CN109567838A - A kind of X ray absorption spectrometry lesion detector - Google Patents
A kind of X ray absorption spectrometry lesion detector Download PDFInfo
- Publication number
- CN109567838A CN109567838A CN201811267769.4A CN201811267769A CN109567838A CN 109567838 A CN109567838 A CN 109567838A CN 201811267769 A CN201811267769 A CN 201811267769A CN 109567838 A CN109567838 A CN 109567838A
- Authority
- CN
- China
- Prior art keywords
- ray
- detector
- translation stage
- absorption spectrometry
- ray source
- 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
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
- A61B6/48—Diagnostic techniques
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
- A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
Abstract
A kind of X ray absorption spectrometry lesion detector, including optical platform, X-ray source, X-ray detector, servo-system, control system and data processing system；The X-ray source is located at translation stage one end；The servo-system includes tri-axial motion controller and translation stage, lifting platform and turntable on optical platform, the translation stage is close to the optical platform other end and its top surface is equipped with lifting platform, the turntable is between translation stage and X-ray source, tri-axial motion controller connection driving translation stage, lifting platform and turntable movement；The control system is connected with X-ray source, X-ray detector and tri-axial motion controller；The data processing system is connected to pre-process to collected X-ray spectrum with control system, and the feature extraction of known sample is carried out using machine learning algorithm, is then assessed using assessment models classification prediction, provides testing result.The hardware cost of present invention saving cirrhosis detector, detection speed are fast, and recognition accuracy is high.
Description
Technical field
The present invention relates to X-ray absorption spectral technique, especially a kind of X ray absorption spectrometry lesion detector.
Background technique
Computed tomography (CT) has high-resolution to human body soft tissue, and imaging is more clear and stablizes, more can be complete
Face objectively shows the shape of liver, the size of spleen and whether there is or not ascites, can overcome the problems, such as the artifact that respiratory movement generates, can
For early diagnosing, help to distinguish low level fibrosis.But simultaneously because CT technology needs repeatedly irradiation, therefore dose of radiation
It is biggish.And the image that CT scan goes out cannot be directly used to diagnose, and need more complicated image segmentation extraction process.This
Outside, it be widely used it can not using contrast agent, limited scanning coverage area.
Nuclear magnetic resonance check (MRI) is to utilize the imaging skill for collecting signal caused by electromagnetic induction phenomenon and reconstruction image
Art, can be used for measuring hepatocyte function, and enhancing degree and heterogeneity of the hepatic tissue in the liver cell stage have with degree of hepatic fibrosis
It closes, but is limited to using contrast agent and nuclear magnetic resonance time, efficiency and effect are to be improved.
Ultrasonic detection technology is the diagnosis common method of disease in the liver and gallbladder, has economical and practical, atraumatic, is repeated
Property, highly-safe, patient is easy to the advantages that receiving, can be widely used for screening fibrosis in clinical practice, is suitable for fatty scorching
Or the patient of fat class, it is ideal to the patient's effect for having narrow intercostal space and ascites symptom.But diagnostic result is checked
Person's qualification subjective factor and ultrasonic instrument are affected, the inspection to chronic, diffuse liver disease and early-phase hepatocirrhosis
It is bad to survey effect.
X ray absorption spectrometry is based on lambert-Bell (lambert-beer) law.When X-ray penetrating material, it
Intensity can decay, and the degree of decaying depends on the substance being pierced, and penetrate X-ray intensity and the thickness for being pierced substance
The attenuation coefficient exponent function relation of degree and substance.It penetrates X-ray intensity I and incident intensity I0 meets following formula attenuation relation: I-
I0e(-μd), it is a constant, d is thickness of sample that μ, which is the x-ray attenuation coefficient of the substance,.It is penetrated in the X-ray of Continuous Energy
During sample, sample is different to the degree of absorption of different-energy X-ray, and it is exactly object that X-ray absorption, which is composed represented by (XAS),
It verifies in the variation of the μ of different-energy X-ray.Collected initial data is incident X-rays spectrum and transmission X-ray in experiment
Spectrum, therefore the XAS of available sample:X-ray spectrum technology is a kind of analysis method of maturation, has energy
The features such as amount is high, penetration power is strong and non-destructive testing.The technology suffers from important in fields such as medicine, chemistry, materialogy, physics
Using.Therefore X-ray absorption spectral technique is applied to cirrhosis pathological examination can yet be regarded as a kind of complementary medicine diagnosing and treating
New method.
Summary of the invention
It is a primary object of the present invention to overcome drawbacks described above in the prior art, a kind of disease applied to cirrhosis is proposed
Reason detection, the X ray absorption spectrometry lesion for combining machine learning algorithm progress Classification and Identification normal hepatocytes and hardening liver detect
Instrument.
The present invention adopts the following technical scheme:
A kind of X ray absorption spectrometry lesion detector, it is characterised in that: including optical platform, X-ray source, X-ray
Detector, servo-system, control system and data processing system；The X-ray source is located at translation stage one end；The servo-system packet
Tri-axial motion controller and translation stage, lifting platform and turntable on optical platform are included, the translation stage is close to optical platform
The other end and its top surface are equipped with lifting platform to place X-ray detector, the turntable between translation stage and X-ray source with
Test substance, tri-axial motion controller connection driving translation stage, lifting platform and turntable movement are placed, and makes X-ray detection
The Be window center of device and the Be window center of X-ray source are on same vertical guide and horizontal plane；The control system and X-ray light
Source, X-ray detector and tri-axial motion controller are connected to control and generate X-ray, acquire X-ray spectrum；The data processing system
System is connected to pre-process to collected X-ray spectrum with control system, carries out known sample using machine learning algorithm
Feature extraction, then using assessment models to classification prediction assess, provide testing result.
Further include radioprotector, is equipped with cabinet body, lead glass door and lead door；The platform, X-ray source, X are penetrated
Line detector, servo-system are located in cabinet body, which is located at cabinet body front end, which is located at cabinet body rear end and left and right
End.
It further include having high voltage power supply, which is connected to power with the X-ray source.
It further include intelligent temperature controller, which is connected with the X-ray source and the high voltage power supply.
It further include optical path adjustment system comprising two groups of infrared laser mould groups and gimbal, two groups of infrared laser mould groups
It is mounted on gimbal at the Be window center of any angle shoot laser to the X-ray source.
It further include video camera and display screen, the video camera and display screen are connected with the control system.
The pretreatment includes shearing, removes background, normalization and PCA dimensionality reduction.
The machine learning algorithm is deep neural network.
The assessment models are confusion matrix, Receiver operating curve or accuracy rate-recall rate curve.
By the above-mentioned description of this invention it is found that compared with prior art, the invention has the following beneficial effects:
1, detector of the present invention can produce the broadband X-ray of tungsten (74W) as incident light spectrum, by photon counting detector
X-ray absorption spectral technique is applied to pathologic finding for the first time by the absorption spectrum of collecting sample, acquires normal hepatocytes and hardening liver
The X-ray transmission spectrum of sample.
2, the X-ray absorption curve of spectrum of sample can be obtained by lambert-Bell (lambert-beer) law by the present invention.
After the methods of Principal Component Analysis, 0-1 normalization pre-processed spectrum data, classify by deep neural network model
Identification, to realize Normal mouse liver and harden the identification of liver sample.
3, the present invention saves the hardware cost of cirrhosis detector.
4, detection speed of the invention is fast, and recognition accuracy is high.
5, testing radiation dosage of the present invention is small, can be used for a variety of different pathological examinations.
Detailed description of the invention
Fig. 1 is overall structure block diagram of the present invention；
Fig. 2 is invention software structural framing figure；
Fig. 3 is invention software flow chart
Fig. 4 is A block process in Fig. 3；
Fig. 5 is B block process in Fig. 3；
Fig. 6 is C block process in Fig. 3；
Fig. 7 is D-module process in Fig. 3；
Fig. 8 detector primary structure figure of the present invention；
Fig. 9 is the original spectral data of Software on Drawing；
Figure 10 is the displaying of the two respective mean absorption coefficients of class sample；
Figure 11 is the first two principal component X-Y scheme in principal component analysis；
Figure 12 is the confusion matrix for assessment models classification prediction effect (before normalization and after normalization)；
Wherein: 10, optical platform, 20, X-ray source, 30, X-ray detector, 40, servo-system, 41, translation stage,
42, lifting platform, 43, turntable, 50, radioprotector, 60, test substance, 70, display screen.
Specific embodiment
Below by way of specific embodiment, the invention will be further described.
Referring to Fig. 1 to Fig. 8, a kind of X ray absorption spectrometry lesion detector, by the broadband X-ray for generating tungsten (74W)
As incident light spectrum, by the absorption spectrum of photon counting detector collecting sample, for the first time by X-ray absorption spectral technique application
In pathologic finding, acquires normal hepatocytes and harden the X-ray transmission spectrum of liver sample.Pass through lambert-Bell (lambert-
Beer) the X-ray absorption curve of spectrum of sample can be obtained in law.Utilize the pretreatment of the methods of Principal Component Analysis, 0-1 normalization
After spectroscopic data, classification identification is carried out by deep neural network model, to realize Normal mouse liver and harden liver sample
Identification.
It is mainly formed including optical platform 10, X-ray source 20, X-ray detector 30, servo-system 40, control system
System, data processing system etc..The X-ray source 20 is located at 41 one end of translation stage, can be used what Shanghai section chin or cheek dimension electronics provided
KYW800 type small-power fixed anode generating positive and negative voltage X-ray tube.The offer of AmpTek company, the U.S. can be used in X-ray detector 30
X-123 type CdTe detector, the type detector are integrated with X-100T-CdTe crystal probe and its preamplifier, DP5 number arteries and veins
Rush processor, multichannel analyzer and power supply and PC interface.The input voltage of system power supply is about+5V DC, and electric current is
200mA.Required DC low-voltage is provided for digital processing unit and preamplifier using Switching Power Supply in PC5, while also being wrapped
A Cockcroft-Walton accelerator is contained, to generate the bias of 400V high required for detector, and closed-loop control thermoelectric cooler is to provide
Up to 85 DEG C of the temperature difference.Entire detector is encapsulated in 7cm × 10cm × 2.5cm aluminium box, it is only necessary to two connecting lines: electricity
Source line (+5V DC) and data line (USB or RS232).
Due to the radiation characteristic of X-ray, it is necessary to assure system cannot cause health to damage operator when running,
Therefore it must be provided with corresponding radioprotector 50 and completely cut off X-ray.The radioprotector 50 be equipped with cabinet body, glass lead door and
Lead door etc., optical platform 10, X-ray source 20, X-ray detector 30, servo-system 40 are located in cabinet body, the lead glass door position
In cabinet body front end, which is located at cabinet body rear end, front and back side.The bottom of cabinet bulk be equipped with bracket, for place host, power supply etc. its
His component.
It is additionally provided with intelligent temperature controller and high voltage power supply, high voltage power supply provides stable DC high voltage and steady for X-ray tube
Determine DC current.The intelligent temperature controller is connected with X-ray source 20, high voltage power supply.(thermocouple passes the magnetic head of intelligent temperature controller
Sensor) it is affixed on X-ray tube brass package surface, for detecting X-ray tube operating temperature；If X-ray tube due to power is excessive,
Heat dissipation keeps temperature excessively high not in time, and intelligent temperature controller shuts off high voltage power supply.
The XRB series 150W high voltage power supply that this graceful high voltage power supply provides can be used in the high voltage power supply, can provide High voltage output,
Complete control signal and a variety of miscellaneous functions, exportable stable small ripple voltage and electric current.In the actual operation process, by
The double lead doors opened of the lead glass door that front end is singly opened and rear end are provided in radioprotector 50, lead is opened in maloperation in order to prevent
Door and make radiation leakage, be provided with high voltage power supply interlock system: use two groups of Schneider XCKN2145P20C type travel switches, peace
At the lead glass door of radioprotector 50, the interlock circuit of high voltage power supply is connected, it can be in the case where opening by mistake lead door
Cutoff high power supply immediately.
The servo-system 40 is used to adjust the spatial position of X-ray detector 30 and test substance 60, including three-axis moving
Controller and translation stage 41, lifting platform 42 and turntable 43 on optical platform 10, the translation stage 41 is close to optical platform
10 other ends and its top surface are equipped with lifting platform 42 to place X-ray detector 30, which is located at translation stage 41 and X-ray
To place test substance 60 between light source 20, tri-axial motion controller connection driving translation stage 41, lifting platform 42 and turntable
43 movements, and the Be window center of X-ray detector 30 and the Be window center of X-ray source 20 is made to be in same vertical guide and level
On face.
The 7S three axis servo motion platforms of series that servo-system 40 is provided using all photoelectricity is matched, including 7STA04A herd horses and are
Serial three axis of column motorized precision translation stage, 7SVA160 jack series lifting platform, 7SRA1 worm gear series electric rotary table and 7SC3
Motion controller.
It further include optical path adjustment system, which uses two groups of infrared laser mould groups, it is mounted on gimbal,
It, can be with any angle shoot laser positioned at 42 side suitable position of lifting platform.Optical path adjusting process is as follows: adjustment is two groups infrared to swash
The focal length and gimbal of optical mode group keep two infrared laser lines difference of outgoing vertical and horizontal, and hand over their cross
The intersection point of cross wires projects at the Be window center of X-ray tube；Then the space of X-ray detector 30 is adjusted by servo-system 40
The Be window center of position, the Be window center and X-ray tube that make detector is in same vertical plane and same level plane.This
The optical path adjustment of sample can guarantee that the photon of X-ray outgoing is collimatedly received by detector Be window as far as possible.
It further include video camera and display screen 70, the video camera and display screen 70 are connected with control system.The control system and X
Ray source 20, X-ray detector 30 are connected with tri-axial motion controller.Control system assigns the control command of light source, detection
The control command of the configuration information of device, data processing system and servo-system 40, makes the X-ray spectrum after sample is decayed
Constantly by detector acquire and pass through data processing system processing returns to display and save to control terminal.
Control system realizes the order transmitting and information exchange with each system: condition needed for control light source parameters generate
X-ray configures parameter detector and acquires X-ray spectrum, the collected X-ray information of control data processing system processing, drive
Dynamic servo-system 40 positions the spatial position of detector and sample；Meanwhile light source parameters information, detector match confidence
Breath, X-ray spectrum information, 40 location information of servo-system are real-time transmitted to control system and show, and can be to these letters
Breath is saved and is exported
Entire hardware system is needed by USB and serially logical under 7 system environments of Windows using PC as controlling terminal
Believe that interface is realized to the control of subsystems and the immediate feedback of subsystems information.
Control system of the present invention mainly realizes the function in terms of following four: (1) acquisition, display of X-ray absorption spectrum
And it saves；(2) control of high voltage power supply；(3) operation of video camera real-time monitoring system；The essence of (4) three axis servo motion platforms
Really control.Using Visual Studio 2010C# programmed environment, the PC control software based on Windows7 is worked out, has been made
Use universal serial bus (USB) and serial communication interface (com interface) as the communication interface between host computer and hardware system.
With reference to Fig. 2, X-ray absorption spectral measurement specifically include that communication between each hardware subsystem and control system and on
The exploitation of position machine interactive interface.User by host computer to high voltage power supply and tri-axial motion controller send serial port control signal into
Row configuration change and experimental implementation send USB control signal to X-ray acquisition device and video camera and carry out configuration change and experiment
Operation, each hardware subsystem by serial ports and USB to the instant return information of user,
With reference to Fig. 3, software development flow, which is divided into, to be logged in, communicates, controlling three parts.Log-in interface is arranged in host computer, into
Row purview certification；Enter communication interface after authenticating successfully, passes through serial ports and USB and high voltage power supply, detector, video camera, three axis
Motion controller establishes communication connection；Enter main control interface after communicating successfully, it at this time can be to the parameters of subsystems
It is configured.
Referring to fig. 4, sub-process is controlled for the high voltage power supply of A module: aobvious in high voltage power supply sub-interface after high pressure is opened in confirmation
Show current associated data, then relevant parameter is set, then run high pressure, power for X-ray tube, and constantly inspection in the process of running
It surveys and whether mistake occurs, if there is just detecting whether in sub-interface notification error state and constantly correctly to handle, finally terminate
High-voltage operation.
Referring to Fig. 5, sub-process is controlled for the X-ray detector 30 of B module: after communicating successfully, being first confirmed whether to modify
Parameter detector then carries out MCA multi-channel data analysis parameter, peaking parameter (Shaping), acquisition gain ginseng if you need to modify
The configuration of number (Gain), power parameter (Power), Misc parameter, are then turned on acquisition mode, in acquisition X-ray absorption spectrum
During simultaneously show operating status and parameter configuration state, finally terminate spectra collection.
Be the camera control sub-process of C module referring to Fig. 6: after opening video camera, confirmation carries out picture monitoring, starting
Function is acquired, single frames or continuous acquisition mode are selected, then modification current camera parameter is optimal to picture, finally terminates to supervise
Control.
It is that the tri-axial motion controller of D-module controls sub-process: after main interface communicates successfully, to X, Y, Z axis referring to Fig. 7
It is selected, modifies the kinematic parameter of each axis, then control the movement of each axis, return zero-bit wait choose whether after the completion of acting,
Finally terminate to move.
Data processing system is based on the third party libraries such as Pandas, Numpy, Scikit-learn and Matplotlib and writes light
Modal data analysis and processing program, and software operation interface is write with TKinter, exe file is packaged into pyinstaller.It should
Software can do different disposal to for different types of experiment sample.
Its data processing mainly includes being sheared to collected original spectral data, going background, normalization, PCA drop
The pretreatment operations such as dimension, referring to Fig. 9-Figure 12；Selection uses deep neural network (deep neural network, DNN) machine
Learning algorithm carries out the feature extraction and classification prediction of known sample；With confusion matrix, Receiver operating curve
(receiver operating characteristic curve, ROC), accuracy rate-recall rate curve (Precision-
Recall curve, PRC) etc. assessment modes assessment models classify prediction effect；To given collected original spectral data,
Prediction label information can be provided, which represents the corresponding classification of test substance 60.
The present invention mainly uses deep neural network algorithm as Classification and Identification model, to be based on back propagation learning algorithm
Single hidden layer artificial neural network for, illustrate the working principle of neural network.Artificial neuron (Artificial Neural,
Abbreviation neuron) be ANN essential information processing unit, the same with biological neuron, it can receive one or several inputs
Then signal carries out processing to input signal and generates output, and output is transmitted to next neuron or output most to terminate
Fruit.One complete neuron has following basic element:
Weight, each input that neuron receives require just be further processed after being weighted.
Adder constitutes a linear combiner, carries out summation operation to the input after weighting.
Threshold value, the effect of threshold value are to increase or decrease corresponding numerical value to the result of adder, by the input of activation primitive
Limitation is within a reasonable range.
Activation primitive generates output for calculate to result of the adder after threshold value, also functions to limitation nerve
The effect of first output area, it is crucial that activation primitive is the non-linear source of network.In general, the output area quilt of neuron
It is limited between 0 to 1 or -1 to 1.
The data handling procedure of neuron can be indicated with following formula:
Y=f (z) (2)
In formula, x is input signal, and w is weight, and b is threshold value, and z is the input of activation primitive, and f is activation primitive, and y is defeated
Signal out.Formula (1) and formula (2) are merged, the signal processing formula of neuron is obtained:
The task of each neuron is exactly to be then output to next neuron or output after handling input as most
Terminate fruit.
Activation primitive is one of basic element of neuron, limit neuron output while return network joined it is non-
Linear factor, so as to solve the nonlinear problem that linear model cann't be solved.Hidden layer and output layer be not usually using
Same activation primitive.For the deep-neural-network under two classification problems, the activation primitive that hidden layer and output layer use divides
It is not line rectification function (Rectified Linear Unit, ReLU) and Sigmoid function.ReLU function is in neural network
The number of plies can effectively avoid gradient extinction tests when deepening, and formula is as follows.
F (x)=max (x, 0) (4)
Sigmoid function most common output layer activation primitive, formula when being Processing with Neural Network binary classification problems
It is as follows:
The derivative of sigmoid function can be indicated by itself are as follows:
The learning rules of neural network are training algorithm, by study priori knowledge to the weight and threshold value of neural network
Training updates.Learning rules have two major classes other: supervised learning and unsupervised learning.For supervised learning, it is desirable to provide first
Knowledge (i.e. training sample to) is tested for training network, weight and threshold value are constantly updated according to the output of network and desired value, made
The output of network increasingly levels off to desired value.And for unsupervised learning, the no desired value of the output of network, weight and threshold
The update of value is only related with network inputs.Error backpropagation algorithm used in BP neural network training is that one kind typically has
Supervised learning, basic thought are to achieve the purpose that Accurate Prediction by study priori knowledge (input/desired output to).That is sample
This input is sent to network input layer, and after being calculated by hidden layer and output layer, output layer exports corresponding predicted value.Such as
Error between fruit predicted value and desired output is unsatisfactory for required precision, then reversely adjusts power by the error condition between them
Weight and threshold value again carry out input after adjustment to calculate generation output.It is gradually reduced between output and desired output repeatedly
Error, until error meets required precision.
The detailed process using error backpropagation algorithm adjustment network weight and threshold value is derived below.
By taking network structure is three layers of BP neural network of m × p × n as an example.Sample exports O to (X, Y) and hidden neuron
Shown in following expression:
Weight connection matrix W between input layer and hidden layer(1), weight connection matrix W between hidden layer and output layer(2)It is as follows:
In formula, wji represent from i-th of neuron of preceding layer to j-th of neuron of later layer when weight, i.e. the jth row of W
Indicate the weight vector that j-th of neuron of later layer is output to from all neurons of preceding layer.
The threshold vector b of hidden layer(1)With the threshold vector b of output layer(2)It is respectively as follows:
The output expression formula of j-th of neuron of hidden layer can be obtained by above-mentioned formula are as follows:
The then output O of hidden layer are as follows:
It can similarly obtain, the output of k-th of neuron of output layer and the output expression formula of output layer are respectively as follows:
In formula, function f and function g are the activation primitive of hidden layer and output layer respectively.
By input X to outputProcess be called positive process, during which the weight of network and threshold value are fixed, input letter
It number propagates in layer in a network, until reaching output end.It is reverse procedure after positive process, by comparing outputNetwork error is obtained with desired output Y, then utilizes the weight and threshold value of the reversed layer-by-layer correction network of network error.
The error of network output and desired output is defined as:
Error E is to weightPartial derivative are as follows:
In formula,
Error E is to weightPartial derivative are as follows:
In formula,
Similarly, error E is to threshold valueAnd threshold valuePartial derivative be respectively
Assuming that Δ W(1)With Δ W(2)It is as follows:
The then adjustment formula of weight are as follows:
In formula, η is learning rate, and learning rate determines the gradient decline i.e. speed of network training.
Similarly, it is assumed that Δ b(1)With Δ b(2)It is as follows:
The then adjustment formula of threshold value are as follows:
The essence of error backpropagation algorithm be the error of network is attributed to weight and threshold value setting it is unreasonable, pass through
Backpropagation calculates the relationship between error and each weight and threshold value, weighs then along making error decline most fast direction adjustment
Value and threshold value, to complete the training process of neural network.
The number of plies of hidden layer and the number of neuron have a great impact to the performance of network, if hidden layer structure is excessively simple
It is single, then non-linear relation that cannot sufficiently between simulation input and output；If structure is excessively complicated, it is existing to will appear over-fitting
As it is elongated to also result in the e-learning time.The size of hidden layer structure can be carried out unlike weight threshold by study
How adjustment determines the theoretical method of none fixation of hidden layer structure.The determination of hidden layer structure is usually to establish passing through
On the basis of testing, by repeatedly attempting influence of the comparison different structure to network performance, to be selected.
The above is only a specific embodiment of the present invention, but the design concept of the present invention is not limited to this, all to utilize this
Design makes a non-material change to the present invention, and should all belong to behavior that violates the scope of protection of the present invention.
Claims (9)
1. a kind of X ray absorption spectrometry lesion detector, it is characterised in that: visited including optical platform, X-ray source, X-ray
Survey device, servo-system, control system and data processing system；The X-ray source is located at translation stage one end；The servo-system includes
Tri-axial motion controller and translation stage, lifting platform and turntable on optical platform, the translation stage are another close to optical platform
One end and its top surface are equipped with lifting platform to place X-ray detector, and the turntable is between translation stage and X-ray source to put
Test substance is set, tri-axial motion controller connection driving translation stage, lifting platform and turntable movement, and make X-ray detector
Be window center and the Be window center of X-ray source be on same vertical guide and horizontal plane；The control system and X-ray light
Source, X-ray detector and tri-axial motion controller are connected to control and generate X-ray, acquire X-ray spectrum；The data processing system
System is connected to pre-process to collected X-ray spectrum with control system, carries out known sample using machine learning algorithm
Feature extraction, then using assessment models to classification prediction assess, provide testing result.
2. a kind of X ray absorption spectrometry lesion detector as described in claim 1, it is characterised in that: further include that radiation is anti-
Protection unit is equipped with cabinet body, lead glass door and lead door；The platform, X-ray source, X-ray detector, servo-system position
In in cabinet body, which is located at cabinet body front end, which is located at cabinet body rear end and left and right end.
3. a kind of X ray absorption spectrometry lesion detector as described in claim 1, it is characterised in that: further include having high pressure
Power supply, the high voltage power supply are connected to power with the X-ray source.
4. a kind of X ray absorption spectrometry lesion detector as claimed in claim 3, it is characterised in that: further include intelligent temperature
Device is controlled, which is connected with the X-ray source and the high voltage power supply.
5. a kind of X ray absorption spectrometry lesion detector as described in claim 1, it is characterised in that: further include optical path tune
Calibration system comprising two groups of infrared laser mould groups and gimbal, two groups of infrared laser mould groups are mounted on gimbal with any
At angle shoot laser to the Be window center of the X-ray source.
6. a kind of X ray absorption spectrometry lesion detector as described in claim 1, it is characterised in that: further include video camera
And display screen, the video camera and display screen are connected with the control system.
7. a kind of X ray absorption spectrometry lesion detector as described in claim 1, it is characterised in that: the pretreatment packet
It includes shearing, remove background, normalization and PCA dimensionality reduction.
8. a kind of X ray absorption spectrometry lesion detector as described in claim 1, it is characterised in that: the machine learning
Algorithm is deep neural network.
9. a kind of X ray absorption spectrometry lesion detector as described in claim 1, it is characterised in that: the assessment models
For confusion matrix, Receiver operating curve or accuracy rate-recall rate curve.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811267769.4A CN109567838A (en) | 2018-10-29 | 2018-10-29 | A kind of X ray absorption spectrometry lesion detector |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811267769.4A CN109567838A (en) | 2018-10-29 | 2018-10-29 | A kind of X ray absorption spectrometry lesion detector |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109567838A true CN109567838A (en) | 2019-04-05 |
Family
ID=65921033
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811267769.4A Pending CN109567838A (en) | 2018-10-29 | 2018-10-29 | A kind of X ray absorption spectrometry lesion detector |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109567838A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110648311A (en) * | 2019-09-03 | 2020-01-03 | 南开大学 | Acne image focus segmentation and counting network model based on multitask learning |
| CN113295722A (en) * | 2021-05-21 | 2021-08-24 | 厦门大学 | X-ray spectral data correction method and device based on deep learning algorithm |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN205758588U (en) * | 2016-05-25 | 2016-12-07 | 厦门大学 | A kind of X-ray digital imaging system |
| CN106821410A (en) * | 2017-03-20 | 2017-06-13 | 厦门大学 | Using the substance identification and system of EO-1 hyperion CT functional imagings |
-
2018
- 2018-10-29 CN CN201811267769.4A patent/CN109567838A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN205758588U (en) * | 2016-05-25 | 2016-12-07 | 厦门大学 | A kind of X-ray digital imaging system |
| CN106821410A (en) * | 2017-03-20 | 2017-06-13 | 厦门大学 | Using the substance identification and system of EO-1 hyperion CT functional imagings |
Non-Patent Citations (2)
| Title |
|---|
| 王倩等: "基于X射线吸收谱的不同生物组织的辨识", 《高等学校化学学报》 * |
| 胡波: "X射线吸收光谱测量系统设计与开发", 《万方硕士论文数据库》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110648311A (en) * | 2019-09-03 | 2020-01-03 | 南开大学 | Acne image focus segmentation and counting network model based on multitask learning |
| CN113295722A (en) * | 2021-05-21 | 2021-08-24 | 厦门大学 | X-ray spectral data correction method and device based on deep learning algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11501438B2 (en) | Cone-beam CT image enhancement using generative adversarial networks | |
| EP3226766B1 (en) | System and method for image calibration | |
| JP3775992B2 (en) | Radiation therapy inspection system | |
| US11249160B2 (en) | Image segmentation using reference gray scale values | |
| EP3689241B1 (en) | X-ray image processing method and x-ray image processing apparatus | |
| CN109381212A (en) | A kind of image formation control method and system | |
| US20050267348A1 (en) | Methods and apparatus for automatic protocol selection | |
| US20040066978A1 (en) | Image processing method and image processing apparatus | |
| US20190350550A1 (en) | Method and system for determining irradiation dose | |
| CN109452947A (en) | For generating positioning image and the method to patient's imaging, x-ray imaging system | |
| CN109646112B (en) | Ultrasonic probe positioning system and ultrasonic probe positioning method | |
| CN104125841A (en) | Control method and control system | |
| CN111199566A (en) | Medical image processing method, medical image processing device, storage medium and computer equipment | |
| CN109567838A (en) | A kind of X ray absorption spectrometry lesion detector | |
| US20130053689A1 (en) | Method and system for design of spectral filter to classify tissue and material from multi-energy images | |
| KR20160107528A (en) | Apparatus and method for providing reliability for computer aided diagnosis | |
| US20210398329A1 (en) | Artificial intelligence (ai)-based standardized uptake vaule (suv) correction and variation assessment for positron emission tomography (pet) | |
| EP3685750A1 (en) | X-ray image processing method and x-ray image processing apparatus using the same | |
| CN108805954B (en) | Projection chromatography three-dimensional blood flow velocity measurement device and method | |
| CN111554383B (en) | Neural network for breast tumor detection and detection system thereof | |
| US20230377724A1 (en) | Temporal prediction in anatomic position monitoring using artificial intelligence modeling | |
| US20220130520A1 (en) | System and methods for radiographic image quality assessment and protocol optimization | |
| CN116832345B (en) | Electronic device, radiation therapy system and related devices | |
| CN111402231A (en) | Automatic evaluation system and method for lung CT image quality | |
| US20210290977A1 (en) | Irradiation control apparatus, radiotherapy system, and irradiation control method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190405 |