CN109191442A - Ultrasound image assessment and screening technique and device - Google Patents
Ultrasound image assessment and screening technique and device Download PDFInfo
- Publication number
- CN109191442A CN109191442A CN201810990318.7A CN201810990318A CN109191442A CN 109191442 A CN109191442 A CN 109191442A CN 201810990318 A CN201810990318 A CN 201810990318A CN 109191442 A CN109191442 A CN 109191442A
- Authority
- CN
- China
- Prior art keywords
- ultrasound image
- assessment
- head circumference
- module
- fetus head
- 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
- G06T7/0014—Biomedical image inspection using an image reference approach
-
- 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
- G06T2207/30016—Brain
Abstract
The present invention provides a kind of assessment of ultrasound image and screening technique and device, and ultrasound image assessment and screening technique include: acquisition fetus head circumference ultrasound image, carries out enhancing processing to fetus head circumference ultrasound image, generates ultrasound image training sample；The assessment score of ultrasound image training sample is formulated according to assessment agreement, wherein assessment agreement is there are an all brain structures clearly then to increase by one point in multiple all brain structures in ultrasound image training sample；The Faster R-CNN model pre-established is trained using the corresponding assessment score of ultrasound image training sample and sample；By the Faster R-CNN model after the fetus head circumference ultrasound image measured input training, obtains and corresponding with the fetus head circumference ultrasound image measured assess score；The ultrasound image that score is assessed in the fetus head circumference ultrasound image measured in default score range is filtered out, for ideal ultrasound image.Ultrasound image assessment of the invention and screening technique, can be improved the accuracy rate and efficiency for obtaining assessment agreement ultrasound image.
Description
Technical field
The present invention relates to image identification technical field, in particular to a kind of assessment of ultrasound image and screening technique and
Device.
Background technique
In antenatal exaination, it is often necessary to estimate the growth and development situation of fetus by ultrasound image, and obtain and meet
The ultrasound image of quality standard is the premise that doctor accurately estimates.
And the acquisition of the current ultrasound image met the quality standard relies on the clinical experience of sonographer more, exists very big
Subjectivity and be easy to get second-rate ultrasound image, accuracy rate is lower.Also, the ultrasound of new hand's sonographer is schemed
The quality of picture is always to be checked by veteran sonographer, and there are errors for the result of inspection, and efficiency is relatively low.
Summary of the invention
In view of the above problems, it the present invention provides a kind of assessment of ultrasound image and screening technique and device, is obtained with improving
The accuracy rate and efficiency of the ideal ultrasound image met the quality standard.
To achieve the goals above, the present invention adopts the following technical scheme that:
A kind of assessment of ultrasound image and screening technique, for obtaining fetus head circumference ultrasound image in pre-natal diagnosis, comprising:
Fetus head circumference ultrasound image is acquired, enhancing processing is carried out to the fetus head circumference ultrasound image, generates ultrasound image
Training sample；
The assessment score of ultrasound image training sample is formulated according to assessment agreement, wherein the assessment agreement is described super
There are an all brain structures clearly then to increase by one point in acoustic image training sample in multiple all brain structures；
Using the corresponding assessment score of the ultrasound image training sample and sample to the Faster R-CNN pre-established
Model is trained；
By the Faster R-CNN model after the fetus head circumference ultrasound image measured input training, obtain and the survey
The fetus head circumference ultrasound image obtained assesses score accordingly；
Ultrasound image of the assessment score in default score range in the fetus head circumference ultrasound image measured is filtered out,
For ideal ultrasound image.
Preferably, described " acquisition fetus head circumference ultrasound image, carries out enhancing processing to the fetus head circumference ultrasound image,
Generate ultrasound image training sample " include:
Acquire at least one described fetus head circumference ultrasound image；
Mirror image, left bank, right bank and skull exterior clipping are carried out at least one described fetus head circumference ultrasound image
Image processing operations, generating multiple ultrasonic subgraphs is the ultrasound image training sample, and to each subgraph of generation
Carry out Labeling Coordinate.
Preferably, the multiple all brain structures include lateral fissure, symmetrical thalamus, choroid plexus, cavity of septum pellucidum and diacele
Clear display standard.
Preferably, the Faster R-CNN model includes Fast R-CNN module and RPN module, wherein the RPN mould
The input terminal of block connects the feature inclusion layer of the Fast R-CNN module, and output end connects the Fast R-CNN module
Pooling layers of ROI.
Preferably, described " using the corresponding assessment score of the ultrasound image training sample and sample to pre-establishing
Faster R-CNN model is trained " include:
Using the feature inclusion layer for having trained Fast R-CNN module described in netinit, and utilize the ultrasound image
Training sample individually trains the RPN module, generates first candidate frame；
Again using the feature inclusion layer for having trained Fast R-CNN module described in netinit, and utilize described first
It criticizes candidate frame and the corresponding assessment score of the sample individually trains the Fast R-CNN module, obtain training knot for the first time
Fruit；
The feature inclusion layer of the Fast R-CNN module is initialized using first time training result, and utilizes the ultrasound
Image training sample and the corresponding assessment score of the sample train the RPN module and the Fast R-CNN mould again
Block；
It repeats above three step and carries out the alternating training of above three step, until reaching default frequency of training.
Preferably, it is described trained network including the use of the trained VGG16 network of ImageNet visible database and
Resnet50 network.
The present invention also provides a kind of assessment of ultrasound image and screening plants, and the fetus head circumference for obtaining in pre-natal diagnosis is super
Acoustic image, comprising:
Sample generation module enhances the fetus head circumference ultrasound image for acquiring fetus head circumference ultrasound image
Processing generates ultrasound image training sample；
Sample grading module, for formulating the assessment score of ultrasound image training sample according to assessment agreement, wherein described
Assessment agreement is to have an all brain structures clearly then to increase by one point in multiple all brain structures in the ultrasound image training sample；
Training module, for using the corresponding assessment score of the ultrasound image training sample and sample to pre-establishing
Faster R-CNN model is trained；
Evaluation module, the Faster R-CNN mould after the input training of fetus head circumference ultrasound image for will measure
Type, acquisition is corresponding with the fetus head circumference ultrasound image measured to assess score；
Optical sieving module assesses score in default score value for filtering out in the fetus head circumference ultrasound image measured
Ultrasound image in range, for ideal ultrasound image.
Preferably, the sample generation module includes:
Image acquisition units, for acquiring at least one described fetus head circumference ultrasound image；
Image processing unit, for carrying out mirror image, left bank, right bank at least one described fetus head circumference ultrasound image
And the image processing operations of skull exterior clipping, generating multiple ultrasonic subgraphs is the ultrasound image training sample, and to life
At each subgraph carry out Labeling Coordinate.
Preferably, the multiple all brain structures include lateral fissure, symmetrical thalamus, choroid plexus, cavity of septum pellucidum and diacele
Clear display standard.
Preferably, the Faster R-CNN model includes Fast R-CNN module and RPN module, wherein the RPN mould
The input terminal of block connects the feature inclusion layer of the Fast R-CNN module, and output end connects the Fast R-CNN module
Pooling layers of ROI.
The present invention provides a kind of assessment of ultrasound image and screening technique, and this method is applied to the fetus obtained in pre-natal diagnosis
Head circumference ultrasound image, comprising: acquisition fetus head circumference ultrasound image carries out enhancing processing to the fetus head circumference ultrasound image, raw
At ultrasound image training sample；The assessment score of ultrasound image training sample is formulated according to assessment agreement, wherein the assessment association
View is to have an all brain structures clearly then to increase by one point in multiple all brain structures in the ultrasound image training sample；Using described
Ultrasound image training sample and the corresponding assessment score of sample are trained the Faster R-CNN model pre-established；It will survey
The Faster R-CNN model after the fetus head circumference ultrasound image input training obtained, obtains and the fetus head circumference measured
Ultrasound image assesses score accordingly；It filters out and assesses score in the fetus head circumference ultrasound image measured in default score value model
Interior ultrasound image is enclosed, for ideal ultrasound image.Ultrasound image assessment of the invention and screening technique, can be improved acquisition and meet
The accuracy rate and efficiency of the ultrasound image of quality standard.
To enable the above objects, features and advantages of the present invention to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate
Appended attached drawing, is described in detail below.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only certain embodiments of the present invention, therefore is not construed as pair
The restriction of the scope of the invention.
Fig. 1 is the flow chart of a kind of assessment of ultrasound image and screening technique that the embodiment of the present invention 1 provides；
Fig. 2 is a kind of Faster R-CNN model structure schematic diagram that the embodiment of the present invention 1 provides；
Fig. 3 is the stream of the generation training sample of a kind of assessment of ultrasound image and screening technique that the embodiment of the present invention 2 provides
Cheng Tu；
Fig. 4 is the flow chart of the training pattern of a kind of assessment of ultrasound image and screening technique that the embodiment of the present invention 3 provides；
Fig. 5 is the structural schematic diagram of a kind of assessment of ultrasound image and screening plant that the embodiment of the present invention 4 provides；
Fig. 6 is the knot of the sample generation module of a kind of assessment of ultrasound image and screening plant that the embodiment of the present invention 4 provides
Structure schematic diagram.
Specific embodiment
Below in conjunction with attached drawing in the embodiment of the present invention, technical solution in the embodiment of the present invention carries out clear, complete
Ground description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Usually exist
The component of the embodiment of the present invention described and illustrated in attached drawing can be arranged and be designed with a variety of different configurations herein.Cause
This, is not intended to limit claimed invention to the detailed description of the embodiment of the present invention provided in the accompanying drawings below
Range, but it is merely representative of selected embodiment of the invention.Based on the embodiment of the present invention, those skilled in the art are not doing
Every other embodiment obtained under the premise of creative work out, shall fall within the protection scope of the present invention.
Embodiment 1
Fig. 1 is the flow chart of a kind of assessment of ultrasound image and screening technique that the embodiment of the present invention 1 provides, this method application
The fetus head circumference ultrasound image obtained in pre-natal diagnosis, includes the following steps:
Step S11: acquisition fetus head circumference ultrasound image carries out enhancing processing to fetus head circumference ultrasound image, generates ultrasound
Image training sample.
In the embodiment of the present invention, fetus head circumference ultrasound image can come from various big hospital, super by different model in hospital
Sound instrument is acquired, wherein the gestational period for receiving the fetus of acquiring ultrasound image is 14 weeks to 28 weeks etc., is using Ultrasound Instrument
When device carries out the acquisition of ultrasound image, pseudo- color ultrasound image can also be obtained by pseudo- color technology.The ultrasound image of acquisition can
It is stored with being sent in server, and in storing process, can select the higher image of mass in advance, such as can be with
By working experience doctor abundant, when carrying out ultrasound image observation to fetus, standard, the high ultrasound figure of quality are picked out
Picture, and store in the server, so as to the training sample as intern or learning model.Wherein, which can connect
More ultrasonic instruments are connect, the ultrasound image that more ultrasonic instruments generate is obtained.
In the embodiment of the present invention, server can pass through algorithm after receiving the ultrasound image that numerous ultrasonic instruments are sent
Or application program carries out enhancing processing to ultrasound image, schemes for example, can use application program to the ultrasound that different instruments generate
As carrying out size adjusting, all ultrasonic pictures are unified for a size, the plane coordinates etc. of ultrasonic picture can also be generated
Processing.By enhancing, treated that ultrasound image is ultrasound image training sample, and the ultrasound image training sample is for training
Learning model, to improve the ability that learning model differentiates the fetus head circumference ultrasound image of high quality.
Step S12: the assessment score of ultrasound image training sample is formulated according to assessment agreement, wherein assessment agreement is super
There are an all brain structures clearly then to increase by one point in acoustic image training sample in multiple all brain structures.
In the embodiment of the present invention, the quality standard of fetus head circumference ultrasound image is that can clearly observe tire in ultrasound image
Youngster's all brain structures, so that doctor will be seen that by ultrasound image the brain development situation of fetus.Wherein, multiple all brain structures
Including lateral fissure, symmetrical thalamus, choroid plexus, cavity of septum pellucidum and diacele.Wherein, learnt in order to facilitate learning model,
Scoring agreement can be formulated for the readability of above-mentioned all brain structures, such as is 1 point when clear display lateral fissure, clear display pair
It is referred to as 1 point when thalamus, is 1 point when clearly showing choroid plexus, is 1 point when clearly showing cavity of septum pellucidum, and clearly show third
It is 1 point when the ventricles of the brain, the total score of a fetus head circumference ultrasound image is 5 points, and the ultrasound image that only total score is 5 points is to meet quality
The picture of standard.
Step S13: using the corresponding assessment score of ultrasound image training sample and sample to the Faster R- pre-established
CNN model is trained.
In the embodiment of the present invention, as shown in Fig. 2, the Faster R-CNN model 200 includes 210 He of Fast R-CNN module
RPN module 220, wherein the feature inclusion layer 211 of the input terminal connection Fast R-CNN module 210 of RPN module 220, output end
Connect the ROI Pooling layer 212 of Fast R-CNN module 210.That is, the Faster R-CNN model 200 is based on Fast
The improvement that R-CNN model 210 and RPN model 220 are made, wherein the feature inclusion layer 211 of the Fast R-CNN module 210
That is convolutional layer can carry out the multiple convolution operation of image after inputting ultrasound image, such as carry out 5 convolution operations, and can be with
Generate the candidate frame coordinate of ultrasound image, candidate frame here refers to intercepting different more of multiple sizes on former ultrasound image
It is a may be containing the rectangular area of target detection object, such as can be the candidate frame of lateral fissure structure, and the candidate frame
Quantity is set as 300.
It is trained end to end since Fast R-CNN module 210 is unable to reach, namely after obtaining candidate frame coordinate, every
The candidate frame of ultrasound image has to obtain by searching algorithm, and then candidate frame is made to enter the complete of Fast R-CNN module 210
Articulamentum carries out the extraction of feature, therefore RPN model 220 is added after its sharing feature layer and provides the ability of search candidate frame,
Train the entire realization of Faster R-CNN model 200 end to end.After sharing feature layer and RPN module 220 also
The ROI Pooling layer 212 of Fast R-CNN module 210.The ROI layers can be according to candidate frame coordinate directly from sharing feature layer
The characteristic dimension in the corresponding original image region of coordinate is extracted in the last layer, and the feature of these fixed dimension sizes is inputted Fast
The extraction that feature is carried out in the full articulamentum of R-CNN module 210, can thus allow 210 needs of Faster R-CNN model
Carrying out a convolution to ultrasound image can all learn its candidate frame, greatly improve the speed of detection and training.
It is exactly two full articulamentums after the ROI Pooling layer 212 of Fast R-CNN module 210 in the embodiment of the present invention,
One is used for ROI layers of class probability Score on Prediction, another is used for ROI layer region coordinate0≤u≤
The offset of U optimizes, and wherein U is the quantity of object class.The total losses function of Fast R-CNN module is above-mentioned two full articulamentum
Weighted sum:
Wherein, LclsIt is the loss function of classification branch, LlocIt is the loss function of ROI layers of coordinate.
Step S14: it by the Faster R-CNN model after the fetus head circumference ultrasound image measured input training, obtains and surveys
The fetus head circumference ultrasound image obtained assesses score accordingly.
Step S15: the ultrasound that score is assessed in the fetus head circumference ultrasound image measured in default score range is filtered out
Image, for ideal ultrasound image.
In the embodiment of the present invention, it is super to can use every test fetus head circumference of trained Faster R-CNN model inspection
5 anatomical structures of sound spectrogram piece, to obtain the assessment score of test fetus head circumference ultrasound image, and according to default score range
The ultrasound image for meeting the quality standard assessment score is filtered out, for example, the default score value is 5 points.Wherein, the result after test is also
It can save in the server, then carry out the qualitative analysis of test result, it can it is fixed to calculate Faster R-CNN model
The Duplication for the result that the test result and medical practitioner of position are positioned manually positions for example, it can be set to Duplication is greater than 0.5
Accurately, mistake is positioned less than 0.5.In addition, can also pass through for the performance indicator of the Faster R-CNN model after training
It precision, recall rate and is judged with accuracy rate, here without limitation.
Embodiment 2
Fig. 3 is the stream of the generation training sample of a kind of assessment of ultrasound image and screening technique that the embodiment of the present invention 2 provides
Cheng Tu includes the following steps:
Step S31: at least one fetus head circumference ultrasound image is acquired.
Step S32: at least one fetus head circumference ultrasound image cut out outside mirror image, left bank, right bank and skull
The image processing operations cut, generating multiple ultrasonic subgraphs is ultrasound image training sample, and to each subgraph of generation
Carry out Labeling Coordinate.
In the embodiment of the present invention, Faster R-CNN model is a kind of deep learning model, and late detection meets quality
Data are influenced when the smart rate of the ultrasound image of standard is trained by it, that is, the Faster R-CNN model needs a large amount of instruction
Practice sample.Here making a series of enhancing processing for the fetus head circumference ultrasound image of the high quality of acquisition includes: mirror image, a left side
The processing such as inclination, right bank and skull exterior clipping, to obtain multiple ultrasonic subgraphs.In practical applications, a ultrasound
Last available 11 ultrasonic subgraphs of image.
Embodiment 3
Fig. 4 is the flow chart of the training pattern of a kind of assessment of ultrasound image and screening technique that the embodiment of the present invention 3 provides,
Include the following steps:
Step S41: using the feature inclusion layer for having trained netinit Fast R-CNN module, and ultrasound image is utilized
Training sample individually trains RPN module, generates first candidate frame.
Step S42: again using the feature inclusion layer for having trained netinit Fast R-CNN module, and first is utilized
It criticizes candidate frame and the corresponding assessment score of sample and individually trains Fast R-CNN module, obtain first time training result.
Step S43: using the feature inclusion layer of first time training result initialization Fast R-CNN module, and ultrasound is utilized
Image training sample and the corresponding assessment score of sample train RPN module and Fast R-CNN module again.
Step S44: repeating above three step and carries out the alternating training of above three step, until reaching default training
Number.
In the embodiment of the present invention, the process of the Faster R-CNN model training is the process of repetitive exercise, RPN module
Individually training is separated with Fast R-CNN module, and alternately the training of search candidate frame is trained with detection, it is default until reaching
Frequency of training constitutes a unified Faster R-CNN model after the training.Trained network including the use of
The trained VGG16 network of ImageNet visible database and Resnet50 network.
Embodiment 4
Fig. 5 is the structural schematic diagram of a kind of assessment of ultrasound image and screening plant that the embodiment of the present invention 4 provides.
Ultrasound image assessment and screening plant 500 include:
Sample generation module 510 carries out at enhancing fetus head circumference ultrasound image for acquiring fetus head circumference ultrasound image
Reason generates ultrasound image training sample；
Sample grading module 520, for formulating the assessment score of ultrasound image training sample according to assessment agreement, wherein
Assessment agreement is to have an all brain structures clearly then to increase by one point in multiple all brain structures in ultrasound image training sample；
Training module 530, for using the corresponding assessment score of ultrasound image training sample and sample to pre-establishing
Faster R-CNN model is trained；
Evaluation module 540, the Faster R-CNN model after the input training of fetus head circumference ultrasound image for will measure,
It obtains and corresponding with the fetus head circumference ultrasound image measured assesses score；
Optical sieving module 550 assesses score in default score value for filtering out in the fetus head circumference ultrasound image measured
Ultrasound image in range, for ideal ultrasound image.
As shown in fig. 6, the sample generation module 510 includes:
Image acquisition units 511, for acquiring at least one fetus head circumference ultrasound image.
Image processing unit 512, for carrying out mirror image, left bank, the right side at least one described fetus head circumference ultrasound image
The image processing operations of inclination and skull exterior clipping, generating multiple ultrasonic subgraphs is the ultrasound image training sample, and
Labeling Coordinate is carried out to each subgraph of generation.
In the embodiment of the present invention, multiple all brain structures include lateral fissure, symmetrical thalamus, choroid plexus, cavity of septum pellucidum and third
The clear display standard of the ventricles of the brain.The Faster R-CNN model includes Fast R-CNN module and RPN module, wherein RPN module
Input terminal connection Fast R-CNN module feature inclusion layer, output end connect Fast R-CNN module ROI Pooling
Layer.
In the embodiment of the present invention, above-mentioned modules and the more detailed function description of unit can refer to previous embodiment
The content of middle corresponding portion, details are not described herein.
In addition, the server includes memory and processor the present invention also provides a kind of server, memory can be used for
Computer program is stored, processor is by running the computer program, so that server be made to execute the above method or above-mentioned
The function of ultrasound image assessment and the modules in screening plant.
Memory may include storing program area and storage data area, wherein storing program area can storage program area, at least
Application program needed for one function (such as sound-playing function, image player function etc.) etc.；Storage data area can store root
Created data (such as audio data, phone directory etc.) etc. are used according to server.In addition, memory may include high speed with
Machine access memory, can also include nonvolatile memory, a for example, at least disk memory, flush memory device or its
His volatile solid-state part.
The present embodiment additionally provides a kind of computer storage medium, for storing computer journey used in above-mentioned server
Sequence.
In several embodiments provided herein, it should be understood that disclosed device and method can also pass through
Other modes are realized.The apparatus embodiments described above are merely exemplary, for example, flow chart and structure in attached drawing
Figure shows the system frame in the cards of the device of multiple embodiments according to the present invention, method and computer program product
Structure, function and operation.In this regard, each box in flowchart or block diagram can represent a module, section or code
A part, a part of the module, section or code includes one or more for implementing the specified logical function
Executable instruction.It should also be noted that function marked in the box can also be to be different from the implementation as replacement
The sequence marked in attached drawing occurs.For example, two continuous boxes can actually be basically executed in parallel, they are sometimes
It can execute in the opposite order, this depends on the function involved.It is also noted that in structure chart and/or flow chart
The combination of each box and the box in structure chart and/or flow chart, can function or movement as defined in executing it is dedicated
Hardware based system realize, or can realize using a combination of dedicated hardware and computer instructions.
In addition, each functional module or unit in each embodiment of the present invention can integrate one independence of formation together
Part, be also possible to modules individualism, an independent part can also be integrated to form with two or more modules.
It, can be with if the function is realized and when sold or used as an independent product in the form of software function module
It is stored in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be embodied in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be intelligence
Can mobile phone, personal computer, server or network equipment etc.) execute each embodiment the method for the present invention whole or
Part steps.And storage medium above-mentioned include: USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory),
Random access memory (RAM, Random Access Memory), magnetic or disk etc. be various to can store program code
Medium.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can easily think of the change or the replacement, and should all contain
Lid is within protection scope of the present invention.Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. a kind of ultrasound image assessment and screening technique, for obtaining fetus head circumference ultrasound image, feature in pre-natal diagnosis
It is, comprising:
Fetus head circumference ultrasound image is acquired, enhancing processing is carried out to the fetus head circumference ultrasound image, generates ultrasound image training
Sample；
The assessment score of ultrasound image training sample is formulated according to assessment agreement, wherein the assessment agreement is the ultrasound figure
As there is an all brain structures clearly then to increase by one point in all brain structures multiple in training sample；
Using the corresponding assessment score of the ultrasound image training sample and sample to the Faster R-CNN model pre-established
It is trained；
By the Faster R-CNN model after the input training of the fetus head circumference ultrasound image that measures, obtains and measured with described
Fetus head circumference ultrasound image assesses score accordingly；
Ultrasound image of the assessment score in default score range in the fetus head circumference ultrasound image measured is filtered out, for reason
Think ultrasound image.
2. ultrasound image assessment according to claim 1 and screening technique, which is characterized in that " the acquisition fetus head circumference
Ultrasound image carries out enhancing processing to the fetus head circumference ultrasound image, generates ultrasound image training sample " include:
Acquire at least one described fetus head circumference ultrasound image；
The image of mirror image, left bank, right bank and skull exterior clipping is carried out at least one described fetus head circumference ultrasound image
Processing operation, generating multiple ultrasonic subgraphs is the ultrasound image training sample, and is carried out to each subgraph of generation
Labeling Coordinate.
3. ultrasound image assessment according to claim 1 and screening technique, which is characterized in that the multiple all brain structures packet
Include lateral fissure, symmetrical thalamus, choroid plexus, cavity of septum pellucidum and diacele.
4. ultrasound image assessment according to claim 1 and screening technique, which is characterized in that the Faster R-CNN mould
Type includes Fast R-CNN module and RPN module, wherein the input terminal of the RPN module connects the Fast R-CNN module
Feature inclusion layer, output end connects Pooling layers of ROI of the Fast R-CNN module.
5. ultrasound image assessment according to claim 4 and screening technique, which is characterized in that described " to utilize the ultrasound
Image training sample and the corresponding assessment score of sample are trained the Faster R-CNN model pre-established " include:
Using the feature inclusion layer for having trained Fast R-CNN module described in netinit, and utilize ultrasound image training
Sample individually trains the RPN module, generates first candidate frame；
Again using the feature inclusion layer for having trained Fast R-CNN module described in netinit, and utilize first described time
It selects frame and the corresponding assessment score of the sample individually to train the Fast R-CNN module, obtains first time training result；
The feature inclusion layer of the Fast R-CNN module is initialized using first time training result, and utilizes the ultrasound image
Training sample and the corresponding assessment score of the sample train the RPN module and the Fast R-CNN module again；
It repeats above three step and carries out the alternating training of above three step, until reaching default frequency of training.
6. ultrasound image according to claim 5 assessment and screening technique, which is characterized in that described to have trained the network to include
Utilize the trained VGG16 network of ImageNet visible database and Resnet50 network.
7. a kind of ultrasound image assessment and screening plant, the fetus head circumference ultrasound image for being obtained in pre-natal diagnosis, feature
It is, comprising:
Sample generation module carries out enhancing processing to the fetus head circumference ultrasound image for acquiring fetus head circumference ultrasound image,
Generate ultrasound image training sample；
Sample grading module, for formulating the assessment score of ultrasound image training sample according to assessment agreement, wherein the assessment
Agreement is to have an all brain structures clearly then to increase by one point in multiple all brain structures in the ultrasound image training sample；
Training module, for using the corresponding assessment score of the ultrasound image training sample and sample to pre-establishing
Faster R-CNN model is trained；
Evaluation module, the Faster R-CNN model after the input training of fetus head circumference ultrasound image for will measure, is obtained
It takes and corresponding with the fetus head circumference ultrasound image measured assesses score；
Optical sieving module assesses score in default score range for filtering out in the fetus head circumference ultrasound image measured
Interior ultrasound image, for ideal ultrasound image.
8. ultrasound image assessment according to claim 7 and screening plant, which is characterized in that the sample generation module packet
It includes:
Image acquisition units, for acquiring at least one described fetus head circumference ultrasound image；
Image processing unit, at least one described fetus head circumference ultrasound image carry out mirror image, left bank, right bank and
The image processing operations of skull exterior clipping, generating multiple ultrasonic subgraphs is the ultrasound image training sample, and to generation
Each subgraph carries out Labeling Coordinate.
9. ultrasound image assessment according to claim 7 and screening plant, which is characterized in that the multiple all brain structures packet
Include the clear display standard of lateral fissure, symmetrical thalamus, choroid plexus, cavity of septum pellucidum and diacele.
10. ultrasound image assessment according to claim 7 and screening plant, which is characterized in that the Faster R-CNN
Model includes Fast R-CNN module and RPN module, wherein the input terminal of the RPN module connects the Fast R-CNN mould
The feature inclusion layer of block, output end connect Pooling layers of ROI of the Fast R-CNN module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810990318.7A CN109191442B (en) | 2018-08-28 | 2018-08-28 | Ultrasonic image evaluation and screening method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810990318.7A CN109191442B (en) | 2018-08-28 | 2018-08-28 | Ultrasonic image evaluation and screening method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109191442A true CN109191442A (en) | 2019-01-11 |
| CN109191442B CN109191442B (en) | 2021-04-13 |
Family
ID=64916699
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810990318.7A Active CN109191442B (en) | 2018-08-28 | 2018-08-28 | Ultrasonic image evaluation and screening method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109191442B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110338842A (en) * | 2019-07-11 | 2019-10-18 | 北京市朝阳区妇幼保健院 | A kind of image optimization method of newborn's lungs ultrasonic image-forming system |
| CN110464380A (en) * | 2019-09-12 | 2019-11-19 | 李肯立 | A kind of method that the ultrasound cross-section image of the late pregnancy period fetus of centering carries out quality control |
| CN111062953A (en) * | 2019-12-17 | 2020-04-24 | 北京化工大学 | Method for identifying parathyroid hyperplasia in ultrasonic image |
| CN112070119A (en) * | 2020-08-11 | 2020-12-11 | 长沙大端信息科技有限公司 | Ultrasonic tangent plane image quality control method and device and computer equipment |
| CN112446862A (en) * | 2020-11-25 | 2021-03-05 | 北京医准智能科技有限公司 | Dynamic breast ultrasound video full-focus real-time detection and segmentation device and system based on artificial intelligence and image processing method |
| CN113378831A (en) * | 2021-05-07 | 2021-09-10 | 太原理工大学 | Mouse embryo organ identification and scoring method and system |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005278995A (en) * | 2004-03-30 | 2005-10-13 | Japan Science & Technology Agency | Childbirth diagnostic supporting program and recording medium storing the same, and method and apparatus for supporting childbirth diagnosis |
| US20070081705A1 (en) * | 2005-08-11 | 2007-04-12 | Gustavo Carneiro | System and method for fetal biometric measurements from ultrasound data and fusion of same for estimation of fetal gestational age |
| US20080240532A1 (en) * | 2007-03-30 | 2008-10-02 | Siemens Corporation | System and Method for Detection of Fetal Anatomies From Ultrasound Images Using a Constrained Probabilistic Boosting Tree |
| CN103222879A (en) * | 2012-01-25 | 2013-07-31 | 通用电气公司 | System and method for identifying an optimal image frame for ultrasound imaging |
| CN103263278A (en) * | 2013-01-23 | 2013-08-28 | 郑末晶 | Image processing method for automatically measuring thickness of fetal nuchal translucency from ultrasonic image |
| CN106408566A (en) * | 2016-11-10 | 2017-02-15 | 深圳大学 | Fetal ultrasound image quality control method and system |
| CN107798678A (en) * | 2017-12-11 | 2018-03-13 | 成都优途科技有限公司 | Type B ultrasonoscopy the best visual effect Automatic Optimal system and control method |
| CN108056789A (en) * | 2017-12-19 | 2018-05-22 | 飞依诺科技(苏州)有限公司 | A kind of method and apparatus for the configuration parameter value for generating ultrasound scanning device |
| CN108378869A (en) * | 2017-12-26 | 2018-08-10 | 珠海艾博罗生物技术股份有限公司 | A kind of image processing method and processing system of the automatic measurement fetal head girth degree from ultrasonoscopy |
-
2018
- 2018-08-28 CN CN201810990318.7A patent/CN109191442B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005278995A (en) * | 2004-03-30 | 2005-10-13 | Japan Science & Technology Agency | Childbirth diagnostic supporting program and recording medium storing the same, and method and apparatus for supporting childbirth diagnosis |
| US20070081705A1 (en) * | 2005-08-11 | 2007-04-12 | Gustavo Carneiro | System and method for fetal biometric measurements from ultrasound data and fusion of same for estimation of fetal gestational age |
| US20080240532A1 (en) * | 2007-03-30 | 2008-10-02 | Siemens Corporation | System and Method for Detection of Fetal Anatomies From Ultrasound Images Using a Constrained Probabilistic Boosting Tree |
| CN103222879A (en) * | 2012-01-25 | 2013-07-31 | 通用电气公司 | System and method for identifying an optimal image frame for ultrasound imaging |
| CN103263278A (en) * | 2013-01-23 | 2013-08-28 | 郑末晶 | Image processing method for automatically measuring thickness of fetal nuchal translucency from ultrasonic image |
| CN106408566A (en) * | 2016-11-10 | 2017-02-15 | 深圳大学 | Fetal ultrasound image quality control method and system |
| CN107798678A (en) * | 2017-12-11 | 2018-03-13 | 成都优途科技有限公司 | Type B ultrasonoscopy the best visual effect Automatic Optimal system and control method |
| CN108056789A (en) * | 2017-12-19 | 2018-05-22 | 飞依诺科技(苏州)有限公司 | A kind of method and apparatus for the configuration parameter value for generating ultrasound scanning device |
| CN108378869A (en) * | 2017-12-26 | 2018-08-10 | 珠海艾博罗生物技术股份有限公司 | A kind of image processing method and processing system of the automatic measurement fetal head girth degree from ultrasonoscopy |
Non-Patent Citations (2)
| Title |
|---|
| SHAOQING REN: "《Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks》", 《ARXIV:1506.01497V3》 * |
| 吴凌云: "《胎儿腹部超声图像自动质量控制与参数测量》", 《中国优秀硕士学位论文全文数据库医药卫生科技辑》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110338842A (en) * | 2019-07-11 | 2019-10-18 | 北京市朝阳区妇幼保健院 | A kind of image optimization method of newborn's lungs ultrasonic image-forming system |
| CN110464380A (en) * | 2019-09-12 | 2019-11-19 | 李肯立 | A kind of method that the ultrasound cross-section image of the late pregnancy period fetus of centering carries out quality control |
| CN111062953A (en) * | 2019-12-17 | 2020-04-24 | 北京化工大学 | Method for identifying parathyroid hyperplasia in ultrasonic image |
| CN112070119A (en) * | 2020-08-11 | 2020-12-11 | 长沙大端信息科技有限公司 | Ultrasonic tangent plane image quality control method and device and computer equipment |
| CN112446862A (en) * | 2020-11-25 | 2021-03-05 | 北京医准智能科技有限公司 | Dynamic breast ultrasound video full-focus real-time detection and segmentation device and system based on artificial intelligence and image processing method |
| CN112446862B (en) * | 2020-11-25 | 2021-08-10 | 北京医准智能科技有限公司 | Dynamic breast ultrasound video full-focus real-time detection and segmentation device and system based on artificial intelligence and image processing method |
| CN113378831A (en) * | 2021-05-07 | 2021-09-10 | 太原理工大学 | Mouse embryo organ identification and scoring method and system |
| CN113378831B (en) * | 2021-05-07 | 2022-05-10 | 太原理工大学 | Mouse embryo organ identification and scoring method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109191442B (en) | 2021-04-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109191442A (en) | Ultrasound image assessment and screening technique and device | |
| US11229419B2 (en) | Method for processing 3D image data and 3D ultrasonic imaging method and system | |
| CN108022238A (en) | Method, computer-readable storage medium and the system being detected to object in 3D rendering | |
| CN105389589B (en) | A kind of chest X ray piece rib cage detection method returned based on random forest | |
| US10762630B2 (en) | System and method for structures detection and multi-class image categorization in medical imaging | |
| CN107895367A (en) | A kind of stone age recognition methods, system and electronic equipment | |
| CN107644419A (en) | Method and apparatus for analyzing medical image | |
| CN109544536A (en) | The fast automatic analysis method of hip joint x-ray image | |
| CN107072637A (en) | The apparatus and method detected for automatic pneumothorax | |
| CN110245657A (en) | Pathological image similarity detection method and detection device | |
| Cheng et al. | Automatic Dent-landmark detection in 3-D CBCT dental volumes | |
| CN109817339A (en) | Patient's group technology and device based on big data | |
| US20220261997A1 (en) | Method and apparatus for acquiring blood vessel evaluation parameter based on physiological parameter, and storage medium | |
| CN108564044A (en) | A kind of method and device of determining Lung neoplasm density | |
| CN110580217A (en) | software code health degree detection method, processing method and device and electronic equipment | |
| WO2020242341A1 (en) | Method for selecting and classifying blood cell types by means of deep convolutional neural networks | |
| CN114708543B (en) | Examination student positioning method in examination room monitoring video image | |
| US20220249060A1 (en) | Method for processing 3d image data and 3d ultrasonic imaging method and system | |
| KR20220123518A (en) | Method and device for generating improved surgical report using machine learning | |
| CN112308820A (en) | Rib positioning method and device, computer equipment and storage medium | |
| CN112084889A (en) | Image behavior recognition method and device, computing equipment and storage medium | |
| CN109523547A (en) | A kind of method and device of image tubercle detection | |
| Giorgi et al. | Morphological Analysis of 3D Faces for Weight Gain Assessment. | |
| CN111062935A (en) | Breast tumor detection method, storage medium and terminal device | |
| CN115619835B (en) | Heterogeneous three-dimensional observation registration method, medium and equipment based on depth phase correlation |
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 |