CN110288615A - A kind of sloped position frame localization method based on deep learning - Google Patents
A kind of sloped position frame localization method based on deep learning Download PDFInfo
- Publication number
- CN110288615A CN110288615A CN201910573065.8A CN201910573065A CN110288615A CN 110288615 A CN110288615 A CN 110288615A CN 201910573065 A CN201910573065 A CN 201910573065A CN 110288615 A CN110288615 A CN 110288615A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- training
- parameter
- sloped position
- position frame
- 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
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10104—Positron emission tomography [PET]
-
- 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
- G06T2207/30008—Bone
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Radiology & Medical Imaging (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention provides a kind of sloped position frame localization method based on deep learning of medical imaging field of locating technology, the DICOM image for including the following steps: step S10, obtaining a large amount of sloped position frames；Step S20, DICOM image is pre-processed；Step S30, pretreated DICOM image input deep learning network is trained；Step S40, the generalization ability for the deep learning network completed to training is verified；Step S50, sloped position frame is positioned based on the deep learning network after being verified.The present invention has the advantages that improving the positioning accuracy of posting.
Description
Technical field
The present invention relates to medical imaging field of locating technology, refers in particular to a kind of sloped position based on deep learning and confine position
Method.
Background technique
Medical imaging is the extremely important branch of modern medicine, emits signal and patient's body by medical imaging devices
The physical mechanisms such as the effect of body tissue show the image of each organ structure inside patient body, and disclosing each organ, whether there is or not lesions, and
Qualitative and quantitative analysis is carried out to lesion region in time, so that effectively auxiliary doctor carries out condition-inference.Medical imaging is with X
Ray computed tomography (CT), Positron emission tomography (PET), Magnetic resonance imaging (MR) etc. are representative, are cured in the modern times
It is with fastest developing speed in imaging technique, it makes one's way in life and also most attracts people's attention, Grade III Class A hospital generally possesses these medicine at present
Imaging device.
Since the figure of different patients, the position of institute's recumbency and position to be scanned are not quite similar, doctor can not be quasi-
Really precognition needs the specific coordinate value scanned, it is therefore desirable to which the scanning for carrying out primary substantially position by posting in advance passes through
The scanning result of posting, doctor carries out manually selecting the suitable size of posting and range, then these parameters are inputted medicine
Imaging device, medical imaging devices determine the length and width of follow up scan, angle, the work such as thickness just with the parameter of these postings
Parameter.Therefore, the selection of posting is particularly significant for the scanography of subsequent medical imaging devices.
Current target detection network is mainly divided to two classes, and one kind is R-CNN system algorithm, it needs first to use heuristic
Or region recommendation network (Region Proposal Network, RPN) generates candidate region, passes through depth on candidate region
Degree learning network extracts area characteristic information, carrys out classification described in critical region further according to these characteristic informations, finally carries out
The fine amendment of candidate region；It is defeated to have benefited from good CNN compared to R-CNN system algorithm for another kind of such as Yolo, SSD scheduling algorithm
The output of mentality of designing out, final full articulamentum is special based on the mixing of image location information+confidence level+picture material classification
Vector is levied, detection target position and classification can be thus placed in the same CNN network, the speed of network training can be accelerated
Degree, it might even be possible to reach the speed of real-time detection, and is only slightly inferior to R-CNN system algorithm in the accuracy of detection, it is very suitable
It closes and needs quickly detection, and the very specific posting image of object.Above-mentioned two classes algorithm of target detection can realize mesh
Target positioning and identification, but the posting that they export all is rectangle frame, without angle information, does not have universality, for
Head, interverbebral disc, the posting that the physical feelings of some angle tilts such as Bones and joints needs are parallelogram, cannot achieve compared with
Good object recognition task.
Therefore, how a kind of sloped position frame localization method based on deep learning is provided, realizes and improves determining for posting
Position precision, becomes a urgent problem to be solved.
Summary of the invention
The technical problem to be solved in the present invention is to provide a kind of sloped position frame localization method based on deep learning,
Realize the positioning accuracy for improving posting.
The present invention is implemented as follows: a kind of sloped position frame localization method based on deep learning, the method includes
Following steps:
Step S10, the DICOM image of a large amount of sloped position frames is obtained；
Step S20, DICOM image is pre-processed；
Step S30, pretreated DICOM image input deep learning network is trained；
Step S40, the generalization ability for the deep learning network completed to training is verified；
Step S50, sloped position frame is positioned based on the deep learning network after being verified.
