CN112137581A - Cerenkov fluorescence tomography reconstruction method based on multilayer perception network - Google Patents
Cerenkov fluorescence tomography reconstruction method based on multilayer perception network Download PDFInfo
- Publication number
- CN112137581A CN112137581A CN202010873011.6A CN202010873011A CN112137581A CN 112137581 A CN112137581 A CN 112137581A CN 202010873011 A CN202010873011 A CN 202010873011A CN 112137581 A CN112137581 A CN 112137581A
- Authority
- CN
- China
- Prior art keywords
- network
- sample
- light source
- cherenkov
- fluorescent light
- 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
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/0059—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence
- A61B5/0071—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence by measuring fluorescence emission
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/0033—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2503/00—Evaluating a particular growth phase or type of persons or animals
- A61B2503/40—Animals
Abstract
The invention belongs to the technical field of optical molecular imaging, and discloses a Cerenkov fluorescence tomography reconstruction method based on a multilayer perception network, which is used for generating a gridded sample model and generating a training sample; constructing a multilayer perception network of the Cerenkov fluorescence tomography and training, wherein the multilayer perception network can be divided into a forward network A and a reverse network B; collecting a Cherenkov fluorescence signal on the surface of the organism, and reconstructing to obtain the three-dimensional distribution information of a Cherenkov fluorescence light source in the organism; and mapping the obtained preliminary reconstruction result to the constructed gridded sample model, and inputting the preliminary reconstruction result to the multilayer perception network to obtain an accurate reconstruction result. The reconstruction result obtained by the method is closer to the real light source in shape and position, and the Cerenkov fluorescent light source can be accurately reconstructed.
Description
Technical Field
The invention belongs to the technical field of optical molecular imaging, and particularly relates to a Cerenkov fluorescence tomography reconstruction method based on a multilayer perception network.
Background
Currently, the closest prior art: cerenkov fluorescence Imaging (CLI) is an optical molecular Imaging technique that has recently emerged. The CCD camera with high sensitivity is used for collecting visible light and near infrared light which are generated by a radionuclide probe in the nuclear decay process and have the wavelength of 400-900 mm and are transmitted to the surface of biological tissue for imaging. Compared with other optical molecular imaging technologies, the CLI technology has the advantages of multiple probe types, high signal intensity, rich acquired information and the like because a large number of nuclide probes are approved by the food and drug administration of America for clinical diagnosis and treatment.
Cheenkov Luminescence Tomography (CLT) is a three-dimensional imaging based on CLI technology for the distribution of a radionuclide probe in a biological tissue, and at this time, the radionuclide probe is a cheenkov fluorescence light source. The system can reflect the three-dimensional spatial position information of the nuclide probe in the small animal body, can accurately and quantitatively observe the nuclide probe, and has higher spatial resolution, so the system is widely concerned by scholars. The CLT reconstruction method comprises two processes of a forward problem and a reverse problem. The forward problem is to describe a complex physical process of transmitting fluorescence from a light source to the surface of an organism through a mathematical model, and establish a mapping relation between a Cerenkov fluorescence signal on the surface of the organism and three-dimensional distribution of a Cerenkov fluorescence light source in the organism by establishing a propagation model of the light in biological tissues and then solving the model. The inverse problem belongs to the optimization problem, namely, the optimal solution is obtained from the mapping equation obtained in the forward problem, and the obtained optimal solution is the three-dimensional distribution of the Cerenkov fluorescent light source in the organism. In order to improve the reconstruction quality of the CLT, an accurate model of propagation of the Cerenkov fluorescence photons in a living body needs to be constructed, and an accurate and rapid reconstruction method is needed. Whereas traditional CLT reconstruction methods rely on building photon propagation models, as well as optical parametric properties of biological organs. Therefore, the reconstruction result is influenced by factors such as inaccurate organ region segmentation and inaccurate organ optical parameters, and the reconstruction result is often different from a real light source, so that the reconstruction precision is low.
In summary, the problems of the prior art are as follows: traditional CLT reconstruction methods rely on building photon propagation models, as well as optical parametric properties of biological organs. Therefore, the reconstruction method is influenced by factors such as inaccurate organ region segmentation and inaccurate organ optical parameters, and the reconstruction precision is low.
