A method, system and device for automatic fat quantification based on vertebral segmentation

Abstract

The present application relates to the field of intelligent medical treatment, and specifically to a method, system and device for automatic quantification of fat based on vertebral segmentation. The method comprises obtaining a medical image of a subject; segmenting the vertebrae in the medical image to obtain vertebrae; obtaining body segments based on the structural positioning of the vertebrae; and calculating the fat area and / or volume of the corresponding layer of the body segment. The present invention can be used to automatically segment and measure the volume of various types of fat throughout the body and determine the fat area at each vertebral layer, and has great clinical value.

Description

Technical Field

[0001] The present application relates to the field of intelligent medical care, and specifically to an automatic fat quantification method, system, device and readable storage medium based on vertebral segmentation. Background Art

[0002] Changes in fat content and distribution are related to the occurrence and development of various diseases. Previous studies have confirmed that visceral fat in the abdominal and pelvic regions is a risk factor for various cardiovascular and metabolic diseases, increased pericardial fat is related to the presence and severity of coronary artery disease, and bone marrow fat fraction is helpful in the diagnosis of bone marrow diseases.

[0003] Clinically, non-invasive assessment of fat content mainly uses CT and MRI. The magnetic resonance Dixon technique uses the chemical shift difference between water and fat to obtain in-phase, anti-phase and water-fat separation images, which has the advantage of no ionizing radiation compared to CT. The traditional method of measuring the fat content of various parts such as subcutaneous fat and abdominal and pelvic visceral fat from magnetic resonance Dixon images is for doctors to manually mark the fat areas of each layer, multiply the area of ​​each layer of fat by the layer thickness and add them up layer by layer to obtain the fat volume. This process is time-consuming and labor-intensive, and difficult to apply in clinical practice. Therefore, in actual work, the fat area of ​​a specific vertebral level is often measured. For example, the visceral fat area at the L3 level is often used to reflect the visceral fat content of the abdomen and pelvis.

[0004] In recent years, deep learning has been widely used in the field of medical image segmentation. Some methods have been tried on the problem of fat segmentation and quantification in MRI Dixon images. However, the existing methods have some defects: (1) Existing methods mainly focus on the segmentation of subcutaneous fat and visceral fat in the abdomen, and cannot simultaneously segment and measure mediastinal fat, supraclavicular fat, and bone marrow; (2) Existing methods cannot locate the anatomical position of this layer and cannot provide information on the fat area of ​​specific vertebral layers commonly used in clinical practice; (3) Since the fat area annotation process often relies on a threshold set according to the image grayscale, deep learning models that directly predict the fat area in the image may deviate when migrating to out-of-domain data from different devices or scanning parameters. Summary of the invention

[0005] In view of the above problems, the present invention proposes an automatic fat quantification method based on vertebral segmentation, comprising:

[0006] Obtain medical images of the subjects;

[0007] Segmenting the vertebrae in the medical image to obtain vertebrae;

[0008] Obtaining body segments based on the structural positioning of the vertebral body;

[0009] Calculate the fat area and / or volume of the corresponding layer of the body segment.

[0010] Furthermore, the positioning is to use a connected domain labeling algorithm to perform body segment positioning from top to bottom;

[0011] Preferably, the fat at the corresponding level of the body segment is divided into chest fat and abdominal fat;

[0012] Preferably, the chest fat is the fat from the upper edge of the first thoracic vertebra to the upper edge of the tenth thoracic vertebra;

[0013] Preferably, the abdominal fat is the fat from the upper edge of the 10th thoracic vertebra to the upper edge of the pelvis;

[0014] Preferably, the method further comprises segmenting the medical image to obtain a pelvis and / or a femur, and based on the body segments obtained by locating the pelvis and / or the femur, calculating the fat at a corresponding level of the body segments, wherein the fat at the corresponding level is pelvic fat and / or thigh fat;

[0015] Preferably, the pelvic fat is the fat from the upper edge of the pelvis to the lower edge of the pelvis;

[0016] Preferably, the thigh fat is the fat from the lower edge of the pelvis to the lower edge of the femur.

[0017] The segmentation also includes segmenting other fat areas of the medical image, and the other fat areas include one or more of the following areas: subcutaneous fat area, neck-supraclavicular-axillary fat area, mediastinum area, and abdominal and pelvic visceral area.

[0018] Furthermore, the segmentation process also includes fine segmentation, removing the portion of the other fat region image whose grayscale value is less than a preset threshold to obtain a fat image;

[0019] Preferably, the subcutaneous fat area, the neck-supraclavicular-axillary fat area, the mediastinum area, and the abdominal-pelvic visceral area in the other fat areas are finely segmented to obtain subcutaneous fat, neck-supraclavicular-axillary fat, mediastinum fat, and abdominal-pelvic visceral fat;

[0020] Preferably, the preset threshold is adaptively determined or self-set using a grayscale histogram.

[0021] The fat area is calculated by multiplying the actual area of ​​each pixel in the fat axial image by the number of pixels;

[0022] Preferably, the fat area includes one or more of the following: subcutaneous fat area, neck-supraclavicular-axillary fat area, mediastinal fat area, abdominal and pelvic visceral fat area, bone marrow fat area, and thigh fat area;

[0023] Preferably, the area calculation further includes calculating the fat area of ​​the vertebral center layer, by first calculating the ordinal number i of the vertebral center layer; and then multiplying the actual area of ​​each pixel in the axial image of the ordinal layer i by the number of pixels to obtain the area;

[0024] Preferably, the calculation formula of the vertebral center level number is:

[0025]

[0026] Among them, the maximum layer number on the vertical axis is i max , the minimum level number is i min , Round(·) means taking an integer.

[0027] The volume is calculated by multiplying the fat area of ​​each layer by the layer thickness and adding them layer by layer;

[0028] Preferably, the volume calculation formula is:

[0029]

[0030] Among them, the maximum layer number on the vertical axis is i max , the minimum level number is i min , S ATi

[0031] is the fat area, h is the layer thickness;

[0032] Preferably, the volume includes one or more of the following: subcutaneous fat volume, neck-supraclavicular-axillary fat volume, mediastinal fat volume, abdominal and pelvic visceral fat volume, bone marrow fat volume, thigh fat volume;

[0033] Preferably, the volume also includes thigh intermuscular fat volume, which is calculated by subtracting thigh subcutaneous fat volume from bone marrow fat volume from thigh fat volume.

[0034] The calculation also includes fat fraction, which is obtained by dividing the signal intensity of the fat phase image by the signal intensity of the same phase image;

[0035] Preferably, the parts involved in the fat fraction include: vertebral body, pelvis, humerus, and femur.

[0036] The method further comprises averaging the chest fat area and the abdominal fat area to obtain the average upper body fat area; summing the chest fat volume and the abdominal fat volume to obtain the upper body fat volume;

[0037] Optionally, the method further comprises averaging the mediastinal fat area and the abdominal and pelvic visceral fat area to obtain the average internal fat area of ​​the upper body; summing the mediastinal fat volume and the abdominal and pelvic visceral fat volume to obtain the internal fat volume of the upper body;

[0038] Optionally, the method further comprises averaging the chest fat area, the abdominal fat area, and the pelvic fat area to obtain an average trunk fat area; and summing the chest fat volume, the abdominal fat volume, and the pelvic fat volume to obtain the trunk fat volume;

[0039] Optionally, the method further comprises averaging the chest fat area, the abdominal fat area, the pelvic fat area, and the thigh fat area to obtain the average fat area of ​​the trunk and lower limbs; summing the chest fat volume, the abdominal fat volume, the pelvic fat volume, and the thigh fat volume to obtain the fat volume of the trunk and lower limbs;

[0040] Optionally, the method further includes averaging the chest fat area, abdominal fat area, pelvic fat area, thigh fat area, and neck-supraclavicular-axillary fat area to obtain the average body fat area; and summing the chest fat volume, abdominal fat volume, pelvic fat volume, thigh fat volume, and neck-supraclavicular-axillary fat volume to obtain the body fat volume.

[0041] The object of the present invention is to provide an automatic fat quantification system based on vertebral segmentation, comprising:

[0042] Acquisition module: acquiring medical images of subjects;

[0043] Segmentation module: segmenting the vertebrae in the medical image to obtain vertebrae;

[0044] Positioning module: obtaining body segments based on the structural positioning of the vertebral body;

[0045] Calculation module: calculates the fat area and / or volume of the corresponding layer of the body segment.

[0046] The object of the present invention is to provide an automatic fat quantification device based on vertebral segmentation, comprising:

[0047] A memory and a processor, wherein the memory is used to store program instructions; the processor is used to call the program instructions, and when the program instructions are executed, any one of the automatic fat quantification methods based on vertebral segmentation is implemented.

[0048] The object of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, comprising:

[0049] When the computer program is executed by a processor, any one of the automatic fat quantification based on vertebral segmentation is implemented.

[0050] Advantages of the present invention:

[0051] 1. This method can output the fat volume of multiple parts, the fat fraction of each major bone, and the fat area of ​​each vertebral layer, which comprehensively reflects the distribution of body fat and creates more possibilities for in-depth research on the relationship between fat distribution and disease;

[0052] 2. This method has good robustness and transferability, and can be well applied to patients with diseases with altered fat distribution, and can also be well applied to data from different devices or scanning parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0054] Figure 1 A schematic diagram of a flow chart of an automatic fat quantification method based on vertebral segmentation provided in an embodiment of the present invention;

[0055] Figure 2 A schematic diagram of an automatic fat quantification system based on vertebral segmentation provided by an embodiment of the present invention;

[0056] Figure 3 A schematic diagram of an automatic fat quantification device based on vertebral segmentation provided by an embodiment of the present invention;

[0057] Figure 4 A schematic diagram of a whole body fat segmentation process provided by an embodiment of the present invention;

[0058] Figure 5 This is a schematic diagram of whole body fat segmentation results provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0060] In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The sequence numbers of the operations, such as S101, S102, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0061] Figure 1 A schematic diagram of an automatic fat quantification method based on vertebral segmentation provided by an embodiment of the present invention specifically includes:

[0062] S101: Obtain medical images of the subject;

[0063] In one embodiment, the medical image adopts one or more of the following: CT image, magnetic resonance imaging image.

[0064] In one embodiment, in magnetic resonance imaging, the molecular environments of hydrogen protons in fat tissue and other tissues in the human body are different, so their resonance frequencies are different; when hydrogen protons in fat and other tissues are excited by radio frequency pulses at the same time, their relaxation times are also different. When signals are collected at different echo times, fat tissue and non-fat tissue show different signal intensities. Using the above characteristics of different tissues in the human body, a variety of pulse sequences for suppressing fat signals have been developed.

[0065] In one embodiment, Dixon, i.e. magnetic resonance water-fat separation technology, is based on the fact that fat and water are at different resonance frequencies. Dixon is based on TSE or 3DGRE sequences and can generate four contrast images, including positive phase (water and fat phases are consistent), negative phase (water and fat phases are opposite), and fat phase map and water phase map generated by post-processing calculation.

[0066] S102: Segmenting the vertebrae in the medical image to obtain vertebrae;

[0067] In one embodiment, the segmentation model uses one or more of the following: U-Net, Faster R-

[0068] CNN, Deeplab V3, YOLACT, Mask R-CNN, RGA U-Net, PSPNet, SegNet,

[0069] FCN, SegGPT, SAM, SEEM.

[0070] In one embodiment, U-Net is a semantic segmentation algorithm based on deep learning. It allows the computer to automatically identify objects in different areas of an image and classify each pixel, thereby achieving pixel-level image segmentation. The name U-Net comes from the U-shaped shape of its network structure, and its design purpose is to solve the problems existing in medical image segmentation.

[0071] The network structure of U-Net mainly consists of two parts, one is the contraction path and the other is the expansion path. The contraction path is used to extract the feature information of the image, while the expansion path is used to achieve pixel-level segmentation. In the contraction path, through some convolution and pooling operations, the resolution of the input image is reduced from the original high resolution to low resolution, and the feature information is extracted. In the expansion path, the network restores the high-resolution feature map through deconvolution operations, thereby obtaining a segmentation result with the same resolution as the original image.

[0072] There is also a span connection between the contraction path and the expansion path of U-Net. The span connection is designed to retain high-quality feature information. After the features have been convolved and pooled, their position information has been lost. At this time, if the deconvolution operation is performed directly, the segmentation result may be inaccurate. The use of span connection can avoid this problem. It will splice the feature maps at the same position between the contraction path and the expansion path, so that the position information can be retained during the deconvolution operation, improving the accuracy of segmentation.

[0073] U-Net has some other features, such as the activation function it uses is ReLU instead of sigmoid or tanh, which can speed up the network training process. In addition, U-Net also uses data enhancement techniques such as rotation, flipping and scaling to increase the diversity of the training set, thereby improving the robustness of the model.

[0074] In general, the principle of U-Net is a convolutional neural network based on deep learning. It uses contraction paths and expansion paths to extract and restore feature information, and uses span connections to retain position information, thereby achieving high-precision semantic segmentation. It has a wide range of applications, such as medical image segmentation, remote sensing image segmentation, etc.

[0075] In one embodiment, the segmentation also includes segmenting other fat areas of the medical image, and the other fat areas include one or more of the following areas: subcutaneous fat area, neck-supraclavicular-axillary fat area, mediastinum area, and abdominal and pelvic visceral area.

[0076] In one embodiment, the segmentation process further includes fine segmentation, removing the portion of the other fat region image whose grayscale value is less than a preset threshold to obtain a fat image;

[0077] Preferably, the subcutaneous fat area, the neck-supraclavicular-axillary fat area, the mediastinum area, and the abdominal-pelvic visceral area in the other fat areas are finely segmented to obtain subcutaneous fat, neck-supraclavicular-axillary fat, mediastinum fat, and abdominal-pelvic visceral fat;

[0078] Preferably, the preset threshold is adaptively determined or self-set using a grayscale histogram.

[0079] In a specific embodiment, the whole body fat segmentation process is as follows Figure 4 As shown, the segmentation scheme is as follows:

[0080] (1) Establish a whole-body magnetic resonance image dataset: A sample of the dataset is all Dixon water phase, fat phase, in-phase, and anti-phase images collected from the same patient during the same examination. The Dixon in-phase image is used as a reference to align other images. The dataset is divided into training set, validation set, and test set. The subcutaneous fat area, cervical-supraclavicular-axillary fat area, mediastinum area, abdominal and pelvic visceral area, and vertebral, pelvic, humeral, and femoral bone marrow are manually annotated.

[0081] (2) Coarse segmentation: A deep learning model was constructed and the image segmentation network was trained using all axial images in the training set. The optimal hyperparameters of the training model were determined based on the results of the validation set. Four images of each axial layer of each sample were predicted in the test set, and the category with the highest average prediction probability was adopted to segment the subcutaneous fat area, the cervical-supraclavicular-axillary fat area, the mediastinum area, the abdominal and pelvic visceral area, and the bone marrow.

[0082] (3) Fine segmentation: For the segmentation results of the subcutaneous fat area, cervical-supraclavicular-axillary fat area, mediastinal area, and abdominal-pelvic visceral area given by the model, according to the threshold set by the user or adaptively determined based on the grayscale histogram, the pixels in the Dixon fat phase image with a grayscale greater than the threshold are defined as subcutaneous fat, cervical-supraclavicular-axillary fat, mediastinal fat, and abdominal-pelvic visceral fat, respectively.

[0083] In one embodiment, a segmentation model is constructed based on the U-Net architecture. The segmentation model consists of an input layer, N groups of downsampling modules, N groups of upsampling modules, and an output layer. N is a natural number greater than 1. The downsampling module consists of a convolution layer and a pooling layer, and the upsampling module consists of a transposed convolution. After the whole body image is input through the input layer, the upsampling module performs feature extraction to obtain a feature map. The feature map is restored to the size of the image through the upsampling module, and the feature maps of different sizes obtained by each downsampling module are fused with the features of the corresponding upsampling module through jump connections to obtain the final feature map, and the final feature map is input to the output layer to obtain the segmentation result. The output layer uses a 1×1 convolution kernel to perform a linear transformation on the final feature map to generate the category probability of each pixel, and finally uses a softmax activation function to convert the probability into a category label for each pixel.

[0084] Furthermore, the loss function of the segmentation model adopts Dice loss and the optimizer adopts Adam.

[0085] In a specific embodiment, the whole body fat visualization result of the fat quantification method based on vertebral segmentation of the present invention is as follows: Figure 5 The Dice coefficients of various categories are shown in Table 1.

[0086] Table 1

[0087]

[0088] S103: obtaining body segments based on the structural positioning of the vertebral body;

[0089] In one embodiment, the positioning is to use a connected domain labeling algorithm to perform body segment positioning from top to bottom;

[0090] Preferably, the fat at the corresponding level of the body segment is divided into chest fat and abdominal fat;

[0091] Preferably, the chest fat is the fat from the upper edge of the first thoracic vertebra to the upper edge of the tenth thoracic vertebra;

[0092] Preferably, the abdominal fat is the fat from the upper edge of the 10th thoracic vertebra to the upper edge of the pelvis;

[0093] Preferably, the method further comprises segmenting the medical image to obtain a pelvis and / or a femur, and based on the body segments obtained by locating the pelvis and / or the femur, calculating the fat at a corresponding level of the body segments, wherein the fat at the corresponding level is pelvic fat and / or thigh fat;

[0094] Preferably, the pelvic fat is the fat from the upper edge of the pelvis to the lower edge of the pelvis;

[0095] Preferably, the thigh fat is the fat from the lower edge of the pelvis to the lower edge of the femur.

[0096] In a specific embodiment, the portion of subcutaneous fat between the upper edge of the 1st thoracic vertebra (inclusive) and the upper edge of the 10th thoracic vertebra (excluding) is defined as the chest subcutaneous fat volume, the portion between the upper edge of the 10th thoracic vertebra (inclusive) and the upper edge of the pelvis (excluding) is defined as the abdominal subcutaneous fat volume, the portion between the upper edge of the pelvis (inclusive) and the lower edge of the pelvis (inclusive) is defined as the pelvic subcutaneous fat volume, and the portion between the lower edge of the pelvis (excluding) and the lower edge of the femur (inclusive) is defined as the thigh subcutaneous fat volume.

[0097] S104: Calculate the fat area and / or volume of the corresponding layer of the body segment.

[0098] In one embodiment, the fat area is calculated by multiplying the actual area of ​​each pixel in the fat axial image by the number of pixels;

[0099] Preferably, the fat area includes one or more of the following: subcutaneous fat area, neck-supraclavicular-axillary fat area, mediastinal fat area, abdominal and pelvic visceral fat area, bone marrow fat area, and thigh fat area;

[0100] In one embodiment, the area calculation further includes calculating the fat area of ​​the vertebral center layer, by first calculating the ordinal number i of the vertebral center layer; and then multiplying the actual area of ​​each pixel in the axial image of the ordinal layer i by the number of pixels to obtain the fat area;

[0101] Preferably, the calculation formula of the vertebral center level number is:

[0102]

[0103] Among them, the maximum layer number on the vertical axis is i max , the minimum level number is i min , Round(·) means taking an integer.

[0104] In one embodiment, the volume is calculated by multiplying the area of ​​each layer of fat by the layer thickness and adding them layer by layer;

[0105] Preferably, the volume calculation formula is:

[0106]

[0107] Among them, the maximum layer number on the vertical axis is i max , the minimum level number is i min , S ATi

[0108] is the fat area, h is the layer thickness;

[0109] In one embodiment, the volume includes one or more of the following: subcutaneous fat volume, neck-supraclavicular-axillary fat volume, mediastinal fat volume, abdominal and pelvic visceral fat volume, bone marrow fat volume, and thigh fat volume;

[0110] Preferably, the volume also includes thigh intermuscular fat volume, which is calculated by subtracting thigh subcutaneous fat volume from bone marrow fat volume from thigh fat volume.

[0111] In one embodiment, the calculation further includes a fat fraction, and the fat fraction is obtained by dividing the signal intensity of the fat phase image by the signal intensity of the same phase image;

[0112] Preferably, the parts involved in the fat fraction include: vertebral body, pelvis, humerus, and femur.

[0113] In a specific embodiment, for the subcutaneous fat segmentation result given by the model, the area of ​​each layer of fat is multiplied by the layer thickness and the layers are added up layer by layer to obtain the subcutaneous fat volume of the whole body.

[0114] For the cervical-supraclavicular-axillary fat, mediastinal fat, and abdominal-pelvic visceral fat, the area of ​​each layer of fat was multiplied by the layer thickness and added up layer by layer to obtain the corresponding fat volume.

[0115] For the bone marrow, the fat fraction was obtained by dividing the grayscale of the aqueous phase image by the grayscale of the in-phase image, and the average fat fraction and standard deviation of each vertebra, pelvis, humerus, and femur were output.

[0116] Calculate the various fat areas at the level where the centroid of each vertebra is located and obtain the results.

[0117] In a specific embodiment, the fat area is calculated. In an axial image, let the number of pixels predicted by the model as a certain type of fat be x AT , the actual area corresponding to each pixel is S pixel , then the fat area S of this type of image AT for:

[0118] S AT =x AT ×S pixel (1)

[0119] S pixel =x_spacing×y_spacing (2)

[0120] Wherein, x_spacing and y_spacing are both image parameters of Dixon image. According to the need, the area of ​​subcutaneous adipose tissue (SAT), the area of ​​visceral adipose tissue (VAT), the area of ​​bone marrow adipose tissue (BMAT), etc. can be calculated by formulas (1) to (2).

[0121] In a specific embodiment, the calculation of the fat area at the center of the vertebral body. Assume that the maximum layer number of the model's segmentation result of a vertebral body on the vertical axis is i max , the minimum level number is i min , then the vertebral center level number i is:

[0122]

[0123] Wherein, Round(·) indicates taking an integer. Obtain the axial image of layer i, and calculate the subcutaneous fat area and visceral fat area at the vertebral center layer according to formulas (1) to (2).

[0124] In one specific embodiment, the calculation of fat volume:

[0125] Chest, abdomen, pelvis: Let the maximum layer number of the part to be measured (such as chest) on the vertical axis be i max , the minimum level number is i min . Let the layer thickness of the Dixon image in the vertical axis direction be h, and for the ordinal number i min ~i max The fat area S of the category to be measured (such as subcutaneous fat) is calculated according to formulas (1) to (2) for each layer i between ATi , then the fat volume V of the tested part (such as subcutaneous fat in the chest) AT for:

[0126]

[0127] Thigh: Thigh fat is mainly divided into three parts: thigh subcutaneous fat, bone marrow fat and intermuscular fat. Refer to formula (4) to calculate the thigh subcutaneous fat volume V thighSAT , bone marrow fat volume V thighBMAT Input the thigh area into the fine segmentation module to obtain the total thigh fat. Refer to formula (4) to calculate the total thigh fat volume V thighAT The volume of thigh intermuscular fat is:

[0128] V thighIMAT =V thighAT -V thighSAT -VthighBMAT (5)

[0129] In a specific embodiment, the calculation of fat fraction is as follows: Fat fraction: Taking the vertebral fat fraction as an example, assuming that the model has given a segmentation result for a vertebral body, there are n voxels in total, and the fat fraction FF of any voxel i of the vertebral body is i for:

[0130]

[0131] Among them, FP i represents the signal intensity of voxel i in the fat phase image, IP i represents the signal intensity in the inphase image. The average value of the vertebral body fat fraction is and standard deviation FF σ They are:

[0132]

[0133]

[0134] In a specific embodiment, the fat volume measures the size of the space occupied by adipose tissue, and the fat fraction measures the degree (percentage) of fat infiltration in tissues and organs (such as bone marrow, liver, spleen, kidney, pancreas).

[0135] In a specific embodiment, the main goal of the present invention is to evaluate the whole body fat distribution from the Dixon image as comprehensively as possible. Existing studies usually only evaluate the distribution of subcutaneous fat and visceral fat, but bone marrow fat is also an important component of body fat, which is one of the significant high signal areas in the Dixon fat phase image. Therefore, the model that simultaneously segments the bone marrow can achieve more comprehensive whole body fat segmentation and quantification, and reduce the risk of misidentifying bone marrow fat as visceral fat.

[0136] In a specific embodiment, due to the cumbersomeness of fat volume measurement, most current fat quantitative studies only measure the fat area at a specific vertebral level, and the measurement of fat volume is currently mainly involved in scientific research, and no widely used normal reference value has been established. In addition, doctors have large differences in the selection of vertebral levels and the labeling of fat areas. Therefore, obtaining objective quantitative fat area at a specific vertebral anatomical level based on vertebral segmentation results has better compatibility with clinical practice.

[0137] By using skeletal structures such as vertebrae as anatomical landmarks, the body can be divided into the chest (upper edge of the 1st thoracic vertebra (inclusive) to upper edge of the 10th thoracic vertebra (exclusive)), abdomen (upper edge of the 10th thoracic vertebra (inclusive) to upper edge of the pelvis (exclusive)), pelvis (upper edge of the pelvis (inclusive) to lower edge of the pelvis (inclusive)), and thigh (lower edge of the pelvis (exclusive) to lower edge of the femur (inclusive)), and the total fat volume of each part can be calculated.

[0138] In a specific embodiment, the body segments can be located through the vertebrae, and then the axial images of the specific segments are intercepted and the organ segmentation model is applied. For example, the abdominal organs (liver, pancreas, spleen, kidney) are segmented using the layers between T6 and L3, and the organ fat score is calculated; the heart is segmented using the layers between T1 and T10, and the epicardial fat volume and pericardial fat volume are measured.

[0139] In one embodiment, the method further comprises averaging the chest fat area and the abdominal fat area to obtain the average upper body fat area; summing the chest fat volume and the abdominal fat volume to obtain the upper body fat volume;

[0140] Optionally, the method further comprises averaging the mediastinal fat area and the abdominal and pelvic visceral fat area to obtain the average internal fat area of ​​the upper body; summing the mediastinal fat volume and the abdominal and pelvic visceral fat volume to obtain the internal fat volume of the upper body;

[0141] Optionally, the method further comprises averaging the chest fat area, the abdominal fat area, and the pelvic fat area to obtain an average trunk fat area; and summing the chest fat volume, the abdominal fat volume, and the pelvic fat volume to obtain the trunk fat volume;

[0142] Optionally, the method further comprises averaging the chest fat area, the abdominal fat area, the pelvic fat area, and the thigh fat area to obtain the average fat area of ​​the trunk and lower limbs; summing the chest fat volume, the abdominal fat volume, the pelvic fat volume, and the thigh fat volume to obtain the fat volume of the trunk and lower limbs;

[0143] Optionally, the method further includes averaging the chest fat area, abdominal fat area, pelvic fat area, thigh fat area, and neck-supraclavicular-axillary fat area to obtain the average body fat area; and summing the chest fat volume, abdominal fat volume, pelvic fat volume, thigh fat volume, and neck-supraclavicular-axillary fat volume to obtain the body fat volume.

[0144] In one embodiment, chest fat includes chest subcutaneous fat, mediastinal fat, and vertebral bone marrow fat from the 1st thoracic vertebra to the upper edge of the 10th thoracic vertebra (excluding); abdominal fat includes abdominal subcutaneous fat, abdominal visceral fat, and vertebral bone marrow fat from the upper edge of the 10th thoracic vertebra (inclusive) to the upper edge of the pelvis (excluding); pelvic fat includes pelvic subcutaneous fat, pelvic visceral fat, and bone marrow fat from the upper edge of the pelvis (inclusive) to the lower edge of the pelvis (inclusive); thigh fat includes thigh subcutaneous fat, thigh intermuscular fat, and bone marrow fat from the lower edge of the pelvis (excluding) to the lower edge of the femur (inclusive).

[0145] In one embodiment, the calculation of fat area of ​​the present invention includes one or more of the following:

[0146] 1) Calculation of fat area of ​​chest and / or abdominal body segments based on vertebral segmentation and positioning;

[0147] 2) Calculate the fat area of ​​the pelvic body segment based on pelvic segmentation and positioning;

[0148] 3) Calculation of fat area of ​​the thigh body segment based on femoral segmentation and positioning;

[0149] 4) Calculate the fat area of ​​other regions (subcutaneous fat area, neck-supraclavicular-axillary fat area, mediastinal fat area, abdominal and pelvic visceral fat area, and bone marrow fat area).

[0150] In one embodiment, the calculation of fat volume of the present invention includes one or more of the following:

[0151] 1) Calculation of fat volume of chest and / or abdominal body segments based on vertebral segmentation and positioning;

[0152] 2) Obtaining the fat area of ​​the chest and / or abdominal body segments based on vertebral segmentation and positioning, and then calculating the fat volume based on the fat area;

[0153] 3) Calculate the fat area of ​​the pelvic body segment based on pelvic segmentation and positioning, and then calculate the fat volume based on the fat area;

[0154] 4) Calculate the fat area of ​​the thigh body segment based on the femur segmentation positioning, and then calculate the fat volume based on the fat area;

[0155] 5) Calculate the fat area of ​​other regions (subcutaneous fat area, neck-supraclavicular-axillary fat area, mediastinal fat area, abdominal and pelvic visceral fat area, and bone marrow fat area), and then calculate the fat volume based on the fat area;

[0156] 6) Calculate the fat volume of the pelvic body segment based on pelvic segmentation and positioning;

[0157] 7) Calculation of fat volume of the thigh body segment based on femoral segmentation and positioning;

[0158] 8) For other areas (subcutaneous fat area, neck-supraclavicular-axillary fat area, mediastinal fat

[0159] In one embodiment, the fat area / volume summation of the present invention includes one or more of the following: 1) averaging / summing the fat area / volume of the chest body segment and the fat area / volume of the abdomen body segment;

[0160] 2) averaging / summing the fat area / volume of the thoracic body segment and the fat area / volume of the pelvic body segment;

[0161] 3) averaging / summing the fat area / volume of the chest body segment and the fat area / volume of the thigh body segment;

[0162] 4) averaging / summing the fat area / volume of the abdominal body segment and the fat area / volume of the pelvic body segment;

[0163] 5) averaging / summing the fat area / volume of the abdominal body segment and the fat area / volume of the thigh body segment;

[0164] 6) averaging / summing the fat area / volume of the pelvic body segment and the fat area / volume of the thigh body segment;

[0165] 7) averaging / summing the fat area / volume of the thoracic body segment, the fat area / volume of the abdominal body segment, and the fat area / volume of the pelvic body segment;

[0166] 8) averaging / summing the fat area / volume of the chest body segment, the fat area / volume of the abdomen body segment, and the fat area / volume of the thigh body segment;

[0167] 9) averaging / summing the fat area / volume of the abdominal body segment, the fat area / volume of the pelvic body segment, and the fat area / volume of the thigh body segment;

[0168] 10) Average the area / volume of chest fat and neck-supraclavicular-axillary fat /

[0169] Sum; average / sum of fat area / volume of abdominal fat and neck-supraclavicular-axillary fat; average / sum of fat area / volume of pelvic fat and neck-supraclavicular-axillary fat; average / sum of fat area / volume of thigh fat and neck-supraclavicular-axillary fat; average / sum of fat area / volume of chest fat, abdominal fat, and neck-supraclavicular-axillary fat; average / sum of fat area / volume of chest fat, abdominal fat, pelvic fat, and neck-supraclavicular-axillary fat; average / sum of fat area / volume of chest fat, abdominal fat, pelvic fat, thigh fat, and neck-supraclavicular-axillary fat; average / sum of fat area / volume of abdominal and pelvic visceral fat and mediastinal fat

[0170] Averaging / summing of volumes; averaging / summing of fat areas / volumes of chest fat, abdominal fat, abdominal-pelvic visceral fat, and mediastinal fat; averaging / summing of fat areas / volumes of chest fat, abdominal fat, pelvic fat, abdominal-pelvic visceral fat, and mediastinal fat; averaging / summing of fat areas / volumes of chest fat, abdominal fat, pelvic fat, thigh fat, abdominal-pelvic visceral fat, and mediastinal fat; averaging / summing of fat areas / volumes of chest fat, abdominal fat, pelvic fat, thigh fat, neck-supraclavicular-axillary fat, abdominal-pelvic visceral fat, and mediastinal fat.

[0171] In one embodiment, the fat area, fat volume, and fat fraction are obtained by inputting the medical image of the subject into the method of the present invention. The fat area, fat volume, and fat fraction can quantify the whole body fat or a certain segment of the whole body or a certain type of fat in the whole body, which can not only conduct research on diseases related to fat, but also promote research on the intrinsic connection of human body fat, which is conducive to improving the research possibility between diseases (involving fat in different parts). In addition, the vertebral segmentation result obtains the objective quantitative fat area of ​​a specific vertebral anatomical level, which has better compatibility with clinical practice.

[0172] In one embodiment, the whole body fat quantification method of the present invention obtains the fat volume and / or area of ​​different body segments based on vertebral segmentation, such as obtaining the fat volume and / or area of ​​the abdominal segment, and the fat of the abdominal segment further includes the volume and / or area of ​​abdominal subcutaneous fat, abdominal visceral fat, and vertebral bone marrow fat of the abdominal segment.

[0173] In one embodiment, the method of the present invention can simultaneously obtain the fat volume and / or area of ​​multiple body segments, such as obtaining the fat volume and / or area of ​​the abdominal segment and the pelvic segment.

[0174] Figure 2 A schematic diagram of an automatic fat quantification system based on vertebral segmentation provided by an embodiment of the present invention specifically includes:

[0175] Acquisition module: acquiring medical images of subjects;

[0176] Segmentation module: segmenting the vertebrae in the medical image to obtain vertebrae;

[0177] Positioning module: obtaining body segments based on the structural positioning of the vertebral body;

[0178] Calculation module: calculates the fat area and / or volume of the corresponding layer of the body segment.

[0179] Figure 3 A schematic diagram of an automatic fat quantification device based on vertebral segmentation provided by an embodiment of the present invention specifically includes:

[0180] A memory and a processor; the memory is used to store program instructions; the processor is used to call the program instructions, and when the program instructions are executed, any one of the above-mentioned automatic fat quantification methods based on vertebral segmentation is performed.

[0181] A computer-readable storage medium stores a computer program, wherein the computer program is executed by a processor to perform any one of the above-mentioned methods for automatic fat quantification based on vertebral segmentation.

[0182] The verification results of this verification embodiment show that assigning inherent weights to indications can improve the performance of the method relative to the default setting. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms. The unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units. A person of ordinary skill in the art may understand that all or part of the steps in the various methods of the above-mentioned embodiments may be completed by instructing the relevant hardware through a program, and the program may be stored in a computer-readable storage medium, and the storage medium may include: a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0183] A person skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be implemented by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. The above-mentioned medium storage can be a read-only memory, a disk or an optical disk, etc.

[0184] The above is a detailed introduction to a computer device provided by the present invention. For a person skilled in the art, according to the concept of the embodiments of the present invention, there may be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for automatic fat quantification based on vertebral segmentation, characterized in that: The method comprises: Obtain medical images of the subjects; Segmenting the vertebrae of the medical image to obtain the vertebrae; and segmenting the medical image to obtain the pelvis and femur; Based on the structural positioning of the vertebrae, a body segment is obtained; the fat at the corresponding level of the body segment is chest fat and abdominal fat; based on the positioning of the pelvis and femur, the fat at the corresponding level is pelvic fat and thigh fat; wherein the chest fat includes the portion from the upper edge of the first thoracic vertebra to the upper edge of the tenth thoracic vertebra, excluding the vertebral marrow fat, chest subcutaneous fat, and mediastinal fat at the upper edge of the tenth thoracic vertebra; the abdominal fat includes the portion from the upper edge of the tenth thoracic vertebra to the upper edge of the pelvis, excluding the vertebral marrow fat, abdominal subcutaneous fat, and abdominal visceral fat at the upper edge of the pelvis; the pelvic fat includes the marrow fat, pelvic subcutaneous fat, and pelvic visceral fat from the upper edge of the pelvis to the lower edge of the pelvis; the thigh fat includes the lower edge of the pelvis, excluding the marrow fat, thigh subcutaneous fat, and thigh intermuscular fat from the lower edge of the pelvis to the lower edge of the femur; Calculating the fat area and volume of the corresponding layer of the body segment; The segmentation is performed by a segmentation model, and the construction process of the segmentation model is as follows: Obtain a whole-body magnetic resonance image set, including Dixon water phase, fat phase, in-phase, and anti-phase images, and annotate the subcutaneous fat area, cervical-supraclavicular-axillary fat area, mediastinum area, abdominal and pelvic visceral area, and vertebral, pelvic, humeral, and femoral bone marrow; The whole-body magnetic resonance image is input into the deep learning model for segmentation training to obtain a segmentation model, wherein the segmentation process also includes fine segmentation, removing the portion of the image with a gray value of other fat regions less than a preset threshold to obtain a fat image, wherein the other fat regions include subcutaneous fat regions, neck-supraclavicular-axillary fat regions, mediastinum regions, and abdominal and pelvic visceral regions; The positioning is to use a connected domain labeling algorithm to perform body segment positioning from top to bottom.

2. The method for automatic fat quantification based on vertebral segmentation according to claim 1, characterized in that: The preset threshold is adaptively determined or self-set using a grayscale histogram.

3. The method for automatic fat quantification based on vertebral segmentation according to any one of claims 1 to 2, characterized in that: The fat area is calculated by multiplying the actual area of ​​each pixel in the fat axial image by the number of pixels.

4. The method for automatic fat quantification based on vertebral segmentation according to claim 3, characterized in that: The area calculation also includes calculating the fat area of ​​the vertebral center layer by first calculating the vertebral center layer sequence ; Then the ordinal It is calculated by multiplying the actual area of ​​each pixel in the axial image of the slice by the number of pixels.

5. The method for automatic fat quantification based on vertebral segmentation according to claim 4, characterized in that: The calculation formula of the vertebral center level number is: Among them, the maximum layer number on the vertical axis is , the minimum level ordinal number is , Indicates integer.

6. The method for automatic fat quantification based on vertebral segmentation according to claim 1, characterized in that: The volume is calculated by multiplying the fat area of ​​each layer by the layer thickness and adding the sum layer by layer.

7. The method for automatic fat quantification based on vertebral segmentation according to claim 3, characterized in that: The volume is calculated by multiplying the fat area of ​​each layer by the layer thickness and adding the sum layer by layer.

8. The method for automatic fat quantification based on vertebral segmentation according to claim 6 or 7, characterized in that: The volume calculation formula is: Among them, the maximum layer number on the vertical axis is , the minimum level ordinal number is , is the fat area, is the layer thickness.

9. The method for automatic fat quantification based on vertebral segmentation according to claim 6 or 7, characterized in that: The volume includes one or more of the following: subcutaneous fat volume, neck-supraclavicular-axillary fat volume, mediastinal fat volume, abdominal and pelvic visceral fat volume, bone marrow fat volume, and thigh fat volume.

10. The method for automatic fat quantification based on vertebral segmentation according to claim 6 or 7, characterized in that: The volume also includes the intermuscular fat volume of the thigh, which is calculated by subtracting the thigh subcutaneous fat volume from the bone marrow fat volume from the thigh fat volume.

11. The method for automatic fat quantification based on vertebral segmentation according to claim 1, characterized in that: The calculation also includes calculation of fat fraction, and the fat fraction is obtained by dividing the signal intensity of the same-phase image by the signal intensity of the fat phase image.

12. The method for automatic fat quantification based on vertebral segmentation according to claim 11, characterized in that: The fat fraction involves the following parts: vertebral body, pelvis, humerus, and femur.

13. The method for automatic fat quantification based on vertebral segmentation according to any one of claims 1 to 2, characterized in that: The method further comprises averaging the chest fat area and the abdominal fat area to obtain the average upper body fat area; and summing the chest fat volume and the abdominal fat volume to obtain the upper body fat volume.

14. The method for automatic fat quantification based on vertebral segmentation according to claim 3, characterized in that: The method further comprises averaging the chest fat area and the abdominal fat area to obtain the average upper body fat area; and summing the chest fat volume and the abdominal fat volume to obtain the upper body fat volume.

15. The method for automatic fat quantification based on vertebral segmentation according to claim 6, characterized in that: The method further comprises averaging the chest fat area and the abdominal fat area to obtain the average upper body fat area; and summing the chest fat volume and the abdominal fat volume to obtain the upper body fat volume.

16. The method for automatic fat quantification based on vertebral segmentation according to claim 7, characterized in that: The method further comprises averaging the chest fat area and the abdominal fat area to obtain the average upper body fat area; and summing the chest fat volume and the abdominal fat volume to obtain the upper body fat volume.

17. The method for automatic fat quantification based on vertebral segmentation according to any one of claims 1 to 2, characterized in that: The method further comprises averaging the mediastinum fat area and the abdominal and pelvic visceral fat area to obtain the average area of ​​the upper body internal fat; and summing the mediastinum fat volume and the abdominal and pelvic visceral fat volume to obtain the upper body internal fat volume.

18. The method for automatic fat quantification based on vertebral segmentation according to claim 3, characterized in that: The method further comprises averaging the mediastinum fat area and the abdominal and pelvic visceral fat area to obtain the average area of ​​the upper body internal fat; and summing the mediastinum fat volume and the abdominal and pelvic visceral fat volume to obtain the upper body internal fat volume.

19. The method for automatic fat quantification based on vertebral segmentation according to claim 6, characterized in that: The method further comprises averaging the mediastinum fat area and the abdominal and pelvic visceral fat area to obtain the average area of ​​the upper body internal fat; and summing the mediastinum fat volume and the abdominal and pelvic visceral fat volume to obtain the upper body internal fat volume.

20. The method for automatic fat quantification based on vertebral segmentation according to claim 7, characterized in that: The method further comprises averaging the mediastinum fat area and the abdominal and pelvic visceral fat area to obtain the average area of ​​the upper body internal fat; and summing the mediastinum fat volume and the abdominal and pelvic visceral fat volume to obtain the upper body internal fat volume.

21. The method for automatic fat quantification based on vertebral segmentation according to any one of claims 1 to 2, characterized in that: The method further comprises averaging the chest fat area, the abdominal fat area, and the pelvic fat area to obtain the average trunk fat area; and summing the chest fat volume, the abdominal fat volume, and the pelvic fat volume to obtain the trunk fat volume.

22. The method for automatic fat quantification based on vertebral segmentation according to claim 3, characterized in that: The method further comprises averaging the chest fat area, the abdominal fat area, and the pelvic fat area to obtain the average trunk fat area; and summing the chest fat volume, the abdominal fat volume, and the pelvic fat volume to obtain the trunk fat volume.

23. The method for automatic fat quantification based on vertebral segmentation according to claim 6, characterized in that: The method further comprises averaging the chest fat area, the abdominal fat area, and the pelvic fat area to obtain the average trunk fat area; and summing the chest fat volume, the abdominal fat volume, and the pelvic fat volume to obtain the trunk fat volume.

24. The method for automatic fat quantification based on vertebral segmentation according to claim 7, characterized in that: The method further comprises averaging the chest fat area, the abdominal fat area, and the pelvic fat area to obtain the average trunk fat area; and summing the chest fat volume, the abdominal fat volume, and the pelvic fat volume to obtain the trunk fat volume.

25. The method for automatic fat quantification based on vertebral segmentation according to any one of claims 1 to 2, characterized in that: The method also includes averaging the chest fat area, abdominal fat area, pelvic fat area, and thigh fat area to obtain the average fat area of ​​the trunk and lower limbs; and summing the chest fat volume, abdominal fat volume, pelvic fat volume, and thigh fat volume to obtain the fat volume of the trunk and lower limbs.

26. The method for automatic fat quantification based on vertebral segmentation according to claim 3, characterized in that: The method also includes averaging the chest fat area, abdominal fat area, pelvic fat area, and thigh fat area to obtain the average fat area of ​​the trunk and lower limbs; and summing the chest fat volume, abdominal fat volume, pelvic fat volume, and thigh fat volume to obtain the fat volume of the trunk and lower limbs.

27. The method for automatic fat quantification based on vertebral segmentation according to claim 6, characterized in that: The method also includes averaging the chest fat area, abdominal fat area, pelvic fat area, and thigh fat area to obtain the average fat area of ​​the trunk and lower limbs; and summing the chest fat volume, abdominal fat volume, pelvic fat volume, and thigh fat volume to obtain the fat volume of the trunk and lower limbs.

28. The method for automatic fat quantification based on vertebral segmentation according to claim 7, characterized in that: The method also includes averaging the chest fat area, abdominal fat area, pelvic fat area, and thigh fat area to obtain the average fat area of ​​the trunk and lower limbs; and summing the chest fat volume, abdominal fat volume, pelvic fat volume, and thigh fat volume to obtain the fat volume of the trunk and lower limbs.

29. The method for automatic fat quantification based on vertebral segmentation according to any one of claims 1 to 2, characterized in that: The method also includes averaging the chest fat area, abdominal fat area, pelvic fat area, thigh fat area, and neck-supraclavicular-axillary fat area to obtain the average body fat area; and summing the chest fat volume, abdominal fat volume, pelvic fat volume, thigh fat volume, and neck-supraclavicular-axillary fat volume to obtain the body fat volume.

30. The method for automatic fat quantification based on vertebral segmentation according to claim 3, characterized in that: The method also includes averaging the chest fat area, abdominal fat area, pelvic fat area, thigh fat area, and neck-supraclavicular-axillary fat area to obtain the average body fat area; and summing the chest fat volume, abdominal fat volume, pelvic fat volume, thigh fat volume, and neck-supraclavicular-axillary fat volume to obtain the body fat volume.

31. The method for automatic fat quantification based on vertebral segmentation according to claim 6, characterized in that: The method also includes averaging the chest fat area, abdominal fat area, pelvic fat area, thigh fat area, and neck-supraclavicular-axillary fat area to obtain the average body fat area; and summing the chest fat volume, abdominal fat volume, pelvic fat volume, thigh fat volume, and neck-supraclavicular-axillary fat volume to obtain the body fat volume.

32. The method for automatic fat quantification based on vertebral segmentation according to claim 7, characterized in that: The method also includes averaging the chest fat area, abdominal fat area, pelvic fat area, thigh fat area, and neck-supraclavicular-axillary fat area to obtain the average body fat area; and summing the chest fat volume, abdominal fat volume, pelvic fat volume, thigh fat volume, and neck-supraclavicular-axillary fat volume to obtain the body fat volume.

33. An automatic fat quantification system based on vertebral segmentation, characterized in that: include: Acquisition module: acquiring medical images of subjects; Segmentation module: segmenting the vertebrae in the medical image to obtain vertebrae; The segmentation is performed by a segmentation model, and the construction process of the segmentation model is as follows: Obtain a whole-body magnetic resonance image set, including Dixon water phase, fat phase, in-phase, and anti-phase images, and annotate the subcutaneous fat area, cervical-supraclavicular-axillary fat area, mediastinum area, abdominal and pelvic visceral area, and vertebral, pelvic, humeral, and femoral bone marrow; The whole-body magnetic resonance image is input into the deep learning model for segmentation training to obtain a segmentation model, wherein the segmentation process also includes fine segmentation, removing the portion of the image with a gray value of other fat regions less than a preset threshold to obtain a fat image, wherein the other fat regions include subcutaneous fat regions, neck-supraclavicular-axillary fat regions, mediastinum regions, and abdominal and pelvic visceral regions; Positioning module: body segments are obtained based on the structural positioning of the vertebrae; the fat at the corresponding level of the body segments is chest fat and abdominal fat; the fat at the corresponding level of the body segments obtained based on the positioning of the pelvis and femur is pelvic fat and thigh fat; wherein, the chest fat includes the portion from the upper edge of the 1st thoracic vertebra to the upper edge of the 10th thoracic vertebra, excluding the vertebral marrow fat, chest subcutaneous fat, and mediastinal fat at the upper edge of the 10th thoracic vertebra; the abdominal fat includes the portion from the upper edge of the 10th thoracic vertebra to the upper edge of the pelvis, excluding the vertebral marrow fat, abdominal subcutaneous fat, and abdominal visceral fat at the upper edge of the pelvis; the pelvic fat includes the marrow fat, pelvic subcutaneous fat, and pelvic visceral fat from the upper edge of the pelvis to the lower edge of the pelvis; the thigh fat includes the lower edge of the pelvis, excluding the marrow fat, thigh subcutaneous fat, and thigh intermuscular fat from the lower edge of the pelvis to the lower edge of the femur; The positioning is to use a connected domain labeling algorithm to locate the body segment from top to bottom; the calculation module is to calculate the fat area and volume of the corresponding layer of the body segment.

34. An automatic fat quantification device based on vertebral segmentation, characterized in that: include: A memory and a processor, wherein the memory is used to store program instructions; the processor is used to call the program instructions, and when the program instructions are executed, the automatic fat quantification method based on vertebral segmentation described in any one of claims 1-32 is implemented.

35. A computer-readable storage medium having a computer program stored thereon, characterized in that: include: When the computer program is executed by a processor, the automatic fat quantification based on vertebral segmentation as described in any one of claims 1 to 32 is implemented.

Images