Further, the step S10 specifically:
Obtain the data set of the DICOM images of a large amount of sloped position frames as training, and by the data set random division
For training set T, verifying collection V and test set U.
Further, the step S20 is specifically included:
It step S21, is the picture format of deep learning network support by each DICOM image format conversion；
Step S22, parameter mark is carried out to the image after format conversion；
Step S23, the parameter after mark is normalized.
Further, in the step S22, the parameter includes the centre coordinate (x, y) of sloped position frame, length h, width
Spend w and horizontal angle theta and classification information.
Further, the step S30 is specifically included:
Step S31, the deep learning network of deep learning is constructed, a stability bandwidth is set, V points of training set T and verifying collection
Input not as the deep learning network is marked the parameter after mark as the training of the deep learning network, right
The network model is trained, and respectively obtains training set loss function curve and verifying collection loss function curve；
Step S32, whether the stability bandwidth of training of judgement collection loss function curve is less than the stability bandwidth of setting, is then to enter
Step S33, it is no, then enter step S31；
Step S33, judge whether verifying collection loss function curve starts to increase, be, then deconditioning to obtain training ginseng
Number, and enter step S34；It is no, then enter step S31；
Step S34, the deep learning network is run with training parameter on test set U, obtains the ginseng of sloped position frame
Number, according to the demand of medical imaging devices reconstruction parameter, is finely adjusted obtained parameter, completes deep learning.
Further, the step S40 specifically:
A similarity is set, the parameter and training label after comparison fine tuning carry out the detection of generalization ability, if after fine tuning
Parameter and training label similarity be greater than setting similarity, then enter step S50；If the parameter and training after fine tuning are marked
The similarity of note is less than or equal to the similarity of setting, then the DICOM image of new increasing hole angle posting, and enters step S20.
The present invention has the advantages that
1, in view of the posting of rectangle is when tilting certain angle, the posting that medical imaging devices scan is to incline
Oblique, as parallelogram greatly improves positioning by the way that sloped position frame input deep learning network to be trained
The positioning accuracy of frame.
2, by the way that sloped position frame input deep learning network to be trained, the positioning accurate of posting is greatly improved
Degree, doctor only need to be finely adjusted posting, greatly improve the working efficiency of doctor, and be avoided as much as possible due to doctor
Experience level, the clinical influence for playing situation, emotional status and physical fatigue degree.
Detailed description of the invention
The present invention is further illustrated in conjunction with the embodiments with reference to the accompanying drawings.
Fig. 1 is a kind of flow chart of the sloped position frame localization method based on deep learning of the present invention.
Fig. 2 is sloped position frame schematic diagram of the present invention.
Fig. 3 is the structure chart of deep learning network of the present invention.
Fig. 4 is the data structure diagram of output layer of the present invention.
Specific embodiment
Please refer to figs. 1 to 4, a kind of preferable reality of the sloped position frame localization method based on deep learning of the present invention
Example is applied, is included the following steps:
Step S10, the DICOM image of a large amount of sloped position frames is obtained；Between each sloped position frame centre coordinate (x,
Y), length h, width w or different from horizontal angle theta；DICOM, that is, digital imaging and communications in medicine, be medical image and
The international standard of relevant information defines the Medical Image Format that can be used for data exchange that quality is able to satisfy clinical needs.
Step S20, DICOM image is pre-processed；
Step S30, pretreated DICOM image input deep learning network is trained；
Step S40, the generalization ability for the deep learning network completed to training is verified；Generalization ability promotes and applies
Ability, whether precision can also meet the requirements after popularization；
Step S50, sloped position frame is positioned based on the deep learning network after being verified.
In view of the posting of rectangle is when tilting certain angle, the posting that medical imaging devices scan is inclination
, as parallelogram greatly improves posting by the way that sloped position frame input deep learning network to be trained
Positioning accuracy.
The step S10 specifically:
Obtain the data set of the DICOM images of a large amount of sloped position frames as training, and by the data set random division
For training set T, verifying collection V and test set U.
The step S20 is specifically included:
It step S21, is the picture format of deep learning network support, such as jpeg format by each DICOM image format conversion
Image；
Step S22, artificial parameter mark is carried out to the image after format conversion；
Step S23, the parameter after mark is normalized.Normalization is a kind of mode of simplified calculating, i.e., will
There is the expression formula of dimension, by transformation, turns to nondimensional expression formula, become scalar.
In the step S22, the parameter includes centre coordinate (x, y), length h, width w and the water of sloped position frame
The angle theta and classification information of horizontal line.The classification information is the corresponding organ of sloped position frame, such as heart, cranium brain, interverbebral disc
Deng.
The step S30 is specifically included:
Step S31, the deep learning network of deep learning is constructed, a stability bandwidth is set, V points of training set T and verifying collection
Input not as the deep learning network is marked the parameter after mark as the training of the deep learning network, right
The network model is trained, and respectively obtains training set loss function curve and verifying collection loss function curve；Loss function
(loss function) is that chance event or its value in relation to stochastic variable are mapped as nonnegative real number to indicate the Random event
The function of " risk " or " loss " of part.In the application, loss function is associated usually as learning criterion with optimization problem, i.e.,
By minimizing loss function solution and assessment models.
Step S32, whether the stability bandwidth of training of judgement collection loss function curve is less than the stability bandwidth of setting, is then to enter
Step S33, it is no, then enter step S31；
Step S33, judge whether verifying collection loss function curve starts to increase, be, then deconditioning to obtain training ginseng
Number, and enter step S34；It is no, then enter step S31；
Step S34, the deep learning network is run with training parameter on test set U, obtains the ginseng of sloped position frame
Number, according to the demand of medical imaging devices reconstruction parameter, is finely adjusted obtained parameter, completes deep learning.
By the way that sloped position frame input deep learning network to be trained, the positioning accurate of posting is greatly improved
Degree, doctor only need to be finely adjusted posting, greatly improve the working efficiency of doctor, and be avoided as much as possible due to doctor
Experience level, the clinical influence for playing situation, emotional status and physical fatigue degree.
Deep learning network of the invention passes through convolution unit and residual unit network consisting main structure, convolution unit are
By convolutional layer (Convolution Layer), batch normalizing layer (Batch Normalization Layer), excitation layer
(Active Layer) is formed, and residual unit by results added before and after two convolution units by being formed.Network inputs
Image passes through several convolution units and residual unit, extracts characteristics of image, and obtain image by up-sampling layer and superimposed layer
Feature under different scale, forms several output layers, and the structure of deep learning network is as shown in Figure 3.
The data structure of output layer is w × h × (Ns×Na) × (5+1+C), wherein w, h indicate the size of output layer, indicate
Image is divided into w × h grid；Ns,NaRefer respectively to the prediction of different length-width ratios and different angle that central point is fallen in the grid
The quantity of posting；(5+1+C) indicates that each anchor prediction block needs x, y, w, and coordinate information, the object of h, θ include object
Probability and classes class probability.
It indicates accuracy of the prediction posting relative to true posting, needs to define criterion distance between the two,
Generally indicated with IOU (Intersection over Union).In the training process, pick out with true posting IOU compared with
Big prediction posting is alternative as positive example.However, introduce with after horizontal angle theta, two sloped position frames it
Between IOU for angle theta be not be monotonically changed, if directly apply original IOU definition, may choose with it is true
Real posting angle difference is larger but the IOU still biggish positive example of puppet, this will cause the concussion of training process loss function value
It even dissipates, it is therefore desirable to a new IOU standard is defined, so that output result and be all dull in respective domain
, so that the positive example chosen is relatively reasonable, such as:
Wherein A' indicates there is an identical x, y, w with A, h and has phase with B
With the sloped position frame (parallelogram) of θ；Indicate the IOU of A' and B；Optionally,Indicate coefficient entry relevant to θ, so that output result is monotonically changed about θ, coef (θA-
θB) it can also be so that other function that output result is monotonically changed about θ.
When training, needs to predict the relevant parameter designing boundary regression equation of posting for each, be parameter designing
Boundary regression equation is similar with the method in general objectives detection network, and θ parameter is also required to design its regression equation: optionally,
tθ=φ (θtruth-θanchor)=tan (θtruth-θanchor), wherein φ (θtruth-θanchor) indicate θ regression equation so that in advance
Survey θ meets when close to true θ or close to the condition for meeting linear transformation, and the regression equation of θ can also be expressed as other form.
The step S40 specifically:
A similarity is set, the parameter and training label after comparison fine tuning carry out the detection of generalization ability, if after fine tuning
Parameter and training label similarity be greater than setting similarity, then enter step S50；If the parameter and training after fine tuning are marked
The similarity of note is less than or equal to the similarity of setting, then the DICOM image of new increasing hole angle posting, and enters step S20.
In conclusion the present invention has the advantages that
1, in view of the posting of rectangle is when tilting certain angle, the posting that medical imaging devices scan is to incline
Oblique, as parallelogram greatly improves positioning by the way that sloped position frame input deep learning network to be trained
The positioning accuracy of frame.
2, by the way that sloped position frame input deep learning network to be trained, the positioning accurate of posting is greatly improved
Degree, doctor only need to be finely adjusted posting, greatly improve the working efficiency of doctor, and be avoided as much as possible due to doctor
Experience level, the clinical influence for playing situation, emotional status and physical fatigue degree.
Although specific embodiments of the present invention have been described above, those familiar with the art should be managed
Solution, we are merely exemplary described specific embodiment, rather than for the restriction to the scope of the present invention, it is familiar with this
The technical staff in field should be covered of the invention according to modification and variation equivalent made by spirit of the invention
In scope of the claimed protection.
Claims (6)
1. a kind of sloped position frame localization method based on deep learning, it is characterised in that: described method includes following steps:
Step S10, the DICOM image of a large amount of sloped position frames is obtained；
Step S20, DICOM image is pre-processed；
Step S30, pretreated DICOM image input deep learning network is trained；
Step S40, the generalization ability for the deep learning network completed to training is verified；
Step S50, sloped position frame is positioned based on the deep learning network after being verified.
2. a kind of sloped position frame localization method based on deep learning as described in claim 1, it is characterised in that: the step
Rapid S10 specifically:
Data set of the DICOM image of a large amount of sloped position frames as training is obtained, and is instruction by the data set random division
Practice collection T, verifying collection V and test set U.
3. a kind of sloped position frame localization method based on deep learning as described in claim 1, it is characterised in that: the step
Rapid S20 is specifically included:
It step S21, is the picture format of deep learning network support by each DICOM image format conversion；
Step S22, parameter mark is carried out to the image after format conversion；
Step S23, the parameter after mark is normalized.
4. a kind of sloped position frame localization method based on deep learning as claimed in claim 3, it is characterised in that: the step
In rapid S22, the parameter include the centre coordinate (x, y) of sloped position frame, length h, width w, with horizontal angle theta and
Classification information.
5. a kind of sloped position frame localization method based on deep learning as claimed in claim 2, it is characterised in that: the step
Rapid S30 is specifically included:
Step S31, the deep learning network of deep learning is constructed, a stability bandwidth is set, training set T and verifying collection V are made respectively
For the input of the deep learning network, marked the parameter after mark as the training of the deep learning network, to described
Network model is trained, and respectively obtains training set loss function curve and verifying collection loss function curve；
Step S32, whether the stability bandwidth of training of judgement collection loss function curve is less than the stability bandwidth of setting, is then to enter step
S33, it is no, then enter step S31；
Step S33, judge whether verifying collection loss function curve starts to increase, be, then deconditioning, acquisition training parameter, and
Enter step S34；It is no, then enter step S31；
Step S34, the deep learning network is run with training parameter on test set U, obtains the parameter of sloped position frame, root
According to the demand of medical imaging devices reconstruction parameter, obtained parameter is finely adjusted, completes deep learning.
6. a kind of sloped position frame localization method based on deep learning as claimed in claim 5, it is characterised in that: the step
Rapid S40 specifically:
A similarity is set, the parameter and training label after comparison fine tuning carry out the detection of generalization ability, if the ginseng after fine tuning
Several similarities with training label are greater than the similarity of setting, then enter step S50；What if the parameter and training after fine tuning marked
Similarity is less than or equal to the similarity of setting, then the DICOM image of new increasing hole angle posting, and enters step S20.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910573065.8A CN110288615B (en) | 2019-06-28 | 2019-06-28 | Inclined positioning frame positioning method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910573065.8A CN110288615B (en) | 2019-06-28 | 2019-06-28 | Inclined positioning frame positioning method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110288615A true CN110288615A (en) | 2019-09-27 |
| CN110288615B CN110288615B (en) | 2020-11-13 |
Family
ID=68019898
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910573065.8A Active CN110288615B (en) | 2019-06-28 | 2019-06-28 | Inclined positioning frame positioning method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110288615B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111105347A (en) * | 2019-11-19 | 2020-05-05 | 贝壳技术有限公司 | Method, device and storage medium for generating panoramic image with depth information |
| CN112734740A (en) * | 2021-01-18 | 2021-04-30 | 推想医疗科技股份有限公司 | Method for training target detection model, method for detecting target and device thereof |
| US11055835B2 (en) | 2019-11-19 | 2021-07-06 | Ke.com (Beijing) Technology, Co., Ltd. | Method and device for generating virtual reality data |
| WO2022007943A1 (en) * | 2020-07-10 | 2022-01-13 | 中国水利水电科学研究院 | Improved inclined rectangular range box annotation method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107563412A (en) * | 2017-08-09 | 2018-01-09 | 浙江大学 | A real-time detection method of infrared image power equipment based on deep learning |
| CN109325504A (en) * | 2018-09-07 | 2019-02-12 | 中国农业大学 | A kind of underwater sea cucumber identification method and system |
| CN109785298A (en) * | 2018-12-25 | 2019-05-21 | 中国科学院计算技术研究所 | A kind of multi-angle object detecting method and system |
-
2019
- 2019-06-28 CN CN201910573065.8A patent/CN110288615B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107563412A (en) * | 2017-08-09 | 2018-01-09 | 浙江大学 | A real-time detection method of infrared image power equipment based on deep learning |
| CN109325504A (en) * | 2018-09-07 | 2019-02-12 | 中国农业大学 | A kind of underwater sea cucumber identification method and system |
| CN109785298A (en) * | 2018-12-25 | 2019-05-21 | 中国科学院计算技术研究所 | A kind of multi-angle object detecting method and system |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111105347A (en) * | 2019-11-19 | 2020-05-05 | 贝壳技术有限公司 | Method, device and storage medium for generating panoramic image with depth information |
| CN111105347B (en) * | 2019-11-19 | 2020-11-13 | 贝壳找房(北京)科技有限公司 | Method, device and storage medium for generating panoramic image with depth information |
| US11055835B2 (en) | 2019-11-19 | 2021-07-06 | Ke.com (Beijing) Technology, Co., Ltd. | Method and device for generating virtual reality data |
| US11721006B2 (en) | 2019-11-19 | 2023-08-08 | Realsee (Beijing) Technology Co., Ltd. | Method and device for generating virtual reality data |
| WO2022007943A1 (en) * | 2020-07-10 | 2022-01-13 | 中国水利水电科学研究院 | Improved inclined rectangular range box annotation method |
| CN112734740A (en) * | 2021-01-18 | 2021-04-30 | 推想医疗科技股份有限公司 | Method for training target detection model, method for detecting target and device thereof |
| CN112734740B (en) * | 2021-01-18 | 2024-02-23 | 推想医疗科技股份有限公司 | Method for training target detection model, target detection method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110288615B (en) | 2020-11-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11455732B2 (en) | Knowledge-based automatic image segmentation | |
| CN110288615A (en) | A kind of sloped position frame localization method based on deep learning | |
| CN109074500B (en) | System and method for segmenting medical images of the same patient | |
| EP3268931B1 (en) | Method and apparatus for assessing image registration | |
| US8634616B2 (en) | Method, apparatus, and computer program product for acquiring medical image data | |
| WO2018119766A1 (en) | Multi-modal image processing system and method | |
| CN111640118B (en) | Tumor tissue characterization using multiparameter magnetic resonance imaging | |
| CN108447063B (en) | Multimodal MRI segmentation method for brain glioblastoma | |
| Kieselmann et al. | Cross‐modality deep learning: contouring of MRI data from annotated CT data only | |
| CN107680107B (en) | Automatic segmentation method of diffusion tensor magnetic resonance image based on multiple maps | |
| RU2745400C2 (en) | Registration of medical map | |
| US9600856B2 (en) | Hybrid point-based registration | |
| CN109472263A (en) | A brain magnetic resonance image segmentation method combining global and local information | |
| EP2646978A1 (en) | Longitudinal monitoring of pathology | |
| US9727975B2 (en) | Knowledge-based automatic image segmentation | |
| US11244472B2 (en) | Method, system and computer program for determining position and/or orientation parameters of an anatomical structure | |
| CN113327224B (en) | System and method for automatic field of view (FOV) definition | |
| US8938107B2 (en) | System and method for automatic segmentation of organs on MR images using a combined organ and bone atlas | |
| CN106462974A (en) | Optimization of parameters for segmenting an image | |
| US20080037848A1 (en) | System and Method for Corpus Callosum Segmentation in Magnetic Resonance Images | |
| CN113962953B (en) | Image processing method and system | |
| JP2021515326A (en) | Systems and methods for accelerated clinical workflows | |
| He et al. | Quantitative comparisons of deep-learning-based and atlas-based auto-segmentation of the intermediate risk clinical target volume for nasopharyngeal carcinoma | |
| CN115147603A (en) | Brain tumor MRI image segmentation method based on multi-scale distillation | |
| CN117078703B (en) | CT image segmentation method and system based on MRI guidance |
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 |