The difficulty of solving the technical problems is as follows:
the dependence on more parameters, more complex calculation and low accuracy.
The significance of solving the technical problems is as follows:
a higher reconstruction accuracy may provide a more accurate reference to the physician during the application of CLT to clinical diagnosis, thereby reducing the likelihood of false positives.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a multilayer perception network-based Cerenkov fluorescence tomography reconstruction method.
The invention is realized in such a way that a multilayer perception network-based Cerenkov fluorescence tomography reconstruction method comprises the following steps:
firstly, generating a gridded sample model and generating a training sample;
secondly, constructing and training a multilayer perception network of the Cerenkov fluorescence tomography, wherein the multilayer perception network can be divided into a forward network A and a reverse network B;
thirdly, acquiring a Cherenkov fluorescence signal on the surface of the organism, and reconstructing to obtain the three-dimensional distribution information of a Cherenkov fluorescence light source in the organism;
and fourthly, mapping the obtained preliminary reconstruction result to the constructed gridded sample model, and inputting the preliminary reconstruction result to the multilayer perception network to obtain an accurate reconstruction result.
Further, the first generating training samples comprises:
(1) constructing a sample model with the size of 5 multiplied by 5mm3The constructed sample model is meshed by utilizing a finite element theory, and further the meshed sample model is obtained;
(2) constructing a simulation sample of the single-Cherenkov fluorescent light source, arranging a single spherical Cherenkov fluorescent light source in the gridded sample model in the step (1), wherein the radius of the light source is 0.1mm, and generating a simulation training sample of the single-Cherenkov fluorescent light source by using a Monte Carlo simulation MOSE platform;
(3) and (3) expanding the sample to obtain a multi-Cherenkov fluorescent light source simulation sample, and expanding the sample by using a sample combination method on the basis of obtaining the single-Cherenkov fluorescent light source simulation sample set in the step (2), so as to obtain the multi-Cherenkov fluorescent light source simulation training sample.
Further, the second step of constructing and training a multilayer perception network for Cerenkov fluorescence tomography comprises:
(1) constructing a forward network A, wherein the forward network A comprises 1 input layer, 4 hidden layers and 1 output layer, the number of nodes of the input layer and the number of nodes of the hidden layers are the same as the number of nodes of a sample model grid, and the number of nodes of the output layer is the same as the number of nodes on the surface of the sample model grid;
(2) training a forward network A by using the obtained multi-Cherenkov fluorescent light source simulation sample, inputting distribution data of the Cherenkov fluorescent light source of the multi-Cherenkov fluorescent light source simulation sample in a sample model by the network, and outputting the predicted distribution data of the multi-Cherenkov fluorescent light source simulation sample on the surface of a grid by the network;
(3) constructing a reverse network B, wherein the reverse network B comprises 1 input layer, 4 hidden layers and 1 output layer, the number of nodes of the input layer is the same as that of nodes on the surface of a sample model grid, and the number of nodes of the output layer and that of nodes of the hidden layers are the same as that of nodes of the sample model grid;
(4) training a reverse network B by using the obtained multi-Cherenkov fluorescent light source simulation sample, inputting the distribution data of the multi-Cherenkov fluorescent light source simulation sample on the surface of a grid by the network, and outputting the predicted distribution data of the Cherenkov fluorescent light source of the multi-Cherenkov fluorescent light source simulation sample in a sample model by the network;
(5) and combining the forward network A and the reverse network B, and taking the output layer of the trained forward network A as the input layer of the trained reverse network B to obtain the final multilayer perception network.
Further, the output result of each layer in (2) and (4) is corrected by using a correction function; the negative values in the output results of the linear units of the hidden and output layers are modified by:
wherein X represents the output result of the linear unit of the current layer, and ReLu represents a correction function; and when the output result is a negative value or zero, the correction function returns the negative value to zero.
Further, in (2) and (4), the relationship between the current layer and the previous layer is as follows:
Xi＝Dropout0.4(ReLu(WiXi-1+bi))i≥2；
wherein XiNode value, W, representing the i-th layeriRepresents the weight of the ith layer, biIndicating the bias, Dropout, of the ith layer0.4Is a random function that indicates that the nodes of each layer have a 40% probability of being zeroed out.
Further, in the (2) and (4), the multi-layer perceptual network is constraint-trained by the following formula:
wherein | · | purple2Denotes the second order norm, minypredRepresenting y satisfying a minimum second order normpred(ii) a In (2), ytrueFor training known information on the distribution of the Cerenkov fluorescence signal in the sample, ypredCorresponding predicted Cerenkov fluorescence signal distribution information output for the network; in (4), ytrueFor training the known information of the three-dimensional distribution of the Cerenkov fluorescent light source in the sample, ypredAnd outputting the three-dimensional distribution information of the corresponding predicted Cherenkov fluorescent light source for the network.
Further, the third step of acquiring the cerenkov fluorescence signal on the surface of the organism, and reconstructing to obtain the three-dimensional distribution information of the cerenkov fluorescence light source in the organism specifically comprises:
(1) collecting a Cerenkov fluorescence signal on the surface of a biological body;
(2) and (4) carrying out reconstruction by using a reconstruction method to obtain a preliminary distribution result of the Cerenkov fluorescent light source in the organism.
Further, the fourth step maps the obtained preliminary reconstruction result to the constructed gridded sample model; inputting the data into a multilayer perception network, and obtaining an accurate reconstruction result specifically comprises the following steps:
(1) mapping the primary reconstruction result to a sample model;
(2) and inputting the data mapped into the sample model into the combined multilayer perception network to obtain an accurate distribution result of the Cerenkov fluorescent light source in the organism.
The invention also aims to provide an application of the multilayer perception network-based Cerenkov fluorescence tomography reconstruction method in optical molecular image processing.
The invention further aims to provide an information data processing terminal applying the multilayer perception network-based Cerenkov fluorescence tomography reconstruction method.
In summary, the advantages and positive effects of the invention are: the training samples obtained by the invention are simulation samples which are obtained by MOSE, and the simulation samples for research in the direction are generated by the MOSE. And building a multilayer perception network, training by using the obtained simulation sample, and storing the trained network after the training is finished. And (3) performing a real in-vivo experiment, injecting a nuclide probe into the mouse body to obtain a Cerenkov fluorescence light source, collecting Cerenkov fluorescence formed on the mouse body surface, and reconstructing the Cerenkov fluorescence by using the conventional method. And inputting the obtained preliminary reconstruction result into a gridding sample model in the first step, carrying out gridding treatment on the preliminary reconstruction result, and then inputting the gridded data into the multi-layer perception network trained in the second step to obtain a final reconstruction result. Because photons need to pass through a long distance from the light source position to the surface of the organism, the difference of the result obtained by the existing reconstruction method is larger than that of a real light source, and the result is inaccurate. On the basis, a more refined model is constructed, and a more accurate result can be obtained by reconstructing the model.
The traditional method has the following defects: due to the fact that the Cerenkov fluorescence signal is weak and the attenuation coefficient of the biological tissue to the Cerenkov fluorescence signal is large, the constructed system matrix for reconstruction has large ill-conditioned performance, the reconstruction result is not accurate enough, and the situation that the position and the size of the reconstruction light source have certain deviation compared with the real light source is reflected. The reconstruction result obtained by the method is closer to the real light source in shape and position, and the Cerenkov fluorescent light source is reconstructed more accurately.
Drawings
Fig. 1 is a flowchart of a multilayer perception network-based cerenkov fluorescence tomography reconstruction method provided by the embodiment of the invention.
Fig. 2 is a flowchart of an implementation of a multilayer perception network-based cerenkov fluorescence tomography reconstruction method according to an embodiment of the present invention.
Fig. 3 is a diagram of a forward network structure of a multi-layer aware network according to an embodiment of the present invention.
Fig. 4 is a reverse network structure diagram of a multi-layer aware network according to an embodiment of the present invention.
Fig. 5 is a schematic distribution diagram of the cerenkov fluorescence light source provided by the embodiment of the invention in the real organism.
Fig. 6 is a schematic diagram of a preliminary reconstruction result provided by the embodiment of the present invention.
Fig. 7 is a schematic diagram of a reconstruction result obtained by the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Aiming at the problems in the prior art, the invention provides a multilayer perception network-based Cerenkov fluorescence tomography reconstruction method, and the detailed description of the invention is provided below by combining with the accompanying drawings.
