CN111986216B - RSG liver CT image interactive segmentation algorithm based on neural network improvement - Google Patents
RSG liver CT image interactive segmentation algorithm based on neural network improvement Download PDFInfo
- Publication number
- CN111986216B CN111986216B CN202010907881.0A CN202010907881A CN111986216B CN 111986216 B CN111986216 B CN 111986216B CN 202010907881 A CN202010907881 A CN 202010907881A CN 111986216 B CN111986216 B CN 111986216B
- Authority
- CN
- China
- Prior art keywords
- image
- liver
- neural network
- pixels
- region
- 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
- 230000011218 segmentation Effects 0.000 title claims abstract description 42
- 210000004185 liver Anatomy 0.000 title claims abstract description 38
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 36
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 19
- 230000002452 interceptive effect Effects 0.000 title claims abstract description 16
- 230000006872 improvement Effects 0.000 title description 3
- 238000000034 method Methods 0.000 claims abstract description 31
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 14
- 239000013598 vector Substances 0.000 claims abstract description 10
- 230000000877 morphologic effect Effects 0.000 claims abstract description 7
- 238000003708 edge detection Methods 0.000 claims abstract description 4
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 210000005228 liver tissue Anatomy 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 210000000056 organ Anatomy 0.000 claims description 4
- 206010028980 Neoplasm Diseases 0.000 claims description 3
- 230000003187 abdominal effect Effects 0.000 claims description 3
- 210000004204 blood vessel Anatomy 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000000295 complement effect Effects 0.000 claims description 3
- 230000009849 deactivation Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 description 7
- 230000003993 interaction Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000005295 random walk Methods 0.000 description 2
- 208000003174 Brain Neoplasms Diseases 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000000338 in vitro Methods 0.000 description 1
- 208000019423 liver disease Diseases 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000007170 pathology Effects 0.000 description 1
- 238000001959 radiotherapy Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06T7/13—Edge detection
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
-
- 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
Abstract
The invention provides an improved region growing algorithm based on a one-dimensional convolutional neural network for interactive segmentation of a liver CT image, which takes various information such as gray values, spatial information, different gradient values and the like of pixels into overall consideration as a growing rule through the neural network, so that the stability of the region growing method is improved, and the segmentation capability of the algorithm on an edge complex structure is enhanced. The method comprises the following specific steps: firstly, preprocessing an image, extracting a slice containing liver in a CT image sequence set, and converting the CT image into a gray level image by using a window algorithm; then detecting the image edge, calculating gradient values of pixels under different edge detection operators as the characteristics of the pixels, and forming pixel characteristic vectors; next, constructing a network model, extracting a training data set, and training the network model; and finally, segmenting, taking the trained convolutional neural network model as a growth criterion of a region growing algorithm, clicking a liver region by using a mouse to generate an initial segmentation result, and filling holes by using a morphological method to obtain a final result.
Description
Technical Field
The invention provides an improved region growing algorithm (Region Seeds Growing, RSG) based on a one-dimensional convolutional neural network for interactively segmenting a liver CT image, and the gray value, spatial information, different gradient values and other various information of pixels are comprehensively considered as a growing rule through the neural network, so that the stability of the region growing method is improved, and the segmentation capability of the algorithm on an edge complex structure is enhanced.
Background
CT is an noninvasive organ in-vitro imaging means, has higher imaging speed, higher resolution and better effect, and has become an essential means for clinical diagnosis, and the combination of a visualization technology and medical image analysis has dominant in the diagnosis of liver diseases. By segmenting the liver CT image, extracting liver tissues and obtaining corresponding characteristic information, a doctor can intuitively know the details of the inside of the liver of a patient, and plays a key role in diagnosis and the establishment of a next treatment plan.
Current segmentation methods can be divided into three categories: manual, semi-automatic and fully automatic. Manual segmentation methods are cumbersome, time consuming, and may be affected by inter-observer and intra-observer variability. Each pixel of an image needs to be assigned manually to its class, and although very accurate results can be obtained with this technique, the time required will limit some tasks to be translated into clinical practice. For some tasks, manual segmentation of a single case may take hours. The fully automated method requires no human effort and in the last decades researchers have developed many automated segmentation methods. However, fully automated segmentation methods rarely achieve sufficiently accurate, robust results that are clinically impractical. This is often due to poor image quality (with noise, partial volume effects, artifacts and low contrast), large patient-to-patient variability, uneven appearance from pathology and differences in protocols between clinicians leading to different definitions of a given structural boundary.
To address the limitations of the fully automated segmentation method, the interactive segmentation method is viable in clinical practice because it can provide higher accuracy and robustness in many applications, such as planning radiation therapy of brain tumors. Since providing manual annotations for segmentation is time consuming and laborious, an efficient interactive segmentation method is very important for practical use. The good interactive segmentation method should obtain accurate results with as little user interaction as possible, thereby improving interaction efficiency. Although there are a large number of interactive segmentation methods, most methods require a large amount of user interaction and take a long time for the user, or the learning ability of the underlying model is limited. For example, the widely used ITK-SNAP starts with user-supplied seed pixels or blobs and employs an active contour model for segmentation. It requires a large amount of user interaction initially, and once the initial subdivision is obtained, it is difficult to refine the base model through other user interactions. The SlicSeg accepts user-provided graffiti in a single starting slice to train an online random forest for 3D segmentation, but lacks flexibility to do further user editing. Random Walks and Graph Cuts learn from graffiti and allow users to provide other graffiti for refinement. They used a random walk and Gaussian Mixture Model (GMM) as the base model. However, they require a lot of graffiti to achieve a satisfactory segmentation. The method utilizes a convolutional neural network to improve the growth rule of a conventional region growth algorithm, and can complete interactive generation of segmented images through mouse clicking.
Disclosure of Invention
The invention aims to solve the problems of low accuracy and weak stability of the traditional region growing method for segmenting a liver CT image, proposes to interactively segment the liver CT image by using a region growing algorithm based on the improvement of a one-dimensional convolutional neural network, and comprehensively considers various information such as gray values, spatial information, different gradient values and the like of pixels as a growing rule by using the neural network, and the method comprises the following steps of:
step one: image preprocessing, namely extracting slices containing livers in a CT image sequence set, and converting the CT images into gray-scale images by using a window algorithm (W/L);
step two: detecting the image edge, calculating gradient values of pixels under different edge detection operators as the characteristics of the pixels, and forming pixel characteristic vectors;
step three: constructing a network model, extracting a training data set, and training the network model, wherein the network takes a pair of pixel characteristic vectors as input and takes a correlation coefficient of two pixels as output;
step four: and (3) segmentation, namely taking the trained convolutional neural network model as a growth criterion of a region growing algorithm, clicking a liver region by using a mouse to generate an initial segmentation result, and filling holes by using a morphological method to obtain a final result.
The specific case in the first step is as follows:
(1) Extracting slices:
the dataset comprises an original CT image and a segmentation label in which the practitioner has associated 13 abdominal organs one-to-one with a number, wherein the liver corresponds to a number of 6. Slice T satisfies: start+5< T < end-5. Wherein Start represents the sequence number of the earliest digit 6 in the tag image sequence set, end represents the sequence number of the last digit 6 in the tag image sequence set;
(2) Image conversion:
the value g (i) of the pixel point after being processed by using a Window-level (W/L) Window algorithm is as follows:
wherein:，/>the CT value of liver tissue is typically between 50 and 250, we=200, wl=150.
The specific case in the second step is as follows:
respectively filtering the image by a Sobel operator, a Roberts operator, a Canny operator, a Gabor operator, a sobel_h operator, a sobel_v operator and a robert_neg_diag operator, and taking the obtained value as a characteristic value of the pixel to form a pixel characteristic vector:whereinIs the gray value of the pixel.
The specific case in the third step is as follows:
(1) Extracting data:
defining a value area, and enabling the boundary of the liver to be outwards within a city block distance of 10 pixels:
the region comprises two parts: an internal region of the liver and an external 10-pixel distance region of the liver. Two pixel combinations are selected arbitrarily in the region to form an input sample X of the neural network,the corresponding output tag Y is provided with a plurality of output tags,
(2) Training network model
The last layer of the network model uses a sigmoid activation function to output valuesNormalized to (0, 1), the probability that two pixels are input to the same region: />Wherein Z represents the output value before deactivation; a binary cross entropy function (binary cross entropy) is used as a loss function for the network:
only whenAnd->When the probabilities are equal, the loss is 0, otherwise, the loss is a positive number, and the larger the probability difference is, the larger the loss is.
The specific cases in the fourth step are as follows:
(1) Taking the trained convolutional neural network model as a growth criterion of a region growth algorithm, and f when judging seed pixels 1 Pixel f in four neighborhoods 2 If it is incorporated into the growth area represented by the seed pixel, f 1 、f 2 Is used as the input of the neural network to obtain the output result y ’ When y is ’ >0.9, combining; otherwise, the two components are not combined. Repeating the step until all the pixels in the four fields of seed pixels do not meet the condition; the initial seed pixels are selected through mouse clicking;
(2) Since the liver tissue contains blood vessels, tumors and the like, holes exist in the liver region in the segmentation result. The basic principle of morphological filling of holes is:wherein->Is the starting point of hole filling, B is the structural element used for filling the hole, ">Is the complement of A. Iterative calculation is continuously carried out on the formula until +.>The final filling result is +.>And the union of the boundaries, i.e. the final segmentation result.
The invention also includes such features:
comparing the growth rule of the traditional region growth algorithm, and only comparing the gray value of the adjacent pixels to form a single dimension; the gray value, the spatial information, different gradient values and other various information of the pixels are comprehensively considered as the growth rules through the neural network, so that the stability of the algorithm is improved, and the processing capacity of the algorithm on the edge complex structure is enhanced. Although only the pixels in the region near the liver are trained, the present invention can also effectively segment the untrained region.
Compared with other interactive methods, the interactive method is simple to operate, and the edges of the segmentation result are finer. The invention is suitable for medical image segmentation with single internal structure, and has less obvious segmentation effect on natural images with complex semantics.
Drawings
FIG. 1 is a flow chart of the method of the present invention
FIG. 2 is a diagram of a one-dimensional convolutional neural network architecture
Detailed Description
It will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted. The invention is further described below with reference to the drawings and implementation steps.
The invention provides an improved region growing algorithm based on a one-dimensional convolutional neural network for interactive segmentation of a liver CT image, which takes various information such as gray values, spatial information, different gradient values and the like of pixels into overall consideration as a growing rule through the neural network, so that the stability of the region growing method is improved, and the segmentation capability of the algorithm on an edge complex structure is enhanced.
FIG. 1 is a flow chart of the method of the present invention, which comprises the steps of firstly, image preprocessing, extracting a slice containing liver in a CT image sequence set, and converting the CT image into a gray level image by using a window algorithm; then detecting the image edge, calculating gradient values of pixels under different edge detection operators as the characteristics of the pixels, and forming pixel characteristic vectors; next, constructing a network model, extracting a training data set, and training the network model; and finally, segmenting, taking the trained convolutional neural network model as a growth criterion of a region growing algorithm, clicking a liver region by using a mouse to generate an initial segmentation result, and filling holes by using a morphological method to obtain a final result.
The specific implementation steps are as follows:
step1.1 extraction of the sections:
the dataset comprises an original CT image and a segmentation label in which the practitioner has associated 13 abdominal organs one-to-one with a number, wherein the liver corresponds to a number of 6. Slice T satisfies: start+5< T < end-5. Wherein Start represents the sequence number of the earliest digit 6 in the tag image sequence set, end represents the sequence number of the last digit 6 in the tag image sequence set;
step1.2 image conversion:
the value g (i) of the pixel point after being processed by using a Window-level (W/L) Window algorithm is as follows:
wherein:，/>the CT value of liver tissue is typically between 50 and 250, we=200, wl=150.
Step2 filters the image by a Sobel operator, a Roberts operator, a Canny operator, a Gabor operator, a sobel_h operator, a sobel_v operator and a robert_neg_diag operator respectively, and the obtained value is used as a characteristic value of the pixel to form a pixel characteristic vector:whereinIs the gray value of the pixel.
Step3.1 extraction data:
defining a value area, and enabling the boundary of the liver to be outwards within a city block distance of 10 pixels:
the region comprises two parts: an internal region of the liver and an external 10-pixel distance region of the liver. Two pixel combinations are selected arbitrarily in the region to form an input sample X of the neural network,the corresponding output tag Y is provided with a plurality of output tags,
step3.2 trains the web model:
the last level of the network model uses a sigmoid activation function,will output the valueNormalized to (0, 1), the probability that two pixels are input to the same region: />Wherein Z represents the output value before deactivation; a binary cross entropy function (binary cross entropy) is used as a loss function for the network:
only whenAnd->When the probabilities are equal, the loss is 0, otherwise, the loss is a positive number, and the larger the probability difference is, the larger the loss is.
Step4.1, taking a trained convolutional neural network model as a growth criterion of a region growth algorithm, and f when judging seed pixels 1 Pixel f in four neighborhoods 2 If it is incorporated into the growth area represented by the seed pixel, f 1 、f 2 Is used as the input of the neural network to obtain the output result y ’ When y is ’ >0.9, combining; otherwise, the two components are not combined. Repeating the step until all the pixels in the four fields of seed pixels do not meet the condition; the initial seed pixels are selected through mouse clicking;
step4.2 the liver region in the segmentation result has holes due to blood vessels, tumors and the like contained in the liver tissue. The basic principle of morphological filling of holes is:wherein->Is the starting point of hole filling, B is the structural element used for filling the hole, ">Is the complement of A. Iterative calculation is continuously carried out on the formula until +.>The final filling result is +.>And the union of the boundaries, i.e. the final segmentation result.
Fig. 2 is a diagram of a one-dimensional convolutional neural network architecture. The neural network architecture of the invention is shown in fig. 2, and is similar to a convolutional neural network, a convolutional layer is firstly carried out, two-dimensional input is unidimensionally carried out through a flat layer, the two-dimensional input is transited to a full-connection layer, and finally a constant probability value is output through a sigmoid activation function. But the convolutional layers of the network are different and one-dimensional convolution is used. The step length of the convolution kernel is 1, namely, each convolution, the convolution kernel corresponds to a whole row of the vector, adjacent rows are mutually independent, and cross combination is not carried out.
Claims (5)
1. The liver CT image interactive segmentation algorithm based on the neural network improved RSG region growing algorithm is characterized by comprising the following steps of:
step1: image preprocessing, namely extracting slices containing livers in a CT image sequence set, and converting the CT image into a gray-scale image by using a Window algorithm Window-level;
step2: detecting the image edge, calculating gradient values of pixels under different edge detection operators as the characteristics of the pixels, and forming pixel characteristic vectors;
step3: constructing a network model, extracting a training data set, and training the network model, wherein the network takes a pair of pixel characteristic vectors as input and takes a correlation coefficient of two pixels as output;
step4: dividing, taking the trained convolutional neural network model as a growth criterion of a region growth algorithm, and judging a seed pixel f 1 Pixel f in four neighborhoods 2 If it is incorporated into the growth area represented by the seed pixel, f 1 、f 2 The neural network is input to obtain an output result y ', when y'>0.9, combining; otherwise, not merging; repeating the step until all the pixels in the four fields of seed pixels do not meet the condition; the initial seed pixels are selected through mouse clicking, initial segmentation results are generated by clicking the liver region through a mouse, and holes are filled through a morphological method to obtain final results.
2. The liver CT image interactive segmentation algorithm based on the neural network improved RSG region growing algorithm of claim 1, wherein the specific process in Step1 is as follows:
step1.1 extraction of the sections:
the dataset comprises an original CT image and a segmentation label in which the practitioner has associated 13 abdominal organs one to one with numbers, wherein the liver corresponds to a number of 6, and the slice T satisfies: start+5< T < end-5;
wherein Start represents the sequence number of the earliest digit 6 in the tag image sequence set, end represents the sequence number of the last digit 6 in the tag image sequence set;
step1.2 image conversion:
the value g (i) of the pixel point after being processed by using a Window-level Window algorithm is as follows:
wherein: min=wl-0.5×ww, max=wl+0.5×ww, and CT values of liver tissue are typically between 50 and 250, with ww=200, wl=150.
3. The liver CT image interactive segmentation algorithm based on the neural network improved RSG region growing algorithm of claim 1, wherein the specific process in Step2 is as follows:
the step2.1 filters the image by a Sobel operator, a Roberts operator, a Canny operator, a Gabor operator, a sobel_h operator, a sobel_v operator and a robert_neg_diag operator respectively, and the obtained value is used as a characteristic value of the pixel to form a pixel characteristic vector:
f＝[α 1 ,α 2 ,α 3 ...α 8 ]
where a1 is the gray value of the pixel.
4. The liver CT image interactive segmentation algorithm based on the neural network improved RSG region growing algorithm of claim 1, wherein the specific process in Step3 is as follows:
step3.1 extraction data:
defining a value area, and enabling the boundary of the liver to be outwards within a city block distance of 10 pixels:
disf(p(x 1 ,y 1 ),P(x 2 ,x 2 ))＝|x 1 -x 2 |+|y 1 -y 2 |<10
the region comprises two parts: an inner region of the liver and an outer 10-pixel distance region of the liver; two pixel combinations are selected arbitrarily in the region to form an input sample X of the neural network,
X i ＝[f 1 ,f 2 ]
the corresponding output tag Y is provided with a plurality of output tags,
step3.2 trains the web model:
the final level of the network model uses a sigmoid activation function to output a value y' i Normalized to (0, 1), the probability that two pixels are input to the same region:
wherein Z represents the output value before deactivation; a binary cross entropy function (binary cross entropy) is used as a loss function for the network,
only when y' i And y i When the probabilities are equal, the loss is 0, otherwise, the loss is a positive number, and the larger the probability difference is, the larger the loss is.
5. The liver CT image interactive segmentation algorithm based on the neural network improved RSG region growing algorithm of claim 1, wherein the specific process in Step4 is as follows:
step4.1, because liver tissues contain blood vessels, tumors and the like, holes exist in liver areas in the segmentation result; the basic principle of morphological filling of holes is:
wherein X is 0 Is the starting point of hole filling, B is the structural element used for filling the holes, A c Is the complement of A; iterative calculation is continuously carried out on the formula until X k ＝X k-1 The final filling result is X k And the union of the boundaries, i.e. the final segmentation result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010907881.0A CN111986216B (en) | 2020-09-02 | 2020-09-02 | RSG liver CT image interactive segmentation algorithm based on neural network improvement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010907881.0A CN111986216B (en) | 2020-09-02 | 2020-09-02 | RSG liver CT image interactive segmentation algorithm based on neural network improvement |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111986216A CN111986216A (en) | 2020-11-24 |
| CN111986216B true CN111986216B (en) | 2024-01-12 |
Family
ID=73448212
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010907881.0A Active CN111986216B (en) | 2020-09-02 | 2020-09-02 | RSG liver CT image interactive segmentation algorithm based on neural network improvement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111986216B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117744232A (en) * | 2024-02-19 | 2024-03-22 | 中铁十六局集团第一工程有限公司 | Concrete pouring method for beam joints of steel reinforced concrete columns |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056596A (en) * | 2015-11-30 | 2016-10-26 | 浙江德尚韵兴图像科技有限公司 | Fully-automatic three-dimensional liver segmentation method based on local apriori information and convex optimization |
| CN106127819A (en) * | 2016-06-30 | 2016-11-16 | 上海联影医疗科技有限公司 | Medical image extracts method and the device thereof of vessel centerline |
| CN107016683A (en) * | 2017-04-07 | 2017-08-04 | 衢州学院 | The level set hippocampus image partition method initialized based on region growing |
-
2020
- 2020-09-02 CN CN202010907881.0A patent/CN111986216B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056596A (en) * | 2015-11-30 | 2016-10-26 | 浙江德尚韵兴图像科技有限公司 | Fully-automatic three-dimensional liver segmentation method based on local apriori information and convex optimization |
| CN106127819A (en) * | 2016-06-30 | 2016-11-16 | 上海联影医疗科技有限公司 | Medical image extracts method and the device thereof of vessel centerline |
| CN107016683A (en) * | 2017-04-07 | 2017-08-04 | 衢州学院 | The level set hippocampus image partition method initialized based on region growing |
Non-Patent Citations (2)
| Title |
|---|
| Radiomics based detection and characterization of suspicious lesions on full field digital mammograms;Suhas G. Sapate et al.;《Computer Methods and Programs in Biomedicine》;第1-20页 * |
| 基于全卷积神经网络和动态自适应区域 生长法的红外图像目标分割方法;任志淼;《光电技术及应用》;第564-569页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111986216A (en) | 2020-11-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3639240B1 (en) | A system and computer-implemented method for segmenting an image | |
| Xu et al. | DW-Net: A cascaded convolutional neural network for apical four-chamber view segmentation in fetal echocardiography | |
| Zhao et al. | An overview of interactive medical image segmentation | |
| Kim et al. | Machine-learning-based automatic identification of fetal abdominal circumference from ultrasound images | |
| CN112150428A (en) | Medical image segmentation method based on deep learning | |
| El-Regaily et al. | Lung nodule segmentation and detection in computed tomography | |
| CN113420826B (en) | Liver focus image processing system and image processing method | |
| EP2401719B1 (en) | Methods for segmenting images and detecting specific structures | |
| Lameski et al. | Skin lesion segmentation with deep learning | |
| CN115496771A (en) | Brain tumor segmentation method based on brain three-dimensional MRI image design | |
| CN114972362A (en) | Medical image automatic segmentation method and system based on RMAU-Net network | |
| Shan et al. | SCA-Net: A spatial and channel attention network for medical image segmentation | |
| Bi et al. | Hyper-fusion network for semi-automatic segmentation of skin lesions | |
| H Khan et al. | Classification of skin lesion with hair and artifacts removal using black-hat morphology and total variation | |
| Kriti et al. | A review of Segmentation Algorithms Applied to B-Mode breast ultrasound images: a characterization Approach | |
| CN111986216B (en) | RSG liver CT image interactive segmentation algorithm based on neural network improvement | |
| Banerjee et al. | A CADe system for gliomas in brain MRI using convolutional neural networks | |
| Chatterjee et al. | A survey on techniques used in medical imaging processing | |
| CN116958705A (en) | Medical image classifying system based on graph neural network | |
| Lawankar et al. | Segmentation of liver using marker watershed transform algorithm for CT scan images | |
| Hoori et al. | Automatic Deep Learning Segmentation and Quantification of Epicardial Adipose Tissue in Non-Contrast Cardiac CT scans | |
| Tu et al. | Segmentation of lesion in dermoscopy images using dense-residual network with adversarial learning | |
| Xue et al. | Region-of-interest aware 3D ResNet for classification of COVID-19 chest computerised tomography scans | |
| CN110706209B (en) | Method for positioning tumor in brain magnetic resonance image of grid network | |
| Hatture et al. | Clinical diagnostic systems based on machine learning and deep learning |
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 | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20231207 Address after: No. 333 Wuxi Avenue, Wuxi City, Jiangsu Province, 214000 Applicant after: Wuxi University Address before: 130012 No. 2055 Yan'an Street, Chaoyang District, Changchun City, Jilin Province Applicant before: Changchun University of Technology |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |