CN107622516A - The after-treatment system and post-processing approach of electrical impedance tomography art image - Google Patents
The after-treatment system and post-processing approach of electrical impedance tomography art image Download PDFInfo
- Publication number
- CN107622516A CN107622516A CN201610557467.5A CN201610557467A CN107622516A CN 107622516 A CN107622516 A CN 107622516A CN 201610557467 A CN201610557467 A CN 201610557467A CN 107622516 A CN107622516 A CN 107622516A
- Authority
- CN
- China
- Prior art keywords
- image
- electrical impedance
- tomography
- impedance tomography
- post
- 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
Abstract
A kind of after-treatment system and post-processing approach of electrical impedance tomography art image.The after-treatment system of electrical impedance tomography art image includes processing unit and after-treatment device, after-treatment device coupling processing unit.Processing unit is to according to metric data and through the calculation method generation anti-tomography image of first resistor.Metric data is to utilize electrical impedance tomography apparatus measures.After-treatment device is post-processed to produce the anti-tomography image of second resistance to receive the anti-tomography image of first resistor, and through neural network algorithm to the anti-tomography image of first resistor.The present invention is able to pass through above-described embodiment, electrical impedance tomography art image caused by calculation method is post-processed through neural network algorithm to produce the higher electrical impedance tomography art image of the degree of accuracy.In addition, the present invention also can pass through neural network directly produces the high electrical impedance tomography art image of the degree of accuracy by metric data.
Description
Technical field
The present invention is on a kind of Imaging processing technology, and especially with regard to a kind of electrical impedance tomography art image
After-treatment system and post-processing approach.
Background technology
Electrical impedance tomography art (Electrical impedance tomography, EIT) is one kind by body
The conductivity distribution of a part produces the medical science shadowgraph technique of fault image.With other traditional shadowgraph technique such as positron emissions
Computed tomography art (Positron emission tomography, PET), computed tomography art (Computed
Tomography, CT) compare, electrical impedance tomography with magnetic resonance imaging (Magnetic resonance imaging, MRI)
Art is a kind of not expensive price, non-intrusion type and without the tomography of free radiation.But electrical impedance tomography art
The shortcomings that be the resolution relative deficiency of image, its reason is typically to be limited by obtain the number of electrodes of data；
However, cost also improves therewith while increase number of electrodes, it is meant that not lifts video recording analysis by increasing number of electrodes
The good method of degree.
The content of the invention
It is an aspect of the present invention to provide a kind of after-treatment system of electrical impedance tomography art image, and it includes processing and filled
Put and after-treatment device, after-treatment device couple processing unit.Processing unit is according to metric data and passing through calculation method
Produce the anti-tomography image of first resistor.Metric data is to utilize electrical impedance tomography apparatus measures.After-treatment device
To receive the anti-tomography image of first resistor, and through neural network algorithm to the anti-tomography of first resistor
Art image is post-processed to produce the anti-tomography image of second resistance.
In one embodiment of the invention, after-treatment device is more to through neural network algorithm and according to measurement number
According to being post-processed to the anti-tomography image of first resistor to produce the anti-tomography image of 3rd resistor.
In one embodiment of the invention, neural network algorithm includes input layer, output layer and an at least hidden layer,
After-treatment device inputs at least one reality of correspondingly at least one training image more to input at least one training image to input layer
Border image to output layer with determine multiple at least between a hidden layer and input layer, at least between a hidden layer and output layer plus
Weight parameter.
In one embodiment of the invention, processing unit is more to according to an at least training data and through calculation method production
Raw at least one training image, and at least one training image is sent to after-treatment device.An at least training data is to utilize electricity
Impedance tomography apparatus measures.
In one embodiment of the invention, after-treatment device is more to according to noise data decision weighting parameters.
In one embodiment of the invention, calculation method is linear algorithm.
In one embodiment of the invention, calculation method is Nonlinear Iterative Method.
Another aspect of the present invention is to provide a kind of post-processing approach of electrical impedance tomography art image, and it includes following
Step.Through processing unit, the anti-tomography image of first resistor is produced according to metric data and through calculation method.Measure
Data are to utilize electrical impedance tomography apparatus measures.Through after-treatment device, using neural network algorithm to the first electricity
Impedance tomography art image is post-processed to produce the anti-tomography image of second resistance.
