CN110270016A - Proton therapeutic monitoring method, device and system neural network based - Google Patents
Proton therapeutic monitoring method, device and system neural network based Download PDFInfo
- Publication number
- CN110270016A CN110270016A CN201910445746.6A CN201910445746A CN110270016A CN 110270016 A CN110270016 A CN 110270016A CN 201910445746 A CN201910445746 A CN 201910445746A CN 110270016 A CN110270016 A CN 110270016A
- Authority
- CN
- China
- Prior art keywords
- proton
- neural network
- peak position
- dosage
- simulation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61N—ELECTROTHERAPY; MAGNETOTHERAPY; RADIATION THERAPY; ULTRASOUND THERAPY
- A61N5/00—Radiation therapy
- A61N5/10—X-ray therapy; Gamma-ray therapy; Particle-irradiation therapy
- A61N5/1048—Monitoring, verifying, controlling systems and methods
- A61N5/1064—Monitoring, verifying, controlling systems and methods for adjusting radiation treatment in response to monitoring
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61N—ELECTROTHERAPY; MAGNETOTHERAPY; RADIATION THERAPY; ULTRASOUND THERAPY
- A61N5/00—Radiation therapy
- A61N5/10—X-ray therapy; Gamma-ray therapy; Particle-irradiation therapy
- A61N2005/1085—X-ray therapy; Gamma-ray therapy; Particle-irradiation therapy characterised by the type of particles applied to the patient
- A61N2005/1087—Ions; Protons
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Pathology (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Nuclear Medicine (AREA)
Abstract
The present invention relates to a kind of proton therapeutic monitoring method, device and systems neural network based to determine tumor focus region according to the said three-dimensional body mould constructed in advance；After proton beam radiation is to tumor focus region, positron radionuclide is measured within a preset time, obtains the distributed intelligence of positron radionuclide；PET image is constructed using image reconstruction algorithm according to distributed intelligence；PET image is input to the neural net regression model and neural network classification model of building, determines dosage distribution and the range of proton beam；It is distributed according to range and dosage, determines bragg peak position；Whether detection bragg peak position and dosage distribution meet preset proton therapeutic requirement；If it is not, adjustment proton beam goes out beam parameter.Using technical solution of the present invention, human organ respiratory movement influence caused by measurement can reduce, improve to the range of proton and the precision of dosage distribution measuring, improve the accuracy of proton therapeutic.
Description
Technical field
The present invention relates to nucleus medical image technical fields, and in particular to a kind of proton therapeutic monitoring side neural network based
Method, device and system.
Background technique
Proton therapeutic is a kind of accurate tip radiotherapy technology for the treatment of malignant tumour (cancer), during proton therapeutic, matter
Beamlet first slowly rises in the intracorporal exposure dose curve of people, then gradually becomes faster, maximum dose is generated at bragg peak
Amount deposition, curve rapid decrease and goes to zero later.It is maximum that the Bragg peak character of proton beam can be such that tumor tissues receive
Exposure dose, while the normal organ for being avoided that tumor tissues rear is reduced the side effect for the treatment of by radiation injury.
The advantage of proton therapeutic is: 1) dosage that the dose ratio at proton bragg peak has just enter at human body is higher by three and arrives
Four times, the dosage after bragg peak is about 0；2) the swollen of different depth can be irradiated by adjusting the energy of proton beam
Tumor, to make the tumour of proton therapeutic adaptation different sizes and shapes；3) when proton transport, there is lesser scattering and background,
So that irradiation field edge clear, therefore can treat the close tumour of distance sensitive organ.
During proton therapeutic, need to be monitored the dosage distribution after proton range and proton irradiation, but existing
There is the monitoring method in technology to be affected by factors such as the uncertainties of human organ respiratory movement and range, causes pair
The range of proton and the precision of dosage distribution measuring are lower, so that the accuracy of proton therapeutic is lower.
Summary of the invention
In view of this, the purpose of the present invention is to provide a kind of proton therapeutic monitoring methods neural network based, device
It is lower to the precision of the range of proton and dosage distribution measuring in the prior art to solve with system, so that proton therapeutic
The lower problem of accuracy.
In order to achieve the above object, the present invention adopts the following technical scheme:
A kind of proton therapeutic monitoring method neural network based, comprising:
According to the said three-dimensional body mould constructed in advance, tumor focus region is determined；
After proton beam radiation is to the tumor focus region, positron radionuclide is measured within a preset time, is obtained described
The distributed intelligence of positron radionuclide；
Positron e mission computed tomography PET figure is constructed using image reconstruction algorithm according to the distributed intelligence
Picture；
The PET image is input to the neural network classification model of building, determines the range of the proton beam；
The PET image is input to the neural net regression model of building, determines the dosage distribution of the proton beam；
It is distributed according to the range and the dosage, determines bragg peak position；
It detects the bragg peak position and whether dosage distribution meets preset proton therapeutic requirement；
If the bragg peak position and/or the dosage, which are distributed, does not meet preset proton therapeutic requirement, described in adjustment
Proton beam goes out beam parameter.
Further, it is pre- whether above-mentioned method, the detection bragg peak position and dosage distribution meet
If proton therapeutic requirement, comprising:
Detect whether the bragg peak position falls in the tumor focus region, and, detect the dosage distribution
Whether radiotherapy planning content in the proton therapeutic requirement is met；
If the bragg peak position is not fallen in the tumor focus region, detect that the bragg peak position is not inconsistent
Close the proton therapeutic requirement；
If the dosage distribution does not meet the radiotherapy planning content in the proton therapeutic requirement, the dosage point is detected
Cloth does not meet the proton therapeutic requirement.
Further, method described above, the neural network classification mould that the PET image is input to building
Type, before the range for determining the proton beam, further includes:
The reaction process of scheduled human body target organ is squeezed by simulating proton, constructs classification based training collection and regression training
Collection；
Based on neural network algorithm, first object neural network model is trained using the classification based training collection, is obtained
To the neural network classification model；
Based on the neural network algorithm, the second target nerve network model is instructed using the regression training collection
Practice, obtains the neural net regression model.
Further, in method described above, the reaction that scheduled human body target organ is squeezed by simulating proton
Process constructs classification based training collection and regression training collection, comprising:
The reaction process of the scheduled human body target organ is squeezed by simulating the proton, obtain the distribution of simulation dosage and
Simulate range；
PET imaging simulation is carried out using described image algorithm for reconstructing according to the reaction process, constructs simulation PET figure
Picture；
The simulation PET image and the simulation range are built into the classification based training collection；
The simulation PET image and simulation dosage distribution are built into the regression training collection.
Further, in method described above, it is described go out beam parameter include: out the beam moment, go out Shu Sudu, go out beam trajectory
Beam dose out.
The present invention also provides a kind of proton therapeutic monitoring devices neural network based, comprising:
First determining module, for determining tumor focus region according to the said three-dimensional body mould constructed in advance；
Measurement module, for measuring positive electron within a preset time after proton beam radiation is to the tumor focus region
Nucleic obtains the distributed intelligence of the positron radionuclide；
First building module, for constructing positron emission meter using image reconstruction algorithm according to the distributed intelligence
Calculation machine tomography PET image；
Second determining module determines the matter for the PET image to be input to the neural network classification model of building
The range of beamlet；The PET image is input to the neural net regression model of building, determines the dosage point of the proton beam
Cloth；
Third determining module determines bragg peak position for being distributed according to the range and the dosage；
Detection module, for detecting whether the bragg peak position and dosage distribution meet preset proton therapeutic
It is required that；
Module is adjusted, if not meeting preset proton therapeutic for the bragg peak position and/or dosage distribution
It is required that adjusting the beam parameter out of the proton beam.
Further, device described above, the detection module are specifically used for:
Detect whether the bragg peak position falls in the tumor focus region, and, detect the dosage distribution
Whether radiotherapy planning content in the proton therapeutic requirement is met；
If the bragg peak position is not fallen in the tumor focus region, detect that the bragg peak position is not inconsistent
Close the proton therapeutic requirement；
If the dosage distribution does not meet the radiotherapy planning content in the proton therapeutic requirement, the dosage point is detected
Cloth does not meet the proton therapeutic requirement.
Further, device described above, further includes:
Second building module, for squeezing into the reaction process of scheduled human body target organ, building classification by simulating proton
Training set and regression training collection；
Training module, for being based on neural network algorithm, using the classification based training collection to first object neural network mould
Type is trained, and obtains the neural network classification model；Based on the neural network algorithm, the regression training collection pair is utilized
Second target nerve network model is trained, and obtains the neural net regression model.
Further, in device described above, the second building module includes: acquiring unit, analog image building
Unit and training set construction unit；
The acquiring unit, for squeezing into the reaction process of the scheduled human body target organ by simulating the proton,
Obtain the distribution of simulation dosage and simulation range；
The analog image construction unit, for being carried out according to the reaction process using described image algorithm for reconstructing
PET imaging simulation constructs simulation PET image；
The training set construction unit, for the simulation PET image and the simulation range to be built into the classification
Training set；The simulation PET image and simulation dosage distribution are built into the regression training collection.
The present invention also provides a kind of proton therapeutics neural network based to monitor system, comprising: CT machine and proton therapeutic prison
Measurement equipment；
The CT machine is connected with the proton therapeutic monitoring device；
The CT machine is used to construct the said three-dimensional body mould of human body, and the said three-dimensional body mould is sent to the proton therapeutic and is monitored
Equipment；
The proton therapeutic monitoring device is at least used to execute any of the above-described proton therapeutic monitoring side neural network based
Method.
Proton therapeutic monitoring method, device and system neural network based of the invention, according to the three-dimensional constructed in advance
Body mould determines tumor focus region；After proton beam radiation is to tumor focus region, positive electricity daughter nucleus is measured within a preset time
Element obtains the distributed intelligence of positron radionuclide；Positron emission meter is constructed using image reconstruction algorithm according to distributed intelligence
Calculation machine tomography PET image；PET image is input to the neural network classification model of building, determines the range of proton beam；It will
PET image is input to the neural net regression model of building, determines the dosage distribution of proton beam；It is distributed according to range and dosage,
Determine bragg peak position；Whether detection bragg peak position and dosage distribution meet preset proton therapeutic requirement；If Bradley
Lattice peak position and/or dosage distribution do not meet proton therapeutic requirement, and adjust proton beam goes out beam parameter.Using technology of the invention
In proton therapeutic nuclear reaction occurs for scheme, generates various positron radionuclides, is deposited on the path and end of line, passes through
PET imaging technique measures positron radionuclide, and constructs PET image, can reduce in this way human organ respiratory movement and
The factors such as the uncertainty of range are influenced caused by measurement, are improved to the range of proton and the precision of dosage distribution measuring, from
And the accuracy of proton therapeutic is improved, guarantee the therapeutic effect that proton therapeutic is carried out to tumor patient.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
It can the limitation present invention.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is the flow chart of proton therapeutic monitoring method embodiment one neural network based of the invention；
Fig. 2 is the flow chart of proton therapeutic monitoring method embodiment two neural network based of the invention；
Fig. 3 is the structural schematic diagram of proton therapeutic monitoring device embodiment one neural network based of the invention；
Fig. 4 is the structural schematic diagram of proton therapeutic monitoring device embodiment two neural network based of the invention；
Fig. 5 is the structural schematic diagram of proton therapeutic neural network based monitoring system embodiment of the invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, technical solution of the present invention will be carried out below
Detailed description.Obviously, described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Base
Embodiment in the present invention, those of ordinary skill in the art are obtained all without making creative work
Other embodiment belongs to the range that the present invention is protected.
Fig. 1 is the flow chart of proton therapeutic monitoring method embodiment one neural network based of the invention.Such as Fig. 1 institute
Show, the proton therapeutic monitoring method neural network based of the present embodiment can specifically include following steps:
The said three-dimensional body mould that S101, basis construct in advance, determines tumor focus region；
Firstly, scanning the whole body of patient by CT machine, multiframe CT image is obtained, then multiframe CT image is combined, from
And it is built into a said three-dimensional body mould.The said three-dimensional body mould built according to CT mechanism is first passed through in advance, sketches out tumor focus region.In addition,
When determining tumor focus region, can also be determined in CT image.
S102, after proton beam radiation is to tumor focus region, measure positron radionuclide within a preset time, obtain positive electricity
The distributed intelligence of daughter nuclide；
When proton beam therapy head using proton beam radiation to tumor focus region after, within a preset time to positron radionuclide
It measures, to obtain the distributed intelligence of positron radionuclide.Wherein, preset time 60s is needed in proton beam radiation knot
The distribution of positron radionuclide is measured in 60s after beam, In-beam can be used by measuring to positron radionuclide distribution
PET system.
S103, positron e mission computed tomography is constructed using image reconstruction algorithm according to distributed intelligence
(Positron Emission Tomography, PET) image；
Through the above steps, after the distributed intelligence for obtaining positron radionuclide, using image reconstruction algorithm, according to distributed intelligence
Construct PET image.Wherein, PET is the picture reproducer of the gene for reflecting lesion, molecule, metabolism and functional status.It is using just
The human metabolites such as electron species labelled glucose reflect that its metabolism becomes to the intake of imaging agent as imaging agent, by lesion
Change, to provide the biological metabolism information of disease for clinic.In the present embodiment, the image reconstruction algorithm of use is preferably three-dimensional two
Secondary algorithm for reconstructing (3D Reprojection, 3DRP).
S104, the neural network classification model that PET image is input to building, determine the range of proton beam；
Through the above steps, after being built into PET image, PET image is input to the neural network classification model of building, is passed through
The processing for crossing neural network classification model obtains the range of proton beam.Wherein, neural network classification model is built in advance
Model.
S105, the neural net regression model that PET image is input to building determine the dosage distribution of proton beam；
Through the above steps, after being built into PET image, PET image is input to the neural net regression model of building, is passed through
The processing for crossing neural net regression model obtains the dosage distribution of proton beam.Wherein, neural net regression model is to construct in advance
Good model.
S106, it is distributed according to range and dosage, determines bragg peak position；
Through the above steps, after the range and the dosage distribution that obtain proton beam, in the range section of proton beam, dosage point
The position of dose maximum is chosen in cloth, and using the position as bragg peak position.Medically, heavy ion beam enters body
After will not largely release energy at once, only can just discharge its most of energy in the position that heavy ion stops, form one
Sharp energy peak, this energy peak are just called bragg peak.
Whether S107, detection bragg peak position and dosage distribution meet preset proton therapeutic requirement；
Through the above steps, after the dosage distribution for obtaining bragg peak position and proton beam, to bragg peak position and matter
The dosage distribution of beamlet is detected, and determines whether the dosage distribution of bragg peak position and proton beam meets preset proton and control
It treats and requires.Wherein, the dosage distribution of proton beam described herein refers to that major dose is distributed, if detecting, major dose distribution is fallen
In tumor focus region, and there is a small amount of dosage distribution not fall in tumor focus region, that a small amount of dosage distribution
It can be ignored.
Preset proton therapeutic requires to include the radiotherapy position pre-established and radiotherapy planning content, radiotherapy planning treatment
Content includes the dosage etc. of this treatment.Whether detection bragg peak position and dosage distribution meet preset proton therapeutic requirement
Specifically include: whether detection bragg peak position falls in tumor focus region, and, whether detection dosage distribution meets proton
Treat the radiotherapy planning content in requiring.If bragg peak position is not fallen in tumor focus region, Prague is detected
Peak position does not meet proton therapeutic requirement；If dosage distribution does not meet the radiotherapy planning content in proton therapeutic requirement, examine
It measures dosage distribution and does not meet proton therapeutic requirement.
If S108, bragg peak position and/or dosage distribution do not meet proton therapeutic requirement, adjust proton beam goes out beam ginseng
Number.
Detection through the above steps, if the dosage of bragg peak position and/or proton beam distribution do not meet it is preset
Proton therapeutic requirement, i.e., bragg peak position is not fallen in tumor focus region, and/or, dosage distribution does not meet proton therapeutic
Radiotherapy planning content in it is required that, then need to adjust proton beam goes out beam parameter, is then detected again, until Prague peak position
It sets after meeting preset proton therapeutic requirement with the distribution of the dosage of proton beam, then is with the beam parameter that goes out of proton beam at this time
Standard carries out proton therapeutic to patient.Wherein, proton beam go out beam parameter include: out the beam moment, go out Shu Sudu, go out beam trajectory and
Beam dose etc. out.If bragg peak position is not fallen in tumor focus region, but deviate tumor focus region, then can
It can be when carrying out proton beam radiation, the position of irradiation deviates, and need to adjust proton beam goes out beam trajectory, i.e. adjustment proton
The irradiation position of beam treatment head；If the dosage distribution of proton beam does not meet the radiotherapy planning content in proton therapeutic requirement, i.e.,
Dosage distribution is not inconsistent with the dosage required in radiotherapy planning content, then explanation is when carrying out proton beam radiation, dosage is not right, needs
Adjust proton beam goes out beam dose.
If the distribution of the dosage of bragg peak position and proton beam meets preset proton therapeutic requirement, just keep
Proton beam goes out beam parameter constant, and the proton beam being subject at this time goes out beam parameter, carries out proton therapeutic to patient.
The proton therapeutic monitoring method neural network based of the present embodiment is determined according to the said three-dimensional body mould constructed in advance
Tumor focus region；After proton beam radiation is to tumor focus region, positron radionuclide is measured within a preset time, obtains positive electricity
The distributed intelligence of daughter nuclide；Positron e mission computed tomography is constructed using image reconstruction algorithm according to distributed intelligence
PET image；PET image is input to the neural network classification model of building, determines the range of proton beam；PET image is inputted
To the neural net regression model of building, the dosage distribution of proton beam is determined；It is distributed according to range and dosage, determines bragg peak
Position；Whether detection bragg peak position and dosage distribution meet preset proton therapeutic requirement；If bragg peak position and/or
Dosage distribution does not meet proton therapeutic requirement, and adjust proton beam goes out beam parameter.In the present embodiment, core occurs in proton therapeutic
Reaction, generates various positron radionuclides, is deposited on the path and end of line, by PET imaging technique to positron radionuclide
Distribution measure, and construct PET image, can reduce in this way uncertainty of human organ respiratory movement and range etc. because
Element is influenced caused by measurement, is improved to the range of proton and the precision of dosage distribution measuring, to improve the essence of proton therapeutic
Exactness guarantees the therapeutic effect that proton therapeutic is carried out to tumor patient.
Fig. 2 is the flow chart of proton therapeutic monitoring method embodiment two neural network based of the invention.Such as Fig. 2 institute
Show, the proton therapeutic monitoring method neural network based of the present embodiment can specifically include following steps:
The said three-dimensional body mould that S201, basis construct in advance, determines tumor focus region；
The implementation procedure of the step is identical as the implementation procedure of S101 shown in FIG. 1, and details are not described herein again.
S202, after proton beam radiation is to tumor focus region, measure positron radionuclide within a preset time, obtain positive electricity
The distributed intelligence of daughter nuclide；
The implementation procedure of the step is identical as the implementation procedure of S102 shown in FIG. 1, and details are not described herein again.
S203, PET image is constructed using image reconstruction algorithm according to distributed intelligence；
The implementation procedure of the step is identical as the implementation procedure of S103 shown in FIG. 1, and details are not described herein again.
S204, the reaction process that scheduled human body target organ is squeezed by simulating proton, construct classification based training collection and recurrence
Training set；
Firstly, simulation proton squeezes into the reaction process of scheduled human body target organ, to obtain the simulation dosage point of proton
Cloth and simulation range, wherein the reaction process that simulation proton squeezes into scheduled human body target organ can putting by GATE program
Penetrate the realization for the treatment of function.
Secondly as the reaction process that proton squeezes into scheduled human body target organ is generated by simulation, therefore it can be with
Immediately arrive at proton goes out beam trajectory information, using image reconstruction algorithm, imitates the beam trajectory information progress PET imaging that goes out of proton
Very, to construct simulation PET image.The process of building simulation PET image was carried out in line off time, the present embodiment
In, what simulation proton squeezed into scheduled human body target organ presets preferably the line period 3 seconds, includes 1 second beam time
With 2 seconds intermittent times, the preset time of entire therapeutic process was preferably 5 minutes, that is, included 100 pulse periods.
Finally, the simulation range of the simulation PET image and proton that construct is built into a training set, instructed as classification
Practice collection；The simulation dosage distribution of the simulation PET image and proton that construct is built into a training set, as regression training
Collection.
S205, it is based on neural network algorithm, first object neural network model is trained using classification based training collection, is obtained
To neural network classification model；
Through the above steps, after constructing classification based training collection, a neural network model is obtained as first object nerve
Network model is trained the first object neural network model using classification based training collection, in the training process, the model
Parameter is continuously available optimization, when the parameter of the model is optimal, just using the first object neural network model after training as
Neural network classification model.
S206, it is based on neural network algorithm, the second target nerve network model is trained using regression training collection, is obtained
To neural net regression model；
Through the above steps, after constructing regression training collection, a Recognition with Recurrent Neural Network model is obtained as the second target
Neural network model is trained the second target nerve network model using regression training collection, in the training process, the mould
The parameter of type is continuously available optimization, when the parameter of the model is optimal, just by the second target nerve network model after training
As neural net regression model.
Wherein, when being trained using regression training collection to the second target nerve network model, the second target nerve network
Model is Recognition with Recurrent Neural Network model, using the radial direction depth of Proton-Induced Reactions as the time step of the Recognition with Recurrent Neural Network,
The input of each door (gate) is the hidden layer location mode of last moment and the state of the moment input layer in hidden layer, defeated
It is the information weight factor for corresponding to each unit of cell state out.Information weight factor control information input, output or
The percentage of forgetting.The relational expression of door input and output information are as follows:
ht=σ (Whhht-1+Whxxt)
Wherein, htIt is the state of t moment cell；xtIt is the state of t moment input layer；WhxFor xtWeight factor；WhhFor
ht-1Weight factor.
S207, the neural network classification model that PET image is input to building, determine the range of proton beam；
The implementation procedure of the step is identical as the implementation procedure of S104 shown in FIG. 1, and details are not described herein again.
S208, the neural net regression model that PET image is input to building determine the dosage distribution of proton beam；
The implementation procedure of the step is identical as the implementation procedure of S105 shown in FIG. 1, and details are not described herein again.
S209, it is distributed according to range and dosage, determines bragg peak position；
The implementation procedure of the step is identical as the implementation procedure of S106 shown in FIG. 1, and details are not described herein again.
Whether S210, detection bragg peak position and dosage distribution meet preset proton therapeutic requirement；
The implementation procedure of the step is identical as the implementation procedure of S107 shown in FIG. 1, and details are not described herein again.
If S211, bragg peak position and/or dosage distribution do not meet proton therapeutic requirement, adjust proton beam goes out beam ginseng
Number.
The implementation procedure of the step is identical as the implementation procedure of S108 shown in FIG. 1, and details are not described herein again.
The execution sequence of step S205 and S206 are unrestricted in the present embodiment, can both first carry out step S205, then execute
Step S206；Step S206 can also be first carried out, then executes step S205.Step S207's and S208 executes sequence in the present embodiment
It is unrestricted, step S207 can be both first carried out, then execute step S208；Step S208 can also be first carried out, then executes step S207.
The proton therapeutic monitoring method neural network based of the present embodiment is determined according to the said three-dimensional body mould constructed in advance
Tumor focus region；After proton beam radiation is to tumor focus region, positron radionuclide is measured within a preset time, obtains positive electricity
The distributed intelligence of daughter nuclide；PET image is constructed using image reconstruction algorithm according to distributed intelligence；Then, by simulating proton
The reaction process of scheduled human body target organ is squeezed into, classification based training collection and regression training collection are constructed；Based on neural network algorithm, benefit
First object neural network model is trained with classification based training collection, using regression training collection to the second target nerve network mould
Type is trained, to obtain neural network classification model and neural net regression model；PET image is input to the mind of building
Through network class model, the range of proton beam is determined；PET image is input to the neural net regression model of building, determines matter
The dosage of beamlet is distributed；It is distributed according to range and dosage, determines bragg peak position；Detect bragg peak position and dosage distribution
Whether preset proton therapeutic requirement is met；If bragg peak position and/or dosage distribution do not meet proton therapeutic requirement, adjust
Proton beam goes out beam parameter.In the present embodiment, nuclear reaction occurs in proton therapeutic, generates various positron radionuclides, is deposited on
On the path and end of line, the distribution of positron radionuclide is measured by PET imaging technique, and construct PET image, this
Sample can reduce the influence caused by measurement of the factors such as the uncertainty of human organ respiratory movement and range, improve to proton
The precision of range and dosage distribution measuring guarantees to carry out proton therapeutic to tumor patient to improve the accuracy of proton therapeutic
Therapeutic effect.
In order to more comprehensively, correspond to proton therapeutic monitoring method neural network based provided in an embodiment of the present invention, this
Application additionally provides proton therapeutic monitoring device neural network based.
Fig. 3 is the structural schematic diagram of proton therapeutic monitoring device embodiment one neural network based of the invention.Such as Fig. 3
Shown, the proton therapeutic monitoring device neural network based of the present embodiment includes: the first determining module 101, measurement module
102, the first building module 103, the second determining module 104, third determining module 105, detection module 106 and adjustment module 107.
First determining module 101, for determining tumor focus region according to the said three-dimensional body mould constructed in advance；
Measurement module 102, for after proton beam radiation is to tumor focus region, measurement positive electricity daughter nucleus within a preset time
Element obtains the distributed intelligence of positron radionuclide；
First building module 103, for constructing PET image using image reconstruction algorithm according to distributed intelligence；
Second determining module 104 determines proton beam for PET image to be input to the neural network classification model of building
Range；PET image is input to the neural net regression model of building, determines the dosage distribution of proton beam；
Third determining module 105 determines bragg peak position for being distributed according to range and dosage；
Detection module 106, for detecting whether bragg peak position and dosage distribution meet preset proton therapeutic requirement；
Module 107 is adjusted, if not meeting proton therapeutic requirement for bragg peak position and/or dosage distribution, adjusts matter
Beamlet goes out beam parameter.
The proton therapeutic monitoring device neural network based of the present embodiment, the first determining module 101 is according to preparatory building
Said three-dimensional body mould, determine tumor focus region；After proton beam radiation is to tumor focus region, measurement module 102 is when default
Interior measurement positron radionuclide, obtains the distributed intelligence of positron radionuclide；First building module 103 is used according to distributed intelligence
Image reconstruction algorithm constructs PET image；PET image is input to the neural network classification mould of building by the second determining module 104
Type determines the range of proton beam；PET image is input to the neural net regression model of building again, determines the dosage of proton beam
Distribution；Third determining module 105 is distributed according to range and dosage, determines bragg peak position；Detection module 106 detects Prague
Whether peak position and dosage distribution meet preset proton therapeutic requirement；If bragg peak position and/or dosage distribution are not met
Preset proton therapeutic requirement, adjustment module 107 adjust the beam parameter out of proton beam.In the present embodiment, sent out in proton therapeutic
Raw nuclear reaction, generates various positron radionuclides, is deposited on the path and end of line, by PET imaging technique to positive electron
The distribution of nucleic measures, and constructs PET image, can reduce the uncertainty of human organ respiratory movement and range in this way
Etc. factors influenced caused by measurement, improve to the range of proton and the precision of dosage distribution measuring, to improve proton therapeutic
Accuracy, guarantee to tumor patient carry out proton therapeutic therapeutic effect.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
Fig. 4 is the structural schematic diagram of proton therapeutic monitoring device embodiment two neural network based of the invention.Such as Fig. 4
Shown, the proton therapeutic monitoring device neural network based of the present embodiment further includes the second building module 108 and training module
109, the second building module 108 includes acquiring unit 1081, analog image construction unit 1082 and training set construction unit 1083.
Acquiring unit 1081 obtains simulation for squeezing into the reaction process of scheduled human body target organ by simulating proton
Dosage distribution and simulation range；
Analog image construction unit 1082, for it is imitative to carry out PET imaging using image reconstruction algorithm according to reaction process
Very, simulation PET image is constructed；
Training set construction unit 1083 is built into classification based training collection for that will simulate PET image and simulation range；It will simulation
PET image and simulation dosage distribution are built into regression training collection；
Training module 109, for being based on neural network algorithm, using classification based training collection to first object neural network model
It is trained, obtains neural network classification model；Based on neural network algorithm, using regression training collection to the second target nerve net
Network model is trained, and obtains neural net regression model.
The proton therapeutic monitoring device neural network based of the present embodiment, nuclear reaction occurs in proton therapeutic, generates
Various positron radionuclides, are deposited on the path and end of line, are carried out by PET imaging technique to the distribution of positron radionuclide
Measurement, and PET image is constructed, it can reduce human organ respiratory movement influence caused by measurement in this way, improve to proton
The precision of range and dosage distribution measuring guarantees to carry out proton therapeutic to tumor patient to improve the accuracy of proton therapeutic
Therapeutic effect, and in the present embodiment obtain proton beam dosage distribution and range method be using by training obtain
Parameter optimal neural net regression model and neural network classification model, therefore dosage distribution and the range of the proton beam obtained
Accuracy it is higher, so as to improve the accuracy of proton therapeutic, guarantee that the treatment that proton therapeutic is carried out to tumor patient is imitated
Fruit.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
In order to more comprehensively, correspond to proton therapeutic monitoring method neural network based provided in an embodiment of the present invention, this
Application additionally provides proton therapeutic monitoring system neural network based.
Fig. 5 is the structural schematic diagram of proton therapeutic neural network based monitoring system embodiment of the invention.Such as Fig. 5 institute
To show, the proton therapeutic neural network based of the present embodiment monitoring system includes CT machine 201 and proton therapeutic monitoring device 202,
Wherein, CT machine 201 is connected with proton therapeutic monitoring device 202.
CT machine 201 is sent to proton therapeutic monitoring device 202 for constructing the said three-dimensional body mould of human body, and by said three-dimensional body mould；
Proton therapeutic monitoring device 202 is used for the proton therapeutic monitoring method neural network based of above-described embodiment.
The proton therapeutic neural network based of the present embodiment monitors system, and the said three-dimensional body of human body is constructed by CT machine 201
Mould, and the said three-dimensional body mould is sent to proton therapeutic monitoring device 202, proton therapeutic monitoring device 202 according to the said three-dimensional body mould,
It determines tumor focus region, then by the distributed intelligence of measurement positron radionuclide, constructs PET image, squeezed into using simulation proton
The reaction process of scheduled human body target organ, constructs classification based training collection and regression training collection, and training obtains neural network classification mould
Type and neural net regression model；It is obtained based on the neural network classification model and neural net regression model according to PET image
It is distributed to the dosage of bragg peak position and proton beam, realizes the monitoring of the dosage distribution to bragg peak position and proton beam,
Whether meet preset proton therapeutic requirement according to the distribution of the dosage of bragg peak position and proton beam, beam is gone out to proton beam
Parameter is adjusted, and guarantees that the dosage that proton therapeutic carries out proton beam radiation to tumor tissues in the process is normal, position is accurate.?
Nuclear reaction can occur when proton therapeutic, generate various positron radionuclides, be deposited on the path and end of line, it is aobvious by PET
As technology measures the distribution of positron radionuclide, and PET image is constructed, can reduce human organ respiratory movement pair in this way
It influences, is improved to the range of proton and the precision of dosage distribution measuring caused by measurement, so that the accuracy of proton therapeutic is improved,
Guarantee the therapeutic effect that proton therapeutic is carried out to tumor patient.
It is understood that same or similar part can mutually refer in the various embodiments described above, in some embodiments
Unspecified content may refer to the same or similar content in other embodiments.
It should be noted that in the description of the present invention, term " first ", " second " etc. are used for description purposes only, without
It can be interpreted as indication or suggestion relative importance.In addition, in the description of the present invention, unless otherwise indicated, the meaning of " multiple "
Refer at least two.
Any process described otherwise above or method description are construed as in flow chart or herein, and expression includes
It is one or more for realizing specific logical function or process the step of executable instruction code module, segment or portion
Point, and the range of the preferred embodiment of the present invention includes other realization, wherein can not press shown or discussed suitable
Sequence, including according to related function by it is basic simultaneously in the way of or in the opposite order, Lai Zhihang function, this should be of the invention
Embodiment person of ordinary skill in the field understood.
It should be appreciated that each section of the invention can be realized with hardware, software, firmware or their combination.Above-mentioned
In embodiment, software that multiple steps or method can be executed in memory and by suitable instruction execution system with storage
Or firmware is realized.It, and in another embodiment, can be under well known in the art for example, if realized with hardware
Any one of column technology or their combination are realized: having a logic gates for realizing logic function to data-signal
Discrete logic, with suitable combinational logic gate circuit specific integrated circuit, programmable gate array (PGA), scene
Programmable gate array (FPGA) etc..
Those skilled in the art are understood that realize all or part of step that above-described embodiment method carries
It suddenly is that relevant hardware can be instructed to complete by program, described program can store in a kind of computer readable storage medium
In, which when being executed, includes the steps that one or a combination set of embodiment of the method.
It, can also be in addition, each functional unit in each embodiment of the present invention can integrate in a processing module
It is that each unit physically exists alone, can also be integrated in two or more units in a module.Above-mentioned integrated mould
Block both can take the form of hardware realization, can also be realized in the form of software function module.The integrated module is such as
Fruit is realized and when sold or used as an independent product in the form of software function module, also can store in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only memory, disk or CD etc..
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not
Centainly refer to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be any
One or more embodiment or examples in can be combined in any suitable manner.
Although the embodiments of the present invention has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is not considered as limiting the invention, those skilled in the art within the scope of the invention can be to above-mentioned
Embodiment is changed, modifies, replacement and variant.
Claims (10)
1. a kind of proton therapeutic monitoring method neural network based characterized by comprising
According to the said three-dimensional body mould constructed in advance, tumor focus region is determined；
After proton beam radiation is to the tumor focus region, positron radionuclide is measured within a preset time, obtains the positive electricity
The distributed intelligence of daughter nuclide；
Positron e mission computed tomography PET image is constructed using image reconstruction algorithm according to the distributed intelligence；
The PET image is input to the neural network classification model of building, determines the range of the proton beam；
The PET image is input to the neural net regression model of building, determines the dosage distribution of the proton beam；
It is distributed according to the range and the dosage, determines bragg peak position；
It detects the bragg peak position and whether dosage distribution meets preset proton therapeutic requirement；
If the bragg peak position and/or dosage distribution do not meet the proton therapeutic requirement, the proton beam is adjusted
Go out beam parameter.
2. the method according to claim 1, wherein the detection bragg peak position and the dosage point
Whether cloth meets preset proton therapeutic requirement, comprising:
Detect whether the bragg peak position falls in the tumor focus region, and, whether detect the dosage distribution
Meet the radiotherapy planning content in the proton therapeutic requirement；
If the bragg peak position is not fallen in the tumor focus region, detect that the bragg peak position does not meet institute
State proton therapeutic requirement；
If the dosage distribution does not meet the radiotherapy planning content in the proton therapeutic requirement, the dosage distribution is detected not
Meet the proton therapeutic requirement.
3. the method according to claim 1, wherein the nerve net that the PET image is input to building
Network disaggregated model, before the range for determining the proton beam, further includes:
The reaction process of scheduled human body target organ is squeezed by simulating proton, constructs classification based training collection and regression training collection；
Based on neural network algorithm, first object neural network model is trained using the classification based training collection, obtains institute
State neural network classification model；
Based on the neural network algorithm, the second target nerve network model is trained using the regression training collection, is obtained
To the neural net regression model.
4. according to the method described in claim 3, it is characterized in that, described squeeze into scheduled human body target organ by simulating proton
Reaction process, construct classification based training collection and regression training collection, comprising:
The reaction process of the scheduled human body target organ is squeezed by simulating the proton, obtains the distribution of simulation dosage and simulation
Range；
PET imaging simulation is carried out, simulation PET image is constructed using described image algorithm for reconstructing according to the reaction process；
The simulation PET image and the simulation range are built into the classification based training collection；
The simulation PET image and simulation dosage distribution are built into the regression training collection.
5. the method according to claim 1, wherein it is described go out beam parameter include: out the beam moment, go out Shu Sudu,
Beam trajectory and out beam dose out.
6. a kind of proton therapeutic monitoring device neural network based characterized by comprising
First determining module, for determining tumor focus region according to the said three-dimensional body mould constructed in advance；
Measurement module, for measuring positron radionuclide within a preset time after proton beam radiation is to the tumor focus region,
Obtain the distributed intelligence of the positron radionuclide；
First building module, for constructing positron emission computer using image reconstruction algorithm according to the distributed intelligence
Tomography PET image；
Second determining module determines the proton beam for the PET image to be input to the neural network classification model of building
Range；The PET image is input to the neural net regression model of building, determines the dosage distribution of the proton beam；
Third determining module determines bragg peak position for being distributed according to the range and the dosage；
Detection module, for detecting whether the bragg peak position and dosage distribution meet preset proton therapeutic and want
It asks；
Module is adjusted, if not meeting the proton therapeutic requirement for the bragg peak position and/or dosage distribution, is adjusted
The whole proton beam goes out beam parameter.
7. device according to claim 6, which is characterized in that the detection module is specifically used for:
Detect whether the bragg peak position falls in the tumor focus region, and, whether detect the dosage distribution
Meet the radiotherapy planning content in the proton therapeutic requirement；
If the bragg peak position is not fallen in the tumor focus region, detect that the bragg peak position does not meet institute
State proton therapeutic requirement；
If the dosage distribution does not meet the radiotherapy planning content in the proton therapeutic requirement, the dosage distribution is detected not
Meet the proton therapeutic requirement.
8. device according to claim 6, which is characterized in that further include:
Second building module constructs classification based training for squeezing into the reaction process of scheduled human body target organ by simulating proton
Collection and regression training collection；
Training module, for be based on neural network algorithm, using the classification based training collection to first object neural network model into
Row training, obtains the neural network classification model；Based on the neural network algorithm, using the regression training collection to second
Target nerve network model is trained, and obtains the neural net regression model.
9. device according to claim 8, which is characterized in that the second building module includes: acquiring unit, simulation drawing
As construction unit and training set construction unit；
The acquiring unit is obtained for squeezing into the reaction process of the scheduled human body target organ by simulating the proton
Simulate dosage distribution and simulation range；
The analog image construction unit, for according to the reaction process, using described image algorithm for reconstructing, carry out PET at
As emulation, simulation PET image is constructed；
The training set construction unit, for the simulation PET image and the simulation range to be built into the classification based training
Collection；The simulation PET image and simulation dosage distribution are built into the regression training collection.
10. a kind of proton therapeutic neural network based monitors system characterized by comprising CT machine and proton therapeutic monitoring
Equipment；
The CT machine is connected with the proton therapeutic monitoring device；
The CT machine is used to construct the said three-dimensional body mould of human body, and the said three-dimensional body mould is sent to the proton therapeutic monitoring and is set
It is standby；
The proton therapeutic monitoring device is at least used for perform claim and any proton neural network based of 1-5 is required to control
Treat monitoring method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910445746.6A CN110270016A (en) | 2019-05-27 | 2019-05-27 | Proton therapeutic monitoring method, device and system neural network based |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910445746.6A CN110270016A (en) | 2019-05-27 | 2019-05-27 | Proton therapeutic monitoring method, device and system neural network based |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110270016A true CN110270016A (en) | 2019-09-24 |
Family
ID=67960270
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910445746.6A Pending CN110270016A (en) | 2019-05-27 | 2019-05-27 | Proton therapeutic monitoring method, device and system neural network based |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110270016A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111569279A (en) * | 2020-05-26 | 2020-08-25 | 杭州珞珈质子科技有限公司 | Parameter monitoring device and system for proton treatment |
| CN111991711A (en) * | 2020-09-07 | 2020-11-27 | 杭州珞珈质子科技有限公司 | Parameter monitoring device and system for proton treatment |
| CN115957453A (en) * | 2022-12-16 | 2023-04-14 | 中国科学院近代物理研究所 | Three-dimensional in-vivo dose reconstruction method and device for heavy ion radiotherapy |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030150983A1 (en) * | 2002-02-14 | 2003-08-14 | Mitsubishi Denki Kabushiki Kaisha | Apparatus for measuring absorption dose distribution |
| CN108535696A (en) * | 2018-04-18 | 2018-09-14 | 彭浩 | Proton bragg peak tuning on-line method and monitoring device |
| CN108717866A (en) * | 2018-04-03 | 2018-10-30 | 陈辛元 | A kind of method, apparatus, equipment and the storage medium of the distribution of prediction radiotherapy planning dosage |
-
2019
- 2019-05-27 CN CN201910445746.6A patent/CN110270016A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030150983A1 (en) * | 2002-02-14 | 2003-08-14 | Mitsubishi Denki Kabushiki Kaisha | Apparatus for measuring absorption dose distribution |
| CN108717866A (en) * | 2018-04-03 | 2018-10-30 | 陈辛元 | A kind of method, apparatus, equipment and the storage medium of the distribution of prediction radiotherapy planning dosage |
| CN108535696A (en) * | 2018-04-18 | 2018-09-14 | 彭浩 | Proton bragg peak tuning on-line method and monitoring device |
Non-Patent Citations (2)
| Title |
|---|
| 戴甜甜等: "质子和重离子治疗在线监测成像技术的发展", 《中国医学影像技术》 * |
| 李中星: "12C离子质子治疗中In-beam PET成像过程研究", 《中国优秀硕士学位论文全文数据库•医药卫生科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111569279A (en) * | 2020-05-26 | 2020-08-25 | 杭州珞珈质子科技有限公司 | Parameter monitoring device and system for proton treatment |
| CN111569279B (en) * | 2020-05-26 | 2021-08-17 | 杭州珞珈质子科技有限公司 | Parameter monitoring device and system for proton treatment |
| WO2021237869A1 (en) * | 2020-05-26 | 2021-12-02 | 杭州珞珈质子科技有限公司 | Parameter monitoring device and system for proton therapy |
| CN111991711A (en) * | 2020-09-07 | 2020-11-27 | 杭州珞珈质子科技有限公司 | Parameter monitoring device and system for proton treatment |
| CN115957453A (en) * | 2022-12-16 | 2023-04-14 | 中国科学院近代物理研究所 | Three-dimensional in-vivo dose reconstruction method and device for heavy ion radiotherapy |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11633626B2 (en) | Methods for real-time image guided radiation therapy | |
| CN110678226B (en) | Radiation therapy method, system and related computer program | |
| CN109069858B (en) | Radiotherapy system and computer readable storage device | |
| US9789339B2 (en) | System design and method for verifying 3D dosimetric imaging of charged particles in media | |
| US11020615B2 (en) | Computing radiotherapy dose distribution | |
| US10493299B2 (en) | Determining parameters for a beam model of a radiation machine using deep convolutional neural networks | |
| CN107072624B (en) | System and method for automated treatment planning | |
| Jones et al. | Treatment plan comparison using equivalent uniform biologically effective dose (EUBED) | |
| US20210031057A1 (en) | Method for reconstructing x-ray cone-beam CT images | |
| AU2017211683A1 (en) | Therapy control using motion prediction | |
| CN110270016A (en) | Proton therapeutic monitoring method, device and system neural network based | |
| CN109621228A (en) | The calculating unit and storage medium of radiological dose | |
| CN104107062A (en) | Radiotherapeutic effect evaluation method and system | |
| US11141609B2 (en) | Dose guided real-time adaptive radiotherapy | |
| WO2014122567A1 (en) | Apparatus for determining a number of beams in imrt | |
| CN109876310A (en) | Monitoring method, device and the system of proton therapeutic | |
| CN117957574A (en) | Iterative image reconstruction | |
| Fiagan et al. | In vivo dosimetry for patients with prostate cancer to assess possible impact of bladder and rectum preparation | |
| US20220362582A1 (en) | Adaptive dose accumulation algorithm | |
| CN113631223B (en) | Determining the risk of hematologic toxicity following radiation therapy | |
| US20180235569A1 (en) | Body tissue location measurement system | |
| US20230218926A1 (en) | Bed calculation with isotoxic planning | |
| Mahuvava et al. | Monte Carlo evaluation of the dose perturbation effect of hip prostheses for megavoltage photon radiotherapy | |
| Pham | Simulation of the transmitted dose in an EPID using a Monte Carlo method | |
| Jhala et al. | EPSM 2017, Engineering and Physical Sciences in Medicine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190924 |
|
| RJ01 | Rejection of invention patent application after publication |