region segmentation and filtering noise reduction methods;

[0006] S2: Based on the similarities and differences of pixel regions in the liver tissue image, the micro-liver fibrosis region is extracted from the liver tissue image to obtain the micro-liver fibrosis region;

[0007] S3: Quantitatively grade the degree of fibrosis in the micro-liver fibrosis area to obtain the degree of fibrosis in different micro-liver fibrosis areas;

[0008] S4: Comprehensive assessment of liver fibrosis is performed based on the fibrosis degree of the micro-liver fibrosis area to obtain the liver fibrosis degree assessment result.

[0009] As a further improvement method of the present invention:

[0010] Optionally, collecting liver medical imaging data and preprocessing the data in step S1 includes:

[0011] Collect liver medical image data, which are magnetic resonance imaging results of the liver, including imaging results of the cross section, coronal plane and sagittal plane. The collected liver medical image data are expressed in the form of:

[0012] ;

[0013] ;

[0014] in:

[0015] Indicates liver medical imaging data;

[0016] The imaging results of the liver in the first to third planes are shown in sequence, wherein the first to third planes are the transverse plane, the coronal plane and the sagittal plane;

[0017] Indicates imaging results In the form of a pixel matrix, Indicates imaging results The pixel value of the pixel at row x and column y in , , X represents the number of pixel rows of the imaging result, and Y represents the number of pixel columns of the imaging result;

[0018] Liver medical imaging data Perform preprocessing, where the preprocessing process is:

[0019] S11: Calculate the pixel gradient matrix of any imaging result in the liver medical imaging data I, where the imaging result The pixel gradient matrix is :

[0020] ;

[0021] ;

[0022] ;

[0023] in:

[0024] Indicates imaging results The gradient value of the pixel at the xth row and yth column in , Represents the horizontal gradient value, Indicates the vertical gradient value;

[0025] S12: Perform local maximum judgment on any gradient value in the pixel gradient matrix, and set the gradient value of the non-local maximum value to 0, where the gradient value The local maximum judgment formula is:

[0026] ;

[0027] in:

[0028] Indicates imaging results The neighborhood gradient value set of the pixel in the xth row and yth column;

[0029] In the embodiment of the present invention, if the imaging result The gradient direction of the pixel at row x and column y in Close to 0 degrees, include ;

[0030] If the gradient direction Close to 90 degrees, include ;

[0031] If the gradient direction Close to 45 degrees, include ;

[0032] If the gradient direction Close to 135 degrees, include ;

[0033] Indicates imaging results The gradient value of the pixel at row x and column y in The local maximum judgment result of 1 indicates the gradient value is a local maximum, If it is 0, it means the gradient value is not a local maximum, and let ;

[0034] S13: Mark pixels with gradient values ​​higher than a preset threshold as edge pixels, and use a clustering algorithm to cluster the edge pixels to obtain multiple edge pixel clusters in the imaging result; connect all edge pixels in the edge pixel cluster to obtain a set of contour curves corresponding to the edge pixel cluster, where the imaging result The contour curve corresponding to the kth edge pixel cluster in for:

[0035] ;

[0036] in:

[0037] Represents the contour curve The curve position coordinates corresponding to the nth edge pixel in , , Indicates imaging results The total number of edge pixel clusters;

[0038] In the embodiment of the present invention, the clustering algorithm adopted is the Kmeans algorithm;

[0039] S14: Calculate the contour curve The normalized arc length of each curve position coordinate in The normalized arc length is :

[0040] ;

[0041] in:

[0042] ;

[0043] S15: The contour curve is interpolated by using a B-spline interpolation method, and an interpolation curve parameter equation corresponding to the contour curve is generated based on the normalized arc length, where the contour curve The corresponding interpolation curve parameter equation ,in is the interpolation curve parameter equation The variables in , s corresponds to the normalized arc length, , represents the curve position coordinates at the normalized arc length s, , Respectively represent the horizontal coordinate value and the vertical coordinate value;

[0044] S16: Based on the smoothness of the contour curve and the edge energy, the curve position coordinates at different normalized arc length positions in the contour curve are updated, where the contour curve The curve position coordinates at the normalized arc length s The update formula is:

[0045] ;

[0046] ;

[0047] ;

[0048] ;

[0049] ;

[0050] in:

[0051] represents the learning rate, Set to 0.8;

[0052] The differential component representing the normalized arc length represents a tiny arc segment along the contour curve;

[0053] Represents the interpolation curve parameter equation The first derivative of Represents the interpolation curve parameter equation The second derivative of

[0054] Represents the energy function that combines the smoothness of the contour curve and the edge energy, Represents the elastic energy weight that controls the length and stretchability of the contour curve. represents the bending energy weight that controls the smoothness of the contour curve;

[0055] Represents the contour curve The degree of curvature, They are respectively the preset maximum bending degree and the minimum bending degree;

[0056] Indicates the curve position coordinates The gradient value of the pixel at ;

[0057] S17: Repeat step S16 until the position coordinates of all curves in the contour curve do not change;

[0058] S18: Perform filtering and denoising processing on the image within the contour curve, and use the filtering and denoising result as the liver tissue image.

[0059] Optionally, in step S18, filtering and denoising the image within the contour curve, and using the filtering and denoising result as the liver tissue image, comprises:

[0060] S181: Acquire the image within the contour curve, where the contour curve The image inside is :

[0061] ;

[0062] in:

[0063] Representing images Middle The pixel value of the pixel in row b column, , , Representing images The number of pixel rows, Representing images The number of pixel columns;

[0064] S182: Filter the image within the contour curve, where the image The filtering result is :

[0065] ;

[0066] ;

[0067] in:

[0068] c represents the filter radius, set c to 1;

[0069] Represents the filter scale coefficient, Set to 2;

[0070] represents an exponential function with a natural constant as base;

[0071] Representing images Middle The filtered pixel value of the pixel in row b column;

[0072] S183: Using the filtered image as a liver tissue image, obtaining a set of liver tissue images with different imaging results in the liver medical imaging data, wherein the imaging results The liver tissue image set is :

[0073] ;

[0074] in:

[0075] Indicates imaging results The total number of edge pixel clusters is also the imaging result Total number of images of liver tissue.

[0076] Optionally, in step S2, based on the similarities and differences of pixel regions in the liver tissue image, extracting micro-liver fibrosis regions from the liver tissue image comprises:

[0077] Based on the similarity and difference of pixel areas in liver tissue images, the micro-liver fibrosis area is extracted from liver tissue images. The extraction process of micro-liver fibrosis area is as follows:

[0078] S21: Liver tissue image Divide into The size is pixels and non-overlapping pixel areas, Image showing liver tissue The total number of pixel areas divided;

[0079] S22: Calculate the similarity and difference of pixels in each pixel area, where the liver tissue image The mth pixel area The similarities and differences are , :

[0080] S23: Extract multi-scale features from the pixel region to obtain multi-scale features of the pixel region, where the pixel region The multi-scale features are :

[0081] ;

[0082] ;

[0083] in:

[0084] Represents pixel area Convolution features at the zth convolution scale, represents the convolution matrix at the zth convolution scale, * represents the convolution operator, represents the activation function;

[0085] Represents normalization processing, restoring the convolution features to Pixel size;

[0086] S24: Combining similarities and differences to decode and reconstruct multi-scale features to highlight fibrosis features, and obtain multi-scale fusion fibrosis features of pixel areas, where the pixel areas The multi-scale fusion fibrosis characteristics are :

[0087]

[0088] S25: Based on the multi-scale fusion fibrosis features of the pixel area, the micro-liver fibrosis area is identified in the pixel area, where the pixel area The identification formula is:

[0089] ;

[0090] in:

[0091] Represents pixel area If the recognition result of If it is 1, it means the pixel area The corresponding liver area has liver fibrosis, and the pixel area As micro-liver fibrosis areas;

[0092] Represents multi-scale fusion fibrosis features The standard deviation of

[0093] Image showing liver tissue The mean of all multi-scale fused fibrosis features;

[0094] Indicates the preset recognition threshold;

[0095] The micro-liver fibrosis area collection constitutes three imaging results, among which the imaging results The micro-fibrosis area collection is ;

[0096] ;

[0097] in:

[0098] Indicates imaging results The hth micro-fibrosis area in Indicates imaging results The total number of micro-fibrotic areas.

[0099] Optionally, in step S22, the similarity and difference of pixels in each pixel area are calculated, wherein the liver tissue image The mth pixel area The similarities and differences are ,include:

[0100] S221: If the pixel values ​​are If two pixels are the most adjacent in the horizontal or vertical direction, it means that the two pixels appear adjacent to each other and the two pixels constitute the pixel value Corresponding pixel pairs;

[0101] S222: Calculate the frequency of adjacent pixels corresponding to any two pixel values ​​in the pixel area, where the pixel area Medium pixel value The probability of the corresponding pixel appearing nearby is :

[0102] ;

[0103] in:

[0104] Represents pixel area Medium pixel value The number of corresponding pixel pairs;

[0105] S223: Calculate pixel area Similarity of pixels in :

[0106] ;

[0107] S224: Calculate pixel area The difference in pixels:

[0108] ;

[0109] in:

[0110] Represents pixel area The difference in pixels.

