CN110276755A - A kind of knub position positioning system and relevant apparatus - Google Patents
A kind of knub position positioning system and relevant apparatus Download PDFInfo
- Publication number
- CN110276755A CN110276755A CN201910554981.7A CN201910554981A CN110276755A CN 110276755 A CN110276755 A CN 110276755A CN 201910554981 A CN201910554981 A CN 201910554981A CN 110276755 A CN110276755 A CN 110276755A
- Authority
- CN
- China
- Prior art keywords
- image
- data block
- target image
- knub position
- positioning
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- 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
- G06T2207/30096—Tumor; Lesion
Abstract
The application provides a kind of knub position positioning system, comprising: module is obtained, for obtaining target image;Image processing module, for carrying out image preprocessing;Sampling module obtains the sampled images of each amplified different scale, and several pieces data block is extracted from the sampled images of different scale for being amplified sampling operation several times to pretreated target image;Data block processing module is integrated the corresponding location drawing picture of each data block and obtains positioning image for being trained using several pieces data block to DenseNet model;Image output module determines knub position for exporting positioning image, and according to positioning image.Effectively characteristics of image can be reused, reduce number of parameters, thus the speed and precision of lift scheme segmentation.The application also provides a kind of computer readable storage medium and a kind of knub position positioning device, has above-mentioned beneficial effect.
Description
Technical field
This application involves field of medical image processing, in particular to a kind of knub position positioning system and relevant apparatus.
Background technique
With the development of medical technology, the knub position in medical image is not content with and has relied solely on naked eyes point
It distinguishes currently, exist and handle using image processing techniques medical image, and then determines the correlation technique of knub position.So
And currently for the processing of medical image, often poor robustness, processing are coarse, poor for the determination precision of knub position, and
The position of any one tumour and size judgement require extremely accurate in medicine, and otherwise any erroneous judgement or misjudgement will cause not
Appreciable loss.
Summary of the invention
It is swollen that the purpose of the application is to provide a kind of knub position positioning system, a kind of computer readable storage medium and one kind
Tumor position positioning device, can be accurately positioned the knub position in tumor image.
In order to solve the above technical problems, the application provides a kind of knub position positioning system characterized by comprising
Module is obtained, for obtaining target image;
Image processing module, for carrying out image preprocessing to the target image;
Sampling module obtains every time for being amplified sampling operation several times to the pretreated target image
The sampled images of amplified different scale, and several pieces data block is extracted from the sampled images of different scale;
Data block processing module, for data block described in several pieces to be inputted DenseNet model and obtains each data
The corresponding location drawing picture of block integrates the location drawing picture and obtains positioning image;Wherein, the DenseNet model be based on
The DenseNet model of 3DFCN;
Image output module determines knub position for exporting the positioning image, and according to the positioning image.
Wherein, described image processing module includes:
Correcting unit, for carrying out biased field correction to the target image;
Information stick unit, for retaining the marginal information of the target image using anisotropic diffusion filtering;
Unit is denoised, for denoising using Gaussian smoothing to the target image;
Normalization unit is 0 for the pixel value of the target image to be fallen into mean value using gray value normalization, variance
For in 1 normal distribution.
Wherein, the data block processing module includes:
Extraction unit, for extracting key feature from the data block using the feature extractor in convolutional neural networks
Information;
Screening unit is obtained for selecting the key feature information using the down-sampling layer general in the convolutional neural networks
To validity feature information;
Processing unit is used for the target image according to the validity feature information processing, obtains location drawing picture;
Integral unit obtains positioning image for integrating each location drawing picture.
Wherein, the processing unit includes:
Processing subelement will swell for modifying the area pixel in the target image according to the validity feature information
The area pixel of tumor position is designated as 1, and the area pixel of non-knub position is designated as 0, obtains location drawing picture.
Wherein, the integral unit includes:
Subelement is integrated, for integrating each location drawing picture using weighted voting method, obtains positioning image.
Wherein, the data block processing module further include:
Intensive connection submodule, gathers for sorting out the convolutional layer of preset quantity to one.
Wherein, the data block processing module further include:
Gather fixed submodule, for fixing the network parameter of the set and intermediate warp lamination;
Small parameter perturbations submodule, for adjusting the parameter of convolution kernel in convolutional layer in the set.
The application also provides a kind of computer readable storage medium, and calculating is stored on the computer readable storage medium
Machine program, the computer program realize following steps when being executed by processor:
Obtain target image;
Image preprocessing is carried out to the target image;
Sampling operation is amplified to the pretreated target image several times, obtains every time amplified different rulers
The sampled images of degree, and several pieces data block is extracted from the sampled images of different scale;
DenseNet model is trained using data block described in several pieces, integrates the corresponding location drawing picture of each data block
Obtain positioning image;Wherein, the DenseNet model is the DenseNet model based on 3DFCN;
The positioning image is exported, and knub position is determined according to the positioning image.
The application also provides a kind of knub position positioning device, comprising:
Memory, for storing computer program;
Processor realizes following steps when for executing the computer program:
Obtain target image;
Image preprocessing is carried out to the target image;
Sampling operation is amplified to the pretreated target image several times, obtains every time amplified different rulers
The sampled images of degree, and several pieces data block is extracted from the sampled images of different scale;
DenseNet model is trained using data block described in several pieces, integrates the corresponding location drawing picture of each data block
Obtain positioning image;Wherein, the DenseNet model is the DenseNet model based on 3DFCN;
The positioning image is exported, and knub position is determined according to the positioning image.
The application provides a kind of knub position positioning system, comprising: module is obtained, for obtaining target image;At image
Module is managed, for carrying out image preprocessing to the target image;Sampling module, for the pretreated target image
Amplified sampling operation several times, obtains the sampled images of each amplified different scale, and from described in different scale
Several pieces data block is extracted in sampled images;Data block processing module, for utilizing data block described in several pieces to DenseNet
Model is trained, and is integrated the corresponding location drawing picture of each data block and is obtained positioning image;Wherein, the DenseNet model is base
In the DenseNet model of 3DFCN;Image output module, for exporting the positioning image, and it is true according to the positioning image
Determine knub position.
The application is trained using DenseNet model to by pretreated target image, and DenseNet can be more
Effectively characteristics of image is reused, reduces number of parameters, thus the speed and precision of lift scheme segmentation.In addition, will
Different scale network modelings, which carries out integrated fusion, can promote the segmentation precision in brain tumor region.The application also provides a kind of meter
Calculation machine readable storage medium storing program for executing and a kind of knub position positioning device have above-mentioned beneficial effect, and details are not described herein again.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of application for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is a kind of structural schematic diagram of knub position positioning system provided by the embodiment of the present application;
Fig. 2 is the DenseNet model structure schematic diagram that the present invention uses;
Fig. 3 is weighted voting method flow chart provided by the present application.
Specific embodiment
To keep the purposes, technical schemes and advantages of the embodiment of the present application clearer, below in conjunction with the embodiment of the present application
In attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is
Some embodiments of the present application, instead of all the embodiments.Based on the embodiment in the application, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall in the protection scope of this application.
Referring to FIG. 1, Fig. 1 is a kind of structural schematic diagram of knub position positioning system provided by the embodiment of the present application,
The positioning system includes:
Module is obtained, for obtaining target image;
Image processing module, for carrying out image preprocessing to target image;
Sampling module is amplified every time for being amplified sampling operation several times to pretreated target image
The sampled images of different scale afterwards, and several pieces data block is extracted from the sampled images of different scale;
Data block processing module, for data block described in several pieces to be inputted DenseNet model and obtains each data
The corresponding location drawing picture of block integrates the location drawing picture and obtains positioning image;Wherein, the DenseNet model be based on
The DenseNet model of 3DFCN;
Image output module determines knub position for exporting positioning image, and according to positioning image.
After obtaining module and obtaining target image, it usually needs pre-processed to target image.Herein for specific
Pretreatment mode is not construed as limiting.Preferably, image processing module may include following unit:
Correcting unit, for carrying out biased field correction to target image, to eliminate limitation and the people of different imaging devices
The influence of the unevenness of image grayscale caused by body complexity and obscurity boundary.
Information stick unit, for retaining the marginal information of target image using anisotropic diffusion filtering.
Unit is denoised, for denoising using Gaussian smoothing to target image;
Normalization unit is 0 for the pixel value of target image to be fallen into mean value using gray value normalization, variance 1
Normal distribution in, the efficiency of model training can be improved in this way.
It should be noted that correcting unit, information stick unit, denoising unit and normalization unit respectively have its unique work
With, therefore, actually carry out target image pretreatment when, can also pre-process from four according to real image process demand singly
It is optionally one or several in member to be combined collocation.Specifically, in the process of implementation, can retain according to correcting unit → information
Unit → denoising unit → normalization unit execution sequence executes.
Sampling module needs to execute pretreated target image amplifies sampling operation several times, herein for times magnification
Number and amplification number are not construed as limiting, and usual amplification factor is set as 2 times, and amplification number is typically not greater than 10 times.Then it obtains
Each amplified image, and extract data block.In addition, usually data block is three-dimensional isometric data block.With 2 times of amplifications, put
It is number 4 times big, for data block side length is N, first obtain the data block of the N*N*N of target image;After amplifying for the first time, original is obtained
The data block of N*N*N under 2 times of target image;After second is amplified, the number of the N*N*N under 4 times of former target image is obtained
According to block;After third time is amplified, the data block of the N*N*N under 8 times of former target image is obtained.Although with time of down-sampling operation
Number increases, and training data block can lose certain characteristics of image, but also save more spatial informations.
Data block processing module is trained DenseNet model using obtained data block.It should be noted that should
DenseNet model is the DenseNet model based on 3DFCN.FCN will be after CNN (Convolution neural network)
Full articulamentum be replaced with convolutional layer, therefore referred to as full convolutional neural networks.
It is worth noting that, DenseNet model can use normal brain parenchym and brain tumor core by transfer learning
Magnetic resonance image is trained and optimizes to model jointly, the problem of training sample scarcity, pole when effective solution model training
The big segmentation performance and generalization ability for improving model.
Then as a preferred embodiment, the data block processing module may include:
Extraction unit is believed for extracting key feature from data block using the feature extractor in convolutional neural networks
Breath;
Screening unit obtains effective spy for selecting key feature information using the down-sampling layer general in convolutional neural networks
Reference breath;
Processing unit, for obtaining location drawing picture according to validity feature information processing target image;
Integral unit obtains positioning image for integrating each position image.
CNN mainly has input layer, convolutional layer, down-sampling layer (also known as pond layer), full articulamentum and output layer to constitute.Convolution
Layer and down-sampling layer constitute a feature extractor, and the high level that layer-by-layer cascade feature extractor constantly deepens to obtain image is special
Sign.Include several convolution kernels, convolution kernel and input picture convolution in convolutional layer, extracts the local feature of image, obtain the layer
Characteristic spectrum (Feature Map, FM).Weight coefficient in convolution kernel is initially provided by netinit, in network
Training process is constantly middle to be constantly updated until terminating.The data characteristics that different convolution kernels extracts is different, convolution kernel
Quantity is more, and the data characteristics extracted is also more.
In network training process, if the feature of image is only extracted by convolutional layer, when encountering larger-size image
Hour operation quantity can be very huge, and the speed of network training is also relatively slow.In order to reduce operand, net training time, convolution are reduced
Neural network is connected to one layer of down-sampling layer (Subsample Layer) behind convolutional layer to reduce data volume.Down-sampling layer
It can reduce the resolution ratio of image, reduce data operation quantity, while network can be enhanced to the adaptability of image change.
The data that data block processing module uses are the data block of different scale size, thus after training each scale number
According to the corresponding tension position image of block, trained obtained tumor image position under the scale is indicated, so it is easy to understand that do not have to ruler
Obtaining picture position under degree, there may be little bit differents, therefore the tumor image position under different scale size is integrated,
Can preferably knub position marginal information, the position where accurate tumour.
Image output module, for exporting positioning image, positioning image is integrated to obtain according to location drawing picture is not had to, but
It is onesize to position the usual Ying Yuyuan target image of image, convenient for comparing positioning image and target image directly to obtain tumour position
It sets.
The application passes through the multi-scale image that will acquire by DenseNet model training first, is conducive to each model and mentions
Take the image feature information and spatial information of different levels.Finally according to pixel fusion, redundancy is eliminated, is mentioned to greatest extent
Take out effective information, the comprehensive image at high quality, to obtain optimal segmentation result.Secondly, passing through building DenseNet
Model extracts as base parted pattern, using FCN and learns image key feature information, is conducive to training for promotion speed and promotion
Model segmentation performance.
Based on the above embodiment, as preferred embodiment, which may include:
Subelement is handled, for modifying the area pixel in target image according to validity feature information, by knub position
Area pixel is designated as 1, and the area pixel of non-knub position is designated as 0, obtains location drawing picture.
The present embodiment is intended to provide a kind of mode of marked tumor regional location, and certainly, those skilled in the art are in this base
There can also be other mark modes on plinth, a different citing limits herein.
Fig. 2 is the DenseNet model structure schematic diagram that the present invention uses, wherein " C " indicates same sized image in channel
On be overlapped.Compared with general parted pattern, the DenseNet model that the present invention uses is had the following characteristics that
1. parameter is less: compared with full articulamentum, the full convolutional layer that the present invention uses shares receptive field, energy by convolution kernel
Training parameter is enough greatly reduced.
2. the training time is shorter: since, using short-circuit connection mechanism, gradient disappears when being able to solve model training in model
And the problem of gradient explosion, accelerate the training of model.
3. training effect is more excellent: since model is using intensive connection submodule (Dense Block), the spy that convolutional layer extracts
It obtains sufficiently and reuses, be more advantageous to model learning and extract characteristics of image, to promote network model performance.
4. model robustness is stronger: the training method due to using transfer learning, the training samples number of model are protected
Card, the robustness and generalization ability of model get a promotion.
Based on the above embodiment, as preferred embodiment, data block processing module further include:
Intensive connection submodule, gathers for sorting out the convolutional layer of preset quantity to one.
Preset quantity is not construed as limiting herein, typically 3 layers.
Based on the above embodiment, as preferred embodiment, data block processing module further include:
Gather fixed submodule, for fixed set and the network parameter of intermediate warp lamination;
Small parameter perturbations submodule, for adjusting the parameter for gathering convolution kernel in interior convolutional layer.
Specifically, by taking Fig. 2 as an example, in implementation process, can according to the fixed preceding 3 Dense Bolck of convolution sequence and
The network parameter of intermediate warp lamination, continues to train using the nuclear magnetic resonance image of brain tumor, finely tunes Dense Block4
The parameter of convolution kernel in interior convolutional layer, thus further lift scheme image characteristics extraction and brain tumor segmentation ability.
Based on the above embodiment, as preferred embodiment, integral unit includes:
Subelement is integrated, for integrating each position image using weighted voting method, obtains positioning image.
Fig. 3 is weighted voting method flow chart provided by the present application, and the present invention is with weighted voting method (Weight Voting) work
For the main method of pixel fusion.Weighted voting method sets each model according to certain principle of similarity, basic segmentation result
Set different weights.The segmentation result obtained when in the present invention, by each model training calculates Dice with original tag and is overlapped
Degree, and sum-average arithmetic, using this as the similarity weight of weighted voting method.Its specific fusion formula are as follows:
Wherein S (x) is the segmentation label in image to be split at pixel x, Li' (x) be i-th of model segmentation result in picture
Label at plain x, c are the index value of segmentation result, it is generally the case that c takes 1 or 0, and respectively indicating is and be not mesh to be split
Mark.Dicei(x) angle value is overlapped for average Dice of i-th of model in training.Image ruler is obtained by repeatedly more times of up-samplings
Very little size carries out pixel fusion further according to segmentation result of the weighted voting method to different models, obtains final segmentation result.
The application also provides a kind of computer readable storage medium, and calculating is stored on the computer readable storage medium
Machine program, the computer program realize following steps when being executed by processor:
Obtain target image;
Image preprocessing is carried out to the target image;
Sampling operation is amplified to the pretreated target image several times, obtains every time amplified different rulers
The sampled images of degree, and several pieces data block is extracted from the sampled images of different scale;
DenseNet model is trained using data block described in several pieces, integrates the corresponding location drawing picture of each data block
Obtain positioning image;Wherein, the DenseNet model is the DenseNet model based on 3DFCN;
The positioning image is exported, and knub position is determined according to the positioning image.
The storage medium may include: USB flash disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random
Access various Jie that can store program code such as memory (Random Access Memory, RAM), magnetic or disk
Matter.
The application also provides a kind of knub position positioning device, comprising:
Memory, for storing computer program;
Processor realizes following steps when for executing the computer program:
Obtain target image;
Image preprocessing is carried out to the target image;
Sampling operation is amplified to the pretreated target image several times, obtains every time amplified different rulers
The sampled images of degree, and several pieces data block is extracted from the sampled images of different scale;
DenseNet model is trained using data block described in several pieces, integrates the corresponding location drawing picture of each data block
Obtain positioning image;Wherein, the DenseNet model is the DenseNet model based on 3DFCN;
The positioning image is exported, and knub position is determined according to the positioning image.
Certainly, so it is easy to understand that the knub position positioning device can also include various network interfaces, the groups such as power supply
Part.It, can also be using various types of CT imaging devices or component etc. when obtaining target image.
Each embodiment is described in a progressive manner in specification, the highlights of each of the examples are with other realities
The difference of example is applied, the same or similar parts in each embodiment may refer to each other.For embodiment provide system and
Speech, since it is corresponding with the method that embodiment provides, so being described relatively simple, related place is referring to method part illustration
?.
Specific examples are used herein to illustrate the principle and implementation manner of the present application, and above embodiments are said
It is bright to be merely used to help understand the present processes and its core concept.It should be pointed out that for the ordinary skill of the art
For personnel, under the premise of not departing from the application principle, can also to the application, some improvement and modification can also be carried out, these improvement
It is also fallen into the protection scope of the claim of this application with modification.
It should also be noted that, in the present specification, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant meaning
Covering non-exclusive inclusion, so that the process, method, article or equipment for including a series of elements not only includes that
A little elements, but also including other elements that are not explicitly listed, or further include for this process, method, article or
The intrinsic element of equipment.In the absence of more restrictions, the element limited by sentence "including a ...", is not arranged
Except there is also other identical elements in the process, method, article or apparatus that includes the element.
Claims (10)
1. a kind of knub position positioning system characterized by comprising
Module is obtained, for obtaining target image;
Image processing module, for carrying out image preprocessing to the target image;
Sampling module is amplified every time for being amplified sampling operation several times to the pretreated target image
The sampled images of different scale afterwards, and several pieces data block is extracted from the sampled images of different scale;
Data block processing module, for data block described in several pieces to be inputted DenseNet model and obtains each data block pair
The location drawing picture answered integrates the location drawing picture and obtains positioning image;Wherein, the DenseNet model is based on 3DFCN's
DenseNet model;
Image output module determines knub position for exporting the positioning image, and according to the positioning image.
2. knub position positioning system according to claim 1, which is characterized in that described image processing module includes:
Correcting unit, for carrying out biased field correction to the target image;
Information stick unit, for retaining the marginal information of the target image using anisotropic diffusion filtering;
Unit is denoised, for denoising using Gaussian smoothing to the target image;
Normalization unit is 0 for the pixel value of the target image to be fallen into mean value using gray value normalization, variance 1
Normal distribution in.
3. knub position positioning system according to claim 1, which is characterized in that the data block processing module includes:
Extraction unit is believed for extracting key feature from the data block using the feature extractor in convolutional neural networks
Breath;
Screening unit is had for selecting the key feature information using the down-sampling layer general in the convolutional neural networks
Imitate characteristic information;
Processing unit is used for the target image according to the validity feature information processing, obtains location drawing picture;
Integral unit obtains positioning image for integrating each location drawing picture.
4. knub position positioning system according to claim 3, which is characterized in that the processing unit includes:
Subelement is handled, for modifying the area pixel in the target image according to the validity feature information, by tumour position
The area pixel set is designated as 1, and the area pixel of non-knub position is designated as 0, obtains location drawing picture.
5. knub position positioning system according to claim 3, which is characterized in that the integral unit includes:
Subelement is integrated, for integrating each location drawing picture using weighted voting method, obtains positioning image.
6. knub position positioning system according to claim 3, which is characterized in that the data block processing module is also wrapped
It includes:
Intensive connection submodule, gathers for sorting out the convolutional layer of preset quantity to one.
7. knub position positioning system according to claim 6, which is characterized in that the data block processing module is also wrapped
It includes:
Gather fixed submodule, for fixing the network parameter of the set and intermediate warp lamination;
Small parameter perturbations submodule, for adjusting the parameter of convolution kernel in convolutional layer in the set.
8. a kind of computer readable storage medium, which is characterized in that be stored with computer on the computer readable storage medium
Program, the computer program realize following steps when being executed by processor:
Obtain target image;
Image preprocessing is carried out to the target image;
Sampling operation is amplified to the pretreated target image several times, obtains each amplified different scale
Sampled images, and several pieces data block is extracted from the sampled images of different scale;
DenseNet model is trained using data block described in several pieces, the corresponding location drawing picture of each data block is integrated and obtains
Position image;Wherein, the DenseNet model is the DenseNet model based on 3DFCN;
The positioning image is exported, and knub position is determined according to the positioning image.
9. a kind of knub position positioning device characterized by comprising
Image collecting device, for obtaining target image and being sent to processor;
Memory, for storing computer program;
The processor, realizes following steps when for executing the computer program:
Image preprocessing is carried out to the target image;The behaviour of amplification sampling several times is carried out to the pretreated target image
Make, obtains the sampled images of each amplified different scale, and extract several pieces from the sampled images of different scale
Data block;DenseNet model is trained using data block described in several pieces, integrates the corresponding location drawing picture of each data block
Obtain positioning image;Wherein, the DenseNet model is the DenseNet model based on 3DFCN;According to the positioning image
Determine knub position.
10. knub position positioning device according to claim 9, and be characterized in that, described image acquisition device is spiral
CT instrument.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910554981.7A CN110276755B (en) | 2019-06-25 | 2019-06-25 | Tumor position positioning system and related device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910554981.7A CN110276755B (en) | 2019-06-25 | 2019-06-25 | Tumor position positioning system and related device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110276755A true CN110276755A (en) | 2019-09-24 |
CN110276755B CN110276755B (en) | 2021-07-06 |
Family
ID=67962323
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910554981.7A Active CN110276755B (en) | 2019-06-25 | 2019-06-25 | Tumor position positioning system and related device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110276755B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021179485A1 (en) * | 2020-03-11 | 2021-09-16 | 平安科技(深圳)有限公司 | Image rectification processing method and apparatus, storage medium, and computer device |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107424145A (en) * | 2017-06-08 | 2017-12-01 | 广州中国科学院软件应用技术研究所 | The dividing method of nuclear magnetic resonance image based on three-dimensional full convolutional neural networks |
CN107680088A (en) * | 2017-09-30 | 2018-02-09 | 百度在线网络技术(北京)有限公司 | Method and apparatus for analyzing medical image |
CN108389214A (en) * | 2018-03-06 | 2018-08-10 | 青岛海信医疗设备股份有限公司 | The processing method and processing device of ultrasonoscopy, electronic equipment, storage medium |
US20180240235A1 (en) * | 2017-02-23 | 2018-08-23 | Zebra Medical Vision Ltd. | Convolutional neural network for segmentation of medical anatomical images |
CN108986067A (en) * | 2018-05-25 | 2018-12-11 | 上海交通大学 | Pulmonary nodule detection method based on cross-module state |
CN109035251A (en) * | 2018-06-06 | 2018-12-18 | 杭州电子科技大学 | One kind being based on the decoded image outline detection method of Analysis On Multi-scale Features |
CN109544534A (en) * | 2018-11-26 | 2019-03-29 | 上海联影智能医疗科技有限公司 | A kind of lesion image detection device, method and computer readable storage medium |
CN109598728A (en) * | 2018-11-30 | 2019-04-09 | 腾讯科技(深圳)有限公司 | Image partition method, device, diagnostic system and storage medium |
CN109598727A (en) * | 2018-11-28 | 2019-04-09 | 北京工业大学 | A kind of CT image pulmonary parenchyma three-dimensional semantic segmentation method based on deep neural network |
CN109685809A (en) * | 2018-12-18 | 2019-04-26 | 清华大学 | A kind of Bile fistula lesion dividing method neural network based and system |
CN109919145A (en) * | 2019-01-21 | 2019-06-21 | 江苏徐工工程机械研究院有限公司 | A kind of mine card test method and system based on 3D point cloud deep learning |
-
2019
- 2019-06-25 CN CN201910554981.7A patent/CN110276755B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180240235A1 (en) * | 2017-02-23 | 2018-08-23 | Zebra Medical Vision Ltd. | Convolutional neural network for segmentation of medical anatomical images |
CN107424145A (en) * | 2017-06-08 | 2017-12-01 | 广州中国科学院软件应用技术研究所 | The dividing method of nuclear magnetic resonance image based on three-dimensional full convolutional neural networks |
CN107680088A (en) * | 2017-09-30 | 2018-02-09 | 百度在线网络技术(北京)有限公司 | Method and apparatus for analyzing medical image |
CN108389214A (en) * | 2018-03-06 | 2018-08-10 | 青岛海信医疗设备股份有限公司 | The processing method and processing device of ultrasonoscopy, electronic equipment, storage medium |
CN108986067A (en) * | 2018-05-25 | 2018-12-11 | 上海交通大学 | Pulmonary nodule detection method based on cross-module state |
CN109035251A (en) * | 2018-06-06 | 2018-12-18 | 杭州电子科技大学 | One kind being based on the decoded image outline detection method of Analysis On Multi-scale Features |
CN109544534A (en) * | 2018-11-26 | 2019-03-29 | 上海联影智能医疗科技有限公司 | A kind of lesion image detection device, method and computer readable storage medium |
CN109598727A (en) * | 2018-11-28 | 2019-04-09 | 北京工业大学 | A kind of CT image pulmonary parenchyma three-dimensional semantic segmentation method based on deep neural network |
CN109598728A (en) * | 2018-11-30 | 2019-04-09 | 腾讯科技(深圳)有限公司 | Image partition method, device, diagnostic system and storage medium |
CN109685809A (en) * | 2018-12-18 | 2019-04-26 | 清华大学 | A kind of Bile fistula lesion dividing method neural network based and system |
CN109919145A (en) * | 2019-01-21 | 2019-06-21 | 江苏徐工工程机械研究院有限公司 | A kind of mine card test method and system based on 3D point cloud deep learning |
Non-Patent Citations (3)
Title |
---|
LELE CHEN等: "MRI tumor segmentation with densely connected 3D CNN", 《PROCEEDINGS OF THE SPIE - PROGRESS IN BIOMEDICAL OPTICS AND IMAGING》 * |
杨晗: "基于深度学习的肺结节检测与诊断研究", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑》 * |
黄奕晖 等: "基于三维全卷积DenseNet的脑胶质瘤MRI分割", 《南方医科大学学报》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021179485A1 (en) * | 2020-03-11 | 2021-09-16 | 平安科技(深圳)有限公司 | Image rectification processing method and apparatus, storage medium, and computer device |
Also Published As
Publication number | Publication date |
---|---|
CN110276755B (en) | 2021-07-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111553406B (en) | Target detection system, method and terminal based on improved YOLO-V3 | |
CN108647681B (en) | A kind of English text detection method with text orientation correction | |
CN108334847B (en) | A kind of face identification method based on deep learning under real scene | |
CN111680706B (en) | Dual-channel output contour detection method based on coding and decoding structure | |
CN105427298B (en) | Remote sensing image registration method based on anisotropic gradient metric space | |
CN103164692B (en) | A kind of special vehicle instrument automatic identification system based on computer vision and method | |
CN110555399B (en) | Finger vein identification method and device, computer equipment and readable storage medium | |
CN108921166A (en) | Medical bill class text detection recognition method and system based on deep neural network | |
CN111291825A (en) | Focus classification model training method and device, computer equipment and storage medium | |
CN106295613A (en) | A kind of unmanned plane target localization method and system | |
CN110287806A (en) | A kind of traffic sign recognition method based on improvement SSD network | |
CN108875739A (en) | A kind of accurate detecting method of digital displaying meter reading | |
CN115240081B (en) | Method and device for detecting full element change of remote sensing image | |
CN110263790A (en) | A kind of power plant's ammeter character locating and recognition methods based on convolutional neural networks | |
CN107506769A (en) | A kind of extracting method and system of urban water-body information | |
CN111709929A (en) | Lung canceration region segmentation and classification detection system | |
CN107256378A (en) | Language Identification and device | |
CN110826534B (en) | Face key point detection method and system based on local principal component analysis | |
CN113706562A (en) | Image segmentation method, device and system and cell segmentation method | |
Zhou et al. | MSAR‐DefogNet: Lightweight cloud removal network for high resolution remote sensing images based on multi scale convolution | |
CN110276755A (en) | A kind of knub position positioning system and relevant apparatus | |
CN113610109A (en) | Visible light camouflage target identification method based on magnifier observation effect | |
CN105528791B (en) | A kind of quality evaluation device and its evaluation method towards touch screen hand-drawing image | |
CN103473562B (en) | Automatic training and identifying system for specific human body action | |
CN115588196A (en) | Pointer type instrument reading method and device based on machine vision |
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 |