CN111462049B - Automatic lesion area form labeling method in mammary gland ultrasonic radiography video - Google Patents
Automatic lesion area form labeling method in mammary gland ultrasonic radiography video Download PDFInfo
- Publication number
- CN111462049B CN111462049B CN202010159426.7A CN202010159426A CN111462049B CN 111462049 B CN111462049 B CN 111462049B CN 202010159426 A CN202010159426 A CN 202010159426A CN 111462049 B CN111462049 B CN 111462049B
- Authority
- CN
- China
- Prior art keywords
- network
- video
- lesion
- module
- data
- 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
- 230000003902 lesion Effects 0.000 title claims abstract description 36
- 238000002372 labelling Methods 0.000 title claims abstract description 15
- 238000002601 radiography Methods 0.000 title claims abstract description 14
- 210000005075 mammary gland Anatomy 0.000 title claims abstract description 7
- 230000000877 morphologic effect Effects 0.000 claims abstract description 21
- 238000000034 method Methods 0.000 claims abstract description 18
- 238000004364 calculation method Methods 0.000 claims abstract description 8
- 230000008569 process Effects 0.000 claims abstract description 7
- 230000008859 change Effects 0.000 claims abstract description 6
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 5
- 238000002604 ultrasonography Methods 0.000 claims description 23
- 210000000481 breast Anatomy 0.000 claims description 13
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 8
- 230000002123 temporal effect Effects 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 4
- 238000012795 verification Methods 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 238000013145 classification model Methods 0.000 claims description 3
- 238000009792 diffusion process Methods 0.000 claims description 3
- 238000001035 drying Methods 0.000 claims description 3
- 238000013508 migration Methods 0.000 claims description 3
- 230000005012 migration Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 2
- 210000001519 tissue Anatomy 0.000 abstract description 6
- 238000013507 mapping Methods 0.000 abstract description 2
- 230000003014 reinforcing effect Effects 0.000 description 6
- 230000006870 function Effects 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002872 contrast media Substances 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000004660 morphological change Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
-
- G06T5/70—
-
- 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
- G06T2207/10016—Video; Image sequence
-
- 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
- G06T2207/20—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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30068—Mammography; Breast
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/03—Recognition of patterns in medical or anatomical images
- G06V2201/032—Recognition of patterns in medical or anatomical images of protuberances, polyps nodules, etc.
Abstract
The invention discloses a method for automatically labeling the lesion area form in a mammary gland ultrasonic radiography video, wherein an end-to-end network model structure is designed, only data to be identified are sent into a model, the model automatically carries out convolution operation on each frame of image, and discrimination characteristics of classification bases are extracted. The range of a lesion area does not need to be manually drawn in the whole identification process, because some lesion morphological characteristics describe contrast changes, such as enhanced intensity, enhanced time sequence and the like, of the related normal tissues and contrast changes of the lesion tissues, the convolution in the convolutional neural network is used for automatically carrying out convolution calculation on the whole contrast video frame sequence, the calculated characteristic values show mapping data of the normal tissues and the lesion areas, and the comparison is carried out according to network rules to obtain results. In addition, morphological characteristics such as crab feet shape, enhancement sequence and the like are used for automatically calculating the characteristics corresponding to the dynamic change of the morphological characteristics of the continuous frames of the video by using the designed network.
Description
Technical Field
The invention relates to the field of medical ultrasonic image data processing, in particular to an automatic lesion form labeling method in a mammary gland ultrasonic radiography video.
Background
Compared with the processing of natural images, the medical ultrasonic image data has poor characteristic learning effect on the images by using common machine learning and deep learning methods due to the characteristics of large noise, low resolution, small data quantity and the like. The invention designs an intelligent marking method, which can realize automatic marking of lesion morphological characteristics in ultrasonic radiography, and marking results can be used for subsequent data analysis, machine learning, data archiving and medical assistance, and have important application value.
The contrast agent is applied to the traditional ultrasonic imaging, so that the reflection of ultrasonic waves can be effectively enhanced, and the ultrasonic images are relatively clear. Professional doctors often need to visually observe dynamic changes of lesions in an ultrasonic contrast video, record characteristic conditions and morphological changes of the lesion parts, provide auxiliary diagnosis information for lesion nodule diagnosis and provide data information for auxiliary disease observation in the subsequent treatment process. When a doctor works, errors caused by visual judgment under subjective experience cannot ensure the accuracy of lesion characteristic morphology recording.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides an automatic lesion form labeling method in a breast ultrasound contrast video, which solves the problems in the background art.
In order to achieve the purpose, the invention provides the following technical scheme: a method for automatically labeling the lesion form in a mammary gland ultrasonic radiography video adopts a convolutional neural network architecture to automatically extract lesion form characteristic parameter information in the ultrasonic radiography video to complete form recognition and classification, and performs lesion labeling on case data, and comprises the following steps:
s1, constructing a breast ultrasound contrast multi-label data set, processing ultrasound contrast video in each case into continuous images by taking breast ultrasound video data provided by a hospital as a unit and storing ultrasound contrast frame sequence images of each case by using a single folder;
s2, preprocessing the data sorted in the step S1, and performing image drying processing by using an edge-enhanced anisotropic diffusion denoising algorithm to remove speckle noise of the ultrasonic image in the data set and keep detail features and edge features in the ultrasonic image;
s3, storing the storage address of each case file of the data dried in the step S2 and the focus morphological feature text of each case recorded by the corresponding doctor as a sample as a json file, wherein each ultrasonic case video contains the morphological type, so that the labels are required to be made into multiple labels for carrying out category marking;
s4, subtracting 127.5 from 3 channels of each pixel of the training data set in the step S3, and dividing by 128 to obtain a normalized ultrasonic film-making video sequence pixel value;
s5, in order to complete form recognition by using an end-to-end classification model, a 3D network is designed, the 3D network uses a residual error network resnet50 and a C3D network as a basic training network, the network comprises 8 modules, each module comprises a convolution layer, the size of the convolution core is 3 multiplied by 3, and the step length is 1 multiplied by 1; after the residual error network resnet50 is pre-trained by using a migration natural image classification task, the sample of the step S4 is transmitted into a network structure to calculate the network model weight parameters, and the detailed calculation steps are as follows;
s6, inputting 16 ultrasound contrast continuous frames with the size of 224 x 224 into the network, and performing sample space-time feature extraction on input data or feature map of an upper network by using the C3D network according to whether the input data is the first module or not;
s7, using a pre-trained residual error network resnet50, receiving input data or feature map of an upper network according to whether the first module is used, extracting spatial feature information, and averaging results of all feature maps to obtain a whole spatial feature residual block of the first module; meanwhile, according to whether the module is the first module, performing time series feature extraction on input data or a feature map of an upper network by using 3 x 1 one-dimensional time convolution;
s8, in 8 modules of the 3D network, the spatial feature information and the time are respectively fused in each moduleThe inter-sequence characteristic information and the space-time characteristic information are specifically as follows: adding the residual block of the temporal and spatial feature map calculated by the features extracted in S7 to the features output by each module in the C3D network as the input of the next module, wherein the calculation formula is shown in formula 1, where Xt represents the input of a network module unit, Xt +1 represents the output of the network module unit, and S (X) ist) Representing a block of spatial feature residues, T (X)t) Representing a temporal characteristic residual block, ST (X)t) Representing the space-time characteristics extracted by the C3D network;
Xt+1＝S(Xt)+T(Xt)+ST(Xt) (1)
s9, adding a full-connection layer to output 4096-dimensional description features at last in the network, then conducting L2 regularization, judging each note by using a SIGMOD function, and finally outputting a prediction result of each morphological feature of a focus in each ultrasound contrast video data;
s10, comparing the prediction result with the real result recorded by the doctor, and calculating and evaluating each label by using a formula (2) to be used as the network identification accuracy, wherein TP, TN, FP and FN are the numbers of predicted true positive, true negative, false positive and false negative respectively;
s11, using BCEloss as a loss function to train and constrain, repeating S6-S10 to train the network until loss convergence and store the model;
and S12, inputting the weight of the trained model by using the verification part in the data set to obtain an automatic recognition result and accuracy.