[0111] Optionally, the step S3 performs quantitative rating of the degree of fibrosis in the micro-liver fibrosis area, comprising:

[0112] The degree of fibrosis in the micro-liver fibrosis area was quantitatively graded, and the quantitative grading process of the degree of fibrosis in the micro-liver fibrosis area was as follows:

[0113] S31: Obtain multi-scale fusion fibrosis features of micro-liver fibrosis areas, where micro-liver fibrosis areas The multi-scale fusion fibrosis characteristics are ;

[0114] S32: Perform perceptual attention processing on the multi-scale fused fibrosis features to obtain a perceptual attention matrix corresponding to the multi-scale fused fibrosis features, where the multi-scale fused fibrosis features The corresponding perceptual attention matrix is :

[0115] ;

[0116] in:

[0117] Represents the multi-scale fusion fibrosis features Any element q in The operation of OK Column perceptual attention matrix ;

[0118] S33: Use the perceptual attention matrix to mask and perform maximum pooling on the multi-scale fused fibrosis features to obtain a quantitative feature map, where the multi-scale fused fibrosis features The mask and maximum pooling processing formula are:

[0119] ;

[0120] in:

[0121] Represents multi-scale fusion fibrosis features The corresponding quantitative feature map;

[0122] Represents element-wise multiplication;

[0123] Represents maximum pooling processing;

[0124] S34: Quantitatively grade the degree of fibrosis on the quantitative feature map to obtain the degree of fibrosis in the micro-liver fibrosis area, wherein the quantitative feature map The quantitative rating formula for the degree of fibrosis is:

[0125] ;

[0126] in:

[0127] Micro-fibrosis area The degree of fibrosis;

[0128] Represents the quantitative feature map of early liver fibrosis obtained by contrastive learning.

[0129] Optionally, in step S4, comprehensive assessment of liver fibrosis is performed based on the fibrosis degree of the micro-liver fibrosis area, including:

[0130] Comprehensive assessment of liver fibrosis is performed based on the degree of fibrosis in the micro-fibrosis area, and the comprehensive assessment process of liver fibrosis is as follows:

[0131] S41: Obtain the fibrosis grade of all micro-liver fibrosis areas;

[0132] S42: Calculate the liver fibrosis assessment weights of the imaging results in different planes, where the imaging results The liver fibrosis assessment weight is :

[0133] ;

[0134] in:

[0135] Indicates imaging results The fibrosis degree grade of all micro-liver fibrosis areas is the set;

[0136] , Represents the set of fibrosis levels Any degree of fibrosis in

[0137] Indicates the degree of fibrosis The imaging results The frequency of all micro-fibrosis areas in the liver;

[0138] S43: Based on the liver fibrosis assessment weight, the liver fibrosis degree assessment result is calculated:

[0139] ;

[0140] in:

[0141] Indicates the results of liver fibrosis assessment. The higher the level, the higher the degree of liver fibrosis.

[0142] In order to solve the above problem, the present invention provides an electronic device, the electronic device comprising:

[0143] A memory storing at least one instruction;

[0144] Communication interface, enabling electronic equipment to communicate; and

[0145] The processor executes the instructions stored in the memory to implement the above-mentioned intelligent analysis method of liver fibrosis medical images.

[0146] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned intelligent analysis method of liver fibrosis medical images.

[0147] Compared with the prior art, the present invention proposes an intelligent analysis method for liver fibrosis medical images, which has the following advantages:

[0148] First, this scheme proposes a liver tissue image segmentation processing method, which collects liver medical imaging data of different plane imaging, generates the initial contour of the liver area based on the gradient value of the pixel point, and constructs the parameter equation of the contour curve. Based on the smoothness, curvature and gradient energy of the contour, the contour curve is iterated, and the curvature of the contour curve is introduced in the iterative process, and different energies are weighted. The smoothness and curvature of the contour are further controlled to improve the fit between the contour curve and the liver boundary, and a smooth and continuous contour curve close to the liver boundary is obtained, so as to realize the accurate segmentation processing of the liver tissue image;

[0149] At the same time, this scheme proposes a method for identifying micro-fibrosis areas, which divides the liver tissue image into multiple pixel areas, and calculates the similarity and difference of the pixel areas based on the frequency of adjacent co-occurrence of pixel values ​​in the pixel areas. The lower the similarity, the more it indicates that there are structural changes in the liver tissue, resulting in uneven pixel values ​​in the pixel areas. The lower the difference, the lower the uniformity of the image texture. Then, the similarity and difference are used to extract multi-scale fusion fibrosis features from the pixel areas, identify the fibrosis features in the pixel areas, and identify and extract micro-liver fibrosis areas, so as to realize the extraction of micro-image areas where liver fibrosis lesions appear in the liver tissue images, quantitatively rate the degree of fibrosis in the micro-liver fibrosis areas based on the attention matrix and global pooling processing, and use the liver fibrosis assessment weights of different plane imaging results to perform a comprehensive assessment of liver fibrosis, so as to realize the early assessment of the degree of liver fibrosis based on micro-lesion areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0150] Figure 1 A schematic diagram of a flow chart of an intelligent analysis method for medical images of liver fibrosis provided in accordance with an embodiment of the present invention.

[0151] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0152] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0153] The embodiment of the present application provides a method for intelligent analysis of medical images of liver fibrosis. The execution subject of the method for intelligent analysis of medical images of liver fibrosis includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided in the embodiment of the present application. In other words, the method for intelligent analysis of medical images of liver fibrosis can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0154] Reference Figure 1 Embodiment 1 of the present invention is: a method for intelligent analysis of liver fibrosis medical images, comprising the following steps:

[0155] S1: Collect liver medical imaging data and perform preprocessing to obtain liver tissue images.

[0156] In step S1, the medical imaging data of the liver is collected and preprocessed, including:

[0157] Collect liver medical image data, which are magnetic resonance imaging results of the liver, including imaging results of the cross section, coronal plane and sagittal plane. The collected liver medical image data are expressed in the form of:

[0158] ;

[0159] ;

[0160] in:

[0161] Indicates liver medical imaging data;

[0162] The imaging results of the liver in the first to third planes are shown in sequence, wherein the first to third planes are the transverse plane, the coronal plane and the sagittal plane;

[0163] Indicates imaging results In the form of a pixel matrix, Indicates imaging results The pixel value of the pixel at row x and column y in , , X represents the number of pixel rows of the imaging result, and Y represents the number of pixel columns of the imaging result;

[0164] Liver medical imaging data Perform preprocessing, where the preprocessing process is:

[0165] S11: Calculate the pixel gradient matrix of any imaging result in the liver medical imaging data I, where the imaging result The pixel gradient matrix is :

[0166] ;

[0167] ;

[0168] ;

[0169] in:

[0170] Indicates imaging results The gradient value of the pixel at row x and column y in , Represents the horizontal gradient value, Indicates the vertical gradient value;

[0171] S12: Perform local maximum judgment on any gradient value in the pixel gradient matrix, and set the gradient value of the non-local maximum value to 0, where the gradient value The local maximum judgment formula is:

[0172] ;

[0173] in:

[0174] Indicates imaging results The neighborhood gradient value set of the pixel in the xth row and yth column;

[0175] In the embodiment of the present invention, if the imaging result The gradient direction of the pixel at row x and column y in Close to 0 degrees, include ;

[0176] If the gradient direction Close to 90 degrees, include ;

[0177] If the gradient direction Close to 45 degrees, include ;

[0178] If the gradient direction Close to 135 degrees, include ;

[0179] Indicates imaging results The gradient value of the pixel at row x and column y in The local maximum judgment result of 1 indicates the gradient value is a local maximum, If it is 0, it means the gradient value is not a local maximum, and let ;

[0180] S13: Mark pixels with gradient values ​​higher than a preset threshold as edge pixels, and use a clustering algorithm to cluster the edge pixels to obtain multiple edge pixel clusters in the imaging result; connect all edge pixels in the edge pixel cluster to obtain a set of contour curves corresponding to the edge pixel cluster, where the imaging result The contour curve corresponding to the kth edge pixel cluster in for:

[0181] ;

[0182] in:

[0183] Represents the contour curve The curve position coordinates corresponding to the nth edge pixel in , , Indicates imaging results The total number of edge pixel clusters;

[0184] S14: Calculate the contour curve The normalized arc length of each curve position coordinate in The normalized arc length is :

[0185] ;

[0186] in:

[0187] ;

[0188] S15: The contour curve is interpolated by using a B-spline interpolation method, and an interpolation curve parameter equation corresponding to the contour curve is generated based on the normalized arc length, where the contour curve The corresponding interpolation curve parameter equation ,in is the interpolation curve parameter equation The variables in , s corresponds to the normalized arc length, , represents the curve position coordinates at the normalized arc length s, , Respectively represent the horizontal coordinate value and the vertical coordinate value;

[0189] S16: Based on the smoothness of the contour curve and the edge energy, the curve position coordinates at different normalized arc length positions in the contour curve are updated, where the contour curve The curve position coordinates at the normalized arc length s The update formula is:

[0190] ;

[0191] ;

[0192] ;

[0193] ;

[0194] ;

[0195] in:

[0196] represents the learning rate, Set to 0.8;

[0197] The differential component representing the normalized arc length represents a tiny arc segment along the contour curve;

[0198] Represents the interpolation curve parameter equation The first derivative of Represents the interpolation curve parameter equation The second derivative of

[0199] Represents the energy function that combines the smoothness of the contour curve and the edge energy, Represents the elastic energy weight that controls the length and stretchability of the contour curve. represents the bending energy weight that controls the smoothness of the contour curve;

[0200] Represents the contour curve The degree of curvature, They are respectively the preset maximum bending degree and the minimum bending degree;

[0201] Indicates the curve position coordinates The gradient value of the pixel at ;

[0202] S17: Repeat step S16 until the position coordinates of all curves in the contour curve do not change;

[0203] S18: Perform filtering and denoising processing on the image within the contour curve, and use the filtering and denoising result as the liver tissue image.

[0204] The step S18 performs filtering and denoising on the image within the contour curve, and uses the filtering and denoising result as the liver tissue image, including:

[0205] S181: Acquire the image within the contour curve, where the contour curve The image inside is :

[0206] ;

[0207] in:

[0208] Representing images Middle The pixel value of the pixel in row b column, , , Representing images The number of pixel rows, Representing images The number of pixel columns;

[0209] S182: Filter the image within the contour curve, where the image The filtering result is :

[0210] ;

[0211] ;

[0212] in:

[0213] c represents the filter radius, set c to 1;

[0214] Represents the filter scale coefficient, Set to 2;

[0215] represents an exponential function with a natural constant as base;

[0216] Representing images Middle The filtered pixel value of the pixel in row b column;

[0217] S183: Using the filtered image as a liver tissue image, obtaining a set of liver tissue images with different imaging results in the liver medical imaging data, wherein the imaging results The liver tissue image set is :

[0218] ;

[0219] in:

[0220] Indicates imaging results The total number of edge pixel clusters is also the imaging result Total number of images of liver tissue.

[0221] S2: Based on the similarities and differences of pixel regions in the liver tissue image, the micro-liver fibrosis region is extracted from the liver tissue image to obtain the micro-liver fibrosis region.

[0222] In step S2, based on the similarities and differences of pixel regions in the liver tissue image, extracting micro-liver fibrosis regions from the liver tissue image includes:

[0223] Based on the similarity and difference of pixel areas in liver tissue images, the micro-liver fibrosis area is extracted from liver tissue images. The extraction process of micro-liver fibrosis area is as follows:

[0224] S21: Liver tissue image Divide into The size is pixels and non-overlapping pixel areas, Image showing liver tissue The total number of pixel areas divided;

[0225] S22: Calculate the similarity and difference of pixels in each pixel area, where the liver tissue image The mth pixel area The similarities and differences are , ;

[0226] S23: Extract multi-scale features from the pixel region to obtain multi-scale features of the pixel region, where the pixel region The multi-scale features are ;

[0227] ;

[0228] ;

[0229] in:

[0230] Represents pixel area Convolution features at the zth convolution scale, represents the convolution matrix at the zth convolution scale, * represents the convolution operator, represents the activation function;

[0231] Represents normalization processing, restoring the convolution features to Pixel size;

[0232] S24: Combining similarities and differences to decode and reconstruct multi-scale features to highlight fibrosis features, and obtain multi-scale fusion fibrosis features of pixel areas, where the pixel areas The multi-scale fusion fibrosis characteristics are :

[0233] ;

[0234] S25: Based on the multi-scale fusion fibrosis features of the pixel area, the micro-liver fibrosis area is identified in the pixel area, where the pixel area The identification formula is:

[0235] ;

[0236] in:

[0237] Represents pixel area If the recognition result of If it is 1, it means the pixel area The corresponding liver area has liver fibrosis, and the pixel area As micro-liver fibrosis areas;

[0238] Represents multi-scale fusion fibrosis features The standard deviation of

[0239] Image showing liver tissue The mean of all multi-scale fused fibrosis features;

[0240] Indicates the preset recognition threshold;

[0241] The micro-liver fibrosis area collection constitutes three imaging results, among which the imaging results The micro-fibrosis area collection is :

[0242] ;

[0243] in:

[0244] Indicates imaging results The hth micro-fibrosis area in Indicates imaging results The total number of micro-fibrotic areas.

[0245] In the step S22, the similarity and difference of pixels in each pixel area are calculated, wherein the liver tissue image The mth pixel area The similarities and differences are ,include:

[0246] S221: If the pixel values ​​are If two pixels are the most adjacent in the horizontal or vertical direction, it means that the two pixels appear adjacent to each other and the two pixels constitute the pixel value Corresponding pixel pairs;

[0247] S222: Calculate the frequency of adjacent pixels corresponding to any two pixel values ​​in the pixel area, where the pixel area Medium pixel value The probability of the corresponding pixel appearing nearby is :

[0248] ;

[0249] in:

[0250] Represents pixel area Medium pixel value The number of corresponding pixel pairs;

[0251] S223: Calculate pixel area Similarity of pixels in :

[0252] ;

[0253] S224: Calculate pixel area The difference in pixels:

[0254] ;

[0255] in:

[0256] Represents pixel area The difference in pixels.

[0257] S3: Quantitatively grade the degree of fibrosis in the micro-liver fibrosis areas to obtain the degree of fibrosis in different micro-liver fibrosis areas.

[0258] The S3 step quantitatively grades the degree of fibrosis in the micro-liver fibrosis area, including:

[0259] The degree of fibrosis in the micro-liver fibrosis area was quantitatively graded, and the quantitative grading process of the degree of fibrosis in the micro-liver fibrosis area was as follows:

[0260] S31: Obtain multi-scale fusion fibrosis features of micro-liver fibrosis areas, where micro-liver fibrosis areas The multi-scale fusion fibrosis characteristics are ;

[0261] S32: Perform perceptual attention processing on the multi-scale fused fibrosis features to obtain a perceptual attention matrix corresponding to the multi-scale fused fibrosis features, where the multi-scale fused fibrosis features The corresponding perceptual attention matrix is :

[0262] ;

[0263] in:

[0264] Represents the multi-scale fusion fibrosis features Any element q in The operation of OK Column perceptual attention matrix ;

[0265] S33: Use the perceptual attention matrix to mask and perform maximum pooling on the multi-scale fused fibrosis features to obtain a quantitative feature map, where the multi-scale fused fibrosis features The mask and maximum pooling processing formula are:

[0266] ;

[0267] in:

[0268] Represents multi-scale fusion fibrosis features The corresponding quantitative feature map;

[0269] Represents element-wise multiplication;

[0270] Represents maximum pooling processing;

[0271] S34: Quantitatively grade the degree of fibrosis on the quantitative feature map to obtain the degree of fibrosis in the micro-liver fibrosis area, wherein the quantitative feature map The quantitative rating formula for the degree of fibrosis is:

[0272] ;

[0273] in:

[0274] Micro-fibrosis area The degree of fibrosis;

[0275] Represents the quantitative feature map of early liver fibrosis obtained by contrastive learning.

[0276] S4: Comprehensive assessment of liver fibrosis is performed based on the fibrosis degree of the micro-liver fibrosis area to obtain the liver fibrosis degree assessment result.

[0277] In step S4, a comprehensive assessment of liver fibrosis is performed based on the degree of fibrosis in the micro-liver fibrosis area, including:

[0278] Comprehensive assessment of liver fibrosis is performed based on the degree of fibrosis in the micro-fibrosis area, and the comprehensive assessment process of liver fibrosis is as follows:

[0279] S41: Obtain the fibrosis grade of all micro-liver fibrosis areas;

[0280] S42: Calculate the liver fibrosis assessment weights of the imaging results in different planes, where the imaging results The liver fibrosis assessment weight is :

[0281] ;

[0282] in:

[0283] Indicates imaging results The fibrosis degree grade of all micro-liver fibrosis areas is the set;

[0284] , Represents the set of fibrosis levels Any degree of fibrosis in

[0285] Indicates the degree of fibrosis The imaging results The frequency of all micro-fibrosis areas in the liver;

[0286] S43: Based on the liver fibrosis assessment weight, the liver fibrosis degree assessment result is calculated:

[0287] ;

[0288] in:

[0289] Indicates the results of liver fibrosis assessment. The higher the level, the higher the degree of liver fibrosis.

