CN110051376A - A kind of stone age intelligent detecting method - Google Patents
A kind of stone age intelligent detecting method Download PDFInfo
- Publication number
- CN110051376A CN110051376A CN201910162542.1A CN201910162542A CN110051376A CN 110051376 A CN110051376 A CN 110051376A CN 201910162542 A CN201910162542 A CN 201910162542A CN 110051376 A CN110051376 A CN 110051376A
- Authority
- CN
- China
- Prior art keywords
- sample
- detected
- age
- group
- characteristic
- 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
- 238000000034 method Methods 0.000 title claims abstract description 66
- 239000004575 stone Substances 0.000 title claims abstract description 49
- 238000013499 data model Methods 0.000 claims abstract description 47
- 238000005457 optimization Methods 0.000 claims abstract description 16
- 238000012795 verification Methods 0.000 claims abstract description 10
- 238000010276 construction Methods 0.000 claims abstract description 4
- 210000002411 hand bone Anatomy 0.000 claims description 53
- 238000012360 testing method Methods 0.000 claims description 31
- 238000012545 processing Methods 0.000 claims description 30
- 238000012549 training Methods 0.000 claims description 20
- 238000004364 calculation method Methods 0.000 claims description 14
- 238000007781 pre-processing Methods 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 11
- 239000000284 extract Substances 0.000 claims description 9
- 210000000988 bone and bone Anatomy 0.000 claims description 8
- 210000003010 carpal bone Anatomy 0.000 claims description 6
- 230000000366 juvenile effect Effects 0.000 claims description 6
- 210000000236 metacarpal bone Anatomy 0.000 claims description 6
- 238000000638 solvent extraction Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 abstract description 9
- 238000013473 artificial intelligence Methods 0.000 abstract description 7
- 238000011161 development Methods 0.000 description 5
- 230000018109 developmental process Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000002059 diagnostic imaging Methods 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007634 remodeling Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 206010010356 Congenital anomaly Diseases 0.000 description 1
- 241001269238 Data Species 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000001919 adrenal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000014461 bone development Effects 0.000 description 1
- 230000018678 bone mineralization Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000001054 cortical effect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 230000002124 endocrine Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000003054 hormonal effect Effects 0.000 description 1
- 206010020718 hyperplasia Diseases 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 230000001755 vocal effect Effects 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
- A61B6/505—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for diagnosis of bone
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
- A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
- A61B6/5217—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data extracting a diagnostic or physiological parameter from medical diagnostic data
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Pathology (AREA)
- Heart & Thoracic Surgery (AREA)
- Veterinary Medicine (AREA)
- Biophysics (AREA)
- High Energy & Nuclear Physics (AREA)
- Public Health (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Optics & Photonics (AREA)
- General Health & Medical Sciences (AREA)
- Radiology & Medical Imaging (AREA)
- Biomedical Technology (AREA)
- Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Orthopedic Medicine & Surgery (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of stone age intelligent detecting method, including sample collection procedure, sample classification step, sample packet step, data model construction step, verification step, data model Optimization Steps, detected sample collection procedure and stone age judgment step.The present invention carries out stone age detection by the method for artificial intelligence, can improve detection efficiency and accuracy rate to avoid artificial subjective factor.
Description
Technical field
The present invention relates to medical imaging process field, in particular to a kind of stone age intelligent detecting method.
Background technique
Stone age analyzes an important indicator as growth and development degree, in medicine, sport and judicial expertise wind field
It plays an important role.Stone age is determined by the bone calcification degree of children.Stone age can accurately reflect people from birth
To the developmental level of each age level during full maturity.Radiologist passes through the X-ray of comparison children's hand and he
The corresponding age standard state, to measure the stone age of children.
Teenager's stone age is evaluated and tested to play an important role in terms of paediatrics hormonal problems and child growth obstacle diagnosis, is usually used in
The screening of the symptoms such as teenager's endocrine disturbance, growth and development delay, congenital adrenal cortical hyper plasia, can also make technology
Intervention effect is evaluated, in addition, the stone age can also be used for the real age of identification minor, in juvenile deliquency case
It is all applied with confirming in player's age in sports tournament.
Current skeleton development maturity assessment method more common in the world includes being mentioned based on American-European skeleton development feature
G-P Atlas Method and TW point-score out.And since there are larger difference, G-P Atlas Method and TW score for various countries' people's bone bone developmental state
Method is not applied for East Asia ethnic group.The common stone age intelligent detecting method in China has percentage counting method, Gu Shi Atlas Method, Chinese
Ossa carpi development standard CHN method etc..But when these manual evaluation stone age methods are cumbersome and vulnerable to subjective impact, random error is big,
The verbal description practical application of bone development classification standard is got up, and more complicated, systematic error is larger.
With GPU accelerate depth learning technology, completed by artificial intelligence for the stone age it is automatic detect at
In order to possible.And medical imaging technology is constantly promoted, and hospital has numerous high-end imaging devices, can quickly be obtained high-quality
The medical image of amount.But the judgement of medical image generally or by doctor is completed, this is not only time-consuming and laborious, and exists
Many supervisor's factors.And hospital has the medical image of big data rank to need to handle daily.
Therefore, it is badly in need of proposing a kind of method with artificial intelligence to carry out stone age detection and can be with the same stepping of mass
Row.
Summary of the invention
The object of the present invention is to provide a kind of stone age intelligent detecting methods, can efficiently solve traditional stone age intelligence
The problems such as error is big, time-consuming and laborious in energy detection method, can effectively improve the execution efficiency that the stone age is detected by hospital.
In order to solve the above technical problems, the present invention provides a kind of stone age intelligent detecting method, include the following steps: that sample is adopted
Collect step, the more than two digitization samples with gender teenager's hand bone image of acquisition, each digitization sample includes one green few
The group image data that the age and the teen-age hand bone image in year are obtained after calculation system;Sample classification step, will
More than two digitization samples are divided into two class of training sample and test sample;Sample packet step, by the training sample
It is divided into more than two groups, same group of other training sample is identified with identical group of distinguishing label, and described group of distinguishing label represents blueness
The juvenile age;Data model construction step is constructed and is instructed using the image data and group distinguishing label of more than two training samples
Practice primary data model;Verification step tests the primary data model according to the image data of an at least test sample
Card processing;Data model Optimization Steps construct optimization data model according to the result of verifying;It is detected sample collection procedure, is adopted
Collect a detected sample, is the digitization sample of the hand bone image of a detected person, the hand bone image including the detected person
The group image data obtained after calculation system;And stone age judgment step, by the sample typing to be detected to described
Optimize data model, the optimization data model obtains the group distinguishing label of the sample to be detected, judges the detected person's
Stone age.
Further, the sample collection procedure includes the following steps: image capturing step, shoots two using X-ray machine
The teen-age hand bone striograph of the above identical gender, records each teen-age age;Sample preprocessing step, to two with
The upper hand bone striograph carries out calculation system, obtains image data more than two, and each group image data include at least one
Group characteristic;Digitization sample generation step, generates more than two digitization samples, and each digitization sample includes one green few
The group image data that the age and the teen-age hand bone image in year are obtained after calculation system.
Further, the sample preprocessing step, includes the following steps: binarization step, to the hand bone striograph
Binarization of gray value processing is carried out, is limited in the gray scale of the hand bone striograph in the range of 0 ~ 255;Equalization procedures mention
It takes the hand bone striograph gray scale to be greater than or equal to the pixel of a preset threshold, an at least foreground area is formed, before each
The gray scale of scene area according to mean value is 0, the normal distribution that variance is 1 re-starts equalization processing;Partitioning step, by the hand
Bone striograph is divided into more than two characteristic areas, and each characteristic area is a rectangle frame, including an at least foreground area;And it is special
Extraction step is levied, the processing of convolution pondization is carried out to each characteristic area, extracts the characteristic of each characteristic area;All features
The set of the characteristic in region is the image data.
Further, the primary data model includes RPN network and FastR-CNN network, the RPN network with
FastR-CNN network share bottom convolutional layer, bottom convolutional layer include five layers of convolutional layer, include the 6th convolution after bottom convolutional layer
Layer, the 6th convolutional layer connect Liang Ge convolution branch, export prime area classification score and boundary respectively by Liang Ge convolution branch
Frame constitutes RPN network, and the initial interest region of target bone is extracted by RPN network;The bottom convolutional layer passes through the pond ROI
Change layer and connects the first full articulamentum and the second full articulamentum, the first full articulamentum and the second full articulamentum according to described initial
Distinguish output category score and bounding box position coordinates in interest region.
Further, the verification step includes the following steps: that the group distinguishing label of test sample estimates step, and one is surveyed
One group image data of sample sheet are multiplied with the primary data model, carry out the processing that rounds up to its product, obtain the survey
The group distinguishing label discreet value of sample sheet;Alternatively, the mass spectrometric data more than two of a test sample is multiplied with the data model,
Its product is lined up into ordered series of numbers according to numerical values recited, the processing that rounds up is carried out to its median, obtains the group of the test sample
Label discreet value.The group distinguishing label of test sample compares step, by the group distinguishing label discreet value of the test sample and its group mark
Label comparison, judges the accuracy rate of the primary data model.
Further, in the data model Optimization Steps, when the accuracy rate of the primary data model is pre- less than one
If when threshold value, returning to the sample packet step, reselecting training sample.
Further, the detected sample collection procedure includes the following steps: to be detected image capturing step, using X
Ray machine shoots the hand bone striograph of a detected person, and the gender of the detected person is identical as the teen-age gender;It is detected
Sample preprocessing step carries out calculation system to the hand bone striograph of detected person, obtains one group and is detected image data, packet
Include an at least detected feature data;Detected data sample generation step generates a detected data sample, including institute
State detected image data.
Further, the detected sample preprocessing step includes the following steps: detected sample binarization step,
Binarization of gray value processing is carried out to the hand bone striograph of detected person, the gray scale of the hand bone striograph is made to be limited in 0 ~ 255
In the range of;Sample equalization procedures are detected, the picture that the hand bone striograph gray scale is greater than or equal to a preset threshold is extracted
Vegetarian refreshments forms an at least foreground area, according to mean value be 0 to the gray scale of each foreground area, normal distribution that variance is 1 again
Carry out equalization processing;It is detected sample partitioning step, the hand bone striograph of the detected person is divided into more than two features
Region;And detected sample characteristics extraction step, the processing of convolution pondization is carried out to each characteristic area, extracts each characteristic area
The characteristic in domain;The set of the characteristic of all characteristic areas is the detected image data.
Further, teenager's hand bone image include teenager under palm flattened state phalanges and/or metacarpal bone and/
Or the image of carpal bone;The detected manpower bone image includes phalanges and/or metacarpal bone of the detected person under palm flattened state
And/or the image of carpal bone.
Further, the teenager, the detected person age be accurate to percentile;The digitization sample size
It is 40000 ~ 300000;The group number of labels is 2000 groups, respectively represents 0 to 20 year old teenager.
The beneficial effects of the present invention are: the present invention proposes a kind of stone age intelligent detecting method, using artificial intelligence technology
In deep learning algorithm, child to different age group or juvenile stone age picture carry out data acquisition and feature extraction, use
The image data training of magnanimity simultaneously establishes data model.The present invention adopts the hand X-ray picture progress data for needing detected person
It send after collection into data model, Feature Selection is carried out by convolution, is inferred to the detected person's by a large amount of characteristic
Stone age value.
The present invention carries out stone age detection by the method for artificial intelligence, can improve detection to avoid artificial subjective factor
Efficiency and accurate.
The present invention can be by verify the data model put up to the picture images of test, and can lead to
It crosses self feed back and data model is constantly optimized, the prediction rate of medical image is improved.
The present invention is finally desirably integrated into hospital system, can be synchronized batch processing, be mentioned for hospital staff
For helping, working efficiency can be improved.
Detailed description of the invention
The invention will be further described with reference to the accompanying drawings and examples.
Fig. 1 is stone age intelligent detecting method flow chart provided by the invention;
Fig. 2 is the flow chart of sample collection procedure of the present invention;
Fig. 3 is the flow chart of sample of the present invention pre-treatment step;
Fig. 4 is the flow chart of verification step of the present invention;
Fig. 5 is the flow chart of the detected sample collection procedure of the present invention;
Fig. 6 is the flow chart of the detected sample preprocessing step of the present invention.
Specific embodiment
The explanation for being below each embodiment is with reference to additional schema, to illustrate the particular implementation of the invention that can be implemented
Example.The direction term that the present invention is previously mentioned, for example, above and below, front, rear, left and right, inside and outside, side etc., be only the side with reference to accompanying drawings
To.The element title that the present invention mentions, such as first, second etc., it is only to discriminate between different components, can preferably be expressed.
The similar unit of structure is given the same reference numerals in the figure.
Herein with reference to the accompanying drawings to detailed description of the present invention embodiment.The present invention can show as many different forms,
The present invention should not be only interpreted as specific embodiment set forth herein.It is to explain the present invention that the present invention, which provides these embodiments,
Practical application, to make others skilled in the art it will be appreciated that various embodiments of the present invention and being suitable for specific expection
The various modifications scheme of application.
Present invention is generally directed to teenager's hand bone image include phalanges of the teenager under palm flattened state and/or
The image of metacarpal bone and/or carpal bone;The detected manpower bone image include phalanges of the detected person under palm flattened state and/
Or the image of metacarpal bone and/or carpal bone.The teenager, the detected person age be accurate to percentile, pass through artificial intelligence
Model predicts the teen-age stone age.
As shown in Figure 1, the present invention provides a kind of stone age intelligent detecting method, include the following steps S1 ~ S8.
S1, sample collection procedure, the sample collection procedure purpose is mainly by two or more the same as gender teenager's hand bone
The digitization sample of image, each digitization sample include a teen-age age and the teen-age hand bone image by digitization
The group image data obtained after processing.
As shown in Fig. 2, the sample collection procedure includes the following steps S11 ~ S13.
S11, image capturing step, using the teen-age hand bone striograph of the more than two identical genders of X-ray machine shooting, note
Record each teen-age age.Because boy student can be different from stone age structure of the schoolgirl under the identical age, in the present invention
Detection method in need to take gender as characteristic of division of the invention.
S12, sample preprocessing step carry out calculation systems to more than two hand bone striographs, obtain two groups with
Upper image data, each group image data include at least one set of characteristic;As shown in figure 3, the sample preprocessing step tool
Body includes the following steps: S121, binarization step, carries out binarization of gray value processing to the hand bone striograph, makes the hand
The gray scale of bone striograph is limited in the range of 0 ~ 255;It is big to extract the hand bone striograph gray scale by S122, equalization procedures
In or equal to a preset threshold pixel, form an at least foreground area, the gray scale to each foreground area is according to mean value
0, the normal distribution that variance is 1 re-starts equalization processing;The hand bone striograph is divided into two by S123, partitioning step
Features above region, each characteristic area are a rectangle frame, and the size of rectangle frame preferably 256 pixels multiply 256 pixels, in rectangle frame
Including an at least foreground area;S124, characteristic extraction step carry out the processing of convolution pondization to each characteristic area, extract each
The characteristic of characteristic area;The set of the characteristic of all characteristic areas is the image data.
In the sample preprocessing step, the binary conversion treatment step be entire medical image is showed it is apparent black
The pixel grey scale for being greater than some threshold grey scale value is set as gray scale maximum, the pixel grey scale for being less than this value is set by white effect
For gray scale minimum, to realize binaryzation.According to preset threshold choose difference, the algorithm of binaryzation be divided into fixed threshold and
Adaptive threshold.More commonly used binarization method then has: Two-peak method, P parametric method, iterative method and OTSU method, and the present invention uses
Two-peak method.
Preset threshold in the equalization procedures is generally higher than equal to 1, is sieved according to the gray value of each pixel
Choosing carries out image segmentation for the next partitioning step.The feature extraction carries out characteristic extraction using convolution,
The capability of fitting of the vector magnitude control overall model of characteristic.The dimension that feature vector can be reduced in over-fitting,
The output dimension of convolutional layer can be improved when poor fitting.
S13, digitization sample generation step, generate more than two digitization samples, and each digitization sample includes one green
The group image data that the juvenile age and the teen-age hand bone image is obtained after calculation system.
More than two digitization samples are divided into two class of training sample and test sample by S2, sample classification step.Institute
Training sample is stated to training and builds data model of the invention, the test sample be mainly used for the data model into
Row optimization.
Sample classification of the invention according to the ratio of 9:1, i.e., the 90% of training sample selection entirety, is made by the 10% of whole sample
For test sample.
The training sample is divided into more than two groups, same group of other training sample mark by S3, sample packet step
Knowledge has identical group of distinguishing label, and described group of distinguishing label represents the teen-age age;The age of the teenager, the detected person
It is accurate to percentile;The group number of labels is 2000 groups, respectively represents 0 to 20 year old teenager.The group class includes
The correlated characteristic of the Adolescent Medicine image, to carry out Classification and Identification.
S4, data model construction step are constructed and are instructed using the image data and group distinguishing label of more than two training samples
Practice primary data model;The primary data model includes RPN network and FastR-CNN network, the RPN network and FastR-
CNN network share bottom convolutional layer, bottom convolutional layer include five layers of convolutional layer, include the 6th convolutional layer after bottom convolutional layer, the
Six convolutional layers connect Liang Ge convolution branch, export prime area classification score and bounding box, structure respectively by Liang Ge convolution branch
At RPN network, the initial interest region of target bone is extracted by RPN network;The bottom convolutional layer is connected by the pond ROI layer
The first full articulamentum and the second full articulamentum are connect, the first full articulamentum and the second full articulamentum are according to the initial region of interest
Distinguish output category score and bounding box position coordinates in domain.
RPN network of the invention can be end-to-endly trained for the generating detection Suggestion box of the task, can predict simultaneously
The boundary of target area and score out.The input of RPN network can be any size.
Faster-R-CNN network is made of two big modules: PRN candidate frame extraction module and FastR-CNN detect mould
Block.Wherein, RPN network is for extracting candidate frame;FastR-CNN is detected based on the RPN suggestion areas extracted and is identified suggestion area
Target in domain.
S5, verification step, main purpose are the image datas according to an at least test sample to the primary data model
Verification processing is carried out, the data processing model that this method proposes is verified.
Specifically, remaining 10% detection sample is put into the model having had built up and is verified, and obtained just
The accuracy rate of grade data model.
As shown in figure 4, the verification step specifically includes the following steps S51 ~ S52.
S51, test sample group distinguishing label estimate step, by a group image data of a test sample and it is described just series
It is multiplied according to model, the processing that rounds up is carried out to its product, obtain the group distinguishing label discreet value of the test sample;Alternatively, by one
The mass spectrometric data more than two of test sample is multiplied with the data model, its product is lined up ordered series of numbers according to numerical values recited, right
Its median carries out the processing that rounds up, and obtains the group distinguishing label discreet value of the test sample.
The group distinguishing label comparison step of S52, test sample, by the group distinguishing label discreet value of the test sample and its group mark
Label comparison, judges the accuracy rate of the primary data model.
S6, data model Optimization Steps construct optimization data model according to the result of verifying;Optimize in the data model
In step, when the accuracy rate of the primary data model is less than a preset threshold, the sample packet step is returned, is selected again
Select training sample.It is constantly duplicate to carry out the data mould and to the remaining sample in the sample packet as test sample
The optimization of type.
S7, it is detected sample collection procedure, the detected sample collection procedure main purpose is one tested test sample of acquisition
This is the digitization sample of the hand bone image of a detected person, and the hand bone image including the detected person is by calculation system
The group image data obtained afterwards;
As shown in figure 5, the detected sample collection procedure, includes the following steps S71 ~ S73:
S71, it is detected image capturing step, the hand bone striograph of a detected person, the property of the detected person is shot using X-ray machine
It is not identical as the teen-age gender;
S72, it is detected sample preprocessing step, calculation system is carried out to the hand bone striograph of detected person, obtains one group of quilt
Detect image data, including an at least detected feature data.The detected sample preprocessing step, as shown in fig. 6, including
Following steps: S721, being detected sample binarization step, carries out binarization of gray value processing to the hand bone striograph of detected person,
It is limited in the gray scale of the hand bone striograph in the range of 0 ~ 255;S722, sample equalization procedures are detected, extract institute
The pixel that hand bone striograph gray scale is greater than or equal to a preset threshold is stated, an at least foreground area is formed, to each foreground zone
The gray scale in domain according to mean value is 0, the normal distribution that variance is 1 re-starts equalization processing;S723, it is detected sample subregion step
Suddenly, the hand bone striograph of the detected person being divided into more than two characteristic areas, each characteristic area is a rectangle frame, including
An at least foreground area;And S724, detected sample characteristics extraction step, convolution pond Hua Chu is carried out to each characteristic area
Reason, extracts the characteristic of each characteristic area;The set of the characteristic of all characteristic areas is the detected image
Data.
S73, detected data sample generation step generate a detected data sample, including the detected shadow
As data.
S8, stone age judgment step, by the sample typing to be detected to the optimization data model, the optimization data mould
Type obtains the group distinguishing label of the sample to be detected, judges the stone age of the detected person.
The present invention is also desirably integrated into hospital system, can synchronize batch processing, it is only necessary to calculated in bulk
Medical image picture to be tested is inputted in machine system, it can export each patient corresponding stone age.The present invention can be
Hospital staff provides help, and working efficiency can be improved.
The present invention proposes stone age intelligent detecting method, and the present invention is calculated by using the deep learning in artificial intelligence technology
Method carries out data acquisition and feature extraction by child to different age group or juvenile stone age picture, passes through the picture of magnanimity
Data carry out training and establishing data model for model.The present invention adopts the hand X-ray picture progress data for needing detected person
Send after collection into data model, to be inferred to the stone age value of the detected person, staff by the stone age value of detected person with
The real age of detected person compares, and can judge whether detected person develops normally.
It should be pointed out that can also have the embodiment of a variety of transformation and remodeling for the present invention through absolutely proving,
It is not limited to the specific embodiment of above embodiment.Above-described embodiment is as just explanation of the invention, rather than to hair
Bright limitation.In short, protection scope of the present invention should include that those are obvious to those skilled in the art
Transformation or substitution and remodeling.
Claims (10)
1. a kind of stone age intelligent detecting method, which comprises the steps of:
Sample collection procedure, the more than two digitization samples with gender teenager's hand bone image of acquisition, each digitization sample
The group image data obtained after calculation system including a teen-age age and the teen-age hand bone image;
More than two digitization samples are divided into two class of training sample and test sample by sample classification step;
The training sample is divided into more than two groups by sample packet step, and same group of other training sample is identified with phase
Same group distinguishing label, described group of distinguishing label represent the teen-age age;
Data model construction step is constructed using the image data and group distinguishing label of more than two training samples and trains just series
According to model;
Verification step carries out verification processing to the primary data model according to the image data of an at least test sample;
Data model Optimization Steps construct optimization data model according to the result of verifying;
It is detected sample collection procedure, acquires a detected sample, is the digitization sample of the hand bone image of a detected person,
The group image data that hand bone image including the detected person is obtained after calculation system;And
Stone age judgment step, by the sample typing to be detected to the optimization data model, the optimization data model is obtained
The group distinguishing label of the sample to be detected, judges the stone age of the detected person.
2. stone age intelligent detecting method as described in claim 1, which is characterized in that
The sample collection procedure, includes the following steps:
Image capturing step records each blueness using the teen-age hand bone striograph of the more than two identical genders of X-ray machine shooting
The juvenile age;
Sample preprocessing step carries out calculation system to more than two hand bone striographs, obtains image number more than two
According to each group image data include at least one set of characteristic;And
Digitization sample generation step, generates more than two digitization samples, and each digitization sample includes a teen-age year
The group image data that age and the teen-age hand bone image are obtained after calculation system.
3. stone age intelligent detecting method as claimed in claim 2, which is characterized in that
The sample preprocessing step, includes the following steps:
Binarization step carries out binarization of gray value processing to the hand bone striograph, limits the gray scale of the hand bone striograph
System is in the range of 0 ~ 255;
Equalization procedures extract the pixel that the hand bone striograph gray scale is greater than or equal to a preset threshold, form at least one
Foreground area according to mean value is 0 to the gray scale of each foreground area, the normal distribution that variance is 1 re-starts equalization processing;
The hand bone striograph is divided into more than two characteristic areas by partitioning step, and each characteristic area is a rectangle frame, including
An at least foreground area;And
Characteristic extraction step carries out the processing of convolution pondization to each characteristic area, extracts the characteristic of each characteristic area;Institute
The set for having the characteristic of characteristic area is the image data.
4. stone age intelligent detecting method as described in claim 1, which is characterized in that
The primary data model includes RPN network and Fast R-CNN network,
The RPN network and Fast R-CNN network share bottom convolutional layer, bottom convolutional layer include five layers of convolutional layer,
It include the 6th convolutional layer after bottom convolutional layer, the 6th convolutional layer connects Liang Ge convolution branch, passes through Liang Ge convolution branch point
Not Shu Chu prime area classification score and bounding box, constitute RPN network, pass through the initial interest that RPN network extracts target bone
Region;The bottom convolutional layer connects the first full articulamentum and the second full articulamentum, the described first full connection by the pond ROI layer
Layer and the second full articulamentum are according to the initial interest region difference output category score and bounding box position coordinates.
5. stone age intelligent detecting method as described in claim 1, which is characterized in that
The verification step, includes the following steps:
The group distinguishing label of test sample estimates step, by a group image data of a test sample and the primary data model phase
Multiply, the processing that rounds up is carried out to its product, obtains the group distinguishing label discreet value of the test sample;Alternatively, by a test sample
Mass spectrometric data more than two be multiplied with the data model, its product is lined up into ordered series of numbers according to numerical values recited, to its median
The processing that rounds up is carried out, the group distinguishing label discreet value of the test sample is obtained;
The group distinguishing label of test sample compares step, and the group distinguishing label discreet value of the test sample and its group of distinguishing label are compared,
Judge the accuracy rate of the primary data model.
6. stone age intelligent detecting method as described in claim 1, which is characterized in that
In the data model Optimization Steps,
When the accuracy rate of the primary data model is less than a preset threshold,
The sample packet step is returned, training sample is reselected.
7. stone age intelligent detecting method as described in claim 1, which is characterized in that
The detected sample collection procedure, includes the following steps:
Be detected image capturing step, using X-ray machine shoot a detected person hand bone striograph, the gender of the detected person with
The teen-age gender is identical;
It is detected sample preprocessing step, calculation system is carried out to the hand bone striograph of detected person, one group is obtained and is detected
Image data, including at least one set of detected feature data;Detected data sample generation step generates a detected data
Change sample, including the detected image data.
8. stone age intelligent detecting method as claimed in claim 7, which is characterized in that
The detected sample preprocessing step, includes the following steps:
It is detected sample binarization step, binarization of gray value processing is carried out to the hand bone striograph of detected person, makes the hand bone
The gray scale of striograph is limited in the range of 0 ~ 255;
Sample equalization procedures are detected, the pixel that the hand bone striograph gray scale is greater than or equal to a preset threshold is extracted,
An at least foreground area is formed, according to mean value is 0 to the gray scale of each foreground area, the normal distribution that variance is 1 re-starts
Equalization processing;
It is detected sample partitioning step, the hand bone striograph is divided into more than two characteristic areas, each characteristic area is one
Rectangle frame, including an at least foreground area;And
It is detected sample characteristics extraction step, the processing of convolution pondization is carried out to each characteristic area, extracts each characteristic area
Characteristic;The set of the characteristic of all characteristic areas is the detected image data.
9. stone age intelligent detecting method as described in claim 1, which is characterized in that
Teenager's hand bone image includes the shadow of phalanges and/or metacarpal bone and/or carpal bone of the teenager under palm flattened state
Picture;
The detected manpower bone image includes phalanges and/or metacarpal bone and/or carpal bone of the detected person under palm flattened state
Image.
10. stone age intelligent detecting method as described in claim 1, which is characterized in that
The teenager, the detected person age be accurate to percentile;
The digitization sample size is 40000 ~ 300000;
The group number of labels is 2000 groups, respectively represents 0 ~ 20 years old teenager.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910162542.1A CN110051376A (en) | 2019-03-05 | 2019-03-05 | A kind of stone age intelligent detecting method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910162542.1A CN110051376A (en) | 2019-03-05 | 2019-03-05 | A kind of stone age intelligent detecting method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110051376A true CN110051376A (en) | 2019-07-26 |
Family
ID=67316570
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910162542.1A Pending CN110051376A (en) | 2019-03-05 | 2019-03-05 | A kind of stone age intelligent detecting method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110051376A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111402213A (en) * | 2020-03-05 | 2020-07-10 | 北京深睿博联科技有限责任公司 | Bone age evaluation method and device, electronic equipment and computer-readable storage medium |
CN112634193A (en) * | 2020-09-30 | 2021-04-09 | 上海交通大学 | Image anomaly detection method and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100838339B1 (en) * | 2007-08-06 | 2008-06-13 | 주식회사 오스테오시스 | Method for bone age judgment and prediction of adult height and the computer readable recording medium thereof |
CN106340000A (en) * | 2015-07-06 | 2017-01-18 | 司法部司法鉴定科学技术研究所 | Bone age assessment method |
CN107767376A (en) * | 2017-11-02 | 2018-03-06 | 西安邮电大学 | X-ray film stone age Forecasting Methodology and system based on deep learning |
CN107895367A (en) * | 2017-11-14 | 2018-04-10 | 中国科学院深圳先进技术研究院 | A kind of stone age recognition methods, system and electronic equipment |
CN108334899A (en) * | 2018-01-28 | 2018-07-27 | 浙江大学 | Quantify the bone age assessment method of information integration based on hand bone X-ray bone and joint |
CN108968991A (en) * | 2018-05-08 | 2018-12-11 | 平安科技(深圳)有限公司 | Hand bone X-ray bone age assessment method, apparatus, computer equipment and storage medium |
-
2019
- 2019-03-05 CN CN201910162542.1A patent/CN110051376A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100838339B1 (en) * | 2007-08-06 | 2008-06-13 | 주식회사 오스테오시스 | Method for bone age judgment and prediction of adult height and the computer readable recording medium thereof |
CN106340000A (en) * | 2015-07-06 | 2017-01-18 | 司法部司法鉴定科学技术研究所 | Bone age assessment method |
CN107767376A (en) * | 2017-11-02 | 2018-03-06 | 西安邮电大学 | X-ray film stone age Forecasting Methodology and system based on deep learning |
CN107895367A (en) * | 2017-11-14 | 2018-04-10 | 中国科学院深圳先进技术研究院 | A kind of stone age recognition methods, system and electronic equipment |
CN108334899A (en) * | 2018-01-28 | 2018-07-27 | 浙江大学 | Quantify the bone age assessment method of information integration based on hand bone X-ray bone and joint |
CN108968991A (en) * | 2018-05-08 | 2018-12-11 | 平安科技(深圳)有限公司 | Hand bone X-ray bone age assessment method, apparatus, computer equipment and storage medium |
Non-Patent Citations (1)
Title |
---|
刘鸣谦等: "基于多维度特征融合的深度学习骨龄评估模型", 《第二军医大学学报》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111402213A (en) * | 2020-03-05 | 2020-07-10 | 北京深睿博联科技有限责任公司 | Bone age evaluation method and device, electronic equipment and computer-readable storage medium |
CN111402213B (en) * | 2020-03-05 | 2023-10-27 | 北京深睿博联科技有限责任公司 | Bone age evaluation method and device, electronic equipment and computer readable storage medium |
CN112634193A (en) * | 2020-09-30 | 2021-04-09 | 上海交通大学 | Image anomaly detection method and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109035187B (en) | Medical image labeling method and device | |
CN104424482B (en) | Image processing equipment and image processing method | |
CN110175993A (en) | A kind of Faster R-CNN pulmonary tuberculosis sign detection system and method based on FPN | |
CN111161311A (en) | Visual multi-target tracking method and device based on deep learning | |
CN109998577A (en) | A kind of artificial intelligence stone age detection terminal device | |
CN109858540B (en) | Medical image recognition system and method based on multi-mode fusion | |
US9183431B2 (en) | Apparatus and method for providing activity recognition based application service | |
CN109903312A (en) | A kind of football sportsman based on video multi-target tracking runs distance statistics method | |
CN109919961A (en) | A kind of processing method and processing device for aneurysm region in encephalic CTA image | |
KR101227569B1 (en) | Body Segments Localization Device and Method for Analyzing Motion of Golf Swing | |
CN108492272A (en) | Cardiovascular vulnerable plaque recognition methods based on attention model and multitask neural network and system | |
CN110033473A (en) | Motion target tracking method based on template matching and depth sorting network | |
CN104977313A (en) | Method and device for detecting and identifying X-ray image defects of welding seam | |
Song et al. | MSFYOLO: Feature fusion-based detection for small objects | |
CN110264444A (en) | Damage detecting method and device based on weak segmentation | |
CN109998576A (en) | A kind of artificial intelligence stone age detection method | |
CN106340000A (en) | Bone age assessment method | |
CN109858476A (en) | The extending method and electronic equipment of label | |
CN108197636A (en) | A kind of paddy detection and sorting technique based on depth multiple views feature | |
CN108460336A (en) | A kind of pedestrian detection method based on deep learning | |
CN109636846A (en) | Object localization method based on circulation attention convolutional neural networks | |
CN110051376A (en) | A kind of stone age intelligent detecting method | |
CN108288020A (en) | Video shelter detecting system based on contextual information and method | |
CN110399804A (en) | A kind of food inspection recognition methods based on deep learning | |
CN108717522A (en) | A kind of human body target tracking method based on deep learning and correlation filtering |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190726 |