CN107374657A - The method and CT scan system being corrected to CT scan data - Google Patents
The method and CT scan system being corrected to CT scan data Download PDFInfo
- Publication number
- CN107374657A CN107374657A CN201710524197.2A CN201710524197A CN107374657A CN 107374657 A CN107374657 A CN 107374657A CN 201710524197 A CN201710524197 A CN 201710524197A CN 107374657 A CN107374657 A CN 107374657A
- Authority
- CN
- China
- Prior art keywords
- network model
- neural network
- sweep parameter
- training
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000002591 computed tomography Methods 0.000 title claims abstract description 46
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000012937 correction Methods 0.000 claims abstract description 77
- 238000003062 neural network model Methods 0.000 claims abstract description 76
- 238000012549 training Methods 0.000 claims description 87
- 238000012545 processing Methods 0.000 claims description 19
- 230000007935 neutral effect Effects 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 4
- 238000003860 storage Methods 0.000 description 18
- 238000001228 spectrum Methods 0.000 description 11
- 230000008859 change Effects 0.000 description 9
- 230000001537 neural effect Effects 0.000 description 9
- 230000006870 function Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 5
- 241000208340 Araliaceae Species 0.000 description 4
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 4
- 235000003140 Panax quinquefolius Nutrition 0.000 description 4
- 235000008434 ginseng Nutrition 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000000977 initiatory effect Effects 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 210000004218 nerve net Anatomy 0.000 description 2
- 230000000712 assembly Effects 0.000 description 1
- 238000000429 assembly Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 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
- A61B6/58—Testing, adjusting or calibrating thereof
- A61B6/582—Calibration
- A61B6/583—Calibration using calibration phantoms
-
- 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
- A61B6/02—Arrangements for diagnosis sequentially in different planes; Stereoscopic radiation diagnosis
- A61B6/03—Computed tomography [CT]
- A61B6/032—Transmission computed tomography [CT]
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Optics & Photonics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- High Energy & Nuclear Physics (AREA)
- Veterinary Medicine (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Public Health (AREA)
- Pathology (AREA)
- Radiology & Medical Imaging (AREA)
- Physics & Mathematics (AREA)
- Heart & Thoracic Surgery (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Pulmonology (AREA)
- Theoretical Computer Science (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses a kind of method being corrected to CT scan data and CT scan system, methods described to include：Determine the first sweep parameter；First sweep parameter is at least partially based on, obtains characteristic value；Obtain housebroken neural network model；And the characteristic value is input in the housebroken neural network model, obtain correction coefficient corresponding to first sweep parameter.
Description
Technical field
The present invention relates to medical imaging equipment field, more particularly to a kind of method being corrected to CT scan data and CT
Scanning system.
Background technology
What it is due to the actual use of CT (Computed Tomography, computerized tomograph) equipment is that multi-power spectrum X is penetrated
Line, with the increase of ray penetration depth, soft ray (low energy rays) decay is more than hard ray (high-energy rays), i.e. beam
Hardening, causes the power spectrum of ray persistently to change, and decay of the image reconstruction of CT equipment based on Single energy X ray absorptionmetry, therefore needs
Spectrum correction is carried out to compensate the change of power spectrum.The beam hardening correction used at present is mostly to be swept respectively under specific scan parameter
Air and standard body mould are retouched, correction coefficient is then calculated according to special algorithm to correct beam projection value.The school that this method obtains
Positive coefficient is more accurate, but is only applicable to specific scan parameter.When clinical scanning parameter changes, for example, working as X-ray tube
When voltage changes, the X-ray energy spectrum sent can also change, and now need to rescan air and body mould and calculate school
Positive coefficient, it is unfavorable for practical application.
The content of the invention
In view of the above-mentioned problems, the invention provides a kind of side for being capable of quick obtaining correction coefficient under any sweep parameter
Method.
To achieve the above object of the invention, technical scheme provided by the invention is as follows：
The present invention discloses a kind of method being corrected to CT scan data, including：Determine the first sweep parameter；At least
First sweep parameter is based partially on, obtains characteristic value；Obtain housebroken neural network model；And by the characteristic value
It is input in the housebroken neural network model, obtains correction coefficient corresponding to first sweep parameter.The feature
Value includes at least one of first sweep parameter and first projection value, and first projection value is included in the first scanning ginseng
The projection value that several lower scanning air obtain；
According to some embodiments of the present invention, the housebroken neural network model of acquisition includes：Build neutral net
Model, the neural network model is trained, obtains the housebroken neural network model.
It is described that the neural network model is trained according to some embodiments of the present invention, including：Obtain one or
Multiple training samples, the training sample include at least one in the second sweep parameter and the second projection value, second throwing
Shadow value, which is included under second sweep parameter, scans the projection value that air obtains；And use one or more of training samples
This is trained to the neural network model.
According to some embodiments of the present invention, it is described neural network model is trained including：Use min | | { ai}-
{ai}ideal| | it is optimization object function, wherein { aiIt is that the training sample is input to the output after the neural network model
Value, { ai}idealFor the reference output valve of the training sample.
It is described to use one or more of training samples to the neutral net mould according to some embodiments of the present invention
Type be trained including：Whether the neural network model after training of judgement meets preparatory condition, if meeting preparatory condition, stopping pair
The training of the neural network model.
It is further comprising the steps of according to some embodiments of the present invention, methods described：Obtained under first sweep parameter
Take CT scan data；And the CT scan data is corrected using the correction coefficient.
According to some embodiments of the present invention, the CT scan data is corrected using the correction coefficient, including
Use following correction function：
Wherein j is correction exponent number, and x is the CT scan data, aiFor the correction coefficient, y is the CT scan after correction
Data.
Invention further discloses a kind of CT scan system, including：Processing component, it is configured to determine the first sweep parameter；School
Positive coefficient computation module, it is configured to be at least partially based on first sweep parameter, obtains characteristic value, obtain housebroken nerve
Network model, and the characteristic value is input in the housebroken neural network model, obtain the first scanning ginseng
Correction coefficient corresponding to number.
According to some embodiments of the present invention, the system also includes scan components, is configured to first scanning
Under parameter, CT scan data is obtained；Wherein described processing component is additionally configured to using the correction coefficient to the CT scan number
According to being corrected.
According to some embodiments of the present invention, the system also includes sample acquisition module and model training module.It is described
Sample acquisition module is configured as obtaining one or more training samples, and the training sample includes the second sweep parameter and second
At least one in projection value, second projection value, which is included under second sweep parameter, scans the projection that air obtains
Value.The model training module, it is configured with one or more of training samples and the neural network model is carried out
Training.
Brief description of the drawings
Fig. 1 is the structural representation according to a kind of CT scan system of the present invention；
Fig. 2 is a kind of structural representation of the neural network model provided according to some embodiments of the present invention；
Fig. 3 is illustrated according to a kind of flow for the method that correction coefficient is obtained based on neural network model provided by the invention
Figure；
Fig. 4 is to obtain the method for housebroken neural network model according to a kind of training neural network model of the present invention
Schematic flow sheet；
Fig. 1 is marked：100 be CT scan system, and 110 be scan components, and 111 be frame, and 112 be scanning bed, and 113 X are penetrated
Line source, 114 be X-ray detector, and 120 be network, and 130 be processing component, and 140 be storage assembly, and 150 be that correction coefficient calculates
Component, 152 be sample acquisition module, and 154 be memory module, and 156 be model training module, and 158 be correction coefficient computing module；
Fig. 2 is marked：201 be the input of neural network model, and 202 be neural net layer, and 203 be output end.
Embodiment
The present invention is described further below by specific embodiment and with reference to accompanying drawing.
Fig. 1 is a kind of structural representation of the CT scan system 100 provided according to some embodiments of the present invention.Such as Fig. 1
Shown, CT scan system 100 includes scan components 110, processing component 130, storage assembly 140 and correction coefficient computation module
150.It can be interconnected between each component of the CT scan system 100 by network 120.
Scan components 110 can include frame 111, scanning bed 112, x-ray source 113 and X-ray detector 114.Frame
111 can at least partly accommodate scanning bed 112 comprising a hollow chamber as scanning chamber, scanning chamber.Frame 111 can rotate,
X-ray source 113 thereon can produce X ray, and object to be scanned is scanned from different perspectives and obtains projection value, the throwing
Shadow value can be used for image reconstruction, obtain CT images.
Scanning bed 112 can support object to be scanned.The object to be scanned can be patient or body mould or
Other scanned objects.Scanning bed 112 can be parallel to ground.
X-ray source 113 can produce X-ray beam, and the beam reaches X-ray detector 114 after passing through object to be scanned.
X-ray detector 114 receives the X-ray beam after decaying through object to be scanned, obtains actual projection value.
High-voltage pipe (not shown) can be included in x-ray source 113, high-voltage pipe is used to produce X-ray beam.
High-voltage pipe can send multi-power spectrum X ray, and the multi-power spectrum X ray can include soft ray (low energy rays) and hard ray
(high-energy rays).When being scanned to object to be scanned, X-ray beam can penetrate object to be scanned.As X-ray beam penetrates
The increase of depth, soft ray (low energy rays) decay are more than hard ray (high-energy rays) and decayed, i.e. beam hardening, cause X
The power spectrum of beam persistently changes.And decay of the image reconstruction of CT scan system based on Single energy X ray absorptionmetry, it is therefore desirable to enter
Row spectrum is corrected to compensate the change of power spectrum.The target of spectrum correction is to obtain one group of correction coefficient, and the correction coefficient can be with corrected X
The actual projection value that ray detector 114 receives.In certain embodiments, the correction coefficient is to the actual projection value
Correction can be realized by the function shown in formula (1)：
Wherein, in formula (1), j is configurable correction exponent number, such as j can be 3 or 4.In some embodiments
In, the correction exponent number can be determined by user.The actual projection value that x receives for X-ray detector 114, aiFor correction coefficient,
Y is the projection value after correction, and the projection value after correction can be used for CT image reconstructions.
Network 120 can connect each component of CT scan system 100, and making between each component of CT scan system 100 can be with
Carry out data exchange.Network 120 can be cable network or wireless network, or its combination.
Processing component 130 is control and the data processing section of CT scan system 100, is configurable to processing data, production
Raw control signal, to control operation of CT scan system 100 etc..
Storage assembly 140 is configurable to store the data of CT scan system 100.For example, storage assembly 140 can store
CT scan agreement, sweep parameter, scanning projection value, CT images, optimum correction coefficient corresponding to specific scan parameter etc..
Correction coefficient computation module 150 is configurable to calculate beam hardening correction coefficient based on neural network model.God
It is a kind of nonlinear algorithm for including multiple parameters through network model, the feature of input data can be extracted after training, and
Classified according to the feature of extraction, obtain feature output.Can according to the specific descriptions of some neural network models of the application
With referring to Fig. 2 and its corresponding description section.Correction coefficient computation module 150 can include sample acquisition module 152, storage mould
Block 154, model training module 156 and correction coefficient computing module 158, will be described in detail below.In some embodiments
In, correction coefficient computation module 150 can be a stand-alone assembly for being connected to processing component 130 and/or network 120.For example,
Correction coefficient computation module 140 can be a computing device, such as personal computer, server, tablet personal computer, mobile phone or class
Like equipment etc..In certain embodiments, correction coefficient computation module 150 is desirably integrated into the processing component 130 and/or described
In CT scan system 100.
Sample acquisition module 152 can obtain training sample.Training sample is used to be trained neural network model.Instruction
Input value can be included and with reference to output valve by practicing sample.The input value of the training sample can include air projection value and/or
It is at least one in sweep parameter.For example, in certain embodiments, the input value of training sample can include sweep parameter and with
Air projection value corresponding to the sweep parameter.In further embodiments, the input value of training sample can include sweep parameter.
In some other embodiments, the input value of training sample can include air projection value corresponding with a certain sweep parameter.Scanning
Parameter can be high-voltage tube voltage and/or current value.Air projection value can be that CT scan system 100 is joined in certain one scan
The projection value received after several lower scanning air in X-ray detector 114.For example, any thing is not placed on scanning bed 112
Body, then control x-ray source 113 to send X ray and be scanned, and X ray is received with X-ray detector 114, obtain air throwing
Shadow value.The reference output valve of the sample can include one or more optimum correction coefficients.In certain embodiments, preferable school
Positive coefficient can be by being calculated after scanning standard body mould.The method that optimum correction coefficient is obtained by scanning standard body mould exists
Have been described, be not specifically described in the present invention in the prior art.
Sample acquisition module 152 can obtain data from the other assemblies of CT scan system 100 and obtain training sample.Example
Such as, sample acquisition module 152 obtained directly from storage assembly 140 sweep parameter, air projection value and optimum correction coefficient with
Obtain training sample.In another example sample acquisition module 152 can obtain sweep parameter from processing component 130, visited from X ray
Survey device 114 obtains the air projection value under the sweep parameter, and the optimum correction under the sweep parameter is obtained from storage assembly 140
Coefficient, obtain training sample.Sample acquisition module 152 can store obtained training sample into memory module 154, also may be used
So that training sample is stored into storage assembly 140.
Memory module 154 can be with data storage, and the data can include training sample set, neural network model, through instruction
Data used during experienced neural network model and training pattern etc., will hereinafter be described.
Model training module 156 can be trained to neural network model.Neural network model is trained, be by
The input value of training sample is input in neural network model, and neural network model is exported to input value after computing
Value.Then the parameter current of model is reversely adjusted using majorized function.In certain embodiments, it is described to the anti-of model
Can be the process of an iteration to adjustment process.After being trained every time using training sample, the parameter in neural network model is all
It can change, as " initiation parameter " that input training sample is trained next time.For example, in certain embodiments,
The parameter of the function pair model shown in formula (2) can be used reversely to be adjusted：
min||{ai}-{ai}ideal|| (2)
Wherein { aiIt is neural network model output valve, { ai}idealFor the reference output valve in the training sample, the optimization
The target of function is to adjust the parameter of model, makes the output valve of neural network model and reference output valve difference minimum.Work as use
After multiple training samples are trained, the parameter value in model can be optimal, i.e. the output valve and ginseng of neural network model
Output valve difference minimum is examined, training is completed.After the completion of neural network model training, housebroken neural network model is obtained, its
In parameter value fix.Housebroken neural network model can be sent in memory module 154 and/or storage assembly 140
Stored, can also be sent to correction coefficient computing module 158.
Correction coefficient computing module 158 can use housebroken neural network model to calculate correction coefficient.Correction coefficient
Computing module 158 can obtain characteristic value, be input in housebroken neural network model, obtain using characteristic value as input value
Output valve is as correction coefficient.Wherein characteristic value can include the air projection value that the scanning air of CT scan system 100 obtains, also
Sweep parameter can be included.In certain embodiments, correction coefficient computing module 158 can be from scan components 110, processing component
130 and/or storage assembly 140 at obtain the characteristic value.
Fig. 2 is a kind of structural representation of the neural network model provided according to some embodiments of the present invention.Nerve net
Network model can include input 201, neural net layer 202 and output end 203.The input 201 of neural network model is used
In reception input value, and input value is input in neural net layer.Input value can include one or more values, such as xkV1、
xkV2、…、xkVN.Neural net layer 202 can carry out calculation process to input value, extract feature and the identification of input value, output
Operation result.Neural net layer 202 can include one or more operation layers, can include one or more in each operation layer
Node, one or more parameters can be included in each node.Each node in neural net layer 202 can receive one
Input of the output of all nodes of operation layer as the node, exports result to all nodes of latter operation layer after computing
In.The node of wherein the first operation layer receives all input values of input, and the operation result of the node of last operation layer is defeated
Go out to output end 203.Output end 203 is used for the operation result for receiving neural net layer, and using operation result as neutral net
The output valve of model.For example, the operation result of one or more nodes is distinguished in last layer of operation layer of neural net layer 202
It is a1、a2、…、aM, then output end 203 is by a1、a2、…、aMOutput valve as neural network model exports.Obviously
It is that user can determine the neutral net mould by adjusting the number of nodes of last layer of operation layer in neural net layer 202
The output valve quantity of type.
Fig. 3 is according to a kind of flow for the method 300 that correction coefficient is obtained based on neural network model provided by the invention
Schematic diagram.
As shown in figure 3, in the step 310, the first sweep parameter can be set.When user carries out CT scan, it can locate
The first sweep parameter is set in reason component 130.
In step 320, processing component 130 can determine whether the first sweep parameter of setting has corresponding optimum correction
Coefficient.For example, after processing component 130 receives the first sweep parameter of user's setting, can be examined from storage assembly 140
Rope, it is confirmed whether had and the first sweep parameter identical sweep parameter and corresponding optimum correction coefficient.If storage group
It is existing with the first sweep parameter identical sweep parameter and corresponding optimum correction coefficient in part 140, then into step 330.
In step 330, processing component 130 can obtain optimum correction coefficient corresponding with the first sweep parameter from storage assembly 140
As the first correction coefficient.
In step 340, CT scan system 100 can use the first sweep parameter of setting to sweep object to be scanned
Retouch, obtain actual projection value.
If in storage assembly 140 not with the first sweep parameter identical sweep parameter or corresponding optimum correction system
Number, then into step 350.
In step 350, user can control first sweep parameter of the CT scan system 100 based on setting to obtain feature
Value.Characteristic value can include it is at least one in the first sweep parameter and the first projection value, wherein the first projection value is included in the
The air projection value that air obtains is scanned under one scan parameter.For example, in certain embodiments, characteristic value can include first and throw
Shadow value.In further embodiments, characteristic value can include the first sweep parameter.In some other embodiments, characteristic value can be with
Including the first sweep parameter and the first projection value.This feature value can be sent to processing component 130 and/or storage assembly 140
In.
In step 360, the correction coefficient computing module 158 in correction coefficient computation module 150 can be obtained in step
The characteristic value obtained in 350, and be input to as input value in housebroken neural network model, obtain one or more
Second correction coefficient.Wherein described housebroken neural network model is the neural network model obtained after the completion of training.Training
Neural network model obtains the detailed process of housebroken neural network model, will be described below.Correction coefficient calculating group
After part 150 obtains the second correction coefficient, the second correction coefficient can be sent to processing component 130 and/or storage assembly 140
In.
In step 370, processing component 130 can use first correction coefficient or the second correction coefficient to actual throwing
Shadow value is corrected, the projection value after being corrected, for CT image reconstructions.It is appreciated that when the judgement in step 320 is
During "Yes", in step 370, actual projection value can be corrected using the first correction coefficient；When the judgement in step 320
For "No" when, actual projection value can be corrected using the second correction coefficient.
It will be understood by those skilled in the art that above-mentioned flow is only the exemplary illustration being corrected to projection value, also
There can be other changes.For example, in certain embodiments, step 320 and step 330 can be omitted.In addition, in above-mentioned flow
The execution sequence of each step can change, for example, step 340 can perform before step 350.
Fig. 4 is to obtain the method for housebroken neural network model according to a kind of training neural network model of the present invention
400 schematic flow sheet.
As shown in figure 4, in step 410, obtain training sample set.It will be understood by those skilled in the art that to nerve
Network model, which is trained, needs training sample.Training sample is concentrated and includes one or more training samples, it is, for example, possible to use
Sample acquisition module 152 obtains training sample.In certain embodiments, sample acquisition module 152 can obtain the second scanning ginseng
Number, the second projection value and optimum correction coefficient are to obtain training sample；In further embodiments, sample acquisition module 152 can
Correspond to the second projection value and optimum correction coefficient to obtain to obtain training sample；In some other embodiments, sample acquisition
Module 152 can obtain the second sweep parameter and optimum correction coefficient corresponding with second sweep parameter to obtain training sample
This.Wherein the second projection value, which is included under the second sweep parameter, scans the air projection value that air obtains.Sample acquisition module 152
Training sample can be sent into memory module 154 and stored.
Use the second different sweep parameters, such as the voltage or electric current of the change mesohigh ray tube of x-ray source 113
Value, can obtain different the second projection value and optimum correction coefficient, you can to obtain different training samples.Obtain one or
After multiple training samples, training sample set is obtained.Training sample set can be stored in memory module 154 and/or storage assembly 140
In.
At step 420, model training module 156 is concentrated from training sample and obtains a training sample, and it is with reference to output
It is worth for optimum correction coefficient.
In step 430, model training module 156 can be trained using the training sample to neural network model.
Model training module 156 can receive neural network model from memory module 154, can also build neural network model, god
At least part parameter through network model can pass through initialization, for example, random initializtion.Model training module 156 instructs this
The input value for practicing sample is input to neural network model, and neural network model is carried out at computing to the input value of the training sample
Reason, including the process such as feature extraction and feature recognition, finally obtain one group of output valve.Model training module 156 uses optimization letter
Several parameters to model are reversely adjusted.For example, in certain embodiments, the majorized function shown in formula (2) can be used
min||{ai}-{ai}ideaL | | the parameter of model is reversely adjusted, wherein { aiBe neural network model output valve,
{ai}idealFor the reference output valve of the training sample, the target of the majorized function is the parameter of reverse adjustment model, makes nerve net
The output valve of network model and reference output valve difference are minimum.Every time after training, the parameter in neural network model can all become
Change, as " initiation parameter " that input training sample is trained next time.
In step 440, whether the neural network model after the training of judgement of model training module 156 meets preparatory condition,
Wherein preparatory condition can be determined by user.For example, in certain embodiments, preparatory condition can be housebroken training sample
This quantity reaches preset value；In further embodiments, preparatory condition can be that the neural network model after training is surveyed
Examination, test result are qualified.If it is judged that being "Yes", then into step 450, housebroken neural network model is obtained, and
Housebroken neural network model is sent into memory module 154 and/or correction coefficient computing module 158, can also will be through
The neural network model of training is sent into storage assembly 140.If it is judged that being "No", then into step 420, continue to obtain
Take training sample to be trained, be not repeated herein.
If do not include the it is understood that in step 410, in the training sample that sample acquisition module 152 obtains
Two sweep parameters, the then characteristic value inputted in the step 360 in method 300 to housebroken neural network model can not also wrap
Include the first sweep parameter；If the second projection is not included in the training sample that in step 410, sample acquisition module 152 obtains
Value, then it can not also include first to the characteristic value that housebroken neural network model inputs in the step 360 in method 300 and throw
Shadow value.
Being preferable to carry out for the present invention is the foregoing is only, is not intended to limit the invention, for the technology of this area
For personnel, the present invention can have various modifications and variations.Within the spirit and principles of the invention, that is made any repaiies
Change, equivalent substitution, improvement etc., should be included in the scope of the protection.
Claims (10)
- A kind of 1. method being corrected to CT scan data, it is characterised in that including：Determine the first sweep parameter；First sweep parameter is at least partially based on, obtains characteristic value；Obtain housebroken neural network model；AndThe characteristic value is input in the housebroken neural network model, obtains school corresponding to first sweep parameter Positive coefficient, whereinThe characteristic value includes at least one of first sweep parameter and first projection value, and first projection value includes The projection value that air obtains is scanned under the first sweep parameter.
- 2. the method as described in claim 1, the housebroken neural network model of acquisition includes：Build neutral net mould Type, the neural network model is trained, obtains the housebroken neural network model.
- 3. method as claimed in claim 2, it is characterised in that it is described that the neural network model is trained, including：One or more training samples are obtained, the training sample includes at least one in the second sweep parameter and the second projection value Individual, second projection value, which is included under second sweep parameter, scans the projection value that air obtains；AndThe neural network model is trained using one or more of training samples.
- 4. method as claimed in claim 3, it is characterised in that it is described the neural network model is trained including：Make With min | | { ai}-{ai}ideal| | it is optimization object function, wherein { aiIt is that the training sample is input to the neutral net mould Output valve after type, { ai}idealFor the reference output valve of the training sample.
- 5. method as claimed in claim 3, it is characterised in that described to use one or more of training samples to the god Through network model be trained including：Whether the neural network model after training of judgement meets preparatory condition, if meeting default bar Part, stop the training to the neural network model.
- 6. the method as described in claim 1, it is characterised in that also include：Under first sweep parameter, CT scan data is obtained；AndThe CT scan data is corrected using the correction coefficient.
- 7. method as claimed in claim 6, it is characterised in that the CT scan data is corrected using the correction coefficient, Including the use of following correction function：<mrow> <mi>y</mi> <mo>=</mo> <msubsup> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>0</mn> </mrow> <mi>j</mi> </msubsup> <msub> <mi>a</mi> <mi>i</mi> </msub> <msup> <mi>x</mi> <mi>i</mi> </msup> </mrow>Wherein j is correction exponent number, and x is the CT scan data, aiFor the correction coefficient, y is the CT scan data after correction.
- A kind of 8. CT scan system, it is characterised in that including：Processing component, it is configured to determine the first sweep parameter；Correction coefficient computation module, it is configured to be at least partially based on first sweep parameter, obtains characteristic value；Obtain housebroken neural network model；AndThe characteristic value is input in the housebroken neural network model, obtains school corresponding to first sweep parameter Positive coefficient.Wherein, the characteristic value includes at least one of first sweep parameter and first projection value, first projection Value, which is included under the first sweep parameter, scans the projection value that air obtains.
- 9. system as claimed in claim 8, it is characterised in that also include：Scan components, it is configured under first sweep parameter, obtains CT scan data；Wherein described processing component is also matched somebody with somebody It is set to and the CT scan data is corrected using the correction coefficient.
- 10. the system as shown in claim 8, it is characterised in that also include：Sample acquisition module, is configured to obtain one or more training samples, the training sample include the second sweep parameter and At least one in second projection value, second projection value, which is included under second sweep parameter, scans the throwing that air obtains Shadow value；AndModel training module, it is configured so that one or more of training samples are trained to the neural network model.
Priority Applications (4)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710524197.2A CN107374657B (en) | 2017-06-30 | 2017-06-30 | Method for correcting CT scanning data and CT scanning system |
| US15/954,953 US10977843B2 (en) | 2017-06-28 | 2018-04-17 | Systems and methods for determining parameters for medical image processing |
| US17/228,690 US11908046B2 (en) | 2017-06-28 | 2021-04-12 | Systems and methods for determining processing parameter for medical image processing |
| US18/437,210 US20240185486A1 (en) | 2017-06-28 | 2024-02-08 | Systems and methods for determining parameters for medical image processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710524197.2A CN107374657B (en) | 2017-06-30 | 2017-06-30 | Method for correcting CT scanning data and CT scanning system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107374657A true CN107374657A (en) | 2017-11-24 |
| CN107374657B CN107374657B (en) | 2021-05-11 |
Family
ID=60334713
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710524197.2A Active CN107374657B (en) | 2017-06-28 | 2017-06-30 | Method for correcting CT scanning data and CT scanning system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107374657B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109523024A (en) * | 2018-11-22 | 2019-03-26 | 天津大学 | Energy spectrum correction method towards medical X-ray detector |
| CN111684486A (en) * | 2018-01-31 | 2020-09-18 | 皇家飞利浦有限公司 | Virtual non-contrast images with improved image quality generated by spectral Computed Tomography (CT) scanners |
| CN111935892A (en) * | 2019-05-13 | 2020-11-13 | 中科智云科技有限公司 | Method and apparatus for measuring plasma state |
| CN112014870A (en) * | 2019-05-31 | 2020-12-01 | 佳能医疗系统株式会社 | Radiation detection device, energy correction method, and program |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH03102477A (en) * | 1989-06-26 | 1991-04-26 | Fuji Photo Film Co Ltd | Radial image processing device |
| US6422751B1 (en) * | 1998-08-07 | 2002-07-23 | General Electric Company | Method and system for prediction of exposure and dose area product for radiographic x-ray imaging |
| US20050220265A1 (en) * | 2003-06-25 | 2005-10-06 | Besson Guy M | Methods for acquiring multi spectral data of an object |
| CN1743835A (en) * | 2004-07-14 | 2006-03-08 | 株式会社东芝 | X-ray computed tomography photography system and method for correcting data of the system |
| CN101336828A (en) * | 2007-07-06 | 2009-01-07 | Ge医疗系统环球技术有限公司 | Acquisition method and device of CT value correction paper |
| CN102768759A (en) * | 2012-07-04 | 2012-11-07 | 深圳安科高技术股份有限公司 | Intraoperative CT (Computed Tomography) image beam hardening artifact correction method and device |
| CN103961125A (en) * | 2013-01-31 | 2014-08-06 | 东北大学 | CT (Computed Tomography) value correcting method for cone-beam CT |
| CN106327495A (en) * | 2016-08-26 | 2017-01-11 | 穆达文 | Biological bone recognition method, device and system |
| US20170143312A1 (en) * | 2014-09-03 | 2017-05-25 | Contextvision Ab | Methods and systems for automatic control of subjective image quality in imaging of objects |
-
2017
- 2017-06-30 CN CN201710524197.2A patent/CN107374657B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH03102477A (en) * | 1989-06-26 | 1991-04-26 | Fuji Photo Film Co Ltd | Radial image processing device |
| US6422751B1 (en) * | 1998-08-07 | 2002-07-23 | General Electric Company | Method and system for prediction of exposure and dose area product for radiographic x-ray imaging |
| US20050220265A1 (en) * | 2003-06-25 | 2005-10-06 | Besson Guy M | Methods for acquiring multi spectral data of an object |
| CN1743835A (en) * | 2004-07-14 | 2006-03-08 | 株式会社东芝 | X-ray computed tomography photography system and method for correcting data of the system |
| CN101336828A (en) * | 2007-07-06 | 2009-01-07 | Ge医疗系统环球技术有限公司 | Acquisition method and device of CT value correction paper |
| CN102768759A (en) * | 2012-07-04 | 2012-11-07 | 深圳安科高技术股份有限公司 | Intraoperative CT (Computed Tomography) image beam hardening artifact correction method and device |
| CN103961125A (en) * | 2013-01-31 | 2014-08-06 | 东北大学 | CT (Computed Tomography) value correcting method for cone-beam CT |
| US20170143312A1 (en) * | 2014-09-03 | 2017-05-25 | Contextvision Ab | Methods and systems for automatic control of subjective image quality in imaging of objects |
| CN106327495A (en) * | 2016-08-26 | 2017-01-11 | 穆达文 | Biological bone recognition method, device and system |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111684486A (en) * | 2018-01-31 | 2020-09-18 | 皇家飞利浦有限公司 | Virtual non-contrast images with improved image quality generated by spectral Computed Tomography (CT) scanners |
| CN109523024A (en) * | 2018-11-22 | 2019-03-26 | 天津大学 | Energy spectrum correction method towards medical X-ray detector |
| CN111935892A (en) * | 2019-05-13 | 2020-11-13 | 中科智云科技有限公司 | Method and apparatus for measuring plasma state |
| CN111935892B (en) * | 2019-05-13 | 2022-11-22 | 中科智云科技有限公司 | Method and apparatus for measuring plasma state |
| CN112014870A (en) * | 2019-05-31 | 2020-12-01 | 佳能医疗系统株式会社 | Radiation detection device, energy correction method, and program |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107374657B (en) | 2021-05-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107374657A (en) | The method and CT scan system being corrected to CT scan data | |
| CN104318536B (en) | The bearing calibration of CT images and device | |
| CN105854191B (en) | 3-dimensional dose verification system and verification method in a kind of radiotherapy | |
| EP3226766B1 (en) | System and method for image calibration | |
| Hansen et al. | The application of transit dosimetry to precision radiotherapy | |
| US7715522B2 (en) | X-ray CT apparatus | |
| US6848827B2 (en) | Method and apparatus for calibrating detector spectral response | |
| CN107945132A (en) | A kind of artifact correction method and device of the CT images based on neutral net | |
| CN106413236A (en) | Exposure parameter adjusting method and device | |
| US20140226887A1 (en) | Medical image processing device and medical image processing method | |
| CN105894525A (en) | Detector correction method for positive electron emission tomography system | |
| US20040017880A1 (en) | Method and system for low dose image simulation for imaging systems | |
| CN106683146A (en) | Image reconstruction method and parameter determining method of image reconstruction algorithm | |
| CN111627083B (en) | Bone hardening artifact correction method, device, computer equipment and readable storage medium | |
| CN104700438A (en) | Image reconstruction method and device | |
| CN103065340A (en) | Method and system for expanding axial coverage in iterative reconstruction in computer tomography (ct) | |
| CN107530042A (en) | X ray CT device and iterative refined parameters determining method | |
| JP6446361B2 (en) | X-ray CT apparatus and correction processing apparatus | |
| CN109924993A (en) | Image scanning agreement automatic planning, device, electronic equipment and storage medium | |
| CN113167913A (en) | Energy weighting of photon counts for conventional imaging | |
| US20220414832A1 (en) | X-ray imaging restoration using deep learning algorithms | |
| CN107077748A (en) | For beginning image of the spectrogram as iterative approximation | |
| Benson et al. | Synthetic CT noise emulation in the raw data domain | |
| CN106530335A (en) | Computed tomography self-calibration without calibration targets | |
| US8938104B2 (en) | Systems and methods for adaptive 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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 201807 Shanghai City, north of the city of Jiading District Road No. 2258 Applicant after: Shanghai Lianying Medical Technology Co., Ltd Address before: 201807 Shanghai City, north of the city of Jiading District Road No. 2258 Applicant before: SHANGHAI UNITED IMAGING HEALTHCARE Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |