CN114723754B - Ultrasonic hip joint bone age assessment method, system, equipment and storage medium - Google Patents
Ultrasonic hip joint bone age assessment method, system, equipment and storage medium Download PDFInfo
- Publication number
- CN114723754B CN114723754B CN202210638349.2A CN202210638349A CN114723754B CN 114723754 B CN114723754 B CN 114723754B CN 202210638349 A CN202210638349 A CN 202210638349A CN 114723754 B CN114723754 B CN 114723754B
- Authority
- CN
- China
- Prior art keywords
- bone age
- ultrasonic
- image
- hip joint
- bone
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 210000000988 bone and bone Anatomy 0.000 title claims abstract description 102
- 210000004394 hip joint Anatomy 0.000 title claims abstract description 33
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000003860 storage Methods 0.000 title claims abstract description 12
- 238000001514 detection method Methods 0.000 claims abstract description 21
- 239000013598 vector Substances 0.000 claims description 17
- 238000010586 diagram Methods 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 11
- 230000006870 function Effects 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 10
- 230000007246 mechanism Effects 0.000 claims description 9
- 230000004044 response Effects 0.000 claims description 8
- 238000005520 cutting process Methods 0.000 claims description 7
- 238000011156 evaluation Methods 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 230000004927 fusion Effects 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims 2
- 238000004364 calculation method Methods 0.000 claims 1
- 210000003484 anatomy Anatomy 0.000 abstract description 13
- 238000011160 research Methods 0.000 abstract description 5
- 238000011161 development Methods 0.000 abstract description 4
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 230000001737 promoting effect Effects 0.000 abstract description 2
- 210000001624 hip Anatomy 0.000 description 5
- 238000002604 ultrasonography Methods 0.000 description 4
- 230000018109 developmental process Effects 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 238000002372 labelling Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000012488 skeletal system development Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 208000006096 Attention Deficit Disorder with Hyperactivity Diseases 0.000 description 1
- 208000036864 Attention deficit/hyperactivity disease Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 208000015802 attention deficit-hyperactivity disease Diseases 0.000 description 1
- -1 carrier Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000003153 chemical reaction reagent Substances 0.000 description 1
- 239000000306 component Substances 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 210000001981 hip bone Anatomy 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 210000000236 metacarpal bone Anatomy 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000001568 sexual effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T2207/20076—Probabilistic image processing
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30008—Bone
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Ultra Sonic Daignosis Equipment (AREA)
Abstract
The invention discloses an ultrasonic hip joint bone age assessment method, a system, equipment and a storage medium, belonging to a weak supervision ultrasonic hip joint bone age assessment scheme based on anatomical region detection, comprising two stages: in the first stage, the anatomical key area can be accurately positioned, and on one hand, the accuracy of subsequent bone age prediction can be improved; on the other hand, interpretable anatomical key areas can be provided, which is beneficial to promoting the further development of medical research; in the second stage, the bone age is regressed and predicted on the basis of the regions of the anatomy key with interpretability.
Description
Technical Field
The invention relates to the technical field of bone age assessment, in particular to an ultrasonic hip joint bone age assessment method, system, equipment and storage medium.
Background
Bone age assessment is a technique for inferring the maturity of skeletal development in children from skeletal morphology.
Currently, the mainstream bone age assessment method uses left-handed metacarpal X-ray images (radiographic images) to estimate bone age by comprehensively analyzing fine visual features extracted from a plurality of key areas of skeletal anatomy, so anatomical region detection is a key problem in bone age assessment. Many existing bone age assessment algorithms are highly dependent on additional manual labeling information in order to achieve satisfactory assessment results. Under the supervision of a hand marking frame, the detection of the metacarpal bone area can be completed, so that the interference of background noise is eliminated; under the supervision of the bone key points, the bone key areas with anatomical significance can be accurately positioned, and further the morphological structure of the bone key areas is subjected to refined feature extraction and analysis. However, accurate and fine manual labeling information is very expensive to acquire, and practical application of the bone age assessment algorithms is limited to a great extent; meanwhile, compared with ultrasonic imaging, the radiographic imaging technology has great limitations in imaging, safety and cost, and is not favorable for wide basic deployment of bone age assessment equipment.
In recent years, many researches tend to finish bone age assessment by excavating structural features of images without using fine manual labeling information, the researches belong to a weak supervision method, and the weak supervision method has wider application value compared with the weak supervision method, but the scheme for performing ultrasonic bone age assessment based on the weak supervision method at present has poor accuracy and lacks interpretability.
Disclosure of Invention
The invention aims to provide an ultrasonic hip joint bone age assessment method, system, equipment and storage medium, which can obviously reduce the average absolute error of bone age assessment, improve the accuracy of bone age assessment results and provide interpretability for the ultrasonic hip joint bone age assessment.
The purpose of the invention is realized by the following technical scheme:
an ultrasonic hip joint bone age assessment method comprising:
the first stage is as follows: automatically positioning an anatomy key area in the ultrasonic hip joint image by using an attention mechanism, and cutting out an image of the anatomy key area;
and a second stage: and performing feature extraction on the image of the anatomical key area, and performing bone age assessment by using a feature vector obtained by feature extraction.
An ultrasonic hip joint bone age assessment system, the system comprising:
the detection network automatically positions an anatomical key area in the ultrasonic hip joint image by using an attention mechanism and cuts out an image of the anatomical key area;
and the regression network is used for extracting the characteristics of the image of the anatomical key area and evaluating the bone age by using the characteristic vector obtained by characteristic extraction.
A processing device, comprising: one or more processors; a memory for storing one or more programs;
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the aforementioned methods.
A readable storage medium, storing a computer program which, when executed by a processor, implements the aforementioned method.
The technical scheme provided by the invention can be seen that the scheme belongs to a weakly supervised ultrasonic hip joint bone age assessment scheme based on anatomical region detection, and comprises two stages: in the first stage, the key anatomical region can be accurately positioned, and on one hand, the accuracy of subsequent bone age prediction can be improved; on the other hand, interpretable anatomical key areas can be provided, which is beneficial to promoting the further development of medical research; the second stage, regression prediction of bone age based on the interpretable anatomical critical area, with higher accuracy.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the description below are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic diagram of an ultrasonic hip joint bone age assessment method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an automatic detection stage of an anatomical key region according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a bone age assessment phase according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of an ultrasonic hip bone age assessment system according to an embodiment of the present invention;
fig. 5 is a schematic diagram of a processing apparatus according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
The terms that may be used herein are first described as follows:
the terms "comprising," "including," "containing," "having," or other similar terms in describing these terms are to be construed as non-exclusive inclusions. For example: including a feature (e.g., material, component, ingredient, carrier, formulation, material, dimension, part, component, mechanism, device, process, procedure, method, reaction condition, processing condition, parameter, algorithm, signal, data, product, or article of manufacture), is to be construed as including not only the particular feature explicitly listed but also other features not explicitly listed as such which are known in the art.
The present invention provides a method, system, device and storage medium for ultrasonic hip joint bone age assessment. Details which are not described in detail in the embodiments of the invention belong to the prior art which is known to the person skilled in the art. Those not specifically mentioned in the examples of the present invention were carried out according to the conventional conditions in the art or conditions suggested by the manufacturer. The reagents or instruments used in the examples of the present invention are not specified by manufacturers, and are all conventional products available by commercial purchase.
Example one
The embodiment of the invention provides an ultrasonic hip joint bone age assessment method, which is a weak supervision ultrasonic hip joint bone age assessment scheme based on anatomical region detection, and the overall flow of the method is shown in figure 1 and mainly comprises the following steps:
the first stage is as follows: and automatically positioning an anatomical key area in the ultrasonic hip joint image by using an attention mechanism, and cutting out an image of the anatomical key area.
And a second stage: and performing feature extraction on the image of the anatomical key area, and performing bone age assessment by using a feature vector obtained by feature extraction.
For ease of understanding, the following description is directed to a preferred embodiment of the two stages above.
The first stage.
In the embodiment of the invention, the first stage is an automatic detection stage of the anatomical key area, which is realized by a detection network, the detection network automatically positions the most distinctive anatomical area in the ultrasonic hip joint image by using an attention mechanism, performs threshold cutting on the anatomical area to obtain an interpretable key local area, and provides the interpretable key local area for the second stage to perform bone age assessment and prediction.
As shown in fig. 2, the main principle of the automatic detection stage of the anatomical key region is demonstrated. The following is a description of both network training and obtaining anatomical key regions.
(1) And (5) network training.
The input image is an ultrasonic hip joint image, the image is adjusted to a specified size (448 x 448 pixels are set in the invention without affecting generality), and then feature extraction is carried out through a first backbone network to obtain a feature map(ii) a Wherein,Rrepresenting real numbersThe collection of the data is carried out,Cthe number of channels is indicated and indicated,HandWthe height and the width of the feature map extracted by the first backbone network are respectively represented, and the specific size of the feature map varies with the first backbone network.
In the embodiment of the present invention, the first backbone network may be implemented by using a feature extraction network in an existing neural network, and a specific network structure is not limited.
Performing global average pooling on the feature map, and converting the feature map into a feature vector indicating channel responseExpressed as:
And then the first full-connection layer transforms the characteristic vector of the indication channel response to obtain an output scoreExpressed as:
wherein,is the weight parameter of the first fully-connected layer,of 1 attAn elementMeans evaluation of bone age astTime corresponding score,t=1,2,…,T,TIs the maximum value of the bone age label and the unit is day. Subscripts of the symbols of the inventiontIndicating the sequence number in the set of the corresponding element, the sequence numbertEquivalent to the corresponding bone aget。
For a true bone age of(known information) ultrasound hip image using label-smoothed vectorsAs a label for the age of a bone,is thatTo (1) atAn element calculated by the formula:
wherein,eis a natural constant and is a natural constant,a super parameter (the size of which can be set according to actual conditions or experience) for controlling the smoothness degree. In order to obtain a higher response of the detection network on the corresponding bone age of the image, thereby obtaining a more accurate and more explanatory attention diagram, a first loss function is constructed by combining the output score and the bone age label, specifically, the detection network can be constrained by using cross entropy loss as a loss function (first loss function) of a first stage, which is expressed as:
wherein,indicates that bone age is evaluated astThe probability of time correspondence is calculated by the following formula:
when classified according to bone age, the more interesting the area of the detection network is, the higher its attention map response is, the more distinctive the features are. The areas corresponding to the highest response in the attention-deficit hyperactivity disorder are critical anatomical areas. In the embodiment of the invention, the aim of detecting the network training is to obtain the optimal weight parameter of the first full-connection layer. According toAn attention map is generated, and images of the anatomical key regions located are automatically detected according to the attention map response values.
(2) An anatomically critical region is obtained.
The main flow of the part is as follows:
extracting a feature map through a first backbone networkObtaining an output score through global average pooling and a first full link layer(the related processing is the same as the training stage, so it is not repeated, since this part is realized by the trained detection network, so the symbols of the related parameters are distinguished), and the output score is usedCalculate bone age estimate astProbability of time correspondence, t=1,2,…,TThe formula is the same as the training phase and is expressed as:
wherein,andrespectively is an output scoreTo (1)tIs first and secondiIndividual elements, corresponding representations assess bone age astAndithe corresponding score.
Selecting the most probable correspondencestIs marked asAnd obtaining the bone age pseudo label by using the label smoothing formulaIn particular, bone age pseudo-tagWhereinis a bone age pseudo labelTo (1) atAn element calculated by the formula:
meanwhile, the weight parameter of the first full connection layer obtained in the training stage is utilizedFor characteristic diagramPerforming transformation to obtainTZhang attention force chart (weighting characteristic picture)Expressed as:
Thereafter, bone age pseudo-tags are utilizedTo the attention force chartCarrying out weighted fusion to obtain corresponding new attention diagramExpressed as:
will give a new attention mapAdjusted to the same size as the ultrasonic hip image (448 x 448 pixels)Obtaining an adjusted attention map. In thatSelecting a rectangular area, enabling the rectangular area to comprise all pixel points with response values larger than a threshold k and the area of the pixel points to be minimum, cutting out a corresponding image from the ultrasonic hip joint image according to the rectangular area, using the image as an image of an anatomical key area with detail information reserved, and applying the image to training of a related network in a second stage; wherein k is a positive integer and can be set according to actual conditions or experience.
Second, second stage.
In the embodiment of the invention, the second stage is a bone age evaluation stage and is realized through a regression network. And the regression network learns the characteristics of the images of the anatomical key regions according to the images of the anatomical key regions obtained by the attention mechanism, and a final bone age prediction result is obtained by combining the sex information.
As shown in fig. 3, the main principle of the bone age assessment stage includes:
(1) extracting feature maps from images of the anatomical key region through a second backbone network,Andrespectively representing the height and width of the feature map extracted by the second backbone network. Similarly, the second backbone network may be implemented by using a feature extraction network in an existing neural network, and a specific network structure is not limited.
(2) Characteristic diagramWill pass through maximum pooling operationObtaining corresponding feature vectors。
considering that the change of the sexual skeletal development speed of men and women is obviously different, the bones of women mature more quickly in the early development stage compared with men, and if the sex information is not considered, the bone age prediction result has larger error. The existing work converts the gender information into a characteristic vector, and the characteristic vector is directly spliced with the image characteristic vector, but a gap exists between the characteristics of different modes, and the gender information cannot be effectively utilized by direct splicing. Therefore, in the embodiment of the invention, the regression network is set to be a double-branch parallel network structure, and corresponding branches can be selected according to the gender information to carry out feature extraction and bone age assessment. Specifically, in a training stage, a double-branch parallel network in the regression network is trained by using samples of different genders respectively; in the testing stage, the testing sample (i.e. the image of the anatomical key region obtained in the first stage) is sent to the corresponding network branch for prediction according to the gender information.
In the embodiment of the invention, in order to directly constrain the bone age estimation result to obtain higher prediction accuracy, in the training stage, the bone age estimation result is used to construct a second loss function, which is expressed as:
In the embodiment of the invention, the detection network and the regression network are trained independently by using corresponding loss functions respectively.
It should be noted that the contents of the ultrasound hip images and the attention diagrams shown in fig. 1 to fig. 3 are only examples, and are not limited to the examples, and in practical applications, the input images may be ultrasound hip images with different contents, and the attention diagrams with different contents may also be obtained according to different network parameters.
Compared with the existing scheme, the scheme of the embodiment of the invention obviously reduces the average absolute error of bone age assessment: mean absolute error of 16.24 days was achieved on the ultrasound hip image bone age assessment dataset USBAA. On one hand, the accuracy of the bone age assessment result can be improved; on the other hand, the anatomical critical area located in the first stage of the present invention is an interpretable anatomical area result, which may facilitate further development of medical research, compared to the end-to-end approach.
Example two
The invention also provides an ultrasonic hip joint bone age assessment system, which is implemented mainly based on the method provided by the first embodiment, as shown in fig. 4, the system mainly comprises:
the detection network automatically positions an anatomical key area in the ultrasonic hip joint image by using an attention mechanism and cuts out an image of the anatomical key area;
and the regression network is used for extracting the features of the image of the anatomical key area and evaluating the bone age by using the feature vector obtained by feature extraction.
It is obvious to those skilled in the art that, for convenience and simplicity of description, the above division of each functional module is only used for illustration, and in practical applications, the above function distribution may be completed by different functional modules according to needs, that is, the internal structure of the system is divided into different functional modules to complete all or part of the above described functions.
It should be noted that, the main technical details of each network in the system have been described in detail in the first embodiment, and thus are not described again.
EXAMPLE III
The present invention also provides a processing apparatus, as shown in fig. 5, which mainly includes: one or more processors; a memory for storing one or more programs; wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the methods provided by the foregoing embodiments.
Further, the processing device further comprises at least one input device and at least one output device; in the processing device, a processor, a memory, an input device and an output device are connected through a bus.
In the embodiment of the present invention, the specific types of the memory, the input device, and the output device are not limited; for example:
the input device can be a touch screen, an image acquisition device, a physical key or a mouse and the like;
the output device may be a display terminal;
the Memory may be a Random Access Memory (RAM) or a non-volatile Memory (non-volatile Memory), such as a disk Memory.
Example four
The present invention also provides a readable storage medium storing a computer program which, when executed by a processor, implements the method provided by the foregoing embodiments.
The readable storage medium in the embodiment of the present invention may be provided in the foregoing processing device as a computer readable storage medium, for example, as a memory in the processing device. The readable storage medium may be various media that can store program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a magnetic disk, or an optical disk.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (8)
1. An ultrasonic hip joint bone age assessment method, comprising:
the first stage is as follows: automatically positioning an anatomical key area in the ultrasonic hip joint image by using an attention mechanism, and cutting out an image of the anatomical key area;
and a second stage: extracting the characteristics of the image of the anatomical key area, and performing bone age assessment by using the characteristic vector obtained by the characteristic extraction;
the first stage is realized by a detection network; extracting the features of the ultrasonic hip joint image through the detection network to obtain a feature map; performing global average pooling on the feature map, and converting the feature map into a feature vector indicating channel response; transforming the characteristic vector responding to the indication channel through a first full-connection layer to obtain an output score, and calculating a bone age pseudo label by using the output score; meanwhile, transforming the feature map by using a first full-link layer to obtain an attention map, performing weighted fusion on the attention map by using a bone age pseudo label to obtain a new attention map, and cutting out an image of an anatomical key region by using the new attention map;
in the training stage, constructing a first loss function by using the output score and the bone age label and constraining the detection network;
defining an output scoreOf 1 attAn elementMeans evaluation of bone age astThe score of the time-domain image is obtained,t=1,2,…,T,Tthe maximum value of the bone age label; calculating bone age probability by using the output score:
wherein,eis a natural constant and is a natural constant,means evaluation of bone age astThe probability of the time-of-day correspondence,means evaluation of bone age asiTime corresponding scores;
for a true bone age ofUsing the smoothed vectors of the labelsAs a label for the age of a bone,is a bone age labelTo (1) atAn element calculated by the formula:
the constructed first loss function is expressed as:
2. the method of claim 1, wherein the step of calculating the bone age pseudo label using the output score comprises:
using the output scoreCalculate bone age estimate astProbability of time correspondence, t=1,2,…,TExpressed as:
wherein,and withRespectively, the bone age was evaluated astAndithe score corresponding to the time-point is obtained, Tthe maximum value of the bone age label is obtained;
selecting the most probable correspondencestIs marked asAnd obtaining the bone age pseudo label by using a label smoothing formulaWhereinis a bone age pseudo labelTo (1) atThe calculation formula is expressed as:
3. The ultrasonic hip joint bone age assessment method according to claim 1, wherein the step of performing weighted fusion on the attention map by using bone age pseudo labels to obtain a new attention map, and cropping out an image of an anatomical key region by using the new attention map comprises the steps of:
recording bone age pseudo label asTransforming the feature map to obtain an attention mapUsing a bone age pseudo labelTo the attention force chartPerforming weighted fusion to obtain new attention diagramExpressed as:
wherein,is a bone age pseudo labelTo (1) atThe number of the elements is one,to draw attention toTo (1)tTense, i.e. bone agetA corresponding attention map;
will give a new attention mapAdjusting the size of the ultrasonic hip joint image to be the same as that of the ultrasonic hip joint image, selecting a rectangular area from the attention diagram after size adjustment to enable the rectangular area to contain all pixel points with response values larger than a threshold value and to be the minimum in area, and cutting out a corresponding image from the ultrasonic hip joint image according to the rectangular area to serve as an image of an anatomical key area.
4. The ultrasonic hip joint bone age assessment method according to claim 1, wherein the second stage is realized by a regression network, the regression network adopts a two-branch parallel network structure, corresponding branches are selected according to gender information, feature maps are extracted from the images of the anatomical key region by using the corresponding branches, corresponding feature vectors are obtained through maximum pooling operation, and bone age assessment results are obtained through a second full-connection layer.
5. The method for assessing bone age of an ultrasonic hip joint according to claim 4, wherein in the training stage, the two-branch parallel network in the regression network is trained by using samples of different genders, and a second loss function is constructed by using the bone age assessment result, and is represented as:
6. An ultrasonic hip joint bone age assessment system realized based on the method of any one of claims 1 to 5, comprising:
the detection network automatically positions an anatomical key area in the ultrasonic hip joint image by using an attention mechanism and cuts out an image of the anatomical key area;
and the regression network is used for extracting the characteristics of the image of the anatomical key area and evaluating the bone age by using the characteristic vector obtained by characteristic extraction.
7. A processing device, comprising: one or more processors; a memory for storing one or more programs;
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-5.
8. A readable storage medium, storing a computer program, characterized in that the computer program, when executed by a processor, implements the method according to any of claims 1-5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210638349.2A CN114723754B (en) | 2022-06-08 | 2022-06-08 | Ultrasonic hip joint bone age assessment method, system, equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210638349.2A CN114723754B (en) | 2022-06-08 | 2022-06-08 | Ultrasonic hip joint bone age assessment method, system, equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114723754A CN114723754A (en) | 2022-07-08 |
CN114723754B true CN114723754B (en) | 2022-09-02 |
Family
ID=82233022
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210638349.2A Active CN114723754B (en) | 2022-06-08 | 2022-06-08 | Ultrasonic hip joint bone age assessment method, system, equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114723754B (en) |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111882517B (en) * | 2020-06-08 | 2024-06-21 | 杭州深睿博联科技有限公司 | Bone age evaluation method, system, terminal and storage medium based on graph convolution neural network |
EP4161391A4 (en) * | 2020-06-09 | 2024-06-19 | Annalise-AI Pty Ltd | Systems and methods for automated analysis of medical images |
WO2022005091A1 (en) * | 2020-06-29 | 2022-01-06 | 주식회사 뷰노 | Method and apparatus for reading bone age |
CN112006654B (en) * | 2020-07-16 | 2021-07-13 | 中国科学技术大学 | Bone age assessment method based on convolutional neural network and multiple attention mechanism |
CN114549470B (en) * | 2022-02-23 | 2024-03-05 | 合肥工业大学 | Hand bone critical area acquisition method based on convolutional neural network and multi-granularity attention |
-
2022
- 2022-06-08 CN CN202210638349.2A patent/CN114723754B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN114723754A (en) | 2022-07-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wu et al. | Automatic landmark estimation for adolescent idiopathic scoliosis assessment using BoostNet | |
KR101977174B1 (en) | Apparatus, method and computer program for analyzing image | |
US9076197B2 (en) | Probability density function estimation | |
CN105555198B (en) | Method and device for automatically identifying measurement items and ultrasonic imaging equipment | |
KR102045223B1 (en) | Apparatus, method and computer program for analyzing bone age | |
US11386553B2 (en) | Medical image data | |
CN110853011B (en) | Method for constructing convolutional neural network model for pulmonary nodule detection | |
US20220215548A1 (en) | Method and device for identifying abnormal cell in to-be-detected sample, and storage medium | |
CN111179227A (en) | Mammary gland ultrasonic image quality evaluation method based on auxiliary diagnosis and subjective aesthetics | |
CN112949408B (en) | Real-time identification method and system for target fish passing through fish channel | |
CN108256527A (en) | A kind of cutaneous lesions multiclass semantic segmentation method based on end-to-end full convolutional network | |
CN112215217B (en) | Digital image recognition method and device for simulating doctor to read film | |
CN112132166A (en) | Intelligent analysis method, system and device for digital cytopathology image | |
Bano et al. | AutoFB: automating fetal biometry estimation from standard ultrasound planes | |
CN113261012A (en) | Method, device and system for processing image | |
CN114565572A (en) | Cerebral hemorrhage CT image classification method based on image sequence analysis | |
Ning et al. | Multi-scale gradational-order fusion framework for breast lesions classification using ultrasound images | |
CN110163907B (en) | Method and device for measuring thickness of transparent layer of fetal neck and storage medium | |
CN113570594A (en) | Method and device for monitoring target tissue in ultrasonic image and storage medium | |
CN111528918B (en) | Tumor volume change trend graph generation device after ablation, equipment and storage medium | |
CN114723754B (en) | Ultrasonic hip joint bone age assessment method, system, equipment and storage medium | |
CN113344028A (en) | Breast ultrasound sequence image classification method and device | |
CN111325282B (en) | Mammary gland X-ray image identification method and device adapting to multiple models | |
CN112801940A (en) | Model evaluation method, device, equipment and medium | |
CN113222985B (en) | Image processing method, image processing device, computer equipment and medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |