CN111160413B - Thyroid nodule classification method based on multi-scale feature fusion - Google Patents
Thyroid nodule classification method based on multi-scale feature fusion Download PDFInfo
- Publication number
- CN111160413B CN111160413B CN201911271119.1A CN201911271119A CN111160413B CN 111160413 B CN111160413 B CN 111160413B CN 201911271119 A CN201911271119 A CN 201911271119A CN 111160413 B CN111160413 B CN 111160413B
- Authority
- CN
- China
- Prior art keywords
- network
- residual
- data set
- scale
- convolution
- 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
- 230000004927 fusion Effects 0.000 title claims abstract description 36
- 208000009453 Thyroid Nodule Diseases 0.000 title claims abstract description 32
- 208000024770 Thyroid neoplasm Diseases 0.000 title claims abstract description 28
- 238000000034 method Methods 0.000 title claims abstract description 21
- 238000002604 ultrasonography Methods 0.000 claims abstract description 8
- 238000004140 cleaning Methods 0.000 claims abstract description 6
- 230000000694 effects Effects 0.000 claims abstract description 6
- 238000012545 processing Methods 0.000 claims abstract description 6
- 210000001685 thyroid gland Anatomy 0.000 claims abstract description 6
- 238000010276 construction Methods 0.000 claims abstract description 3
- 238000004364 calculation method Methods 0.000 claims description 11
- 230000017531 blood circulation Effects 0.000 claims description 7
- 230000002093 peripheral effect Effects 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 238000003384 imaging method Methods 0.000 claims description 6
- 230000035945 sensitivity Effects 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 2
- 238000013145 classification model Methods 0.000 claims description 2
- 238000011156 evaluation Methods 0.000 claims description 2
- 238000013507 mapping Methods 0.000 claims description 2
- 230000009466 transformation Effects 0.000 claims description 2
- 238000013461 design Methods 0.000 abstract description 3
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000003745 diagnosis Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000003211 malignant effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 210000000750 endocrine system Anatomy 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 201000002510 thyroid cancer Diseases 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10132—Ultrasound image
-
- 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
- G06T2207/30096—Tumor; Lesion
Abstract
The invention relates to a thyroid nodule classification method based on multi-scale feature fusion, which is characterized by comprising the following steps of: the method comprises the following steps: 1) Acquiring an original thyroid ultrasonic image data set, and processing each ultrasonic image; 2) Cleaning the original ultrasonic image data set, and removing the images which do not meet the requirements to obtain a data set containing 2000 Gao Zhiliang thyroid nodule ultrasonic images; 3) Thyroid nodule ultrasound image classification network construction based on residual network; 4) Replacing the residual error module with a multi-scale fusion module; 5) Adding a high-resolution channel on the basis of a residual error network; 6) And analyzing the classification effect of the network model based on the multi-scale feature fusion and the high-resolution channel. The invention has scientific and reasonable design, designs a mechanism combining multi-scale characteristics and high-resolution channels, and improves the network performance.
Description
Technical Field
The invention belongs to the field of deep learning and medical image processing, relates to a data cleaning technology of a thyroid nodule ultrasonic image data set and a convolutional neural network technology, and particularly relates to a thyroid nodule classification method based on multi-scale feature fusion.
Background
Thyroid nodules are common diseases of the endocrine system, carried by 18% of adults. Although most thyroid nodules are benign, 10% of patients develop malignant thyroid nodules, primarily thyroid cancer. Ultrasound diagnosis is a common means of examining thyroid nodules for benign and malignant properties. However, since doctors generally rely on subjective judgment, there is a lack of objective criteria, and errors are liable to occur.
Deep learning, and in particular convolutional neural networks, has demonstrated its effectiveness in solving the problem of practical imaging. On one hand, the convolutional neural network can extract features from medical images through a multi-layer network, and the accuracy of the features far exceeds that of other methods on related medical problems is obtained by utilizing the features; on the other hand, the deep learning method based on the medical image can efficiently assist the imaging doctor, and greatly lighten the workload of the doctor.
The former work has mostly utilized convolutional neural networks for feature extraction or fine tuning on ImageNet pairs. These above approaches ignore the importance of end-to-end training and do not design network structures for the features of thyroid nodule ultrasound images.
By searching for the published patent documents, a published patent document similar to the present patent application is not found.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a thyroid nodule classification method based on multi-scale feature fusion.
The invention solves the technical problems by the following technical proposal:
a thyroid nodule classifying method based on multi-scale feature fusion is characterized by comprising the following steps: the method comprises the following steps:
1) Acquiring an original thyroid ultrasonic image data set, and processing each ultrasonic image;
2) Cleaning the original ultrasonic image data set, and removing the images which do not meet the requirements to obtain a data set containing 2000 Gao Zhiliang thyroid nodule ultrasonic images;
3) Thyroid nodule ultrasound image classification network construction based on residual network;
4) Replacing the residual error module with a multi-scale fusion module;
5) Adding a high-resolution channel on the basis of a residual error network;
6) And analyzing the classification effect of the network model based on the multi-scale feature fusion and the high-resolution channel.
And, the peripheral boundary area in the original ultrasonic image in the step 1) contains privacy information, and the boundary area is cut off.
And, the peripheral boundary area in the original ultrasonic image in the step 1) contains privacy information, the boundary area is cut out, and the color Doppler blood flow image exists in the original ultrasonic image data set in the step 2), and the color Doppler blood flow image needs to be screened out by using a color operator and is divided into a training set and a test set.
And, the peripheral boundary area in the original ultrasonic image in the step 1) contains privacy information, the boundary area is cut out, and the step 3) uses the data set in the step 2), and the end-to-end training is based on a network model of a residual network.
And, the peripheral boundary area in the original ultrasonic image in the step 1) contains privacy information, the boundary area is cut out, and the step 4) replaces a residual module with a multi-scale information fusion module in the residual network in the step 3), and the module uses hole convolution and instance regularization to perform information fusion.
And, the peripheral boundary area in the original ultrasonic image in the step 1) contains privacy information, the step 5) is to add a high resolution channel based on the residual network in the step 3), and the high resolution channel is internally formed by a multi-scale information fusion module.
The invention has the advantages and beneficial effects that:
1. the thyroid nodule classifying method based on the multi-scale feature fusion provides a brand new high-resolution thyroid nodule ultrasonic image data set, can help researchers to build a network model to a great extent, and is favorable for the researchers to intuitively evaluate work.
2. According to the thyroid nodule classification method based on multi-scale feature fusion, the feature extraction or feature engineering is not needed through an end-to-end training method, so that the training difficulty is greatly reduced; the invention provides a multi-scale information fusion module which can be used for extracting global characteristics irrelevant to image styles and appearance in a shallow layer of a network. The domain adaptability and the robustness of the network are improved to a certain extent; meanwhile, features are extracted jointly by combining batch regularization, so that the classification performance of the network is improved.
3. According to the thyroid nodule classifying method based on multi-scale feature fusion, a plurality of cavity convolutions with different cavity rates are used in parallel in the module, more multi-scale context information is captured while the network receptive field is improved, and a plurality of multi-scale information is fused through an information fusion mechanism, so that network new energy is improved on the basis of not greatly improving network parameters.
4. According to the thyroid nodule classifying method based on multi-scale feature fusion, a high-resolution channel is designed on the basis of a residual error network to keep high-resolution information, and meanwhile, a multi-scale feature fusion module is combined, so that richer multi-scale feature information can be obtained, and the performance of the network is greatly improved.
Drawings
FIG. 1 is a flow chart of the classification method of the present invention;
FIG. 2 is a block diagram of a multi-scale feature fusion module of the present invention;
FIG. 3 is a schematic diagram of a high resolution channel according to the present invention.
Detailed Description
The invention is further illustrated by the following examples, which are intended to be illustrative only and not limiting in any way.
A thyroid nodule classifying method based on multi-scale feature fusion is characterized by comprising the following steps: the method comprises the following steps:
1) Acquiring an original thyroid ultrasonic image data set, and processing each ultrasonic image;
2) Removing additional marks of thyroid ultrasonic image boundaries, such as machine model, diagnosis time, ultrasonic probe emission frequency, hospital name and other additional marks which are irrelevant to the diagnosis of the nodules, so that the influence of ultrasonic image peripheral information on the judgment of benign and malignant thyroid nodules is avoided;
3) Cleaning the original ultrasonic image data set, and removing the images which do not meet the requirements to obtain a data set containing 2000 Gao Zhiliang thyroid nodule ultrasonic images;
4) For an ultrasonic image containing N pixels, it is assumed that the R, G, B three-channel values of the pixel P (i, j) are respectivelyAnd->Variance sigma of R, G, B three-channel value of pixel point P (i, j) 2 The calculation formula of (2) is as follows:
to further find the color operator C, the variance σ is used to calculate the color operator C 2 Setting a threshold value eta for the value, and carrying out binarization processing on the pixel point P (i, j) according to the following formula;
4) In the color operator C, a point with a pixel value of 255 is a region of interest. In general, the more points 255 in a graph, the higher the probability that it contains color Doppler flow imaging. Assuming that the total number of pixel points in a binary color operator graph is N and the total number of points with pixel values of 255 is K, the graph is considered to contain color Doppler blood flow imaging when N/K is more than or equal to 0.01.
5) Manually screening out other irrelevant images, and then forming a data set;
6) Constructing a thyroid nodule ultrasonic image classification model based on a residual network: the residual network contains a total of 50 layers, the first of which is a convolution of convolution kernel size 7 x 7, step size 2; followed by a 2 x 2 max pooling layer; then the network is composed of 4 residual groups, which respectively contain 3, 4, 6 and 3 residual modules, each residual module is composed of 1×1, 3×3 and 1×1 convolution layers, each residual module contains identity transformation, and the last full-connection layer outputs a prediction result.
7) The multi-scale feature fusion module is shown in fig. 2. Assuming that the output feature map is obtained after the feature map passes through the first layer 1×1 convolution layer, the output feature map is X, and batch regularization and instance regularization are performed on X respectively. The following is a formula for batch regularization and instance regularization:
X BN ＝BN(X)
X IN ＝IN(X)
8)X BN and X is IN The general convolution calculation formulas for all 3×3 in parallel are:
9) X is to be BN And X is IN Same as X BN1 And X is IN1 After feature fusion, 3×3 hole convolution with a hole rate of 2 is performed, and the calculation formula is:
10 Adding the obtained four feature maps to obtain an output Y with rich multi-scale information, wherein the calculation formula is as follows:
Y＝X BN1 +X BN2 +X IN1 +X IN2
11 Assuming there is an input feature mapRespectively transmitting the signals to a high-resolution channel, and obtaining an output characteristic map +.>The common resolution channel carries out convolution operation with the step length of 2 on the first convolution layer, the width and the height of the feature mapping are reduced to 1/2, and meanwhile, the number of channels is changed to 2 times of the original number; obtaining characteristic output after passing through the normal resolution channel>Pooling the feature map Y to obtain +.>Finally, the Y' and the Z are spliced byThe following formula calculates the output H, wherein +.>Representing feature stitching.
12 Classifying effect of network model based on multi-scale feature fusion and high-resolution channel is analyzed: the trained network is used for analysis of a test set, and main indexes of evaluation include Accuracy (Accumey), sensitivity (Sensitivity), sensitivity (Specificity) and F1 score, wherein the calculation modes are as follows:
where TP represents true positive, FN represents false negative, FP represents false positive, TN represents true negative.
Table 1 shows the experimental results of the network, from which it can be seen that the network model proposed by the present invention achieves the best results and is optimal.
Table 1 network experimental results table
Aiming at the problem that the current thyroid nodule ultrasonic image diagnosis field lacks available public data sets for researchers to use, the invention introduces a method for collecting, cleaning and screening high-quality thyroid nodule images. A high quality thyroid nodule ultrasound image dataset was collected for use by researchers. Meanwhile, the possible application of the multi-scale feature fusion in the field of thyroid ultrasound image classification is discussed. A multi-scale feature fusion module is designed based on a residual error module of a residual error network, so that an excellent multi-scale feature fusion effect is achieved. In addition, a mechanism of fusing a high-resolution channel and a low-resolution channel is designed on the basis of a residual network structure, so that the classification performance of the network model is improved.
Although the embodiments of the present invention and the accompanying drawings have been disclosed for illustrative purposes, those skilled in the art will appreciate that: various substitutions, changes and modifications are possible without departing from the spirit and scope of the invention and the appended claims, and therefore the scope of the invention is not limited to the embodiments and the disclosure of the drawings.
Claims (1)
1. A thyroid nodule classifying method based on multi-scale feature fusion is characterized by comprising the following steps: the method comprises the following steps:
1) Acquiring an original thyroid ultrasonic image data set, and processing each ultrasonic image;
2) Cleaning the original ultrasonic image data set, and removing the images which do not meet the requirements to obtain a data set containing 2000 Gao Zhiliang thyroid nodule ultrasonic images;
3) Thyroid nodule ultrasound image classification network construction based on residual network;
4) Replacing the residual error module with a multi-scale fusion module;
5) Adding a high-resolution channel on the basis of a residual error network;
6) Analyzing classification effects of a network model based on multi-scale feature fusion and high-resolution channels;
the peripheral boundary area in the original ultrasonic image in the step 1) contains privacy information, and the boundary area is cut off;
the color Doppler blood flow image exists in the original ultrasonic image data set in the step 2), and a color operator is required to screen the color Doppler blood flow image and divide the data set into a training set and a testing set;
the step 3) uses the data set of the step 2), and the end-to-end training is based on a network model of a residual network;
step 4) replacing a residual module with a multi-scale information fusion module in the residual network in step 3), wherein the module uses hole convolution and instance regularization to carry out information fusion;
step 5) adds a high-resolution channel on the basis of the residual error network in step 3), wherein the high-resolution channel is internally formed by a multi-scale information fusion module;
specifically, for an ultrasound image containing N pixels, it is assumed that the R, G, B three-channel values of the pixel P (i, j) are respectivelyAnd->Variance sigma of R, G, B three-channel value of pixel point P (i, j) 2 The calculation formula of (2) is as follows:
to further find the color operator C, the variance σ is used to calculate the color operator C 2 The threshold value eta is set according to the following formulaBinarizing the pixel point P (i, j);
in the color operator C, the point with the pixel value of 255 is an interested region, the more the total number of the points of 255 in one image is, the higher the probability of containing color Doppler blood flow imaging is, and if the total number of the pixel points in one binary color operator image is N and the total number of the points with the pixel value of 255 is K, the image is considered to contain color Doppler blood flow imaging when N/K is more than or equal to 0.01;
constructing a thyroid nodule ultrasonic image classification model based on a residual network: the residual network contains a total of 50 layers, the first of which is a convolution of convolution kernel size 7 x 7, step size 2; followed by a 2 x 2 max pooling layer; then the network is composed of 4 residual groups, which respectively contain 3, 4, 6 and 3 residual modules, each residual module is composed of 1×1, 3×3 and 1×1 convolution layers, each residual module contains identical transformation, and the last full-connection layer outputs a prediction result;
assuming that the output feature map is X after the feature map passes through the first layer 1×1 convolution layer, performing batch regularization and instance regularization on X respectively, where the following is a formula of batch regularization and instance regularization:
X BN ＝BN(X)
X IN ＝IN(X)
X BN and X is IN The general convolution calculation formulas for all 3×3 in parallel are:
x is to be BN And X is IN Same as X BN1 And X is IN1 After feature fusion, 3×3 hole convolution with a hole rate of 2 is performed, and the calculation formula is:
and adding the obtained four feature maps to obtain an output Y with rich multi-scale information, wherein the calculation formula is as follows:
Y＝X BN1 +X BN2 +X IN1 +X IN2
assume there is an input feature mapRespectively transmitting the signals to a high-resolution channel, and obtaining an output characteristic map +.>The common resolution channel carries out convolution operation with the step length of 2 on the first convolution layer, the width and the height of the feature mapping are reduced to 1/2, and meanwhile, the number of channels is changed to 2 times of the original number; obtaining characteristic output after passing through the normal resolution channel>Pooling the feature map Y to obtain +.>Finally, the Y' and the Z are spliced, the output H is obtained through calculation according to the following formula, wherein the characteristic is spliced,
classifying effect of the network model based on multi-scale feature fusion and high-resolution channels is analyzed: the trained network is used for analysis of a test set, and main indexes of evaluation include Accuracy (Accumey), sensitivity (Sensitivity), sensitivity (Specificity) and F1 score, wherein the calculation modes are as follows:
where TP represents true positive, FN represents false negative, FP represents false positive, TN represents true negative.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911271119.1A CN111160413B (en) | 2019-12-12 | 2019-12-12 | Thyroid nodule classification method based on multi-scale feature fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911271119.1A CN111160413B (en) | 2019-12-12 | 2019-12-12 | Thyroid nodule classification method based on multi-scale feature fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111160413A CN111160413A (en) | 2020-05-15 |
| CN111160413B true CN111160413B (en) | 2023-11-17 |
Family
ID=70557076
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911271119.1A Active CN111160413B (en) | 2019-12-12 | 2019-12-12 | Thyroid nodule classification method based on multi-scale feature fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111160413B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111724343A (en) * | 2020-05-18 | 2020-09-29 | 天津大学 | Thyroid nodule ultrasonic image data set enhancing method based on antagonistic learning |
| CN112785586A (en) * | 2021-02-04 | 2021-05-11 | 天津大学 | Multi-scale self-attention unsupervised domain self-adaptive algorithm |
Citations (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107610194A (en) * | 2017-08-14 | 2018-01-19 | 成都大学 | MRI super resolution ratio reconstruction method based on Multiscale Fusion CNN |
| CN107680678A (en) * | 2017-10-18 | 2018-02-09 | 北京航空航天大学 | Based on multiple dimensioned convolutional neural networks Thyroid ultrasound image tubercle auto-check system |
| CN108010013A (en) * | 2017-11-03 | 2018-05-08 | 天津大学 | A kind of lung CT image pulmonary nodule detection methods |
| CN108268870A (en) * | 2018-01-29 | 2018-07-10 | 重庆理工大学 | Multi-scale feature fusion ultrasonoscopy semantic segmentation method based on confrontation study |
| CN108537731A (en) * | 2017-12-29 | 2018-09-14 | 西安电子科技大学 | Image super-resolution rebuilding method based on compression multi-scale feature fusion network |
| CN108734659A (en) * | 2018-05-17 | 2018-11-02 | 华中科技大学 | A kind of sub-pix convolved image super resolution ratio reconstruction method based on multiple dimensioned label |
| CN108734660A (en) * | 2018-05-25 | 2018-11-02 | 上海通途半导体科技有限公司 | A kind of image super-resolution rebuilding method and device based on deep learning |
| CN109064405A (en) * | 2018-08-23 | 2018-12-21 | 武汉嫦娥医学抗衰机器人股份有限公司 | A kind of multi-scale image super-resolution method based on dual path network |
| CN109165667A (en) * | 2018-07-06 | 2019-01-08 | 中国科学院自动化研究所 | Based on the cerebral disease categorizing system from attention mechanism |
| CN109242839A (en) * | 2018-08-29 | 2019-01-18 | 上海市肺科医院 | A kind of good pernicious classification method of CT images Lung neoplasm based on new neural network model |
| CN109389556A (en) * | 2018-09-21 | 2019-02-26 | 五邑大学 | The multiple dimensioned empty convolutional neural networks ultra-resolution ratio reconstructing method of one kind and device |
| CN109493333A (en) * | 2018-11-08 | 2019-03-19 | 四川大学 | Ultrasonic Calcification in Thyroid Node point extraction algorithm based on convolutional neural networks |
| CN109598269A (en) * | 2018-11-14 | 2019-04-09 | 天津大学 | A kind of semantic segmentation method based on multiresolution input with pyramid expansion convolution |
| CN110020606A (en) * | 2019-03-13 | 2019-07-16 | 北京工业大学 | A kind of crowd density estimation method based on multiple dimensioned convolutional neural networks |
| CN110047069A (en) * | 2019-04-22 | 2019-07-23 | 北京青燕祥云科技有限公司 | A kind of image detection device |
| CN110059730A (en) * | 2019-03-27 | 2019-07-26 | 天津大学 | A kind of thyroid nodule ultrasound image classification method based on capsule network |
| CN110197468A (en) * | 2019-06-06 | 2019-09-03 | 天津工业大学 | A kind of single image Super-resolution Reconstruction algorithm based on multiple dimensioned residual error learning network |
| CN110276721A (en) * | 2019-04-28 | 2019-09-24 | 天津大学 | Image super-resolution rebuilding method based on cascade residual error convolutional neural networks |
| CN110309880A (en) * | 2019-07-01 | 2019-10-08 | 天津工业大学 | A kind of 5 days and 9 days hatching egg embryo's image classification methods based on attention mechanism CNN |
| KR20190119261A (en) * | 2018-04-12 | 2019-10-22 | 가천대학교 산학협력단 | Apparatus and method for segmenting of semantic image using fully convolutional neural network based on multi scale image and multi scale dilated convolution |
| CN110363134A (en) * | 2019-07-10 | 2019-10-22 | 电子科技大学 | A kind of face blocked area localization method based on semantic segmentation |
| CN110415170A (en) * | 2019-06-24 | 2019-11-05 | 武汉大学 | A kind of image super-resolution method based on multiple dimensioned attention convolutional neural networks |
-
2019
- 2019-12-12 CN CN201911271119.1A patent/CN111160413B/en active Active
Patent Citations (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107610194A (en) * | 2017-08-14 | 2018-01-19 | 成都大学 | MRI super resolution ratio reconstruction method based on Multiscale Fusion CNN |
| CN107680678A (en) * | 2017-10-18 | 2018-02-09 | 北京航空航天大学 | Based on multiple dimensioned convolutional neural networks Thyroid ultrasound image tubercle auto-check system |
| CN108010013A (en) * | 2017-11-03 | 2018-05-08 | 天津大学 | A kind of lung CT image pulmonary nodule detection methods |
| CN108537731A (en) * | 2017-12-29 | 2018-09-14 | 西安电子科技大学 | Image super-resolution rebuilding method based on compression multi-scale feature fusion network |
| CN108268870A (en) * | 2018-01-29 | 2018-07-10 | 重庆理工大学 | Multi-scale feature fusion ultrasonoscopy semantic segmentation method based on confrontation study |
| KR20190119261A (en) * | 2018-04-12 | 2019-10-22 | 가천대학교 산학협력단 | Apparatus and method for segmenting of semantic image using fully convolutional neural network based on multi scale image and multi scale dilated convolution |
| CN108734659A (en) * | 2018-05-17 | 2018-11-02 | 华中科技大学 | A kind of sub-pix convolved image super resolution ratio reconstruction method based on multiple dimensioned label |
| CN108734660A (en) * | 2018-05-25 | 2018-11-02 | 上海通途半导体科技有限公司 | A kind of image super-resolution rebuilding method and device based on deep learning |
| CN109165667A (en) * | 2018-07-06 | 2019-01-08 | 中国科学院自动化研究所 | Based on the cerebral disease categorizing system from attention mechanism |
| CN109064405A (en) * | 2018-08-23 | 2018-12-21 | 武汉嫦娥医学抗衰机器人股份有限公司 | A kind of multi-scale image super-resolution method based on dual path network |
| CN109242839A (en) * | 2018-08-29 | 2019-01-18 | 上海市肺科医院 | A kind of good pernicious classification method of CT images Lung neoplasm based on new neural network model |
| CN109389556A (en) * | 2018-09-21 | 2019-02-26 | 五邑大学 | The multiple dimensioned empty convolutional neural networks ultra-resolution ratio reconstructing method of one kind and device |
| CN109493333A (en) * | 2018-11-08 | 2019-03-19 | 四川大学 | Ultrasonic Calcification in Thyroid Node point extraction algorithm based on convolutional neural networks |
| CN109598269A (en) * | 2018-11-14 | 2019-04-09 | 天津大学 | A kind of semantic segmentation method based on multiresolution input with pyramid expansion convolution |
| CN110020606A (en) * | 2019-03-13 | 2019-07-16 | 北京工业大学 | A kind of crowd density estimation method based on multiple dimensioned convolutional neural networks |
| CN110059730A (en) * | 2019-03-27 | 2019-07-26 | 天津大学 | A kind of thyroid nodule ultrasound image classification method based on capsule network |
| CN110047069A (en) * | 2019-04-22 | 2019-07-23 | 北京青燕祥云科技有限公司 | A kind of image detection device |
| CN110276721A (en) * | 2019-04-28 | 2019-09-24 | 天津大学 | Image super-resolution rebuilding method based on cascade residual error convolutional neural networks |
| CN110197468A (en) * | 2019-06-06 | 2019-09-03 | 天津工业大学 | A kind of single image Super-resolution Reconstruction algorithm based on multiple dimensioned residual error learning network |
| CN110415170A (en) * | 2019-06-24 | 2019-11-05 | 武汉大学 | A kind of image super-resolution method based on multiple dimensioned attention convolutional neural networks |
| CN110309880A (en) * | 2019-07-01 | 2019-10-08 | 天津工业大学 | A kind of 5 days and 9 days hatching egg embryo's image classification methods based on attention mechanism CNN |
| CN110363134A (en) * | 2019-07-10 | 2019-10-22 | 电子科技大学 | A kind of face blocked area localization method based on semantic segmentation |
Non-Patent Citations (2)
| Title |
|---|
| Li Xuewei.Fully Convolutional Networks for Ultrasound Image Segmentation of Thyroid Nodules.《2018 IEEE 20th International Conference on High Performance Computing and Communications》.2018,全文. * |
| 基于多尺度融合的甲状腺结节图像特征提取;王昊;彭博;陈琴;杨燕;;数据采集与处理(05);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111160413A (en) | 2020-05-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109886273B (en) | CMR image segmentation and classification system | |
| CN108446730B (en) | CT pulmonary nodule detection device based on deep learning | |
| CN112489061B (en) | Deep learning intestinal polyp segmentation method based on multi-scale information and parallel attention mechanism | |
| CN112508850B (en) | Deep learning-based method for detecting malignant area of thyroid cell pathological section | |
| CN107610129B (en) | CNN-based multi-modal nasopharyngeal tumor joint segmentation method | |
| CN111160413B (en) | Thyroid nodule classification method based on multi-scale feature fusion | |
| CN108305241B (en) | SD-OCT image GA lesion segmentation method based on depth voting model | |
| EP4198819A1 (en) | Method for detecting and classifying lesion area in clinical image | |
| CN109949297B (en) | Lung nodule detection method based on recection and fast R-CNN | |
| CN108305253A (en) | A kind of pathology full slice diagnostic method based on more multiplying power deep learnings | |
| CN110111895A (en) | A kind of method for building up of nasopharyngeal carcinoma far-end transfer prediction model | |
| CN101403743A (en) | Automatic separating method for X type overlapping and adhering chromosome | |
| CN113768519B (en) | Method for analyzing consciousness level of patient based on deep learning and resting state electroencephalogram data | |
| CN112750132A (en) | White blood cell image segmentation method based on dual-path network and channel attention | |
| CN113343755A (en) | System and method for classifying red blood cells in red blood cell image | |
| CN113538435B (en) | Pancreatic cancer pathological image classification method and system based on deep learning | |
| Fang et al. | A deep neural network for cervical cell classification based on cytology images | |
| CN114140437A (en) | Fundus hard exudate segmentation method based on deep learning | |
| CN111743574B (en) | Method for constructing knowledge graph of breast cancer ultrasonic image high confidence entity relationship | |
| CN117152433A (en) | Medical image segmentation method based on multi-scale cross-layer attention fusion network | |
| CN112200810A (en) | Multi-modal automated ventricular segmentation system and method of use thereof | |
| CN116993699A (en) | Medical image segmentation method and system under eye movement auxiliary training | |
| Verma et al. | Classifying Breast Density in Mammographic Images Using Wavelet-Based and Fine-Tuned Sensory Neural Networks | |
| CN115409812A (en) | CT image automatic classification method based on fusion time attention mechanism | |
| CN114022494A (en) | Automatic segmentation method of traditional Chinese medicine tongue image based on light convolutional neural network and knowledge distillation |
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 |