CN113487568B - Liver surface smoothness measuring method based on differential curvature - Google Patents
Liver surface smoothness measuring method based on differential curvature Download PDFInfo
- Publication number
- CN113487568B CN113487568B CN202110757748.6A CN202110757748A CN113487568B CN 113487568 B CN113487568 B CN 113487568B CN 202110757748 A CN202110757748 A CN 202110757748A CN 113487568 B CN113487568 B CN 113487568B
- Authority
- CN
- China
- Prior art keywords
- liver
- smoothness
- image
- curve
- calculating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/181—Segmentation; Edge detection involving edge growing; involving edge linking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30056—Liver; Hepatic
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method for measuring liver surface smoothness based on differential curvature, and mainly relates to a digital image processing technology. The technical scheme of the invention is as follows: (1) Preprocessing a liver CT image by using a medical image window adjusting algorithm to enhance the contrast of an abdomen liver region; (2) Obtaining a liver segmentation result by using a depth network model; (3) Acquiring the contour of the liver segmentation result by using an edge detection algorithm; (4) Selecting a liver region to be evaluated by utilizing man-machine interaction to obtain a liver profile curve to be evaluated; (5) The degree of smoothness of the liver shape is measured using curvature calculation of the curve. The method effectively quantifies the smoothness of the liver surface, realizes accurate and objective measurement of the smoothness of the liver contour, and further assists doctors in quantifying and evaluating liver functions.
Description
Technical Field
The invention belongs to the field of medical image analysis, and particularly provides a method for measuring liver surface smoothness based on differential curvature.
Background
Lesions of the liver have sudden onset and high morbidity, and currently, the incidence and mortality rate of liver cancer in China are the first world. Therefore, liver cancer seriously threatens the life health of people in China. Analysis of abdominal medical images is the most effective and direct adjuvant therapy for liver disease, and comprehensive medical image analysis has important value for treatment and prognosis of patients. There are several important indicators for assessing liver function in patients using abdominal CT images, mainly concerning liver volume, tumor volume, liver to spleen ratio, and liver smoothness. The segmentation of liver, spleen and liver tumor can be realized by using the image segmentation technology, so as to estimate the volume data. However, the calculation of smoothness of the liver has long relied on subjective judgment of doctors, and there is no unified standard and objective evaluation method. According to clinical anatomy, the normal liver has smooth surface, and the liver with diseases is easy to form lumps with different sizes, so that the surface of the liver is uneven. Generally, clinicians consider that smoothness of liver surface is directly related to liver function, so that research on a measurement method of liver surface smoothness is of great importance to promote intelligent medical treatment.
Since the surface of the liver is an irregularly curved surface and abdominal CT imaging is performed in slices, the problem of estimating the smoothness of the liver surface can be translated into a smoothness estimate of the liver contour curve in the slices. In order to effectively estimate the smoothness of the liver contour curve, it is first necessary to obtain an accurate liver contour. In image processing, there are generally two methods for acquiring the contour of an object: edge detection and image segmentation. Since there is typically a lot of noise in the abdominal CT image and the boundary of the liver with surrounding tissue is blurred, these problems make it difficult to achieve accurate liver edge detection with classical edge detection operators. Compared with the traditional edge detection operator, the object of image segmentation is to extract the closed contour information of the target, so that better liver contour information can be obtained by utilizing the image segmentation.
Traditional image segmentation methods are usually semi-automatic, rely on manually designed feature descriptors to extract features, and generally have low segmentation accuracy. Such as a region growing method, a level set method, a watershed method, a clustering method, and the like. The traditional image segmentation method relies on model driving, ignores the distribution characteristic of data, is difficult to extract high-level semantic information of an image, and has low algorithm robustness and limited segmentation accuracy although a good segmentation result can be realized for a simple image.
Compared with the traditional image segmentation method, the image segmentation method based on deep learning can effectively learn high-level semantic information of images and realize end-to-end full-automatic image segmentation, so that the image segmentation method based on deep learning becomes a popular research field in the field of computer vision in recent years. For image semantic segmentation, the most representative convolutional neural networks are: a fully convolutional neural network (fully convolutional network, FCN), segNet, deepLab V1-V3, and the like. The FCN adopts the encoding and decoding network for the first time, realizes the effective fusion of the low-layer semantic information and the high-layer semantic information of the image through jump connection, greatly improves the image segmentation effect, and particularly improves the contour detail of the object in the image obviously. SegNet can only realize the image segmentation problem of fixed size to FCN, utilize image pyramid algorithm to make the network can accept the input image of arbitrary size. And deep Lab utilizes ResNet to obtain better feature coding, and realizes wider receiving domain by introducing cavity convolution, so as to realize high-efficiency multi-scale feature fusion without increasing network parameters, and realize high-precision image segmentation effect. Although image segmentation methods based on deep convolutional neural networks can realize end-to-end image semantic segmentation, the methods are difficult to directly popularize in medical image segmentation. On one hand, the medical image segmentation is a small sample problem, and on the other hand, the medical image has the problems of low contrast, serious noise pollution, ambiguous target information and the like.
In response to the above problems, ronneeberger et al proposes a U-shaped network structure (U-Net) with a perfectly symmetric structure to achieve efficient segmentation of medical images, which fully considers the small sample problem of data, employing fewer convolution layers to prevent the over-fitting problem. Because the decoder of the U-shaped network adopts deconvolution, the fusion of the high-layer and low-layer characteristics is more beneficial to optimizing edge details, and so far, the mainstream medical image segmentation network depends on the U-shaped structure. Based on this, li et al use a mixed dense connection operation to augment the supplementation of liver edge detail information, suggesting that H-DenseUnet networks are used for liver and liver tumor segmentation. Feng et al propose CPFNet, which adopts a double pyramid network to fuse the global multi-scale context information of the medical image, thereby improving the precision of medical image segmentation.
Although these networks described above enable a high precision segmentation of the liver, how to define the smoothness of the liver contours has been reported to date. On one hand, since the accurate segmentation of the liver is difficult to realize by the traditional image segmentation method, the three-dimensional modeling error of the liver is large, and researchers are difficult to evaluate the smoothness of the liver surface. On the other hand, there is always no uniform metric for the smoothness of the liver contours. For both reasons, clinical staff has so far relied on subjective interpretation to roughly estimate liver smoothness. However, with the rapid development of deep convolutional neural networks, high-precision segmentation of the liver has been achieved, which lays a foundation for accurate estimation of liver smoothness.
Disclosure of Invention
The invention provides a method for measuring liver surface smoothness based on differential curvature, which solves the problem of objectively evaluating liver function without quantitative index.
Firstly, preprocessing a liver CT image by utilizing a medical image window adjusting algorithm to enhance the contrast of an abdomen liver region; then, the network model provided by the invention is utilized to obtain the accurate segmentation result of the liver; secondly, acquiring the contour of a liver segmentation result by using an edge detection algorithm; then, selecting a liver region to be evaluated by utilizing man-machine interaction to obtain a liver profile curve to be evaluated; and finally, measuring the smoothness of the liver surface by using curvature calculation of the curve.
The specific technical scheme of the invention is as follows:
firstly, performing data preprocessing on an abdomen CT image, setting a window level window width, and enhancing the contrast of the liver part image; then acquiring a liver segmentation result according to a segmentation model of the U-Net-based cross attention network; secondly, obtaining a contour curve of a liver segmentation result through an edge detection algorithm; further selecting a measuring and calculating area through man-machine interaction; and finally, measuring the smoothness of the liver profile curve to be measured through the differential curvature.
The specific implementation steps are as follows:
step one, image preprocessing: acquiring an abdomen CT image, preprocessing the abdomen CT image by utilizing a Window-level Window adjusting algorithm, and enhancing the contrast ratio of liver organs and other tissues; different pixel value fields in the CT image correspond to different abdominal organ tissues, and the data value fields are normalized to be within the range of [ -200, 250] to obtain a preprocessed image F;
step two, image segmentation: providing a U-Net-based cross attention Network model, namely Cross Attention U-shape Network, CA-UNet, wherein an encoder and a decoder are formed by using conventional convolution, pooling and deconvolution, skip connection adopts a cross attention module, firstly, spatial position relation of a decoding feature map is obtained, normalization is carried out, normalization weight is multiplied by a feature map of an encoding end at the same stage, the obtained feature map is connected to a decoding end to obtain more accurate detail information, the Network model CA-UNet is trained by utilizing the existing liver data set until convergence, and a preprocessed image F is input into the trained model, so that a segmentation result L is output;
step three, edge extraction: respectively calculating the horizontal and vertical brightness difference approximate value D of the segmentation result L by utilizing a Sobel edge detection operator x And D y Then combining the results obtained in the two directionsBinarizing the gradient image to obtain a binary profile of the liver image>
Step four, selecting a region to be measured: selecting liver region to be detected by human-computer interaction mode, and forming a binary profileObtaining a liver profile curve E to be evaluated;
step five, calculating smoothness: the smoothness of the liver profile E is measured by using curvature calculation of a curve, firstly, an interval N is set, the curve is sampled at equal intervals, then the slope of tangent lines of each sampling point is calculated, the included angle delta theta between the tangent lines is calculated through the slope, then the included angle delta theta is divided by 180, the angle delta theta is normalized to be between [0,1], and finally, the mean square error of all included angles is obtained, so that the smoothness S is obtained.
In the fifth step, the specific method for calculating the smoothness comprises the following steps:
(a) Firstly, sampling a liver contour curve E, and taking the first pixel point of the edge as an initial coordinate point x 1 Sampling is carried out by taking the interval as N, wherein N is the pixel length L of the region to be detected P 1/20 of (C);
(b) Second, calculate the slope K of the curve at the sampling point i Connecting the sampling point with an adjacent point to serve as a tangent, and calculating a slope formula as follows:
wherein x is i And y i Coordinate value x representing the ith pixel point i+1 And y i+1 Coordinate values indicating the i+1th point.
(c) Through slope K i And calculating an included angle delta theta between the tangent lines of the adjacent sampling points to the liver curve, wherein the calculation formula is as follows:
(d) Finally dividing the included angle of the tangent line by 180, normalizing to [0,1], and obtaining the variance of the included angles of all the tangent lines, wherein the calculation formula of the obtained smoothness S is as follows:
wherein I represents the number of equally spaced sampling points, Δθ m Representing the average of the angles.
In the second step, the feature map F' s The specific formula is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,representation of coding feature map->2-dimensional matrix obtained by transformation, +.>Representation of decoding characteristic map->A 2-dimensional matrix obtained by transformation, epsilon is a parameter which can be learned, f softmax (. Cndot.) represents the normalization function.
Compared with the prior art, the invention has the following advantages:
1. in the aspect of image segmentation, the invention provides a CA-UNet network, and the proposed cross attention module can improve the contour segmentation precision of the liver and solve the problem of low edge precision of a liver segmentation result in a CT image in the prior art.
2. In the aspect of liver smoothness estimation, the invention provides a liver smoothness index, and the smoothness of the liver surface is calculated by adopting an image processing technology and a curve curvature principle. Compared with the prior art which relies on manual subjective evaluation of liver smoothness, the method provided by the invention not only gives objective evaluation results, but also gives specific quantitative indexes, thereby laying a foundation for accurate evaluation of liver functions by clinicians.
Drawings
FIG. 1 is a logic flow diagram of a measurement method of the present invention;
FIG. 2 is a block diagram of a segmented network of the liver image based on a U-Net cross-attention network model of the present invention;
FIG. 3 is a schematic diagram of liver smoothness based on curvature calculation according to the present invention;
FIG. 4 is a flow chart of the calculation of the present invention for measuring liver smoothness;
fig. 5 is a graph of liver smoothness experimental results based on the LITS liver dataset according to the present invention.
Detailed Description
Fig. 1 is a logic flow chart of the measurement method of the present invention, fig. 2 is a structural diagram of the split network of the present invention, fig. 3 is a schematic diagram of the liver smoothness calculation of the present invention, and fig. 4 is a flow chart of the liver smoothness calculation of the present invention.
Aiming at the problems that liver segmentation is inaccurate and no quantifiable index exists clinically on the smoothness of the liver surface, the invention provides a method for measuring the smoothness of the liver surface based on differential curvature, which is specifically described as follows:
(1) Image preprocessing: the method comprises the steps of obtaining an abdomen CT image, preprocessing the abdomen CT image by using a W/L window adjusting algorithm, enhancing the contrast ratio of liver organs and other tissues, and ensuring that the image resolution is 512 x 512, wherein the specific algorithm comprises the following steps:
(a) Firstly, a W/L window adjusting algorithm is adopted to convert a CT image into HU values, and a calculation formula is as follows:
HU＝Pixel_val*Rs+Ri
wherein HU represents the density of different tissues, pixel_val is the Pixel value, the value of Rs is 1, and the value of Ri is-1024.
(b) Calculating the maximum value H of tissue density max And a minimum value H min ：
H max ＝(2*wc-ww)/2.0+0.5
H min ＝(2*wc+ww)/2.0+0.5
Wherein ww (window width) is a window width, here 400HU; wc (window center) is the window level, here taken as 100HU.
(c) HU is mapped to gray scale interval [0, 255]:
where F is the image after windowing preprocessing.
(2) Image segmentation: the CA-UNet liver segmentation model is provided, a cross attention module is adopted to obtain the spatial position relation of a decoding feature map, normalization is carried out, the normalized weight is multiplied by a feature map of an encoding end, and then the normalized weight is connected to the decoding end in parallel through jump connection, the network structure is shown in a figure 2, and the specific implementation mode is as follows:
(a) The network coding stage adopts conventional convolution and pooling, and the decoding stage adopts conventional deconvolution operation.
(b) The jump connection stage adopts cross attention, firstly, the spatial relationship of the decoding characteristic diagram is calculated, then normalization is carried out to obtain a spatial position attention weight diagram, then matrix multiplication is carried out on the spatial position attention weight diagram and the coding characteristic diagram of the same stage, and the result is connected in parallel to the decoding stage to obtain a characteristic diagram F' s The specific formula is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,representation of coding feature map->2-dimensional matrix obtained by transformation, +.>Representation of decoding characteristic map->A 2-dimensional matrix obtained by transformation, epsilon is a learnable parameter,f softmax (. Cndot.) represents the normalization function.
(c) Training a network model and segmenting the liver: firstly, initializing a network model, wherein model training parameters are as follows: learning rate of 1×10 -3 The optimization function is Adam. Training is carried out until the network converges, and then the test image F is input into the network model for prediction, so as to obtain a liver segmentation result graph L.
(3) Edge extraction is carried out on the segmentation result: and extracting the edges of the segmentation result L by utilizing a Sobel edge detection operator, wherein the specific implementation mode is as follows:
(a) Firstly, filtering processing is carried out on a segmentation result L in the horizontal direction by utilizing a Sobel edge detection operator, wherein the specific formula is as follows:
D x ＝[f(x+1，y-1)-f(x-1，y-1)]+2[f(x+1，y)-f(x-1，y]+[f(x+1，y+1)-f(x-1，y+1)]
where f (x, y) is the pixel value of the (x, y) position corresponding to the segmentation result L.
(b) And then filtering the segmentation result L in the vertical direction by utilizing a Sobel edge detection operator, wherein the specific formula is as follows:
D y ＝[f(x-1，y+1)-f(x-1，y-1)]+2[f(x，y+1)-f(x，y-1]+[f(x+1，y+1)-f(x+1，y-1)]
(c) And fusing results in the horizontal direction and the vertical direction by utilizing a Sobel edge detection operator to obtain a contour D of the region to be detected, wherein the specific formula is as follows:
(d) Binarizing the obtained profile D:
wherein t represents a set threshold, t=0, d set by the present invention i Is the gray value of the ith position of the contour D.
(4) Selecting a region of liver contour to be measured: selecting a region needing to calculate smoothness by adopting a man-machine interaction picture frame mode, and corresponding the selected region to the binary profileThe liver contour curve E to be evaluated is obtained as shown in fig. 4.
(5) Smoothness is calculated as shown in fig. 3:
(e) Firstly, sampling a liver contour curve, and taking the first pixel point of the edge as an initial coordinate point x 1 Sampling is carried out by taking the interval as N, the setting of the setting N is adaptively determined according to the size of a picture frame, and the pixel length L of the region to be detected is obtained P 1/20 of (C).
(f) Second, calculate the slope K of the curve at the sampling point i Since the relation of the curve function is unknown, in order to solve the problem that the tangent of the curve is difficult to obtain, an approximation algorithm is adopted to obtain the slope, the sampling point and the adjacent point are connected as the tangent, and the calculation slope formula is as follows:
wherein x is i And y i Coordinate value x representing the ith pixel point i+1 And y i+1 Coordinate values indicating the i+1th point.
(g) Through slope K i And calculating an included angle delta theta between the tangent lines of the adjacent sampling points to the liver curve, wherein the calculation formula is as follows:
(h) Finally dividing the included angle of the tangent line by 180, normalizing to [0,1], and obtaining the variance of the included angles of all the tangent lines, wherein the calculation formula of the obtained smoothness S is as follows:
wherein I represents the number of equally spaced sampling points, Δθ m The smaller the value of S indicates a smoother the curve, and the larger the value of S indicates a rougher the curve.
The effects of the present invention can be further illustrated by the following experiments.
In order to verify the effect of the invention on liver smoothness calculation, experiments were performed on the published LITS data set, and the LITS data set has 131 cases, and the invention adopts 90 cases as training sets, 10 cases as verification sets and 31 cases as test sets. The liver segmentation network model is realized based on a PyTorch 1.6.0 framework, and training and evaluation are carried out on the condition that a CPU is Intel Core i9-9900X 10,3.5GHz,GPU video memory 11GB NVIDIA GeForce RTX 2080Ti. We use the following two criteria to evaluate the superiority of the algorithm,wherein A represents a model prediction result, and B represents a gold standard result. Table 1 shows the results of the comparison of the proposed technique with other techniques, and the DICE index was increased from 93.99% to 93.30% for CAU-Net compared to U-Net. CAU-Net improves the IOU by 0.22 compared to the latest CPFNet. The CAU-Net can be obtained to realize high-precision liver segmentation, and lays a foundation for measuring and calculating smoothness.
TABLE 1 comparison of different segmentation models on LITS liver test set
As shown in fig. 5, the liver surface in fig. (a) was smoother, without lumps, and the smoothness was lower by 0.084 by measuring the smoothness of the curve in the selected region. In the graphs (b) and (c), the liver has some bumps, and the smoothness is high. In the graph (d), the liver surface is smoother, and the smoothness value of the calculation method is lower and is 0.0629. In the graph (e), there was a lump on the liver surface, and the smoothness was 0.2 higher than that in the graph (b). In the graph (e), the liver surface is obviously unsmooth, and the smoothness value is high.
Claims (2)
1. The method for measuring the liver surface smoothness based on the differential curvature is characterized by comprising the following steps of: firstly, preprocessing data of an abdomen CT image, setting window level and window width, and enhancing the contrast of an image of a liver part; then acquiring a liver segmentation result according to a segmentation model of the U-Net-based cross attention network; secondly, obtaining a contour curve of a liver segmentation result through an edge detection algorithm; further selecting a measuring and calculating area through man-machine interaction; finally, measuring the smoothness of the liver profile curve to be measured through the differential curvature; the specific implementation steps are as follows:
step one, image preprocessing:
acquiring an abdomen CT image, preprocessing the abdomen CT image by utilizing a Window-level Window adjusting algorithm, and enhancing the contrast ratio of liver organs and other tissues; different pixel value fields in the CT image correspond to different abdominal organ tissues, and the data value fields are normalized to be within the range of [ -200, 250] to obtain a preprocessed image F;
step two, image segmentation:
providing a U-Net-based cross attention Network model, namely Cross Attention U-shape Network, CA-UNet, wherein an encoder and a decoder are formed by using conventional convolution, pooling and deconvolution, skip connection adopts a cross attention module, firstly, spatial position relation of a decoding feature map is obtained, normalization is carried out, normalization weight and the feature map of the same stage coding end are subjected to matrix multiplication, the obtained feature map is connected to a decoding end in parallel to obtain more accurate detail information, the Network model CA-UNet is trained by utilizing the existing liver data set until convergence, and a preprocessed image F is input into the trained model, so that a segmentation result L is output;
step three, edge extraction:
respectively calculating the horizontal and vertical brightness difference approximate value D of the segmentation result L by utilizing a Sobel edge detection operator x And D y Then combining the results obtained in the two directions Binarizing the gradient image to obtain a binary profile of the liver image>
Step four, selecting a region to be measured:
selecting liver region to be detected by human-computer interaction mode, and forming a binary profileObtaining a liver profile curve E to be evaluated;
step five, calculating smoothness:
the smoothness of the liver contour E is measured by means of curvature calculation of the curve, the interval N is first set,
sampling the curve at equal intervals, calculating the slope of the tangent line of each sampling point, calculating the included angle delta theta between the tangent lines through the slope, dividing the included angle delta theta by 180, normalizing the included angle delta theta between the tangent lines to be between 0 and 1, and finally obtaining the variance of all included angles to obtain the smoothness S;
in the second step, the feature map is marked as a feature map F s ' the specific formula is as follows:
wherein F is e C×N Representation pair coding feature map2-dimensional matrix obtained by transformation, +.>Representation of decoding characteristic map->A 2-dimensional matrix obtained by transformation, epsilon is a parameter which can be learned, f softmax (. Cndot.) represents the normalization function.
2. A method for measuring liver surface smoothness based on differential curvature according to claim 1, wherein:
in the fifth step, the specific method for calculating the smoothness comprises the following steps:
(a) Firstly, sampling a liver contour curve, and taking the first pixel point of the edge as an initial coordinate point x 1 Sampling is carried out by taking the interval as N, wherein N is the pixel length L of the region to be detected P 1/20 of (C);
(b) Second, calculate the slope K of the curve at the sampling point i Connecting the sampling point with a point on an adjacent contour curve as a tangent line, and calculating the slope of the tangent line;
(c) Through slope K i And calculating an included angle delta theta between the tangent lines of the adjacent sampling points to the liver curve, wherein the calculation formula is as follows:
(d) Finally dividing the included angle of the tangent line by 180, normalizing to [0,1], obtaining the variance of all included angles,
the calculation formula for obtaining the smoothness S is as follows:
wherein I represents the number of equally spaced sampling points, Δθ m Representing the average of the angles.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110757748.6A CN113487568B (en) | 2021-07-05 | 2021-07-05 | Liver surface smoothness measuring method based on differential curvature |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110757748.6A CN113487568B (en) | 2021-07-05 | 2021-07-05 | Liver surface smoothness measuring method based on differential curvature |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113487568A CN113487568A (en) | 2021-10-08 |
| CN113487568B true CN113487568B (en) | 2023-09-19 |
Family
ID=77940909
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110757748.6A Active CN113487568B (en) | 2021-07-05 | 2021-07-05 | Liver surface smoothness measuring method based on differential curvature |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113487568B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114451870A (en) * | 2022-04-12 | 2022-05-10 | 中南大学湘雅医院 | Pigment nevus malignant change risk monitoring system |
| CN116993719A (en) * | 2023-09-25 | 2023-11-03 | 惠州艺都宇正数码科技有限公司 | OCA optical film surface defect visual detection method and system |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2808844A2 (en) * | 2013-05-31 | 2014-12-03 | Siemens Aktiengesellschaft | Segmentation of a calcified blood vessel |
| CN105631885A (en) * | 2016-01-06 | 2016-06-01 | 复旦大学 | Method for extracting glisson's capsule line and describing characteristics based on superficial tangent plane ultrasonic image |
| CN110503649A (en) * | 2019-07-05 | 2019-11-26 | 陕西科技大学 | One kind is based on Multi scale U-net and the modified liver segmentation method of super-pixel |
| CN111145183A (en) * | 2019-12-30 | 2020-05-12 | 中南大学 | Segmentation system and method for transparent separation cavity ultrasonic image |
| CN111583285A (en) * | 2020-05-12 | 2020-08-25 | 武汉科技大学 | Liver image semantic segmentation method based on edge attention strategy |
| CN111951235A (en) * | 2020-07-31 | 2020-11-17 | 湘潭大学 | Skin image processing method based on deep learning |
| CN112184748A (en) * | 2020-09-30 | 2021-01-05 | 陕西科技大学 | Deformable context coding network model and segmentation method for liver and liver tumor |
| CN112561860A (en) * | 2020-11-23 | 2021-03-26 | 重庆邮电大学 | BCA-UNet liver segmentation method based on prior shape constraint |
| CN112927255A (en) * | 2021-02-22 | 2021-06-08 | 武汉科技大学 | Three-dimensional liver image semantic segmentation method based on context attention strategy |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7995810B2 (en) * | 2005-06-24 | 2011-08-09 | The University Of Iowa Research Foundation | System and methods for image segmentation in n-dimensional space |
-
2021
- 2021-07-05 CN CN202110757748.6A patent/CN113487568B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2808844A2 (en) * | 2013-05-31 | 2014-12-03 | Siemens Aktiengesellschaft | Segmentation of a calcified blood vessel |
| CN105631885A (en) * | 2016-01-06 | 2016-06-01 | 复旦大学 | Method for extracting glisson's capsule line and describing characteristics based on superficial tangent plane ultrasonic image |
| CN110503649A (en) * | 2019-07-05 | 2019-11-26 | 陕西科技大学 | One kind is based on Multi scale U-net and the modified liver segmentation method of super-pixel |
| CN111145183A (en) * | 2019-12-30 | 2020-05-12 | 中南大学 | Segmentation system and method for transparent separation cavity ultrasonic image |
| CN111583285A (en) * | 2020-05-12 | 2020-08-25 | 武汉科技大学 | Liver image semantic segmentation method based on edge attention strategy |
| CN111951235A (en) * | 2020-07-31 | 2020-11-17 | 湘潭大学 | Skin image processing method based on deep learning |
| CN112184748A (en) * | 2020-09-30 | 2021-01-05 | 陕西科技大学 | Deformable context coding network model and segmentation method for liver and liver tumor |
| CN112561860A (en) * | 2020-11-23 | 2021-03-26 | 重庆邮电大学 | BCA-UNet liver segmentation method based on prior shape constraint |
| CN112927255A (en) * | 2021-02-22 | 2021-06-08 | 武汉科技大学 | Three-dimensional liver image semantic segmentation method based on context attention strategy |
Non-Patent Citations (2)
| Title |
|---|
| Attention U-Net:Learning Where to Look for the Pancreas;Ozan Oktay 等;《arXiv》;20180520;论文摘要，第1-2节 * |
| 面向多模态磁共振脑瘤图像的小样本分割方法;董阳 等;《计算机应用》;20210410;第1049-1054页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113487568A (en) | 2021-10-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xie et al. | Dynamic adaptive residual network for liver CT image segmentation | |
| CN111798462B (en) | Automatic delineation method of nasopharyngeal carcinoma radiotherapy target area based on CT image | |
| CN110930416B (en) | MRI image prostate segmentation method based on U-shaped network | |
| CN110503649B (en) | Liver segmentation method based on spatial multi-scale U-net and superpixel correction | |
| CN113487568B (en) | Liver surface smoothness measuring method based on differential curvature | |
| CN112258530A (en) | Neural network-based computer-aided lung nodule automatic segmentation method | |
| CN111738997A (en) | Method for calculating new coronary pneumonia lesion area ratio based on deep learning | |
| CN107766874B (en) | Measuring method and measuring system for ultrasonic volume biological parameters | |
| CN112263217B (en) | Improved convolutional neural network-based non-melanoma skin cancer pathological image lesion area detection method | |
| Ye et al. | Medical image diagnosis of prostate tumor based on PSP-Net+ VGG16 deep learning network | |
| CN111724357B (en) | Arm bone density measurement method based on digital radiological image and support vector regression | |
| CN112767407A (en) | CT image kidney tumor segmentation method based on cascade gating 3DUnet model | |
| CN112071418B (en) | Gastric cancer peritoneal metastasis prediction system and method based on enhanced CT image histology | |
| CN111652871B (en) | Cornea nerve curvature measurement system and method based on IVCM image | |
| CN111783792A (en) | Method for extracting significant texture features of B-ultrasonic image and application thereof | |
| CN115512110A (en) | Medical image tumor segmentation method related to cross-modal attention mechanism | |
| CN115496720A (en) | Gastrointestinal cancer pathological image segmentation method based on ViT mechanism model and related equipment | |
| CN109919216B (en) | Counterlearning method for computer-aided diagnosis of prostate cancer | |
| CN107909577A (en) | Fuzzy C-mean algorithm continuous type max-flow min-cut brain tumor image partition method | |
| CN114581474A (en) | Automatic clinical target area delineation method based on cervical cancer CT image | |
| CN116309551B (en) | Method, device, electronic equipment and readable medium for determining focus sampling area | |
| CN113269799A (en) | Cervical cell segmentation method based on deep learning | |
| CN103606142B (en) | A kind of elastic strain appraisal procedure based on ultrasonic elastic image and system | |
| CN112950611A (en) | Liver blood vessel segmentation method based on CT image | |
| CN117036288A (en) | Tumor subtype diagnosis method for full-slice pathological image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |