CN109671062A - Ultrasound image detection method, device, electronic equipment and readable storage medium storing program for executing - Google Patents
Ultrasound image detection method, device, electronic equipment and readable storage medium storing program for executing Download PDFInfo
- Publication number
- CN109671062A CN109671062A CN201811512039.6A CN201811512039A CN109671062A CN 109671062 A CN109671062 A CN 109671062A CN 201811512039 A CN201811512039 A CN 201811512039A CN 109671062 A CN109671062 A CN 109671062A
- Authority
- CN
- China
- Prior art keywords
- ultrasound image
- ultrasound
- training
- group
- image detection
- 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.)
- Pending
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
- 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]
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
A kind of ultrasound image detection method, device, electronic equipment and readable storage medium storing program for executing provided by the embodiments of the present application, pass through the multiple groups ultrasound image pretreatment to acquisition, ultrasound image described in each group includes multiple ultrasound images, multiple ultrasound images in ultrasound image described in same group are obtained by the same primary inspection of same sufferer, after being divided to pretreated data by preset ratio, it is respectively trained and tests the deep neural network model constructed in advance, until the test effect of ultrasound image detection model achieves the desired results.This programme merges same sufferer with multiple ultrasound image datas once checked, effectively reduces average information loss, keeps testing result more accurate.
Description
Technical field
The present invention relates to application image detection fields, in particular to a kind of ultrasound image detection method, device, electricity
Sub- equipment and readable storage medium storing program for executing.
Background technique
Color ultrasound is one of Imaging Method preferred during thyroid nodular nidus checks, high-resolution thyroid gland is color at present
Super inspection is to evaluate the most sensitive method of thyroid nodule.It is now existing by the method for color ultrasonic image automatic identification thyroid cancer
Many researchs, there are mainly two types of methods, one is to every thyroid gland color ultrasonic image, is first partitioned into thyroid gland knot therein
Section, then judge the good pernicious of tubercle, to judge whether examiner suffers from thyroid cancer, another kind is based on directly on individual first shape
Gland color ultrasonic image, by extracting characteristics of image, directly judgement wherein whether there is thyroid cancer.
The detection mode of existing thyroid cancer is all based on individual color ultrasonic image mostly, and in actual clinical inspection
In, doctor judge thyroid nodule it is good pernicious when, generally require the tubercle from multiple angles, need when necessary combine tubercle
The information such as the blood flow signal of surrounding judge the good pernicious of tubercle.Therefore, individual color ultrasonic image often can not include a first shape
The missing of whole features of gland tubercle, information can cause the good pernicious mistaken diagnosis of tubercle to a certain extent.In addition, another kind passes through
It is first partitioned into tubercle and judges good pernicious intelligent diagnosing method again, dependent on nodule segmentation and tubercle two modules of classification, if point
It cuts algorithm and omits tubercle, directly will affect the accuracy of the good pernicious classification of tubercle.
Summary of the invention
In view of this, the purpose of the application is, a kind of ultrasound image detection method, device, electronic equipment and can are provided
Storage medium is read to improve the above problem.
The embodiment of the present application provides a kind of ultrasound image detection method, comprising:
Multiple groups ultrasound image is chosen, the ultrasound image described in each group is pre-processed to obtain data set, wherein each group institute
Stating ultrasound image includes multiple ultrasound images, multiple ultrasound images in ultrasound image described in same group are the same of same sufferer
Obtained by secondary inspection；
The data set is divided into training set and test set by preset ratio, is constructed in advance with training set training
Deep neural network model obtains ultrasound image and detects training pattern；
The test set is inputted into the ultrasound image and detects training pattern, the ultrasound image is tested and detects training mould
Type obtains ultrasound image detection model.
Further, the selection multiple groups ultrasound image, the ultrasound image described in each group are pre-processed to obtain data
The step of collection includes:
Multiple groups ultrasound image is chosen, disease is carried out to each group ultrasound image according to the report of the pathological examination of acquisition and diagnostic result
Become mark, and obtains lesion annotation results；
Frame select and cut carry out lesion mark after color ultrasound detection part in ultrasound image described in each group to constitute data set.
Further, the deep neural network model constructed in advance includes input layer, hidden layer and output layer, described to hide
Layer includes multiple network layers, and each input layer, hidden layer and output layer include multiple neurons.
Further, described that the data set is divided into training set and test set by preset ratio, with the training set
The deep neural network model that training constructs in advance, obtaining the step of ultrasound image detects training pattern includes:
The data set is divided into training set and test set by preset ratio, the training set includes X={ (X1,d1),
(X2, d2),…,(Xi,di) ..., (Xm,dm), wherein Xi={ xi1,xi2..., xij,…,xiN, N indicates one group of ultrasound image
In include ultrasound image number, xijIndicate that the jth ultrasound image in i-th group of ultrasound image, m indicate ultrasound figure in training set
The group number of picture, diIndicate the corresponding lesion annotation results of i-th group of ultrasound image；
Augmentation operation and normalization operation are carried out to each group ultrasound image in the training set, the augmentation operation includes
It rotates and turn over；
The ultrasound image is inputted to the deep neural network model constructed in advance, is calculated using forward calculation and backpropagation
Method optimizes the connection weight between the neuron, obtains ultrasound image and detects training pattern.
Further, described that the ultrasound image is inputted to the deep neural network model constructed in advance, using preceding to meter
It calculates and back-propagation algorithm optimizes the connection weight between the neuron, obtain ultrasound image detection training pattern
Step includes:
The ultrasound image is inputted to the deep neural network model constructed in advance；
The deep neural network model extracts the characteristic set of each ultrasound image respectivelyIndicate the spy that the jth image of i-th group of ultrasound image is extracted in L layer network layer
Sign；
The characteristic set for merging each ultrasound image obtains the fusion feature in the ultrasound imageWherein, αjFor attention weight；
The fusion feature is input to the output layer to carry out classification processing, obtains network output,；
The connection weight, which is updated, using back-propagation algorithm obtains ultrasound image detection training pattern.
Further, the pretreatment includes that lesion mark is carried out to multiple groups ultrasound image, described that the test set is defeated
Enter the ultrasound image detection training pattern, tests the ultrasound image detection training pattern, obtain ultrasound image detection model
The step of include；
The test set is inputted into the ultrasound image detection training pattern and obtains test result；
The test result and the lesion annotation results are compared, statistical test result and lesion mark are tied
The consistent ultrasound image group number of fruit, acquires the accuracy rate of test result；
Compare the accuracy rate and preset value, if the accuracy rate is less than the preset value, uses the training training
Practice the ultrasound image and detect training pattern, until by the ultrasound image when accuracy rate is more than or equal to the preset value
Training pattern is detected as ultrasound image detection model.
Further, after the step of acquisition ultrasound image detection model, the method also includes:
Obtain raw ultrasound image；
The color ultrasound detection part cut in the raw ultrasound image constitutes testing image, and the raw ultrasound image is same
The same of one sufferer once checks multiple ultrasound images obtained；
Lesion detection is carried out to the testing image using the ultrasound image detection model, to obtain testing result.
The embodiment of the present application also provides a kind of ultrasound image detection device, comprising:
Preprocessing module, for choosing multiple groups ultrasound image, the ultrasound image described in each group is pre-processed to obtain number
According to collection, wherein ultrasound image described in each group includes multiple ultrasound images, multiple ultrasound images in ultrasound image described in same group
Obtained by same primary inspection for same sufferer；
Training module, for the data set to be divided into training set and test set by preset ratio, with the training set
The deep neural network model that training constructs in advance obtains ultrasound image and detects training pattern；
Test module detects training pattern for the test set to be inputted the ultrasound image, tests the ultrasound figure
As detection training pattern, ultrasound image detection model is obtained.
The embodiment of the present application also provides a kind of electronic equipment, and the electronic equipment includes:
Storage medium；
Processor；
Ultrasound image detection device, the ultrasound image detection device are stored in the storage medium and include described
The software function module that processor executes, described device include:
Preprocessing module, for choosing multiple groups ultrasound image, the ultrasound image described in each group is pre-processed to obtain number
According to collection, wherein ultrasound image described in each group includes multiple ultrasound images, multiple ultrasound images in ultrasound image described in same group
Obtained by same primary inspection for same sufferer；
Training module, for the data set to be divided into training set and test set in proportion, with training set training
The deep neural network model constructed in advance obtains ultrasound image and detects training pattern；
Test module detects training pattern for the test set to be inputted the ultrasound image, tests the ultrasound figure
As detection training pattern, ultrasound image detection model is obtained.
The embodiment of the present application also provides a kind of readable storage medium storing program for executing, and computer journey is stored in the readable storage medium storing program for executing
Sequence, the computer program, which is performed, realizes above-mentioned ultrasound image detection method.
A kind of ultrasound image detection method, device, electronic equipment and readable storage medium storing program for executing provided by the embodiments of the present application lead to
The multiple groups ultrasound image pretreatment to acquisition is crossed, ultrasound image described in each group includes multiple ultrasound images, ultrasound described in same group
Multiple ultrasound images in image are to draw to pretreated data by preset ratio obtained by the same primary inspection of same sufferer
After point, it is respectively trained and tests the deep neural network model constructed in advance, until the test effect of ultrasound image detection model
It achieves the desired results.This programme merges same sufferer with multiple ultrasound image datas once checked, effectively reduces average information
It loses, keeps testing result more accurate.
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 range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 is the structural block diagram of electronic equipment provided by the embodiments of the present application.
Fig. 2 is the flow chart of ultrasound image detection method provided by the embodiments of the present application.
Fig. 3 is the flow chart of the sub-step of step S1 in Fig. 2.
Fig. 4 is the schematic diagram that ultrasound image detection method center provided by the embodiments of the present application selects color ultrasound live part.
Fig. 5 is the flow chart of the sub-step of step S2 in Fig. 2.
Fig. 6 is the flow chart of the sub-step of step S3 in Fig. 2.
Fig. 7 is another flow chart of ultrasound image detection method provided by the embodiments of the present application.
Icon: 100- electronic equipment；110- ultrasound image detection device；120- processor；130- storage medium.
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.
Therefore, the detailed description of the embodiment of the present invention provided in the accompanying drawings is not intended to limit below claimed
The scope of the present invention, but be merely representative of selected embodiment of the invention.Based on the embodiment of the present invention, those skilled in the art
Member's every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
It should also be noted that similar label and letter indicate similar terms in following attached drawing, therefore, once a certain Xiang Yi
It is defined in a attached drawing, does not then need that it is further defined and explained in subsequent attached drawing.
As shown in Figure 1, the electronic equipment 100 includes that storage is situated between the embodiment of the invention provides a kind of electronic equipment 100
Matter 130, processor 120 and ultrasound image detection device 110.
Directly or indirectly be electrically connected between the storage medium 130 and processor 120, with realize data transmission or
Interaction.It is electrically connected for example, these elements can be realized between each other by one or more communication bus or signal wire.It is described super
Acoustic image detection device 110 includes that at least one can be stored in the storage medium in the form of software or firmware (firmware)
Software function module in 130.The processor 120 is for executing the executable calculating stored in the storage medium 130
Machine program, for example, software function module and computer program etc. included by the ultrasound image detection device 110, to realize
Ultrasound image detection method.
Wherein, the storage medium 130 may be, but not limited to, random access memory (Random Access
Memory, RAM), read-only memory (Read Only Memory, ROM), programmable read only memory (Programmable
Read-Only Memory, PROM), erasable read-only memory (Erasable Programmable Read-Only
Memory, EPROM), electricallyerasable ROM (EEROM) (Electric Erasable Programmable Read-Only
Memory, EEPROM) etc..Wherein, storage medium 130 is for storing program, and the processor 120 is executed instruction receiving
Afterwards, described program is executed.
The processor 120 may be a kind of IC chip, the processing capacity with signal.Above-mentioned processor
120 can be general processor, including central processing unit (Central Processing Unit, CPU), network processing unit
(Network Processor, NP) etc.；It can also be digital signal processor (DSP), specific integrated circuit (ASIC), scene
Programmable gate array (FPGA) either other programmable logic device, discrete gate or transistor logic, discrete hardware group
Part.It may be implemented or execute disclosed each method, step and the logic diagram in the embodiment of the present invention.General processor can be with
It is that microprocessor or the processor 120 are also possible to any conventional processor etc..
It is appreciated that structure shown in FIG. 1 is only to illustrate, the electronic equipment 100 may also include more than shown in Fig. 1
Perhaps less component or with the configuration different from shown in Fig. 1.Each component shown in Fig. 1 can use hardware, software
Or combinations thereof realize.
Optionally, the concrete type of the electronic equipment 100 is unrestricted, for example, it may be, but be not limited to, intelligent hand
Machine, PC (personal computer, PC), tablet computer, personal digital assistant (personal digital
Assistant, PDA), mobile internet surfing equipment (mobile Internet device, MID), web (website) server, data
The equipment that server etc. has processing function.
In conjunction with Fig. 2, the embodiment of the present invention also provides a kind of ultrasound image detection side that can be applied to above-mentioned electronic equipment 100
Method.Wherein, method and step defined in the related process of the method can be realized by the processor 120.
Detailed process shown in Fig. 2 will be described in detail by taking Thyroid ultrasound image as an example below.
The method includes S1, S2, S3.
S1, chooses multiple groups ultrasound image, and the ultrasound image described in each group is pre-processed to obtain data set, wherein each
The group ultrasound image includes multiple ultrasound images, multiple ultrasound images in ultrasound image described in same group are same sufferer
With obtained by primary check.
In order to detect whether examiner suffers from thyroid cancer, radiographer can be to the thyroid gland of examiner when checking every time
And bilateral neck acquires the color ultrasonic image of multiple different angles, to wherein find there may be the thyroid nodules of lesion to adopt
Collect the color ultrasonic image of the diversified forms such as its blood flow signal, spectrum signal.Detection method is collected more by a sufferer
It opens image to be input in detection model, the ultrasound image of multiple same sufferers is passed in model as a data group to be examined
It surveys.
Deep neural network (Deep Neural Networks, DNNs) in thyroid gland color ultrasonic image data to whether there is
The detection of thyroid cancer is automatically learnt from a large amount of thyroid gland color ultrasonic images to thyroid cancer and other thyroid diseases disease
Become feature.Therefore, it before being trained deep neural network, needs to be labeled a large amount of thyroid gland color ultrasonic image and in advance
Processing, and data set is accordingly divided.
Therefore referring to Fig. 3, the S1 specifically includes S11 and S12.
S11, choose multiple groups ultrasound image, according to the pathological examination of acquisition report and diagnostic result to each group ultrasound image into
Row lesion mark, and obtain lesion annotation results.
When examiner checks every time, radiographer acquires multiple to the thyroid gland of examiner and periphery region of interest respectively
Color ultrasonic image, the single of some examiner check that resulting every color ultrasonic image is one group of image data.The radiologic technologist of profession
The definite position that thyroid gland lesion occurs for every group of image data is labeled and provides the type of lesion.When radiologist not
When whether determine the tubercle of lesion has cancer, it will usually it is recommended that examiner further checks, that is, pass through pathology puncture or pathological section
Etc. modes determine the good pernicious of tubercle.Common thyroid gland lesion nodosity goitre, Hashimoto thyroiditis, thyroid gland
Tumor, thyroid cancer etc..Color ultrasonic image used herein all carries out data according to radiographer and pathological replacement diagnostic result
Mark.
S12, frame select and cut carry out lesion mark after color ultrasound examination part in ultrasound image described in each group to constitute number
According to collection.
Referring to Fig. 4, Fig. 4 is the schematic diagram that frame selects color ultrasound live part.Specifically, every color ultrasonic image is cut out
It cuts, the color ultrasound examination part in every color ultrasonic image is automatically positioned, is marked using outline border A, and cut.
The data set is divided into training set and test set by preset ratio by S2, with the preparatory structure of training set training
The deep neural network model built obtains ultrasound image and detects training pattern.
Specifically, the deep neural network model constructed in advance includes input layer, hidden layer and output layer, the hidden layer
Comprising multiple network layers, each input layer, hidden layer and output layer include multiple neurons.Each adjacent network layer
The neuron between can be full connection and be also possible to local connection.Wherein, the adjacent network layer refers to by data
Flow to the layer successively passed through.
In the embodiment of the present application, the preset ratio for dividing data set can be 4:1, in practical applications, can basis
The ratio is arranged in demand, it is not limited here.
Referring to Fig. 5, the S2 further includes S21, S22 and S23.
The data set is divided into training set and test set by preset ratio by S21, and the training set includes X={ (X1,
d1),(X2,d2),…,(Xi,di),…,(Xm,dm), wherein Xi={ xi1,xi2,…,xij,…,xiN, N indicates one group of ultrasound figure
The ultrasound image number for including as in, xijIndicate that the jth ultrasound image in i-th group of ultrasound image, m indicate ultrasound in training set
The group number of image, diIndicate the corresponding lesion annotation results of i-th group of ultrasound image.
Wherein, it should be noted that N can be the natural numbers such as 2,3,4,5 and 6, and specific ultrasound image number is different
It can change in ultrasound image group, it is not limited here.
S22 carries out augmentation operation and normalization operation, the augmentation operation to each group ultrasound image in the training set
Including rotating, overturning.
In order to allow the deep neural network model that there is better performance, in the training stage, using sides such as rotation, overturnings
Method carries out augmentation operation to each group ultrasound image, to obtain bigger data acquisition system for deep neural network model training.
Specifically, to each group ultrasound image, directly use triple channel (RGB) normalized value of ultrasound image as institute
State the input value of deep neural network model.To the output layer of the deep neural network model, due to the invention solves
It is two classification problem of image, i.e. output two is different classes of, by being encoded to classification, such as i-th of image pattern category
In the 1st class, then, target output is expressed as di=[1,0]T, conversely, the output of its target is expressed as di=[0,1]T。
The ultrasound image is inputted the deep neural network model constructed in advance by S23, uses forward calculation and reversed biography
It broadcasts algorithm to optimize the connection weight between the neuron, obtains ultrasound image and detect training pattern.
Further, the S23 includes S231, S232, S233, S234 and S235.
The ultrasound image is inputted the deep neural network model constructed in advance by S231.
S232, the deep neural network model extract the characteristic set of each ultrasound image respectivelyIndicate the spy that the jth image of i-th group of ultrasound image is extracted in L layer network layer
Sign.
Specifically, in the embodiment of the present invention, for the network layer containing L layers, l layers to l+1 layers of company of network layer
Connecing weight matrix is Wl.If the activation primitive of l layers of epineural member be f (), from input layer to output layer, constantly carry out before to
It calculates, by taking full articulamentum as an example, process are as follows:
Wherein, ai lIndicate the activation value of i-th of neuron of l layer network layer.Activation primitive uses ReLU
(Rectified Linear Unit, line rectification function),
It successively calculates, sample X can be obtainediIn all images characteristic set
S233, the characteristic set for merging each ultrasound image obtain the fusion feature in the ultrasound imageWherein, αjFor attention weight.
The fusion feature is input to the output layer to carry out classification processing by S234, obtains network output.
Specifically, after obtaining the characteristic set, merge using feature of the attention mechanism to multiple images
To sample XiFeatureIts process are as follows:
Wherein,For attention weight set, obtained by attention network.
By fused featureSorter network is inputted, the network output of the last layer is obtainedThe output layer is adopted
With softmax classifier, use cross entropy (Cross Entropy) function as loss function (Loss Function), it is described
Loss function isWherein,For diTransposed matrix.
S235 updates the connection weight using back-propagation algorithm and obtains ultrasound image detection training pattern.S235 is also
Including S2351, S2352, S2353 and S2354.
Specifically, S2351 calculates the susceptibility of output layer
Wherein,For the net input of the output layer, diFor target output.
S2352 successively calculates forward the susceptibility of each layer from output layer by taking full articulamentum as an example
S2353 calculates corresponding gradient by susceptibility
S2354 updates corresponding weight:
Wl←Wl-ηΔWl
Wherein η is the Learning Step of right value update.
The test set is inputted the ultrasound image and detects training pattern by S3, tests the ultrasound image detection training
Model obtains ultrasound image detection model.
Referring to Fig. 6, S3 further includes S31, S32 and S33.
The test set is inputted the ultrasound image detection training pattern and obtains test result by S31.
The test of deep neural network model be the designed neural network model of test after training whether can to from
Can the data not learnt be correctly classified, i.e., correctly judge wherein whether contain first to new thyroid gland color ultrasonic image
Shape gland cancer.It can be used for evaluating designed deep neural network performance.During the test, it inputs in the test set
Multiple groups thyroid gland color ultrasonic image.The activation of network output layer neuron is calculated by trained deep neural network
Value, and this group of image generic is predicted according to activation value.This process detects training pattern by the ultrasound image completely
Complete independently.
S32 compares the test result and the lesion annotation results, statistical test result and the lesion mark
The consistent ultrasound image group number of result is infused, the accuracy rate of test result is acquired.
S33, the accuracy rate and preset value use the training if the accuracy rate is less than the preset value
The collection training ultrasound image detects training pattern, until by the ultrasound when accuracy rate is more than or equal to the preset value
Image detection training pattern is as ultrasound image detection model.
Specifically, the test result of ultrasound image detection training pattern output and the lesion annotation results are carried out
It compares, the correct sample group number of statistical forecast, calculates accuracy rate.When accuracy rate reaches the preset value, then thyroid gland coloured silk is completed
Training of the hypergraph as the deep neural network of classification problem.Conversely, carrying out S2 continues training pattern, until the accuracy rate is greater than
Or using ultrasound image detection training pattern as ultrasound image detection model when being equal to the preset value.
Referring to Fig. 7, actual application method also wraps after the step of acquisition ultrasound image detection model
It includes:
S100 obtains raw ultrasound image.
S200, the color ultrasound examination part cut in the raw ultrasound image constitute testing image, the raw ultrasound figure
As being that the same of same sufferer once checks multiple ultrasound images obtained.
Specifically, the color ultrasound examination part in every color ultrasonic image can be automatically positioned, red frame is used referring again to Fig. 4
It marks, and cuts.
S300 carries out lesion detection to the testing image using the ultrasound image detection model, to obtain detection knot
Fruit.
After this method can be used for training the ultrasound image detection model, the ultrasound to be detected of new unknown result is used
Image judges whether the position in ultrasound image occurs lesion with this, can save a large amount of manpower and material resources for medical diagnosis, save
A part work of radiologic technologist, the development for facilitating the diagnosis and treatment of doctors' resource such as township hospital hospital not abundant to work are gone.
The application also provides a kind of ultrasound image detection device 110, comprising:
Preprocessing module, for choosing multiple groups ultrasound image, the ultrasound image described in each group is pre-processed to obtain number
According to collection, wherein ultrasound image described in each group includes multiple ultrasound images, multiple ultrasound images in ultrasound image described in same group
Obtained by same one-time detection for same sufferer.
Training module, for the data set to be divided into training set and test set by preset ratio, with the training set
The deep neural network model that training constructs in advance obtains ultrasound image and detects training pattern.
Test module detects training pattern for the test set to be inputted the ultrasound image, tests the ultrasound figure
As detection training pattern, ultrasound image detection model is obtained.
It is understood that the concrete operation method of each functional module in the present embodiment can refer to above method embodiment
The detailed description of middle corresponding steps, it is no longer repeated herein.
A kind of readable storage medium storing program for executing is stored with computer program, the computer program quilt in the readable storage medium storing program for executing
Above-mentioned ultrasound image detection method is realized when execution.
In conclusion a kind of ultrasound image detection method provided by the embodiments of the present application, device, electronic equipment 100 and can
Storage medium is read, by the multiple groups ultrasound image pretreatment to acquisition, ultrasound image described in each group includes multiple ultrasound images, together
Multiple ultrasound images in ultrasound image described in one group are obtained by the same primary inspection of same sufferer, to pretreated data
After dividing by preset ratio, it is respectively trained and tests the deep neural network model constructed in advance, until ultrasound image detects mould
The test effect of type achieves the desired results.This programme merges same sufferer with multiple ultrasound image datas once checked, effectively
It reduces average information to lose, keeps testing result more accurate.
In embodiment provided herein, it should be understood that disclosed device and method, it can also be by other
Mode realize.The apparatus embodiments described above are merely exemplary, for example, the flow chart and block diagram in attached drawing are shown
Architectural framework in the cards, function and the behaviour of devices in accordance with embodiments of the present invention, method and computer program product
Make.In this regard, each box in flowchart or block diagram can represent a part of a module, section or code, institute
The a part for stating module, section or code includes one or more executable instructions for implementing the specified logical function.
It should also be noted that function marked in the box can also be to be different from attached drawing in some implementations as replacement
The sequence marked occurs.For example, two continuous boxes can actually be basically executed in parallel, they sometimes can also be by
Opposite sequence executes, and this depends on the function involved.It is also noted that each box in block diagram and or flow chart,
And the combination of the box in block diagram and or flow chart, hardware can be based on the defined function of execution or the dedicated of movement
System realize, or can realize using a combination of dedicated hardware and computer instructions.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, method, article or equipment for including a series of elements not only includes those elements, and
And further include other elements that are not explicitly listed, or further include for this process, method, article or equipment institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including institute
State in the process, method, article or equipment of element that there is also other identical elements.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the invention may be variously modified and varied.All within the spirits and principles of the present invention, made any to repair
Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.It should also be noted that similar label and letter exist
Similar terms are indicated in following attached drawing, therefore, once being defined in a certain Xiang Yi attached drawing, are then not required in subsequent attached drawing
It is further defined and explained.
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, protection scope of the present invention should be based on the protection scope of the described claims.
Claims (10)
1. a kind of ultrasound image detection method characterized by comprising
Multiple groups ultrasound image is chosen, the ultrasound image described in each group is pre-processed to obtain data set, wherein surpass described in each group
Acoustic image includes multiple ultrasound images, multiple ultrasound images in ultrasound image described in same group are the same primary inspection of same sufferer
Obtained by looking into；
The data set is divided into training set and test set by preset ratio, the depth constructed in advance with training set training
Neural network model obtains ultrasound image and detects training pattern；
The test set is inputted into the ultrasound image and detects training pattern, the ultrasound image detection training pattern is tested, obtains
Obtain ultrasound image detection model.
2. ultrasound image detection method according to claim 1, which is characterized in that the selection multiple groups ultrasound image, it is right
The step of ultrasound image described in each group is pre-processed to obtain data set include:
Multiple groups ultrasound image is chosen, lesion mark is carried out to each group ultrasound image according to the report of the pathological examination of acquisition and diagnostic result
Note, and obtain lesion annotation results；
Frame select and cut carry out lesion mark after color ultrasound detection part in ultrasound image described in each group to constitute data set.
3. ultrasound image detection method according to claim 2, which is characterized in that the deep neural network mould constructed in advance
Type includes input layer, hidden layer and output layer, and the hidden layer includes multiple network layers, each input layer, hidden layer and defeated
Layer includes multiple neurons out.
4. ultrasound image detection method according to claim 3, which is characterized in that described to compare the data set by default
Example is divided into training set and test set, and the deep neural network model constructed in advance with training set training obtains ultrasound figure
As the step of detection training pattern includes:
The data set is divided into training set and test set by preset ratio, the training set includes X={ (X1, d1), (X2,
d2) ..., (Xi, di) ..., (XM,dm), wherein Xi={ xi1, xi2... xij..., xiN, N indicate one group of ultrasound image in include
Ultrasound image number, xijIndicate that the jth ultrasound image in i-th group of ultrasound image, m indicate the group of ultrasound image in training set
Number, diIndicate the corresponding lesion annotation results of i-th group of ultrasound image；
Augmentation operation and normalization operation are carried out to each group ultrasound image in the training set, the augmentation operation includes rotation
And overturning；
The ultrasound image is inputted to the deep neural network model constructed in advance, uses forward calculation and back-propagation algorithm pair
Connection weight between the neuron optimizes, and obtains ultrasound image and detects training pattern.
5. ultrasound image detection method according to claim 4, which is characterized in that described to input the ultrasound image in advance
The deep neural network model first constructed, using forward calculation and back-propagation algorithm to the connection weight between the neuron
It optimizes, obtaining the step of ultrasound image detects training pattern includes:
The ultrasound image is inputted to the deep neural network model constructed in advance；
The deep neural network model extracts the characteristic set of each ultrasound image respectively
Indicate the feature that the jth image of i-th group of ultrasound image is extracted in L layer network layer；
The feature for merging each ultrasound image obtains the fusion feature in the ultrasound image
Wherein, αjFor attention weight；
The fusion feature is input to the output layer to carry out classification processing, obtains network output；
The connection weight, which is updated, using back-propagation algorithm obtains ultrasound image detection training pattern.
6. ultrasound image detection method according to claim 1, which is characterized in that the pretreatment includes to multiple groups ultrasound
Image carries out lesion mark, described that the test set is inputted the ultrasound image detection training pattern, tests the ultrasound figure
As the step of detecting training pattern, obtaining ultrasound image detection model includes；
The test set is inputted into the ultrasound image detection training pattern and obtains test result；
The test result and the lesion annotation results are compared, statistical test result and the lesion annotation results one
The ultrasound image group number caused, acquires the accuracy rate of test result；
Compare the accuracy rate and preset value, if the accuracy rate is less than the preset value, uses training set training institute
Ultrasound image detection training pattern is stated, until the accuracy rate detects the ultrasound image when being greater than or equal to the preset value
Training pattern is as ultrasound image detection model.
7. ultrasound image detection method according to claim 1, which is characterized in that described to obtain the ultrasound image detection
After the step of model, the method also includes:
Obtain raw ultrasound image；
It cuts the color ultrasound detection part in the raw ultrasound image and constitutes testing image, the raw ultrasound image is same disease
The same of trouble once checks multiple ultrasound images obtained；
Lesion detection is carried out to the testing image using the ultrasound image detection model, to obtain testing result.
8. a kind of ultrasound image detection device characterized by comprising
Preprocessing module, for choosing multiple groups ultrasound image, the ultrasound image described in each group is pre-processed to obtain data set,
Wherein, ultrasound image described in each group includes multiple ultrasound images, multiple ultrasound images in ultrasound image described in same group are same
Obtained by the same primary inspection of one sufferer；
Training module, for the data set to be divided into training set and test set by preset ratio, with training set training
The deep neural network model constructed in advance obtains ultrasound image and detects training pattern；
Test module detects training pattern for the test set to be inputted the ultrasound image, tests the ultrasound image inspection
Training pattern is surveyed, ultrasound image detection model is obtained.
9. a kind of electronic equipment, which is characterized in that the electronic equipment includes:
Storage medium；
Processor；
Ultrasound image detection device, the ultrasound image detection device are stored in the storage medium and include the processing
The software function module that device executes, described device include:
Preprocessing module, for choosing multiple groups ultrasound image, the ultrasound image described in each group is pre-processed to obtain data set,
Wherein, ultrasound image described in each group includes multiple ultrasound images, multiple ultrasound images in ultrasound image described in same group are same
Obtained by the same primary inspection of one sufferer；
Training module is shifted to an earlier date for the data set to be divided into training set and test set in proportion with training set training
The deep neural network model of building obtains ultrasound image and detects training pattern；
Test module detects training pattern for the test set to be inputted the ultrasound image, tests the ultrasound image inspection
Training pattern is surveyed, ultrasound image detection model is obtained.
10. a kind of readable storage medium storing program for executing, which is characterized in that be stored with computer program, the meter in the readable storage medium storing program for executing
Calculation machine program, which is performed, realizes ultrasound image detection method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811512039.6A CN109671062A (en) | 2018-12-11 | 2018-12-11 | Ultrasound image detection method, device, electronic equipment and readable storage medium storing program for executing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811512039.6A CN109671062A (en) | 2018-12-11 | 2018-12-11 | Ultrasound image detection method, device, electronic equipment and readable storage medium storing program for executing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109671062A true CN109671062A (en) | 2019-04-23 |
Family
ID=66143700
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811512039.6A Pending CN109671062A (en) | 2018-12-11 | 2018-12-11 | Ultrasound image detection method, device, electronic equipment and readable storage medium storing program for executing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109671062A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110136066A (en) * | 2019-05-23 | 2019-08-16 | 北京百度网讯科技有限公司 | Super-resolution method, device, equipment and storage medium towards video |
| CN110378888A (en) * | 2019-07-22 | 2019-10-25 | 新名医(北京)科技有限公司 | A kind of physiology phase monitoring method, device, ultrasonic device and storage medium |
| CN111275689A (en) * | 2020-01-20 | 2020-06-12 | 平安科技（深圳）有限公司 | Medical image identification and detection method and device and computer readable storage medium |
| CN113609971A (en) * | 2021-08-04 | 2021-11-05 | 广州威拓电子科技有限公司 | Method, device and equipment for inspecting microseism observation equipment and storage medium |
| CN113842166A (en) * | 2021-10-25 | 2021-12-28 | 上海交通大学医学院 | Ultrasonic image acquisition method based on ultrasonic imaging equipment and related device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107945870A (en) * | 2017-12-13 | 2018-04-20 | 四川大学 | Retinopathy of prematurity detection method and device based on deep neural network |
| CN108154505A (en) * | 2017-12-26 | 2018-06-12 | 四川大学 | Diabetic retinopathy detection method and device based on deep neural network |
| CN108230311A (en) * | 2018-01-03 | 2018-06-29 | 四川大学 | A kind of breast cancer detection method and device |
| CN108288027A (en) * | 2017-12-28 | 2018-07-17 | 新智数字科技有限公司 | A kind of detection method of picture quality, device and equipment |
| CN108537135A (en) * | 2018-03-16 | 2018-09-14 | 北京市商汤科技开发有限公司 | The training method and device of Object identifying and Object identifying network, electronic equipment |
-
2018
- 2018-12-11 CN CN201811512039.6A patent/CN109671062A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107945870A (en) * | 2017-12-13 | 2018-04-20 | 四川大学 | Retinopathy of prematurity detection method and device based on deep neural network |
| CN108154505A (en) * | 2017-12-26 | 2018-06-12 | 四川大学 | Diabetic retinopathy detection method and device based on deep neural network |
| CN108288027A (en) * | 2017-12-28 | 2018-07-17 | 新智数字科技有限公司 | A kind of detection method of picture quality, device and equipment |
| CN108230311A (en) * | 2018-01-03 | 2018-06-29 | 四川大学 | A kind of breast cancer detection method and device |
| CN108537135A (en) * | 2018-03-16 | 2018-09-14 | 北京市商汤科技开发有限公司 | The training method and device of Object identifying and Object identifying network, electronic equipment |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110136066A (en) * | 2019-05-23 | 2019-08-16 | 北京百度网讯科技有限公司 | Super-resolution method, device, equipment and storage medium towards video |
| CN110136066B (en) * | 2019-05-23 | 2023-02-24 | 北京百度网讯科技有限公司 | Video-oriented super-resolution method, device, equipment and storage medium |
| CN110378888A (en) * | 2019-07-22 | 2019-10-25 | 新名医(北京)科技有限公司 | A kind of physiology phase monitoring method, device, ultrasonic device and storage medium |
| CN110378888B (en) * | 2019-07-22 | 2022-01-25 | 新名医(北京)科技有限公司 | Physiological period monitoring method and device, ultrasonic equipment and storage medium |
| CN111275689A (en) * | 2020-01-20 | 2020-06-12 | 平安科技（深圳）有限公司 | Medical image identification and detection method and device and computer readable storage medium |
| WO2021147218A1 (en) * | 2020-01-20 | 2021-07-29 | 平安科技（深圳）有限公司 | Medical image recognition and analysis method and apparatus, device and storage medium |
| CN113609971A (en) * | 2021-08-04 | 2021-11-05 | 广州威拓电子科技有限公司 | Method, device and equipment for inspecting microseism observation equipment and storage medium |
| CN113842166A (en) * | 2021-10-25 | 2021-12-28 | 上海交通大学医学院 | Ultrasonic image acquisition method based on ultrasonic imaging equipment and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yan et al. | 3D context enhanced region-based convolutional neural network for end-to-end lesion detection | |
| Agaian et al. | A new acute leukaemia-automated classification system | |
| CN112101451B (en) | Breast cancer tissue pathological type classification method based on generation of antagonism network screening image block | |
| CN107280697A (en) | Lung neoplasm grading determination method and system based on deep learning and data fusion | |
| Deng et al. | Classification of breast density categories based on SE-Attention neural networks | |
| CN109770932A (en) | The processing method of multi-modal brain neuroblastoma image feature | |
| CN110021425A (en) | A kind of relatively detector and its construction method and cervical cancer cell detection method | |
| CN109671062A (en) | Ultrasound image detection method, device, electronic equipment and readable storage medium storing program for executing | |
| CN112614133B (en) | Three-dimensional pulmonary nodule detection model training method and device without anchor point frame | |
| CN113808738B (en) | Disease identification system based on self-identification image | |
| CN109934278A (en) | A kind of high-dimensional feature selection method of information gain mixing neighborhood rough set | |
| Sethanan et al. | Double AMIS-ensemble deep learning for skin cancer classification | |
| CN114240874A (en) | Bone age assessment method and device based on deep convolutional neural network and feature fusion and computer readable storage medium | |
| CN114399634B (en) | Three-dimensional image classification method, system, equipment and medium based on weak supervision learning | |
| CN113902669A (en) | Method and system for reading urine exfoliative cell fluid-based smear | |
| Kaliyugarasan et al. | Pulmonary nodule classification in lung cancer from 3D thoracic CT scans using fastai and MONAI | |
| Li et al. | Ct scan synthesis for promoting computer-aided diagnosis capacity of covid-19 | |
| Tang et al. | Detection of COVID-19 using deep convolutional neural network on chest X-ray (CXR) images | |
| Nanthagopal et al. | A region-based segmentation of tumour from brain CT images using nonlinear support vector machine classifier | |
| Wang et al. | MARnet: Multi-scale adaptive residual neural network for chest X-ray images recognition of lung diseases | |
| Ganeshkumar et al. | Two-stage deep learning model for automate detection and classification of lung diseases | |
| CN112784924A (en) | Rib fracture CT image classification method based on grouping aggregation deep learning model | |
| CN107590806A (en) | A kind of detection method and system based on brain medical imaging | |
| CN112690815A (en) | System and method for assisting in diagnosing lesion grade based on lung image report | |
| Abdalla et al. | Transfer learning models comparison for detecting and diagnosing skin cancer |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190423 |
|
| RJ01 | Rejection of invention patent application after publication |