In one embodiment of the invention, through after-treatment device, using the neural network algorithm and according to the amount
Data are surveyed to post-process the anti-tomography image of the first resistor to produce the anti-tomography image of a 3rd resistor.
In one embodiment of the invention, neural network algorithm includes input layer, output layer and an at least hidden layer.
Post-processing approach also comprises the steps of.Through after-treatment device, the training image of input at least one inputs correspondingly to input layer
At least one actual image of at least one training image to output layer to determine at least between a hidden layer and input layer, it is at least one hidden
Hide multiple weighting parameters between layer and output layer.
In one embodiment of the invention, through processing unit, produced according to an at least training data and through calculation method
Raw at least one training image.An at least training data is to utilize electrical impedance tomography apparatus measures.
In one embodiment of the invention, through after-treatment device, weighting parameters are determined according to noise data.
In one embodiment of the invention, calculation method is linear algorithm.
In one embodiment of the invention, calculation method is Nonlinear Iterative Method.
The present invention is able to pass through above-described embodiment, and electrical impedance caused by calculation method is broken through neural network algorithm
Layer photography image is post-processed to produce the higher electrical impedance tomography art image of the degree of accuracy.In addition, the present invention also may be used
The high electrical impedance tomography art image of the degree of accuracy is directly produced by metric data through neural network.
Above-mentioned explanation will be explained in detail with embodiment below, and technical scheme is provided more to enter
The explanation of one step.
Brief description of the drawings
For above and other purpose, feature, advantage and the embodiment of the present invention can be become apparent, appended accompanying drawing is said
It is bright as follows：
Fig. 1 is the schematic diagram of the after-treatment system for the electrical impedance tomography art image for illustrating one embodiment of the invention；
Fig. 2A~Fig. 2 C are the neural network schematic diagrames for illustrating some embodiments of the invention；
Fig. 3 is the neural network schematic diagram for illustrating one embodiment of the invention；
Fig. 4~Fig. 9 is the electrical impedance tomography art image schematic diagram for illustrating one embodiment of the invention；
Figure 10 is the post-processing approach flow chart for the electrical impedance tomography art image for illustrating one embodiment of the invention；
Figure 11 is the post-processing approach flow chart for the electrical impedance tomography art image for illustrating one embodiment of the invention；
Figure 12 is electrical impedance tomography instrumentation diagram according to an embodiment of the invention；And
Figure 13 is electrical impedance tomography instrumentation diagram according to an embodiment of the invention.
Embodiment
In order that the present invention narration it is more detailed with it is complete, can refer to accompanying drawing and various embodiments described below.But
The embodiment that is there is provided simultaneously is not used to the scope that the limitation present invention is covered；The description of step is also not used to limit the suitable of its execution
Sequence, it is any by reconfiguring, it is produced that there is equal and other effects device, it is all the scope that the present invention is covered.
In embodiment and claim, unless be particularly limited in interior text for article, otherwise " one " with
"the" can refer to single one or more.It will be further appreciated that "comprising" used herein, " comprising ", " having " and
Similar vocabulary, feature, region, integer, step, operation, element and/or component described in it are indicated, but be not excluded for described in it
Or extra one or more further feature, region, integer, step, operation, element, component, and/or group therein.
On " about " used herein, " about " or " substantially about " be commonly exponential quantity error or scope about
Within 20 percent, preferably it is within about 10, and is more preferably then within about 5 percent.Wen Zhongruo is without clear and definite
Illustrate, the numerical value mentioned by it all regards as approximation, i.e., error or model as represented by " about ", " about " or " substantially about "
Enclose.
In addition, on " coupling " used herein and " connection ", can refer to two or multiple element mutually directly put into effect
Body contact or it is in electrical contact, body of mutually putting into effect indirectly contact or it is in electrical contact, and " coupling " can also refer to two or multiple element it is mutual
Operation or action.
Fig. 1 is refer to, Fig. 1 is the after-treatment system for the electrical impedance tomography art image for illustrating one embodiment of the invention
100 schematic diagram.After-treatment system 100 includes processing unit 110 and after-treatment device 120, the coupling of after-treatment device 120 processing
Device 110.Processing unit 110 is to according to metric data generation electrical impedance tomography art image.Metric data is to utilize electricity
Impedance tomography art (Electrical impedance tomography, EIT), resistance tomography (Electrical
Resistivity tomography, ERT) or electric capacity tomography (Electrical capacitance
Tomography, ECT) produce.
Specifically, processing unit 110 (such as computer, computer, programmable gate array (Field-
Programmable gate array, FPGA), but the present invention is not limited) to according to metric data and through resolving side
Method produces the anti-tomography image of first resistor.Taken the photograph in general, only transmitting the anti-tomography of first resistor caused by calculation method
The actual image of shadow art image and target object is incomplete same, that is, loses genuine problem.Then, (the example of after-treatment device 120
Such as programmable gate array (FPGA), but the present invention is not limited) passing through neural network (Neural
Network, NN) algorithm post-processed to the anti-tomography image of above-mentioned first resistor to produce the anti-tomography of second resistance
Photography image.In the present embodiment, the neural network algorithm that after-treatment device 120 utilizes can effectively solve measurement number
According to the nonlinear problem between the actual image of object.Therefore, compared to the anti-tomography image of first resistor, second resistance
Anti- tomography image has higher accuracy.It should be noted that above-mentioned after-treatment device 120, which utilizes, has been subjected to training
Neural network algorithm produces the anti-tomography image of second resistance.In other words, after-treatment device 120 utilizes and applies rank
The neural network algorithm of section (Application phase) is to produce the anti-tomography image of second resistance.
In an embodiment, the neural network algorithm that after-treatment device 120 utilizes can be artificial neural network
(Artificial neural network, ANN), as shown in Fig. 2A~Fig. 2 C.Neural network 200 include input layer 210,
Hidden layer 220 and output layer 230.Input layer 210 includes at least one input 211~21n of neuron, and hidden layer 220 includes at least
One 221~22n of hidden neuron, output layer 230 include an at least 231~23n of output neuron.Input layer 210 and hidden layer
Weighting parameters 240 be present between 220, between hidden layer 220 and output layer 230.In the training stage of neural network 200
In (Training phase), after-treatment device 120 can pass through training image and actual image train neural network 200 with
Improve the degree of accuracy of the anti-tomography image of second resistance.Specifically, after-treatment device 120 can input at least one training shadow
As to input layer 210, and input at least one actual image to output layer 230 to determine between hidden layer 220 and input layer 210,
Weighting parameters 240 at least between a hidden layer 220 and output layer 230.It is noted that above-mentioned actual image and above-mentioned instruction
Practicing image has corresponding relation.Therefore, after-treatment device 120 will transmit through the anti-tomography of first resistor that similar measurement condition obtains and take the photograph
Shadow art image input neural network 200 (weighting parameters 240 with above-mentioned decision), can produce and be closer to actual image
The anti-tomography image of second resistance, that is, the anti-tomography image of second resistance that the degree of accuracy is higher.
Furthermore in an embodiment, neural network 200 is emanant basis function (Radial basis
Function, RBF) neural network.Emanant basis function neural network has a hidden layer 220, and hidden layer 220 is
Emanant basis function, formula represent as follows：
In above-mentioned formula：wjFor the weighting parameters of j-th of hidden neuron, x is input vector, and t is emanant substrate
The center vector (being normally set up t as 0) of function, σ is diffusion constant (Spread constant).Training process is in selection one
After Heart vector t, the output of neural network 200 is adapted for input vector x and weighting parameters wjEmanant basis function
Linear combination.If having sufficient amount of hidden neuron, emanant basis function neural network can reach high approximate journey
Degree.
It must supplement, above-mentioned artificial neural network can be implemented as feed forward type neural network (Feedforward
Neural network, as shown in Figure 2 A), feed-back type neural network (Recurrent neural network, such as Fig. 2 B institutes
Show) or convolution neural network (Convolutional neural network, as shown in Figure 2 C), but the present invention not as
Limit.In addition, input layer 210 is also not limited to simple layer or multilayer with output layer 230.
Or in another embodiment, the neural network algorithm that after-treatment device 120 utilizes can be deep layer class god
Through network (Deep neural network, DNN), as shown in Figure 3.Neural network 300 include input layer 310, it is multiple hide
Layer 320,330 and output layer 340.Input layer 310 includes at least one input 311~31n of neuron, and hidden layer 320 includes at least
One 321~32n of hidden neuron, hidden layer 330 include an at least 331~33n of hidden neuron, and output layer 340 includes at least
One 341~34n of output neuron.Between input layer 310 and hidden layer 320, between hidden layer 320 and output layer 340, hidden layer
Weighting parameters 350 between 330 and output layer 340 be present.As described above, in the training stage of neural network 300, post processing
Device 120 can pass through training image with actual image to train neural network 300 to improve the anti-tomography of second resistance
The degree of accuracy of image, is not repeated herein.It is noted that the hiding number of layers of neural network 300 can be other numbers
Mesh, it is not limited with two hidden layers 320,330, and the weighting parameters 350 between hidden layer can also fill via above-mentioned post processing
Put 120 training process decision.
Similarly, above-mentioned deep layer neural network can be implemented as feed forward type neural network (as shown in Figure 3), feed-back type
Neural network or convolution neural network, but the present invention is not limited.In addition, input layer 310 is also unlimited with output layer 340
In simple layer or multilayer.
In an embodiment, above-mentioned calculation method can be that linear calculation method (such as drill by the rapid Gauss-Newton of linear one-step
Algorithm (Linear one-step Gauss-Newton algorithm)) or non-linear calculation method (such as point in former antithesis
Method (Primal-dual interior point method, PDIPM) or other iterative methods).
In an embodiment, above-mentioned after-treatment device 120 is used for training the training image of neural network 200,300 can be saturating
Processing unit 110 is crossed to produce according to training data and via calculation method (such as linear calculation method or non-linear calculation method)
It is raw.Similar to metric data, training data is to utilize electrical impedance tomography apparatus measures.Specifically, user can be first sharp
Target object is measured with electrical impedance tomography art to obtain training data.Then, processing unit 110 is through according to training data
And produce training image through calculation method.In general, only transmit training image caused by calculation method and target object
Actual image is incomplete same, that is, loses genuine problem.Therefore, after-treatment device 120 is through training image and actual image
The neural network 200,300 trained can effectively produce the high electrical impedance tomography art image of accuracy rate (that is, second
Electrical impedance tomography art image).In addition, after-treatment system 100 is according to metric data and through calculation method and class nerve net
Network algorithm produce time needed for the anti-tomography image of second resistance it is very short (such as only about 0.80 second, it is actual to calculate
Time need to be depending on the speed of image size, processing unit 110 with after-treatment device 120).
Or in another embodiment, after-treatment device 120 can pass through neural network algorithm and according to metric data
The anti-tomography image of first resistor is post-processed to produce the anti-tomography image of 3rd resistor.Neural network
200th, 300 training method as described above, is not repeated herein.It is noted that in the present embodiment, after-treatment device 120
The zeta potential of electrical impedance tomography art image is further corrected to actual zeta potential, therefore using metric data
Three electrical impedance tomography art images have the high accuracy of image and zeta potential.In addition, after-treatment system 100 is according to measurement
Data and that the time needed for the anti-tomography image of 3rd resistor is produced through neural network algorithm is very short (such as only
About 0.36 second, the actual calculating time need to be depending on the speed of image size, processing unit 110 with after-treatment device 120).
Consequently, it is possible to using neural network algorithm, after-treatment system 100 of the invention can be anti-disconnected according to first resistor
Layer photography image (it is produced via calculation method), promptly produce the high anti-tomography of second resistance of the image degree of accuracy
Art image (such as functional electric impedance tomography art image (Functional EIT image)).In addition, after the present invention
Processing system also can be according to metric data and the anti-tomography image of first resistor (it is produced via calculation method), rapidly
Ground produces image and the anti-tomography image of the high 3rd resistor of the electrical conductivity degree of accuracy (such as the anti-tomography of absolute resistance
Image (Absolute EIT image)).
Must supplement, the anti-tomography image of second resistance caused by above-mentioned after-treatment device 120 or 3rd resistor
Anti- tomography image can be the electrical impedance tomography art shadow formed with facing conductive angle value or absolute zeta potential
Picture.In other words, after-treatment system of the invention can determine the facing conductive angle value of electrical impedance tomography art image or definitely lead
Electric degree value is with response to different demands.
Due to there may be noise among actual measurement process, in an embodiment, after-treatment device 120 is more to basis
Noise data determine above-mentioned weighting parameters 240,350.Therefore, after-treatment system 100 can further improve electrical impedance tomography
The degree of accuracy of art image.
In order to illustrate compared to prior art, after-treatment system 100 of the invention produces electrical impedance tomography art image
With the higher degree of accuracy, with | Δ RES | error represents target object volume and object body in electrical impedance tomography art image
Long-pending difference.In prior art, electrical impedance tomography art image caused by linear calculation method is only transmitted with 35.95%
| Δ RES | error, electrical impedance tomography art image caused by non-linear calculation method is only transmitted with 26.20% | Δ RES
| error.On the other hand, after-treatment system 100 passes through electrical impedance tomography art image caused by neural network algorithm (also
That is the anti-tomography image of 3rd resistor) with 12.54% | Δ RES | error, and after-treatment system 100 passes through resolving side
Method and electrical impedance tomography art image (that is, the anti-tomography image of second resistance) caused by neural network algorithm
With 13.29% | Δ RES | error, wherein neural network are trained without noise data.
As described above, neural network of the invention training can pass through noise data and train further to lift the degree of accuracy.Afterwards
Electrical impedance tomography art image caused by the transmission neural network algorithm of processing system 100 (that is, the anti-tomography of 3rd resistor
Photography image) with 13.02% | Δ RES | error, and after-treatment system 100 is drilled through calculation method with neural network
Electrical impedance tomography art image caused by algorithm (that is, the anti-tomography image of second resistance) is with 13.16% | Δ
RES | error, wherein neural network have trains by noise data.
In order to illustrate the present invention after-treatment system 100 caused by electrical impedance tomography art image, refer to Fig. 4~figure
9。
Fig. 4 is to only transmit electrical impedance tomography art image caused by linear calculation method.As shown in figure 4, electrical impedance is broken
Layer photography image is substantially very big with target object 410,420 differences, and many wrong images be present, and the degree of accuracy is not high.
Fig. 5 is to only transmit electrical impedance tomography art image caused by non-linear calculation method.As shown in figure 5, electrical impedance
Tomography image substantially display target object 420, but target object 410 is then less accurate, and target object 410,420
Region in addition also shows that the image of mistake.
Fig. 6 for only transmit electrical impedance tomography art image caused by neural network algorithm (that is, it is above-mentioned 3rd electricity
Impedance tomography art image), wherein neural network 200,300 is trained without noise data.As shown in fig. 6, electrical impedance
Tomography image substantially display target object 420 and target object 410, and the region beyond target object 410,420
Wrong image is not shown generally.
Fig. 7 is through electrical impedance tomography art image caused by calculation method and neural network algorithm (on that is,
State the anti-tomography image of second resistance), wherein neural network 200,300 is trained without noise data.Such as Fig. 7 institutes
Show, electrical impedance tomography art image substantially display target object 420 and target object 410, and target object 410,420 with
Outer region does not show wrong image.
Fig. 8 for only transmit electrical impedance tomography art image caused by neural network algorithm (that is, it is above-mentioned 3rd electricity
Impedance tomography art image), wherein neural network 200,300 has trains by noise data.As shown in figure 8, electrical impedance
Tomography image substantially display target object 420 and target object 410, and the region beyond target object 410,420
Wrong image is not shown.
Fig. 9 is through electrical impedance tomography art image caused by calculation method and neural network algorithm (on that is,
State the anti-tomography image of second resistance), wherein neural network 200,300 has trains by noise data.Such as Fig. 9 institutes
Show, electrical impedance tomography art image substantially display target object 420 and target object 410, and target object 410,420 with
Outer region does not show wrong image.
Therefore, compared to electrical impedance tomography art image (as shown in the 4th, 5 figures) caused by calculation method are only transmitted, thoroughly
Crossing electrical impedance tomography art image (as shown in the 6th~9 figure) caused by neural network algorithm has the higher degree of accuracy.
In addition, the after-treatment system 100 of the present invention also can be by neural network algorithm combination calculation method, or utilize noise data
Neural network is trained further to lift the degree of accuracy.
Figure 10 is the flow chart of post-processing approach 1000 for the electrical impedance tomography art image for illustrating one embodiment of the invention.
Post-processing approach 1000 has multiple step S1002~S1006, and it can be applied to after-treatment system 100 as described in Figure 1.So
The those skilled in the art of the present invention is familiar with it will be understood that mentioned step in the above-described embodiments, in addition to its bright order person is especially chatted,
Its tandem can be adjusted according to being actually needed, or even simultaneously or partially can performed simultaneously.Specific implementation discloses as preceding, herein
It is not repeated to describe.
In step S1002, through electrical impedance tomography instrument, object is measured to produce metric data.
In step S1004, through processing unit, the anti-tomography of first resistor is produced according to metric data and through calculation method
Photography image.
In step S1006, through after-treatment device, using neural network algorithm to the anti-tomography of first resistor
Image is post-processed to produce the anti-tomography image of second resistance.
Figure 11 is the flow chart of post-processing approach 1100 for the electrical impedance tomography art image for illustrating one embodiment of the invention.
Post-processing approach 1100 has multiple step S1102~S1106, and it can be applied to after-treatment system 100 as described in Figure 1.So
The those skilled in the art of the present invention is familiar with it will be understood that mentioned step in the above-described embodiments, in addition to its bright order person is especially chatted,
Its tandem can be adjusted according to being actually needed, or even simultaneously or partially can performed simultaneously.Specific implementation discloses as preceding, herein
It is not repeated to describe.
In step S1102, through electrical impedance tomography instrument, object is measured to produce metric data.
In step S1104, through processing unit, the anti-tomography of first resistor is produced according to metric data and through calculation method
Photography image.
In step S1106, through after-treatment device, using neural network algorithm and according to metric data to the first electricity
Impedance tomography art image is post-processed to produce the anti-tomography image of 3rd resistor.
In order to illustrate electrical impedance tomography instrument, Figure 12, Figure 13 refer to.Figure 12 is to observation type with interior
(Inward-looking) as an example.As shown in figure 12, electrical impedance tomography instrument includes multiple electrodes 1232, loading end
1234th, electrode controller 1210 and data fatching apparatus 1220.Loading end 1234 is solid surface, and coats and be intended to be detected
Object under test 1240.Therefore, electrode 1232 is covered on the appearance of object under test 1240.
Electrode controller 1210 is respectively and electrically connected to electrode 1232 with driving electrodes 1232 respectively.Data fatching apparatus
1220 collecting and analyze the detection signal of electrode 1232.Region of variability block 1242 in object under test 1240 be present, and region of variability
Block 1242 is the target of electrical impedance tomography instrument detecting analysis.
To illustrate different measurement modes, Figure 13 refer to.Figure 13 is with export-oriented observation type (Outward-looking) electricity
Impedance tomography instrument is as an example.As shown in figure 13, electrical impedance tomography instrument includes multiple electrodes 1332, loading end
1334th, electrode controller 1210 and data fatching apparatus 1220.Loading end 1334 is the installing of cylindric and its surface by electrode
1332 electrod-arrays 1336 formed, wherein being not repeated to describe with Figure 12 identicals part.
The present invention is able to pass through above-described embodiment, and electrical impedance caused by calculation method is broken through neural network algorithm
Layer photography image is post-processed to produce the higher electrical impedance tomography art image of the degree of accuracy.In addition, the present invention also may be used
The high electrical impedance tomography art image of the degree of accuracy is directly produced by metric data through neural network.
Although the present invention is disclosed above with embodiment, so it is not limited to the present invention, any to be familiar with this skill
Person, without departing from the spirit and scope of the present invention, when can be used for a variety of modifications and variations, therefore protection scope of the present invention is worked as
It is defined depending on the scope that claims are defined.
Claims (14)
1. a kind of after-treatment system of electrical impedance tomography art image, it is characterised in that include：
One processing unit, to produce the anti-tomography shadow of a first resistor according to a metric data and through a calculation method
Picture, the wherein metric data are to utilize electrical impedance tomography apparatus measures；And
One after-treatment device, couple the processing unit and to receive the anti-tomography image of the first resistor, and pass through
One neural network algorithm is post-processed anti-disconnected to produce a second resistance to the anti-tomography image of the first resistor
Layer photography image.
2. the after-treatment system of electrical impedance tomography art image according to claim 1, it is characterised in that the post processing
Device is more to the transmission neural network algorithm and according to the metric data to the anti-tomography image of the first resistor
Post-processed to produce the anti-tomography image of a 3rd resistor.
3. the after-treatment system of electrical impedance tomography art image according to claim 1, it is characterised in that such nerve
Network calculus method includes an at least input layer, an at least output layer and an at least hidden layer, and the after-treatment device is more inputting
At least one training image to an at least input layer, and input to should at least one training image at least one actual image to should
An at least output layer is to determine between an at least hidden layer and an at least input layer, an at least hidden layer and this at least one
Multiple weighting parameters between output layer.
4. the after-treatment system of electrical impedance tomography art image according to claim 3, it is characterised in that the processing fills
Put more to according to an at least training data and through the calculation method produce this at least one training image, and by this at least one
Training image is sent to the after-treatment device, and wherein an at least training data is to utilize electrical impedance tomography apparatus measures.
5. the after-treatment system of electrical impedance tomography art image according to claim 3, it is characterised in that the post processing
Device is more to according to the multiple weighting parameters of noise data decision.
6. the after-treatment system of electrical impedance tomography art image according to claim 1, it is characterised in that the resolving side
Method is a linear algorithm.
7. the after-treatment system of electrical impedance tomography art image according to claim 1, it is characterised in that the resolving side
Method is a Nonlinear Iterative Method.
8. a kind of post-processing approach of electrical impedance tomography art image, it is characterised in that include：
Through a processing unit, the anti-tomography shadow of a first resistor is produced according to a metric data and through a calculation method
Picture, the wherein metric data are to utilize electrical impedance tomography apparatus measures；And
Through an after-treatment device, after being carried out using a neural network algorithm to the anti-tomography image of the first resistor
Handle to produce the anti-tomography image of a second resistance.
9. the post-processing approach of electrical impedance tomography art image according to claim 8, it is characterised in that also include：
Through the after-treatment device, using the neural network algorithm and according to the metric data to the anti-tomography of the first resistor
Photography image is post-processed to produce the anti-tomography image of a 3rd resistor.
10. the post-processing approach of electrical impedance tomography art image according to claim 8, it is characterised in that such god
An at least input layer, at least an output layer and an at least hidden layer, the post-processing approach are included through network calculus method also to include：
Through the after-treatment device, the training image of input at least one is inputted to should at least one instruction to an at least input layer
Practice at least one actual image of image to an at least output layer with determine an at least hidden layer and an at least input layer it
Between, multiple weighting parameters between an at least hidden layer and an at least output layer.
11. the post-processing approach of electrical impedance tomography art image according to claim 10, it is characterised in that also wrap
Contain：
Through the processing unit, at least one training image is produced according to an at least training data and through the calculation method, its
In an at least training data be to utilize electrical impedance tomography apparatus measures.
12. the post-processing approach of electrical impedance tomography art image according to claim 10, it is characterised in that also wrap
Contain：
Through the after-treatment device, the multiple weighting parameters are determined according to a noise data.
13. the post-processing approach of electrical impedance tomography art image according to claim 8, it is characterised in that the resolving
Method is a linear algorithm.
14. the post-processing approach of electrical impedance tomography art image according to claim 8, it is characterised in that the resolving
Method is a Nonlinear Iterative Method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610557467.5A CN107622516A (en) | 2016-07-15 | 2016-07-15 | The after-treatment system and post-processing approach of electrical impedance tomography art image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610557467.5A CN107622516A (en) | 2016-07-15 | 2016-07-15 | The after-treatment system and post-processing approach of electrical impedance tomography art image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107622516A true CN107622516A (en) | 2018-01-23 |
Family
ID=61086593
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610557467.5A Pending CN107622516A (en) | 2016-07-15 | 2016-07-15 | The after-treatment system and post-processing approach of electrical impedance tomography art image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107622516A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113327223A (en) * | 2020-02-13 | 2021-08-31 | 中央大学 | System for analyzing brain tissue components based on computed tomography and operating method thereof |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6522910B1 (en) * | 1997-09-11 | 2003-02-18 | Wisys Technology Foundation | Electrical property enhanced tomography (EPET) apparatus and method |
| US7490085B2 (en) * | 2002-12-18 | 2009-02-10 | Ge Medical Systems Global Technology Company, Llc | Computer-assisted data processing system and method incorporating automated learning |
| CN102435637A (en) * | 2010-08-12 | 2012-05-02 | 通用电气公司 | System and method for performing electrical impedance tomography |
| US20160195484A1 (en) * | 2006-04-20 | 2016-07-07 | Jack S. Emery | Systems and methods for impedance analysis of conductive medium |
-
2016
- 2016-07-15 CN CN201610557467.5A patent/CN107622516A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6522910B1 (en) * | 1997-09-11 | 2003-02-18 | Wisys Technology Foundation | Electrical property enhanced tomography (EPET) apparatus and method |
| US7490085B2 (en) * | 2002-12-18 | 2009-02-10 | Ge Medical Systems Global Technology Company, Llc | Computer-assisted data processing system and method incorporating automated learning |
| US20160195484A1 (en) * | 2006-04-20 | 2016-07-07 | Jack S. Emery | Systems and methods for impedance analysis of conductive medium |
| CN102435637A (en) * | 2010-08-12 | 2012-05-02 | 通用电气公司 | System and method for performing electrical impedance tomography |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113327223A (en) * | 2020-02-13 | 2021-08-31 | 中央大学 | System for analyzing brain tissue components based on computed tomography and operating method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Rymarczyk et al. | Comparison of selected machine learning algorithms for industrial electrical tomography | |
| TWI654965B (en) | Post processing system and post processing method of electrical impedance tomography images | |
| WO2018203470A1 (en) | Learning apparatus, learning method, and learning program | |
| CN109242849A (en) | Medical image processing method, device, system and storage medium | |
| CN108898238B (en) | Medical equipment fault prediction system and related method, device and equipment | |
| CN107480677A (en) | The method and device of area-of-interest in a kind of identification three-dimensional CT image | |
| CN110411580B (en) | Diagnosis method and diagnosis system for heating defects of electric equipment | |
| CN104933698A (en) | Image Processing Apparatus And Computer-Readable Recording Medium | |
| CN105938513A (en) | Apparatus and method for providing reliability for computer aided diagnosis | |
| CN109859204A (en) | Convolutional neural networks Model Checking and device | |
| Hutton et al. | Individualized Calibration of Rotation Gain Thresholds for Redirected Walking. | |
| US20220374768A1 (en) | Apparatus of machine learning, machine learning method, and inference apparatus | |
| JP6791966B2 (en) | Process and measurement system for data acquisition and processing in soft tomography surveys | |
| CN109256205A (en) | Method and system for the clinical decision support carried out using local and remote analytics | |
| EP3657444A1 (en) | Tomographic image processing apparatus and method, and computer program product | |
| CN108231189A (en) | Data processing method and medical diagnosis on disease device based on dual-depth nerve learning network | |
| CN107622516A (en) | The after-treatment system and post-processing approach of electrical impedance tomography art image | |
| CN112183555B (en) | Method and system for detecting welding quality, electronic device and storage medium | |
| CN105354414A (en) | Hierarchical analysis-based human body health condition assessment method | |
| CN112881054A (en) | Hoisting machinery fault diagnosis method and system | |
| CN117195683A (en) | Mechanical arm, training method of fault diagnosis model of mechanical arm and fault diagnosis method of mechanical arm | |
| CN107016440A (en) | The multiresolution deep neural network intelligent diagnosing method of machine driving failure | |
| CN113033845B (en) | Construction method and device for power transmission resource co-construction and sharing | |
| Kowalska et al. | Towards high precision electrical capacitance tomography multilayer sensor structure using 3d modelling and 3d printing method | |
| Li et al. | Human standing posture recognition based on CNN and pressure floor |
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 | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Cai Deming Inventor after: Sebatian Martin Inventor before: Cai Deming |