Preferably, the lesion morphological feature text in the step S3 includes the following 6 types: the intensity is enhanced, the time phase is enhanced, the sequence is enhanced, the enhancement is uniform, the shape is regular after the enhancement, and the crab foot shape is enhanced, each shape type has a value range label, the true value is obtained by evaluating the ultrasonic radiography characteristics of the breast lesion by 2 high-age sonographers, the characteristics are important lesion shape description information, and the change before and after the treatment can be observed in the treatment process to assist the observation of the state of an illness.
Preferably, the tag class flag in step S3 is obtained by using a unique hot coding method to set the tag attribute value field in each type to 1, and set the tag attribute value field in the absence of the tag attribute field to 0. And the data set is divided into 6: 2: 2 into training set, test set and validation set 3.
The invention has the beneficial effects that: the invention designs an end-to-end network model structure, only data to be identified is sent into the model, the model automatically carries out convolution operation on each frame of image, and the distinguishing characteristics of classification bases are extracted. The range of a lesion area does not need to be manually drawn in the whole identification process, because some lesion morphological characteristics describe contrast changes, such as enhanced intensity, enhanced time sequence and the like, of the related normal tissues and contrast changes of the lesion tissues, the convolution in the convolutional neural network is used for automatically carrying out convolution calculation on the whole contrast video frame sequence, the calculated characteristic values show mapping data of the normal tissues and the lesion areas, and the comparison is carried out according to network rules to obtain results. In addition, morphological characteristics such as crab feet shape, enhancement sequence and the like are used for automatically calculating the characteristics corresponding to the dynamic change of the morphological characteristics of the continuous frames of the video by using the designed network.
Drawings
FIG. 1 is a schematic view of a data sample json file;
FIG. 2 is a diagram of spatiotemporal features in conjunction with a network architecture;
FIG. 3 is a block diagram of a temporal feature and spatial feature residual error module;
FIG. 4 is a diagram illustrating predicted results;
FIG. 5 is a flow chart of a labeling method;
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1-5, the present invention provides a technical solution: a method for automatically labeling the lesion form in a mammary gland ultrasonic radiography video adopts a convolutional neural network architecture to automatically extract lesion form characteristic parameter information in the ultrasonic radiography video to complete form recognition and classification, and performs lesion labeling on case data, and comprises the following steps:
s1, constructing a breast ultrasound contrast multi-label data set, processing the ultrasound contrast video in each case into continuous images by taking the breast ultrasound video data provided by a hospital as a unit and storing the ultrasound contrast frame sequence images of each case by using a single folder;
s2, preprocessing the data sorted in the step S1, and performing image drying processing by using an edge-enhanced anisotropic diffusion denoising algorithm to remove speckle noise of the ultrasonic image in the data set and keep detail features and edge features in the ultrasonic image;
s3, storing the storage address of each case file of the data dried in the step S2 and the focus morphological feature text of each case recorded by the corresponding doctor as a sample as a json file,focus of diseaseThe morphological feature text comprises the types shown in the following table 6, the true values are all obtained by evaluating the ultrasound contrast features of the breast lesions by 2 high-age sonographers, the features are important lesion morphological description information, and the change before and after treatment of the breast lesions can be observed in the treatment process to assist the observation of the disease condition. The modality type and value range labels include:
| type of modality | Value range label |
| Enhanced strength | High, equal, low |
| Enhanced time phase | Fast, synchronous and slow advance |
| Order of enhancement | Centripetal and non-centripetal |
| Enhancing uniformity | Uniform and non-uniform |
| Enhanced morphological rules | Yes, no, difficult to distinguish |
| Crab foot shape | Yes, no |
Since each ultrasound case video contains the above morphological types, it is necessary to make the label multi-label for category labeling. And (3) setting the label attribute value field in each type to be 1 by using a one-hot coding mode, and setting the label attribute value field in each type to be 0 without the attribute. And the data set is divided into 6: 2: 2 into a training set, a test set and a verification set, 3, and the data samples are shown in fig. 1, which shows an example of a set of videos after being serialized, and their label attribute values,
s4, subtracting 127.5 from 3 channels of each pixel of the training data set in the step S3, and dividing by 128 to obtain a normalized ultrasonic radiography video sequence pixel value;
s5, in order to complete form recognition by using an end-to-end classification model, the invention designs a 3D network, a 3D network framework uses a residual error network (resnet50) and a C3D network as a basic training network, the network comprises 8 modules, each module comprises a convolution layer, the size of the convolution kernel is 3 multiplied by 3, the step size is 1 multiplied by 1, the size of the convolution kernel is 2 multiplied by 2, and the step size is 2 multiplied by 2. After a residual error network (resnet50) is pre-trained by using a migration natural image classification task, the sample of the step S4 is transmitted into a network structure to calculate a network model weight parameter, and the detailed calculation steps are as follows;
s6, the input of the network is 16 ultrasound contrast continuous frames with size 224 x 224, and the C3D network is used to extract the sample spatio-temporal features from the input data or the feature map (feature map) of the upper network according to whether it is the first module.
S7, using a pre-trained residual error network (resnet50), receiving input data or a feature map (feature map) of an upper network according to whether the first module is used, extracting spatial feature information, averaging results of all feature maps to be used as a whole spatial feature residual block of the layer, and simultaneously using 3 x 1 one-dimensional time convolution to extract time sequence features of the input data or the feature map (feature map) of the upper network according to whether the first module is used;
s8, in 8 modules of the 3D network, spatial feature information, time series feature information, and spatio-temporal feature information are respectively fused in each module, and the specific method includes: adding the residual block of the temporal and spatial feature map calculated by the features extracted in S7 to the features output by each module in the C3D network as the input of the next module, wherein the calculation formula is shown in formula 1, where Xt represents the input of a network module unit, Xt +1 represents the output of the network module unit, and S (X) ist) Representing a block of spatial feature residues, T (X)t) Representing a temporal characteristic residual block, ST (X)t) Representing spatiotemporal features extracted by a 3D network;
Xt+1＝S(Xt)+T(Xt)+ST(Xt) (1)
fig. 3 shows the combination of temporal features and spatial features and spatio-temporal features extracted by a certain module of the C3D network, which constitute the whole residual module structure.
S9, adding a full-connection layer to output 4096-dimensional description features at last in the network, then conducting L2 regularization, judging each note by using a SIGMOD function, and finally outputting a prediction result of each morphological feature of a focus in each ultrasound contrast video data;
and S10, comparing the prediction result with the real result recorded by the doctor, and calculating and evaluating each label by using the formula (2) to obtain the network identification accuracy. Wherein TP, TN, FP and FN are respectively the number of true positive, true negative, false positive and false negative;
s11, using BCEloss as a loss function to train and constrain, repeating S6-S10 to train the network until loss convergence and store the model;
s12, inputting the trained model weight by using a data set verification part to obtain an automatic recognition result and accuracy, wherein the prediction result is 100100101001010, and the decoding result is as follows: the reinforcing strength is high, the reinforcing time phase is rapid, the reinforcing sequence is non-centripetal, the reinforcing is uniform and uneven, the reinforcing shape is regular if not, and the reinforcing shape is crab-foot-shaped if not.
In the whole identification process, the range of a focus area is not required to be manually drawn, and the designed network is used for automatically calculating the characteristics corresponding to the dynamic change of the form of the space-time characteristics of the continuous frames of the video.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that various changes in the embodiments and/or modifications of the invention can be made, and equivalents and modifications of some features of the invention can be made without departing from the spirit and scope of the invention.
Claims (3)
1. A method for automatically labeling the lesion form in a mammary gland ultrasonic radiography video adopts a convolutional neural network architecture to automatically extract lesion form characteristic parameter information in the ultrasonic radiography video to complete form recognition and classification, and performs lesion labeling on case data, and comprises the following steps:
s1, constructing a breast ultrasound contrast multi-label data set, processing ultrasound contrast video in each case into continuous images by taking breast ultrasound video data provided by a hospital as a unit and storing ultrasound contrast frame sequence images of each case by using a single folder;
s2, preprocessing the data sorted in the step S1, and performing image drying processing by using an edge-enhanced anisotropic diffusion denoising algorithm to remove speckle noise of the ultrasonic image in the data set and keep detail features and edge features in the ultrasonic image;
s3, storing the storage address of each case file of the data dried in the step S2 and the focus morphological feature text of each case recorded by the corresponding doctor as a sample as a json file, wherein each ultrasonic case video contains the morphological type, so that the labels are required to be made into multiple labels for carrying out category marking;
s4, subtracting 127.5 from 3 channels of each pixel of the training data set in the step S3, and dividing by 128 to obtain a normalized ultrasonic film-making video sequence pixel value;
s5, in order to complete form recognition by using an end-to-end classification model, a 3D network is designed, the 3D network uses a residual error network resnet50 and a C3D network as a basic training network, the network comprises 8 modules, each module comprises a convolution layer, the size of the convolution core is 3 multiplied by 3, and the step length is 1 multiplied by 1; after the residual error network resnet50 is pre-trained by using a migration natural image classification task, the sample of the step S4 is transmitted into a network structure to calculate the network model weight parameters, and the detailed calculation steps are as follows;
s6, inputting 16 ultrasound contrast continuous frames with the size of 224 x 224 into the network, and performing sample space-time feature extraction on input data or feature map of an upper network by using the C3D network according to whether the input data is the first module or not;
s7, using a pre-trained residual error network resnet50, receiving input data or feature map of an upper network according to whether the first module is used, extracting spatial feature information, and averaging results of all feature maps to obtain a whole spatial feature residual block of the first module; meanwhile, according to whether the module is the first module, performing time series feature extraction on input data or a feature map of an upper network by using 3 x 1 one-dimensional time convolution;
s8, in 8 modules of the 3D network, spatial feature information, time series feature information, and spatio-temporal feature information are respectively fused in each module, and the specific steps are as follows: adding the residual block of the temporal and spatial feature map calculated by the features extracted in S7 to the features output by each module in the C3D network as the input of the next module, wherein the calculation formula is shown in formula 1, where Xt represents the input of a network module unit, Xt +1 represents the output of the network module unit, and S (X) ist) Representing a block of spatial feature residues, T (X)t) Representing a temporal characteristic residual block, ST (X)t) Representing the space-time characteristics extracted by the C3D network;
Xt+1＝S(Xt)+T(Xt)+ST(Xt) (1)
s9, adding a full-connection layer to output 4096-dimensional description features at last in the network, then conducting L2 regularization, judging each note by using a SIGMOD function, and finally outputting a prediction result of each morphological feature of a focus in each ultrasound contrast video data;
s10, comparing the prediction result with the real result recorded by the doctor, and calculating and evaluating each label by using a formula (2) to be used as the network identification accuracy, wherein TP, TN, FP and FN are the numbers of predicted true positive, true negative, false positive and false negative respectively;
s11, using BCEloss as a loss function to train and constrain, repeating S6-S10 to train the network until loss convergence and store the model;
and S12, inputting the weight of the trained model by using the verification part in the data set to obtain an automatic recognition result and accuracy.
2. The method for automatically labeling the focal zone morphology in the breast ultrasound contrast video according to claim 1, characterized in that: the lesion morphological feature text in the step S3 includes the following 6 types: the intensity is enhanced, the time phase is enhanced, the sequence is enhanced, the enhancement is uniform, the shape is regular after the enhancement, and the crab foot shape is enhanced, each shape type has a value range label, the true value is obtained by evaluating the ultrasonic radiography characteristics of the breast lesion by 2 high-age sonographers, the characteristics are important lesion shape description information, and the change before and after the treatment can be observed in the treatment process to assist the observation of the state of an illness.
3. The method for automatically labeling the focal zone morphology in the breast ultrasound contrast video according to claim 1, characterized in that: the tag category flag in step S3 is obtained by using a one-hot encoding method to set the presence in the tag attribute value field of each type to 1, and set the absence of the attribute to 0, and the data set is represented by 6: 2: 2 into training set, test set and validation set 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010159426.7A CN111462049B (en) | 2020-03-09 | 2020-03-09 | Automatic lesion area form labeling method in mammary gland ultrasonic radiography video |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010159426.7A CN111462049B (en) | 2020-03-09 | 2020-03-09 | Automatic lesion area form labeling method in mammary gland ultrasonic radiography video |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111462049A CN111462049A (en) | 2020-07-28 |
| CN111462049B true CN111462049B (en) | 2022-05-17 |
Family
ID=71684216
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010159426.7A Active CN111462049B (en) | 2020-03-09 | 2020-03-09 | Automatic lesion area form labeling method in mammary gland ultrasonic radiography video |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111462049B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111899848B (en) * | 2020-08-05 | 2023-07-07 | 中国联合网络通信集团有限公司 | Image recognition method and device |
| CN113781440B (en) * | 2020-11-25 | 2022-07-29 | 北京医准智能科技有限公司 | Ultrasonic video focus detection method and device |
| CN112488937B (en) * | 2020-11-27 | 2022-07-01 | 河北工业大学 | Medical image feature enhancement method for segmentation task |
| CN112419396B (en) * | 2020-12-03 | 2024-04-26 | 前线智能科技(南京)有限公司 | Automatic thyroid ultrasonic video analysis method and system |
| CN113239951B (en) * | 2021-03-26 | 2024-01-30 | 无锡祥生医疗科技股份有限公司 | Classification method, device and storage medium for ultrasonic breast lesions |
| CN113593707B (en) * | 2021-09-29 | 2021-12-14 | 武汉楚精灵医疗科技有限公司 | Stomach early cancer model training method and device, computer equipment and storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106991445A (en) * | 2017-04-05 | 2017-07-28 | 重庆大学 | A kind of ultrasonic contrast tumour automatic identification and detection method based on deep learning |
| CN107644419A (en) * | 2017-09-30 | 2018-01-30 | 百度在线网络技术（北京）有限公司 | Method and apparatus for analyzing medical image |
| CN108596069A (en) * | 2018-04-18 | 2018-09-28 | 南京邮电大学 | Neonatal pain expression recognition method and system based on depth 3D residual error networks |
| CN108665456A (en) * | 2018-05-15 | 2018-10-16 | 广州尚医网信息技术有限公司 | The method and system that breast ultrasound focal area based on artificial intelligence marks in real time |
| CN208274569U (en) * | 2017-07-06 | 2018-12-25 | 上海长海医院 | A kind of resistance anti-detection devices for detecting glioblastoma boundary in body |
| CN110148113A (en) * | 2019-04-02 | 2019-08-20 | 成都真实维度科技有限公司 | A kind of lesion target area information labeling method based on tomoscan diagram data |
| CN110349141A (en) * | 2019-07-04 | 2019-10-18 | 复旦大学附属肿瘤医院 | A kind of breast lesion localization method and system |
| CN110378259A (en) * | 2019-07-05 | 2019-10-25 | 桂林电子科技大学 | A kind of multiple target Activity recognition method and system towards monitor video |
| CN110378281A (en) * | 2019-07-17 | 2019-10-25 | 青岛科技大学 | Group Activity recognition method based on pseudo- 3D convolutional neural networks |
| CN110458249A (en) * | 2019-10-10 | 2019-11-15 | 点内（上海）生物科技有限公司 | A kind of lesion categorizing system based on deep learning Yu probability image group |
| CN110781830A (en) * | 2019-10-28 | 2020-02-11 | 西安电子科技大学 | SAR sequence image classification method based on space-time joint convolution |
-
2020
- 2020-03-09 CN CN202010159426.7A patent/CN111462049B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106991445A (en) * | 2017-04-05 | 2017-07-28 | 重庆大学 | A kind of ultrasonic contrast tumour automatic identification and detection method based on deep learning |
| CN208274569U (en) * | 2017-07-06 | 2018-12-25 | 上海长海医院 | A kind of resistance anti-detection devices for detecting glioblastoma boundary in body |
| CN107644419A (en) * | 2017-09-30 | 2018-01-30 | 百度在线网络技术（北京）有限公司 | Method and apparatus for analyzing medical image |
| CN108596069A (en) * | 2018-04-18 | 2018-09-28 | 南京邮电大学 | Neonatal pain expression recognition method and system based on depth 3D residual error networks |
| CN108665456A (en) * | 2018-05-15 | 2018-10-16 | 广州尚医网信息技术有限公司 | The method and system that breast ultrasound focal area based on artificial intelligence marks in real time |
| CN110148113A (en) * | 2019-04-02 | 2019-08-20 | 成都真实维度科技有限公司 | A kind of lesion target area information labeling method based on tomoscan diagram data |
| CN110349141A (en) * | 2019-07-04 | 2019-10-18 | 复旦大学附属肿瘤医院 | A kind of breast lesion localization method and system |
| CN110378259A (en) * | 2019-07-05 | 2019-10-25 | 桂林电子科技大学 | A kind of multiple target Activity recognition method and system towards monitor video |
| CN110378281A (en) * | 2019-07-17 | 2019-10-25 | 青岛科技大学 | Group Activity recognition method based on pseudo- 3D convolutional neural networks |
| CN110458249A (en) * | 2019-10-10 | 2019-11-15 | 点内（上海）生物科技有限公司 | A kind of lesion categorizing system based on deep learning Yu probability image group |
| CN110781830A (en) * | 2019-10-28 | 2020-02-11 | 西安电子科技大学 | SAR sequence image classification method based on space-time joint convolution |
Non-Patent Citations (6)
| Title |
|---|
| "A CNN-based methodology for breast cancer diagnosis using thermal image";Juan Zuluaga-Gomez等;《arxiv》;20191030;第1-19页 * |
| "一种基于语义模型的乳腺钙化病灶标注方法";赵可心等;《生物医学工程学杂志》;20120229;第29卷(第1期);第160-163页 * |
| "基于三维残差稠密网络的人体行为识别算法";郭明祥等;《计算机应用》;20191210;第39卷(第12期);第3482-3489页 * |
| "基于深度学习的癌症计算机辅助分类诊断研究进展";肖焕辉等;《国际医学放射学杂志》;20191231;第42卷(第1期);第22-25页及第58页 * |
| "基于深度学习的视频行为识别方法综述";赵朵朵等;《电信科学》;20191231;第35卷(第12期);第1-13页 * |
| "基于级联生成对抗网络的人脸图像修复";陈俊周等;《电子科技大学学报》;20191130;第48卷(第6期);第910-917页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111462049A (en) | 2020-07-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111462049B (en) | Automatic lesion area form labeling method in mammary gland ultrasonic radiography video | |
| CN110232383B (en) | Focus image recognition method and focus image recognition system based on deep learning model | |
| CN107895367B (en) | Bone age identification method and system and electronic equipment | |
| CN109886273B (en) | CMR image segmentation and classification system | |
| WO2022063199A1 (en) | Pulmonary nodule automatic detection method, apparatus and computer system | |
| EP3770850A1 (en) | Medical image identifying method, model training method, and computer device | |
| CN108133476B (en) | Method and system for automatically detecting pulmonary nodules | |
| Kim et al. | Machine-learning-based automatic identification of fetal abdominal circumference from ultrasound images | |
| Ni et al. | Standard plane localization in ultrasound by radial component model and selective search | |
| CN109754361A (en) | The anisotropic hybrid network of 3D: the convolution feature from 2D image is transmitted to 3D anisotropy volume | |
| Zhang et al. | Intelligent scanning: Automated standard plane selection and biometric measurement of early gestational sac in routine ultrasound examination | |
| CN111553892B (en) | Lung nodule segmentation calculation method, device and system based on deep learning | |
| Oghli et al. | Automatic fetal biometry prediction using a novel deep convolutional network architecture | |
| Sinha et al. | Medical image processing | |
| CN112070119A (en) | Ultrasonic tangent plane image quality control method and device and computer equipment | |
| CN108052909B (en) | Thin fiber cap plaque automatic detection method and device based on cardiovascular OCT image | |
| CN112820399A (en) | Method and device for automatically diagnosing benign and malignant thyroid nodules | |
| CN114092450A (en) | Real-time image segmentation method, system and device based on gastroscopy video | |
| CN110570425B (en) | Pulmonary nodule analysis method and device based on deep reinforcement learning algorithm | |
| CN111754485A (en) | Artificial intelligence ultrasonic auxiliary system for liver | |
| Clark et al. | Developing and testing an algorithm for automatic segmentation of the fetal face from three-dimensional ultrasound images | |
| US20230115927A1 (en) | Systems and methods for plaque identification, plaque composition analysis, and plaque stability detection | |
| CN112862786B (en) | CTA image data processing method, device and storage medium | |
| CN112862785B (en) | CTA image data identification method, device and storage medium | |
| CN113177953B (en) | Liver region segmentation method, liver region segmentation device, electronic equipment and storage medium |
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 |