A system for identifying MASH and the degree of MASH inflammation

Abstract

This invention discloses a system for identifying MASH and the degree of MASH inflammation, belonging to the field of medical technology. It includes a fat content distribution matrix acquisition module, a liver elasticity distribution matrix acquisition module, a severe fatty lesion area acquisition module, a liver fibrosis area acquisition module, a severe fatty lesion surrounding area acquisition module, a liver fibrosis surrounding area acquisition module, a weighted IQ envelope signal acquisition module, and a MASH and MASH inflammation degree identification module. This invention can safely, quickly, and accurately identify MASH and the degree of MASH inflammation.

Description

Technical Field

[0001] This invention relates to the field of medical technology, and more specifically to a system for identifying MASH and the degree of MASH inflammation. Background Technology

[0002] Among the spectrum of liver diseases, metabolic dysfunction-associated fatty liver disease (MASLD) has become one of the most common chronic liver diseases worldwide. Its early characteristic is abnormal accumulation of fat in hepatocytes (fat content > 5%). At this stage, there is usually no significant inflammation or fibrosis, and it is a reversible condition that can be normalized (hepatocyte fat content < 5%) through lifestyle interventions. However, if left untreated, MASLD can gradually progress to metabolic dysfunction-associated steatohepatitis (MASH), accompanied by hepatocyte ballooning degeneration and lobular inflammation, eventually leading to liver fibrosis, cirrhosis, and even hepatocellular carcinoma. Therefore, accurate identification of MASH is crucial for halting disease progression and implementing early intervention.

[0003] Currently, liver biopsy remains the gold standard for clinical diagnosis of MASH, allowing direct observation of key pathological features such as steatosis, inflammatory activity, and ballooning degeneration. However, liver biopsy is an invasive procedure, carrying risks of sampling error, bleeding, and infection, making it unsuitable as an ideal method for repeated screening and dynamic monitoring. In recent years, non-invasive diagnostic technologies have become a research hotspot, with liver stiffness measurement being widely used to assess the degree of fibrosis. Traditionally, a liver stiffness value greater than 7.3 kPa is considered the threshold for cirrhosis, but recent research suggests that liver stiffness is already significantly elevated in the early stages of MASH. For example, a patent in the prior art entitled "Tissue Inflammation Activity Detection Device" reveals that liver stiffness in the range of 8.7 kPa to 9.7 kPa indicates that MASH is in a mild inflammatory stage, while stiffness exceeding 9.7 kPa indicates that MASH has progressed to severe, and if left uncontrolled, will rapidly evolve into irreversible cirrhosis. This finding corrects previous understanding of the patterns of liver stiffness changes, indicating that the stiffness threshold alone cannot fully reflect the degree of inflammation in MASH.

[0004] Therefore, how to provide a system for identifying MASH and the degree of MASH inflammation that can safely, quickly, and accurately identify MASH and the degree of MASH inflammation is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the object of the present invention is to provide a system for identifying MASH and the degree of MASH inflammation.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A system for identifying MASH and the degree of MASH inflammation includes a fat content distribution matrix acquisition module, a liver elasticity distribution matrix acquisition module, a severe fatty lesion area acquisition module, a liver fibrosis area acquisition module, a severe fatty lesion surrounding area acquisition module, a liver fibrosis surrounding area acquisition module, a weighted IQ envelope signal acquisition module, and a MASH and MASH inflammation degree identification module.

[0008] The fat content distribution matrix acquisition module is used to extract the fat content distribution matrix from a single-frame IQ envelope signal; wherein, the single-frame IQ envelope signal is the IQ envelope signal corresponding to a single-frame liver ultrasound signal;

[0009] The liver elasticity distribution matrix acquisition module is used to extract the liver elasticity distribution matrix corresponding to a single frame IQ envelope signal;

[0010] The severe lipopathy region acquisition module is used to cluster the fat content distribution matrix and then perform a preset threshold judgment to obtain the severe lipopathy region.

[0011] The liver fibrosis region acquisition module is used to cluster the liver elastic distribution matrix and then perform a preset threshold judgment to obtain the liver fibrosis region.

[0012] The module for obtaining the area surrounding the severe lipolesion is used to expand the area of ​​the severe lipolesion and then subtract the area of ​​the severe lipolesion to obtain the area surrounding the severe lipolesion.

[0013] The liver fibrosis surrounding area acquisition module is used to expand the liver fibrosis area and then subtract the liver fibrosis area to obtain the liver fibrosis surrounding area.

[0014] The weighted IQ envelope signal acquisition module is used to process multiple frames of IQ envelope signals based on the area surrounding the severe fatty lesion and the area surrounding the liver fibrosis to obtain a weighted IQ envelope signal.

[0015] The MASH and MASH inflammation degree recognition module is used to perform channel splicing, feature extraction and classification on the weighted IQ envelope signal and the multi-frame IQ envelope signal to obtain the MASH and MASH inflammation degree recognition results.

[0016] Preferably, the fat content distribution matrix acquisition module includes a Shannon entropy matrix acquisition submodule and an alignment submodule:

[0017] The Shannon entropy matrix acquisition submodule is used to traverse the single-frame IQ envelope signal using a sliding window and calculate the Shannon entropy value of each sliding window to obtain the Shannon entropy matrix of the single-frame IQ envelope signal.

[0018] The alignment submodule is used to sample the Shannon entropy matrix to the same size as the single-frame IQ envelope signal to obtain the fat content distribution matrix.

[0019] Preferably, the Shannon entropy value of each sliding window is obtained based on the following formula:

[0020]

[0021] In the formula, w(x) i ) represents the probability of the i-th gray level appearing in the gray-level histogram of the IQ envelope signal within the sliding window; n represents the total number of gray levels into which the amplitude range of the IQ envelope signal within the sliding window is divided; H c This represents the Shannon entropy value of the sliding window.

[0022] Preferably, the sliding window is square, and the side length of the square is one pulse length of the ultrasonic transducer.

[0023] Preferably, the severe fatty lesion area acquisition module includes a first clustering submodule and a first threshold judgment submodule;

[0024] The first clustering submodule is used to perform K-means clustering on the fat content distribution matrix to obtain healthy liver regions, mild to moderate fatty liver regions, and severe fatty liver regions;

[0025] The first threshold determination submodule is used to obtain the severe fatty lesion region based on a first preset threshold and the cluster center values ​​of the healthy liver region, the mild to moderate fatty liver region, and the severe fatty liver region; wherein, if the cluster center values ​​of the healthy liver region, the mild to moderate fatty liver region, and the severe fatty liver region are all less than the first preset threshold, then the severe fatty lesion region is set to a 0 matrix; if the cluster center values ​​of the healthy liver region, the mild to moderate fatty liver region, and the severe fatty liver region are not all less than the first preset threshold, then the severe fatty lesion region is the region where all cluster center values ​​are greater than or equal to the first preset threshold.

[0026] Preferably, the liver fibrosis region acquisition module includes a second clustering submodule and a second threshold judgment submodule;

[0027] The second clustering submodule is used to perform K-means clustering on the liver elasticity distribution matrix to obtain healthy liver stiffness regions, mild to moderate liver fibrosis regions, and severe liver fibrosis regions.

[0028] The second threshold determination submodule is used to obtain the liver fibrosis region based on a second preset threshold and the cluster center values ​​of the healthy liver stiffness region, the mild to moderate liver fibrosis region, and the severe liver fibrosis region. If the cluster center values ​​of the healthy liver stiffness region, the mild to moderate liver fibrosis region, and the severe liver fibrosis region are all less than the second preset threshold, then the liver fibrosis region is set to a 0 matrix. If the cluster center values ​​of the healthy liver stiffness region, the mild to moderate liver fibrosis region, and the severe liver fibrosis region are not all less than the second preset threshold, then the liver fibrosis region is the region where all cluster center values ​​are greater than or equal to the second preset threshold.

[0029] Preferably, the weighted IQ envelope signal acquisition module includes a 3D MASK region acquisition submodule, a MASK region merging submodule, and a weighted dot product submodule;

[0030] The 3D MASK region acquisition submodule is used to perform binarization processing on the 3D region surrounding severe fatty lesions and the 3D region surrounding liver fibrosis to obtain two 3D MASK regions; wherein, multiple frames of the region surrounding severe fatty lesions constitute the 3D region surrounding severe fatty lesions; and multiple frames of the region surrounding liver fibrosis constitute the 3D region surrounding liver fibrosis.

[0031] The MASK region merging submodule is used to calculate the union of the regions surrounding severe fatty lesions and liver fibrosis in each MASK frame, and then merge the two 3D MASK regions into one, obtaining a merged MASK region. If the number of pixels with a value of 1 in the merged MASK region is greater than a third preset threshold, the merged MASK region is input to the weighted dot multiplication submodule; if it is less than the third preset threshold, the IQ envelope signals of multiple frames are directly input into the convolutional neural network and the fully connected layer in sequence to obtain the MASH and MASH inflammation degree recognition results.

[0032] The weighted dot multiplier submodule is used to obtain a weighted IQ envelope signal based on the merged MASK region, preset weight coefficients, and the multi-frame IQ envelope signal; wherein, the calculation formula for the weighted IQ envelope signal is: M represents the weighted IQ envelope signal. IQ The multi-frame IQ envelope signal represents the signal, α represents the preset weighting coefficient, and M represents the multi-frame IQ envelope signal. ROI The '·' indicates the merged MASK region, and '·' indicates matrix dot product.

[0033] Preferably, the system also includes a 3D acquisition submodule for the area surrounding severe steatosis and a 3D acquisition submodule for the area surrounding liver fibrosis;

[0034] The 3D severe lipopathy surrounding area acquisition submodule is used to process multiple frames of IQ envelope signals sequentially through the fat content distribution matrix acquisition module, the severe lipopathy area acquisition module, and the severe lipopathy surrounding area acquisition module to obtain multiple frames of the severe lipopathy surrounding area.

[0035] The 3D liver fibrosis surrounding area acquisition submodule is used to extract the multi-frame liver elastic distribution matrix corresponding to the multi-frame IQ envelope signal sequentially using the liver elastic distribution matrix acquisition module, and process the multi-frame liver elastic distribution matrix sequentially through the liver fibrosis area acquisition module and the liver fibrosis surrounding area acquisition module to obtain the multi-frame liver fibrosis surrounding area.

[0036] Preferably, the system also includes a 3D acquisition submodule for the area surrounding severe steatosis and a 3D acquisition submodule for the area surrounding liver fibrosis;

[0037] The 3D severe lipopathy surrounding area acquisition submodule is used to process multiple frames of IQ envelope signals sequentially through the fat content distribution matrix acquisition module, the severe lipopathy area acquisition module, and the severe lipopathy surrounding area acquisition module to obtain multiple frames of the severe lipopathy surrounding area.

[0038] The 3D liver fibrosis surrounding region acquisition submodule is used to extract the selected liver elastic distribution matrix corresponding to the selected frame IQ envelope signal using the liver elastic distribution matrix acquisition module. Then, the number of channels of the selected liver elastic distribution matrix is ​​expanded to the same number of frames as the multiple frames of IQ envelope signals. After that, it is processed by the liver fibrosis region acquisition module and the liver fibrosis surrounding region acquisition module respectively to obtain multiple frames of the liver fibrosis surrounding region. The selected frame IQ envelope signal is one frame of IQ envelope signal in the multiple frames of IQ envelope signals.

[0039] Preferably, the MASH and the MASH inflammation degree recognition module include a channel splicing layer, a convolutional neural network, and a fully connected layer;

[0040] The channel splicing layer is used to perform channel splicing on the weighted IQ envelope signal and the multi-frame IQ envelope signal to obtain a spliced ​​IQ envelope signal.

[0041] The convolutional neural network is used to extract features from the spliced ​​IQ envelope signal to obtain CNN features;

[0042] The fully connected layer is used to classify the CNN features to obtain the MASH and MASH inflammation degree recognition results; wherein, the MASH and MASH inflammation degree recognition results include no inflammation, mild inflammation, and severe inflammation.

[0043] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a system for identifying MASH and the degree of MASH inflammation, which can safely, quickly and accurately identify MASH and the degree of MASH inflammation. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0045] Figure 1 This invention provides a schematic diagram of a system structure for identifying MASH and the degree of MASH inflammation. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] like Figure 1 As shown, this embodiment of the invention discloses a system for identifying MASH and the degree of MASH inflammation, including a fat content distribution matrix acquisition module, a liver elasticity distribution matrix acquisition module, a severe fatty lesion area acquisition module, a liver fibrosis area acquisition module, a severe fatty lesion surrounding area acquisition module, a liver fibrosis surrounding area acquisition module, a weighted IQ envelope signal acquisition module, and a MASH and MASH inflammation degree identification module;

[0048] The fat content distribution matrix acquisition module is used to extract the fat content distribution matrix from a single-frame IQ envelope signal; wherein, the single-frame IQ envelope signal is the IQ envelope signal corresponding to a single-frame liver ultrasound signal;

[0049] In one embodiment, the fat content distribution matrix acquisition module includes a Shannon entropy matrix acquisition submodule and an alignment submodule:

[0050] The Shannon entropy matrix acquisition submodule is used to traverse the single-frame IQ envelope signal using a sliding window and calculate the Shannon entropy value of each sliding window to obtain the Shannon entropy matrix of the single-frame IQ envelope signal.

[0051] In one embodiment, the Shannon entropy value of each sliding window is obtained based on the following formula:

[0052]

[0053] In the formula, w(x) i ) represents the probability of the i-th gray level appearing in the gray-level histogram of the IQ envelope signal within the sliding window; n represents the total number of gray levels into which the amplitude range of the IQ envelope signal within the sliding window is divided; H c This represents the Shannon entropy value of the sliding window.

[0054] In one embodiment, the sliding window is a square, and the side length of the square is one pulse length of the ultrasonic transducer.

[0055] It is understandable that during the process of traversing the sliding window from left to right and from top to bottom, each window will obtain a Shannon entropy value. After the sliding window traversal is completed, a Shannon entropy matrix will be obtained. The step size of the traversal can be set to half the size of the sliding window, which can ensure that the generated Shannon entropy matrix has sufficient resolution while the amount of computation is not too large.

[0056] The alignment submodule is used to sample the Shannon entropy matrix to the same size as the single-frame IQ envelope signal to obtain the fat content distribution matrix.

[0057] It is understandable that sampling is for aligning the Shannon entropy matrix with the single-frame IQ envelope data.

[0058] The liver elasticity distribution matrix acquisition module is used to extract the liver elasticity distribution matrix corresponding to a single frame IQ envelope signal;

[0059] It is understandable that because ARFI can obtain the stiffness distribution of a specified area and can qualitatively detect the severity of liver fibrosis, the liver elasticity distribution matrix corresponding to a single frame IQ envelope signal can be extracted using Acoustic Radiation Force Impulse (ARFI).

[0060] The severe lipopathy region acquisition module is used to cluster the fat content distribution matrix and then perform a preset threshold judgment to obtain the severe lipopathy region.

[0061] In one embodiment, the severe fat lesion area acquisition module includes a first clustering submodule and a first threshold judgment submodule;

[0062] The first clustering submodule is used to perform K-means clustering on the fat content distribution matrix to obtain healthy liver regions, mild to moderate fatty liver regions, and severe fatty liver regions;

[0063] The specific steps of K-means clustering are as follows:

[0064] (1) Set the number of clusters K for clustering to 3;

[0065] It is understandable that during the progression of fatty liver, the distribution of fat content is not uniform. Some liver tissue is healthy, while some areas have varying degrees of fatty lesions. Setting the cluster number to 3 indicates that the liver is divided into healthy liver areas, mild to moderate fatty liver areas, and severe fatty liver areas.

[0066] (2) Randomly select elements in the fat content distribution matrix as the initial values ​​of the cluster center values ​​of each cluster;

[0067] (3) Calculate the distance from each element in the fat content distribution matrix to the center value of each cluster (this distance can be Euclidean distance), and then assign each element in the fat content distribution matrix to the cluster with the closest distance to it;

[0068] (4) Update the cluster center value of each cluster; where the updated cluster center value is the average of all elements included in the cluster before the update;

[0069] (5) Repeat steps (3) and (4) until the difference between the cluster center values ​​of the two updates is less than the preset threshold or the preset maximum number of iterations Tmax is reached (e.g., Tmax = 100 times), to obtain the healthy liver region, the mild to moderate fatty liver region and the severe fatty liver region.

[0070] The first threshold determination submodule is used to obtain the severe fatty lesion region based on a first preset threshold and the cluster center values ​​of the healthy liver region, the mild to moderate fatty liver region, and the severe fatty liver region; wherein, if the cluster center values ​​of the healthy liver region, the mild to moderate fatty liver region, and the severe fatty liver region are all less than the first preset threshold, then the severe fatty lesion region is set to a 0 matrix; if the cluster center values ​​of the healthy liver region, the mild to moderate fatty liver region, and the severe fatty liver region are not all less than the first preset threshold, then the severe fatty lesion region is the region where all cluster center values ​​are greater than or equal to the first preset threshold.

[0071] In one embodiment, the first preset threshold is 4.1;

[0072] In one embodiment, if the fat content distribution matrix is ​​constructed by the shape parameter m of the Nakagami distribution, then the first preset threshold is 1;

[0073] It is understandable that for any fat content distribution matrix, K-means clustering will forcibly cluster the matrix into 3 categories (i.e., healthy liver region, mild to moderate fatty liver region, and severe fatty liver region). However, when the cluster center value is higher than the first preset threshold, the region corresponding to the cluster is already considered to belong to the severe fatty lesion region. Therefore, threshold judgment is required for the three regions after clustering in order to obtain a more accurate severe fatty lesion region.

[0074] The liver fibrosis region acquisition module is used to cluster the liver elastic distribution matrix and then perform a preset threshold judgment to obtain the liver fibrosis region.

[0075] In one embodiment, the liver fibrosis region acquisition module includes a second clustering submodule and a second threshold judgment submodule;

[0076] The second clustering submodule is used to perform K-means clustering on the liver elasticity distribution matrix to obtain healthy liver stiffness regions, mild to moderate liver fibrosis regions, and severe liver fibrosis regions.

[0077] The specific steps of K-means clustering are as follows:

[0078] (1) Set the number of clusters K for clustering to 3;

[0079] It is understandable that setting the number of clusters to 3 indicates that the liver is divided into areas of healthy liver stiffness, areas of mild to moderate liver fibrosis, and areas of severe liver fibrosis.

[0080] (2) Randomly select elements in the liver elastic distribution matrix as the initial values ​​of the cluster center values ​​of each cluster;

[0081] (3) Calculate the distance from each element in the liver elasticity distribution matrix to the center value of each cluster (this distance can be Euclidean distance), and then assign each element in the liver elasticity distribution matrix to the cluster with the nearest distance to it.

[0082] (4) Update the cluster center value of each cluster; where the updated cluster center value is the average of all elements included in the cluster before the update;

[0083] (5) Repeat steps (3) and (4) until the difference between the cluster center values ​​of the two updates is less than the preset threshold or the preset maximum number of iterations Tmax is reached, to obtain the healthy liver stiffness region, the mild to moderate liver fibrosis region and the severe liver fibrosis region.

[0084] The second threshold determination submodule is used to obtain the liver fibrosis region based on a second preset threshold and the cluster center values ​​of the healthy liver stiffness region, the mild to moderate liver fibrosis region, and the severe liver fibrosis region. If the cluster center values ​​of the healthy liver stiffness region, the mild to moderate liver fibrosis region, and the severe liver fibrosis region are all less than the second preset threshold, then the liver fibrosis region is set to a 0 matrix. If the cluster center values ​​of the healthy liver stiffness region, the mild to moderate liver fibrosis region, and the severe liver fibrosis region are not all less than the second preset threshold, then the liver fibrosis region is the region where all cluster center values ​​are greater than or equal to the second preset threshold.

[0085] In one embodiment, the second preset threshold is 7.3.

[0086] It is understandable that for any liver elasticity distribution matrix, K-means clustering will forcibly cluster the matrix into 3 categories (i.e., healthy liver stiffness region, mild to moderate liver fibrosis region, and severe liver fibrosis region). However, if the cluster center value is higher than the second preset threshold, the region corresponding to the cluster is already considered to belong to the liver fibrosis region. Therefore, threshold judgment is required for the three regions after clustering in order to obtain a more accurate liver fibrosis region.

[0087] The module for obtaining the area surrounding the severe lipolesion is used to expand the area of ​​the severe lipolesion and then subtract the area of ​​the severe lipolesion to obtain the area surrounding the severe lipolesion.

[0088] It is understandable that the expansion kernel can be set as needed, for example, it can be set to 5×5 and the expansion number can be set to 3 times.

[0089] The liver fibrosis surrounding area acquisition module is used to expand the liver fibrosis area and then subtract the liver fibrosis area to obtain the liver fibrosis surrounding area.

[0090] It is understandable that the expansion kernel can be set as needed, for example, it can be set to 5×5 and the expansion number can be set to 3 times.

[0091] The weighted IQ envelope signal acquisition module is used to process multiple frames of IQ envelope signals based on the area surrounding the severe fatty lesion and the area surrounding the liver fibrosis to obtain a weighted IQ envelope signal.

[0092] In one embodiment, the weighted IQ envelope signal acquisition module includes a 3D MASK region acquisition submodule, a MASK region merging submodule, and a weighted dot product submodule;

[0093] The 3D MASK region acquisition submodule is used to perform binarization processing on the 3D region surrounding severe fatty lesions and the 3D region surrounding liver fibrosis to obtain two 3D MASK regions (with pixels of 0 or 1); wherein, multiple frames of the region surrounding severe fatty lesions constitute the 3D region surrounding severe fatty lesions; and multiple frames of the region surrounding liver fibrosis constitute the 3D region surrounding liver fibrosis.

[0094] The MASK region merging submodule is used to calculate the union of the regions surrounding severe fatty lesions and liver fibrosis in each MASK frame, and then merge the two 3D MASK regions into one, obtaining a merged MASK region. If the number of pixels with a value of 1 in the merged MASK region is greater than a third preset threshold, the merged MASK region is input to the weighted dot multiplication submodule; if it is less than the third preset threshold, the IQ envelope signals of multiple frames are directly input into the convolutional neural network and the fully connected layer in sequence to obtain the MASH and MASH inflammation degree recognition results.

[0095] The weighted dot multiplier module is used to obtain a weighted IQ envelope signal based on the merged MASK region, preset weight coefficients, and the multi-frame IQ envelope signal; wherein, the calculation formula for the weighted IQ envelope signal is: M represents the weighted IQ envelope signal. IQ The multi-frame IQ envelope signal represents the signal, α represents the preset weighting coefficient, and M represents the multi-frame IQ envelope signal. ROI The '·' indicates the merged MASK region, and '·' indicates matrix dot product.

[0096] In one embodiment, the system further includes a 3D acquisition submodule for the area surrounding severe fatty lesions and a 3D acquisition submodule for the area surrounding liver fibrosis.

[0097] The 3D severe lipopathy surrounding area acquisition submodule is used to process multiple frames of IQ envelope signals sequentially through the fat content distribution matrix acquisition module, the severe lipopathy area acquisition module, and the severe lipopathy surrounding area acquisition module to obtain multiple frames of the severe lipopathy surrounding area.

[0098] The 3D liver fibrosis surrounding area acquisition submodule is used to extract the multi-frame liver elastic distribution matrix corresponding to the multi-frame IQ envelope signal sequentially using the liver elastic distribution matrix acquisition module, and process the multi-frame liver elastic distribution matrix sequentially through the liver fibrosis area acquisition module and the liver fibrosis surrounding area acquisition module to obtain the multi-frame liver fibrosis surrounding area.

[0099] In one embodiment, the system further includes a 3D acquisition submodule for the area surrounding severe fatty lesions and a 3D acquisition submodule for the area surrounding liver fibrosis.

[0100] The 3D severe lipopathy surrounding area acquisition submodule is used to process multiple frames of IQ envelope signals sequentially through the fat content distribution matrix acquisition module, the severe lipopathy area acquisition module, and the severe lipopathy surrounding area acquisition module to obtain multiple frames of the severe lipopathy surrounding area.

[0101] The 3D liver fibrosis surrounding region acquisition submodule is used to extract the selected liver elastic distribution matrix corresponding to the selected frame IQ envelope signal using the liver elastic distribution matrix acquisition module. Then, the number of channels of the selected liver elastic distribution matrix is ​​expanded to the same number of frames as the multiple frames of IQ envelope signals. After that, it is processed by the liver fibrosis region acquisition module and the liver fibrosis surrounding region acquisition module in sequence to obtain multiple frames of the liver fibrosis surrounding region (this method can improve the final classification speed); wherein, the selected frame IQ envelope signal is one frame of IQ envelope signal in the multiple frames of IQ envelope signals.

[0102] The MASH and MASH inflammation degree recognition module is used to perform channel splicing, feature extraction and classification on the weighted IQ envelope signal and the multi-frame IQ envelope signal to obtain the MASH and MASH inflammation degree recognition results.

[0103] In one embodiment, the MASH and the MASH inflammation degree recognition module include a channel splicing layer, a convolutional neural network, and a fully connected layer;

[0104] The channel splicing layer is used to perform channel splicing on the weighted IQ envelope signal and the multi-frame IQ envelope signal to obtain a spliced ​​IQ envelope signal.

[0105] The convolutional neural network is used to extract features from the spliced ​​IQ envelope signal to obtain CNN features;

[0106] The fully connected layer is used to classify the CNN features to obtain the MASH and MASH inflammation degree recognition results; wherein, the MASH and MASH inflammation degree recognition results include no inflammation, mild inflammation, and severe inflammation.

[0107] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0108] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A system for identifying MASH and the degree of MASH inflammation, characterized in that, It includes modules for obtaining the fat content distribution matrix, the liver elasticity distribution matrix, the area of ​​severe steatosis, the area of ​​liver fibrosis, the area surrounding severe steatosis, the area surrounding liver fibrosis, the weighted IQ envelope signal, and MASH and MASH inflammation degree recognition. The fat content distribution matrix acquisition module is used to extract the fat content distribution matrix from a single-frame IQ envelope signal; wherein, the single-frame IQ envelope signal is the IQ envelope signal corresponding to a single-frame liver ultrasound signal; The fat content distribution matrix acquisition module includes a Shannon entropy matrix acquisition submodule and an alignment submodule: The Shannon entropy matrix acquisition submodule is used to traverse the single-frame IQ envelope signal using a sliding window and calculate the Shannon entropy value of each sliding window to obtain the Shannon entropy matrix of the single-frame IQ envelope signal. The alignment submodule is used to sample the Shannon entropy matrix to the same size as the single-frame IQ envelope signal to obtain the fat content distribution matrix; The liver elasticity distribution matrix acquisition module is used to extract the liver elasticity distribution matrix corresponding to a single frame IQ envelope signal; The severe lipopathy region acquisition module is used to cluster the fat content distribution matrix and then perform a preset threshold judgment to obtain the severe lipopathy region. The severe fat lesion area acquisition module includes a first clustering submodule and a first threshold judgment submodule; The first clustering submodule is used to perform K-means clustering on the fat content distribution matrix to obtain healthy liver regions, mild to moderate fatty liver regions, and severe fatty liver regions; The liver fibrosis region acquisition module is used to cluster the liver elastic distribution matrix and then perform a preset threshold judgment to obtain the liver fibrosis region. The liver fibrosis region acquisition module includes a second clustering submodule and a second threshold judgment submodule; The second clustering submodule is used to perform K-means clustering on the liver elasticity distribution matrix to obtain healthy liver stiffness regions, mild to moderate liver fibrosis regions, and severe liver fibrosis regions. The module for obtaining the area surrounding the severe lipolesion is used to expand the area of ​​the severe lipolesion and then subtract the area of ​​the severe lipolesion to obtain the area surrounding the severe lipolesion. The liver fibrosis surrounding area acquisition module is used to expand the liver fibrosis area and then subtract the liver fibrosis area to obtain the liver fibrosis surrounding area. The weighted IQ envelope signal acquisition module is used to process multiple frames of IQ envelope signals based on the area surrounding the severe fatty lesion and the area surrounding the liver fibrosis to obtain a weighted IQ envelope signal. The MASH and MASH inflammation degree recognition module is used to perform channel splicing, feature extraction and classification on the weighted IQ envelope signal and the multi-frame IQ envelope signal to obtain the MASH and MASH inflammation degree recognition results.

2. The system for identifying MASH and the degree of MASH inflammation according to claim 1, characterized in that, The Shannon entropy value for each sliding window is obtained based on the following formula: In the formula, This represents the probability of the i-th gray level appearing in the gray-level histogram of the IQ envelope signal within the sliding window; n represents the total number of gray levels into which the amplitude range of the IQ envelope signal within the sliding window is divided. This represents the Shannon entropy value of the sliding window.

3. The system for identifying MASH and the degree of MASH inflammation according to claim 2, characterized in that, The sliding window is square, and the side length of the square is one pulse length of the ultrasonic transducer.

4. The system for identifying MASH and the degree of MASH inflammation according to claim 1, characterized in that, The first threshold determination submodule is used to obtain the severe fatty lesion region based on a first preset threshold and the cluster center values ​​of the healthy liver region, the mild to moderate fatty liver region, and the severe fatty liver region; wherein, if the cluster center values ​​of the healthy liver region, the mild to moderate fatty liver region, and the severe fatty liver region are all less than the first preset threshold, then the severe fatty lesion region is set to a 0 matrix; if the cluster center values ​​of the healthy liver region, the mild to moderate fatty liver region, and the severe fatty liver region are not all less than the first preset threshold, then the severe fatty lesion region is the region where all cluster center values ​​are greater than or equal to the first preset threshold.

5. The system for identifying MASH and the degree of MASH inflammation according to claim 1, characterized in that, The second threshold determination submodule is used to obtain the liver fibrosis region based on a second preset threshold and the cluster center values ​​of the healthy liver stiffness region, the mild to moderate liver fibrosis region, and the severe liver fibrosis region. If the cluster center values ​​of the healthy liver stiffness region, the mild to moderate liver fibrosis region, and the severe liver fibrosis region are all less than the second preset threshold, then the liver fibrosis region is set to a 0 matrix. If the cluster center values ​​of the healthy liver stiffness region, the mild to moderate liver fibrosis region, and the severe liver fibrosis region are not all less than the second preset threshold, then the liver fibrosis region is the region where all cluster center values ​​are greater than or equal to the second preset threshold.

6. The system for identifying MASH and the degree of MASH inflammation according to claim 1, characterized in that, The weighted IQ envelope signal acquisition module includes a 3D MASK region acquisition submodule, a MASK region merging submodule, and a weighted dot product submodule; The 3D MASK region acquisition submodule is used to perform binarization processing on the 3D region surrounding severe fatty lesions and the 3D region surrounding liver fibrosis to obtain two 3D MASK regions; wherein, multiple frames of the region surrounding severe fatty lesions constitute the 3D region surrounding severe fatty lesions; and multiple frames of the region surrounding liver fibrosis constitute the 3D region surrounding liver fibrosis. The MASK region merging submodule is used to calculate the union of the regions surrounding severe fatty lesions and liver fibrosis in each MASK frame, and then merge the two 3D MASK regions into one, obtaining a merged MASK region. If the number of pixels with a value of 1 in the merged MASK region is greater than a third preset threshold, the merged MASK region is input to the weighted dot multiplication submodule; if it is less than the third preset threshold, the IQ envelope signal regions of multiple frames are directly input into the convolutional neural network and the fully connected layer in sequence to obtain the MASH and MASH inflammation degree recognition results. The weighted dot multiplier module is used to obtain a weighted IQ envelope signal based on the merged MASK region, preset weight coefficients, and the multi-frame IQ envelope signal; wherein, the calculation formula for the weighted IQ envelope signal is: ; This represents the weighted IQ envelope signal. This represents the multi-frame IQ envelope signal. This represents the preset weighting coefficient. This refers to the merged mask region. This represents the matrix dot product.

7. A system for identifying MASH and the degree of MASH inflammation according to claim 6, characterized in that, It also includes a 3D submodule for acquiring the area surrounding severe steatosis and a 3D submodule for acquiring the area surrounding liver fibrosis. The 3D severe lipopathy surrounding area acquisition submodule is used to process multiple frames of IQ envelope signals sequentially through the fat content distribution matrix acquisition module, the severe lipopathy area acquisition module, and the severe lipopathy surrounding area acquisition module to obtain multiple frames of the severe lipopathy surrounding area. The 3D liver fibrosis surrounding area acquisition submodule is used to extract the multi-frame liver elastic distribution matrix corresponding to the multi-frame IQ envelope signal sequentially using the liver elastic distribution matrix acquisition module, and process the multi-frame liver elastic distribution matrix sequentially through the liver fibrosis area acquisition module and the liver fibrosis surrounding area acquisition module to obtain the multi-frame liver fibrosis surrounding area.

8. A system for identifying MASH and the degree of MASH inflammation according to claim 6, characterized in that, It also includes a 3D submodule for acquiring the area surrounding severe steatosis and a 3D submodule for acquiring the area surrounding liver fibrosis. The 3D severe lipopathy surrounding area acquisition submodule is used to process multiple frames of IQ envelope signals sequentially through the fat content distribution matrix acquisition module, the severe lipopathy area acquisition module, and the severe lipopathy surrounding area acquisition module to obtain multiple frames of the severe lipopathy surrounding area. The 3D liver fibrosis surrounding region acquisition submodule is used to extract the selected liver elastic distribution matrix corresponding to the selected frame IQ envelope signal using the liver elastic distribution matrix acquisition module. Then, the number of channels of the selected liver elastic distribution matrix is ​​expanded to the same number of frames as the multiple frames of IQ envelope signals. After that, it is processed by the liver fibrosis region acquisition module and the liver fibrosis surrounding region acquisition module respectively to obtain multiple frames of the liver fibrosis surrounding region. The selected frame IQ envelope signal is one frame of IQ envelope signal in the multiple frames of IQ envelope signals.

9. A system for identifying MASH and the degree of MASH inflammation according to claim 1, characterized in that, The MASH and MASH inflammation degree recognition module include a channel splicing layer, a convolutional neural network, and a fully connected layer; The channel splicing layer is used to perform channel splicing on the weighted IQ envelope signal and the multi-frame IQ envelope signal to obtain a spliced ​​IQ envelope signal. The convolutional neural network is used to extract features from the spliced ​​IQ envelope signal to obtain CNN features; The fully connected layer is used to classify the CNN features to obtain the MASH and MASH inflammation degree recognition results; wherein, the MASH and MASH inflammation degree recognition results include no inflammation, mild inflammation, and severe inflammation.

Images