[0290] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0291] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0292] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0293] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for intelligent analysis of liver fibrosis medical images, characterized in that: The method comprises: Collect liver medical imaging data from different planes, generate the initial contour of the liver area based on the gradient value of the pixel points, and construct the parameter equation of the contour curve. Iterate the contour curve based on the smoothness, curvature and gradient energy of the contour, introduce the curvature of the contour curve in the iterative process, and perform weighted calculation on different energies; The liver tissue image is divided into multiple pixel regions. The similarity and difference of the pixel regions are calculated based on the frequency of co-occurrence of adjacent pixel values ​​in the pixel regions. The similarity and difference are used to extract multi-scale fusion fibrosis features of the pixel regions, identify the fibrosis features in the pixel regions, and identify and extract micro-liver fibrosis regions. The micro-image region extraction of liver fibrosis lesions in the liver tissue image is achieved. The fibrosis degree of the micro-liver fibrosis region is quantitatively rated based on the attention matrix and global pooling processing, and the liver fibrosis assessment weights of different plane imaging results are used to perform a comprehensive assessment of liver fibrosis, and an early liver fibrosis degree assessment based on the tiny lesion area is obtained. The specific steps include: S1: Collect liver medical imaging data and perform preprocessing to obtain liver tissue images; S2: Based on the similarities and differences of pixel regions in the liver tissue image, the micro-liver fibrosis region is extracted from the liver tissue image to obtain the micro-liver fibrosis region; Liver tissue image The extraction process of micro-liver fibrosis area is as follows: S21: Liver tissue image Divide into The size is pixels and non-overlapping pixel areas, Image showing liver tissue The total number of pixel areas divided; S22: Calculate the similarity and difference of pixels in each pixel area, where the liver tissue image The mth pixel area The similarities and differences are , : S23: Extract multi-scale features from the pixel region to obtain multi-scale features of the pixel region, where the pixel region The multi-scale features are : in: Represents pixel area Convolution features at the zth convolution scale, represents the convolution matrix at the zth convolution scale, * represents the convolution operator, represents the activation function; Represents normalization processing, restoring the convolution features to Pixel size; S24: Combining similarities and differences to decode and reconstruct multi-scale features to highlight fibrosis features, and obtain multi-scale fusion fibrosis features of pixel areas, where the pixel areas The multi-scale fusion fibrosis characteristics are : S25: Based on the multi-scale fusion fibrosis features of the pixel area, the micro-liver fibrosis area is identified in the pixel area, where the pixel area The identification formula is: in: Represents pixel area If the recognition result of If it is 1, it means the pixel area The corresponding liver area has liver fibrosis, and the pixel area As micro-liver fibrosis areas; Represents multi-scale fusion fibrosis features The standard deviation of Image showing liver tissue The mean of all multi-scale fused fibrosis features; Indicates the preset recognition threshold; represents the L2 norm; The micro-liver fibrosis area collection constitutes three imaging results, among which the imaging results The micro-fibrosis area collection is : in: Indicates imaging results The hth micro-fibrosis area in Indicates imaging results the total number of areas of moderate to mild liver fibrosis; S3: Quantitatively grade the degree of fibrosis in the micro-liver fibrosis area to obtain the degree of fibrosis in different micro-liver fibrosis areas; S4: Comprehensive assessment of liver fibrosis is performed based on the fibrosis degree level of the micro-liver fibrosis area to obtain the liver fibrosis degree assessment result.

2. The method for intelligent analysis of liver fibrosis medical images according to claim 1, characterized in that: The S1 step includes: Collect liver medical image data, which are magnetic resonance imaging results of the liver, including imaging results of the cross section, coronal plane and sagittal plane. The collected liver medical image data are expressed in the form of: in: Indicates liver medical imaging data; The imaging results of the liver in the first to third planes are shown in sequence, wherein the first to third planes are the transverse plane, the coronal plane and the sagittal plane; Indicates imaging results In the form of a pixel matrix, Indicates imaging results The pixel value of the pixel at row x and column y in , , X represents the number of pixel rows of the imaging result, and Y represents the number of pixel columns of the imaging result; Liver medical imaging data Perform preprocessing, where the preprocessing process is: S11: Calculate and obtain the pixel gradient matrix of any imaging result in the liver medical imaging data I; S12: Perform local maximum judgment on any gradient value in the pixel gradient matrix, and set the gradient value of the non-local maximum value to 0; S13: Mark pixels with gradient values ​​higher than a preset threshold as edge pixels, and use a clustering algorithm to cluster the edge pixels to obtain multiple edge pixel clusters in the imaging result; connect all edge pixels in the edge pixel cluster to obtain a set of contour curves corresponding to the edge pixel cluster, where the imaging result The contour curve corresponding to the kth edge pixel cluster in for: in: Represents the contour curve The curve position coordinates corresponding to the nth edge pixel in , , Indicates imaging results The total number of edge pixel clusters; S14: Calculate the contour curve The normalized arc length of each curve position coordinate in The normalized arc length is : in: ； S15: The contour curve is interpolated by using a B-spline interpolation method, and an interpolation curve parameter equation corresponding to the contour curve is generated based on the normalized arc length, where the contour curve The corresponding interpolation curve parameter equation ,in is the interpolation curve parameter equation The variables in , s corresponds to the normalized arc length, , represents the curve position coordinates at the normalized arc length s, , Respectively represent the horizontal coordinate value and the vertical coordinate value; S16: based on the smoothness and edge energy of the contour curve, updating the curve position coordinates at different normalized arc length positions in the contour curve; S17: Repeat step S16 until the position coordinates of all curves in the contour curve do not change; S18: Perform filtering and denoising processing on the image within the contour curve, and use the filtering and denoising result as the liver tissue image.

3. The method for intelligent analysis of liver fibrosis medical images according to claim 2, characterized in that: The step S18 performs filtering and denoising on the image within the contour curve, and uses the filtering and denoising result as the liver tissue image, including: S181: Acquire the image within the contour curve, where the contour curve The image inside is : in: Representing images Middle The pixel value of the pixel in row b column, , , Representing images The number of pixel rows, Representing images The number of pixel columns; S182: Filter the image within the contour curve, where the image The filtering result is : in: c represents the filter radius, set c to 1; Represents the filter scale coefficient, Set to 2; represents an exponential function with a natural constant as base; Representing images Middle The filtered pixel value of the pixel in row b column; S183: Using the filtered image as a liver tissue image, obtaining a set of liver tissue images with different imaging results in the liver medical imaging data, wherein the imaging results The liver tissue image set is : in: Indicates imaging results The total number of edge pixel clusters is also the imaging result Total number of images of liver tissue.

4. The method for intelligent analysis of liver fibrosis medical images according to claim 3, characterized in that: In the step S22, the similarity and difference of pixels in each pixel area are calculated, wherein the liver tissue image The mth pixel area The similarities and differences are ,include: S221: If the pixel values ​​are If two pixels are the most adjacent in the horizontal or vertical direction, it means that the two pixels appear adjacent to each other and the two pixels constitute the pixel value Corresponding pixel pairs; S222: Calculate the frequency of adjacent pixels corresponding to any two pixel values ​​in the pixel area, where the pixel area Medium pixel value The probability of the corresponding pixel appearing nearby is : in: Represents pixel area Medium pixel value The number of corresponding pixel pairs; S223: Calculate pixel area Similarity of pixels in : S224: Calculate pixel area The difference in pixels: in: Represents pixel area The difference in pixels.

5. The method for intelligent analysis of liver fibrosis medical images according to claim 1, characterized in that: The S3 step includes: The degree of fibrosis in the micro-liver fibrosis area was quantitatively graded, and the quantitative grading process of the degree of fibrosis in the micro-liver fibrosis area was as follows: S31: Obtain multi-scale fusion fibrosis features of micro-liver fibrosis areas, where micro-liver fibrosis areas The multi-scale fusion fibrosis characteristics are ; S32: Perform perceptual attention processing on the multi-scale fused fibrosis features to obtain a perceptual attention matrix corresponding to the multi-scale fused fibrosis features, where the multi-scale fused fibrosis features The corresponding perceptual attention matrix is : in: Represents the multi-scale fusion fibrosis features Any element q in The operation of OK Column perceptual attention matrix ; S33: Use the perceptual attention matrix to mask and perform maximum pooling on the multi-scale fused fibrosis features to obtain a quantitative feature map, where the multi-scale fused fibrosis features The mask and maximum pooling processing formula are: in: Represents multi-scale fusion fibrosis features The corresponding quantitative feature map; Represents element-wise multiplication; Represents maximum pooling processing; S34: Quantitatively grade the degree of fibrosis on the quantitative feature map to obtain the degree of fibrosis in the micro-liver fibrosis area, wherein the quantitative feature map The quantitative rating formula for the degree of fibrosis is: in: Micro-fibrosis area The degree of fibrosis; Represents the quantitative feature map of early liver fibrosis obtained by contrastive learning.

6. The method for intelligent analysis of liver fibrosis medical images according to claim 5, characterized in that: The S4 step includes: Comprehensive assessment of liver fibrosis is performed based on the degree of fibrosis in the micro-fibrosis area, and the comprehensive assessment process of liver fibrosis is as follows: S41: Obtain the fibrosis grade of all micro-liver fibrosis areas; S42: Calculate the liver fibrosis assessment weights of the imaging results in different planes, where the imaging results The liver fibrosis assessment weight is : in: Indicates imaging results The fibrosis degree grade of all micro-liver fibrosis areas is the set; , Represents the set of fibrosis levels Any degree of fibrosis in Indicates the degree of fibrosis The imaging results The frequency of all micro-fibrosis areas in the liver; S43: Based on the liver fibrosis assessment weight, the liver fibrosis degree assessment result is calculated: in: Indicates the results of liver fibrosis assessment. The higher the level, the higher the degree of liver fibrosis.

7. The method for intelligent analysis of liver fibrosis medical images according to claim 2, characterized in that: In step S16, based on the smoothness of the contour curve and the edge energy, the curve position coordinates at different normalized arc length positions in the contour curve are updated, including: The contour curve The curve position coordinates at the normalized arc length s The update formula is: in: represents the learning rate, Set to 0.8; The differential component representing the normalized arc length represents a tiny arc segment along the contour curve; Represents the interpolation curve parameter equation The first derivative of Represents the interpolation curve parameter equation The second derivative of Represents the energy function combining the smoothness of the contour curve and the edge energy, Represents the elastic energy weight that controls the length and stretchability of the contour curve. represents the bending energy weight that controls the smoothness of the contour curve; Represents the contour curve The degree of curvature, They are respectively the preset maximum bending degree and the minimum bending degree; Indicates the curve position coordinates The gradient value of the pixel at .

Images