As shown in fig. 1, the method for reconstructing cerenkov fluorescence tomography based on a multilayer sensing network according to the embodiment of the present invention includes the following steps:
s101: generating a gridded sample model and generating a training sample;
s102: constructing a multilayer perception network of the Cerenkov fluorescence tomography and training, wherein the multilayer perception network can be divided into a forward network A and a reverse network B;
s103: collecting a Cherenkov fluorescence signal on the surface of the organism, and reconstructing to obtain the three-dimensional distribution information of a Cherenkov fluorescence light source in the organism;
s104: and mapping the obtained preliminary reconstruction result to the constructed gridded sample model, and inputting the preliminary reconstruction result to the multilayer perception network to obtain an accurate reconstruction result.
The technical solution of the present invention is further described below with reference to the accompanying drawings.
As shown in fig. 2, the method for reconstructing cerenkov fluorescence tomography based on a multilayer sensing network provided by the embodiment of the present invention specifically includes the following steps:
firstly, a Monte Carlo simulation platform is used for generating a training sample set, which can be specifically divided into:
(1) constructing a sample model with the size of 5 multiplied by 5mm3And gridding the constructed sample model by using a finite element theory to further obtain the gridded sample model.
(2) And (2) constructing a simulation sample of the single-Cherenkov fluorescent light source, arranging a single spherical Cherenkov fluorescent light source in the gridded sample model in the step (1), wherein the radius of the light source is 0.1mm, and generating the simulation training sample of the single-Cherenkov fluorescent light source by using a Monte Carlo simulation (MOSE) platform.
(3) And (3) expanding the sample to obtain a multi-Cherenkov fluorescent light source simulation sample, and expanding the sample by using a sample combination method on the basis of obtaining the single-Cherenkov fluorescent light source simulation sample set in the step (2), so as to obtain the multi-Cherenkov fluorescent light source simulation training sample.
The combined bicinchonkov fluorescent light source is taken as a specific embodiment for explanation: in the single-Cherenkov fluorescence light source sample set, 2 training samples are randomly selected, and as the spherical Cherenkov fluorescence light source can emit Cherenkov fluorescence, the Cherenkov fluorescence can be transmitted to the surface of an imaging object, and the Cherenkov fluorescence is distributed on the surface of the imaging object. The Cherenkov fluorescence light sources of the selected samples are added, and the Cherenkov fluorescence distribution on the surface of the imaging object is added, so that the double-Cherenkov fluorescence light source simulation training sample is obtained. By analogy, the same number of the TriCerenkov fluorescent light source simulation samples and the same number of the TetraCerenkov fluorescent light source simulation training samples can be obtained, and then a multi-Cerenkov fluorescent light source simulation training sample set is formed.
And secondly, constructing a multilayer perception network of the Cerenkov fluorescence tomography and training, wherein the multilayer perception network can be divided into a forward network A and a reverse network B, and specifically can be divided into:
(1) and constructing a forward network A, wherein the forward network A comprises 1 input layer, 4 hidden layers and 1 output layer, as shown in FIG. 3, the number of nodes of the input layer and the number of nodes of the hidden layers are the same as the number of nodes of the sample model mesh, and the number of nodes of the output layer is the same as the number of nodes on the surface of the sample model mesh.
(2) Training a forward network A by using the multi-Cherenkov fluorescent light source simulation sample obtained in the first step, inputting the distribution data of the Cherenkov fluorescent light source of the multi-Cherenkov fluorescent light source simulation sample in a sample model by the network, and outputting the predicted distribution data of the multi-Cherenkov fluorescent light source simulation sample on the surface of the grid by the network.
(3) And constructing an inverse network B, wherein the inverse network B comprises 1 input layer, 4 hidden layers and 1 output layer, as shown in FIG. 4, the number of nodes of the input layer is the same as that of nodes on the surface of the sample model mesh, and the number of nodes of the output layer and that of nodes of the hidden layers are the same as that of nodes of the sample model mesh.
(4) And training a reverse network B by using the multi-Cherenkov fluorescent light source simulation sample obtained in the first step, wherein the network input is the distribution data of the multi-Cherenkov fluorescent light source simulation sample on the surface of the grid, and the network output is the predicted distribution data of the Cherenkov fluorescent light source of the multi-Cherenkov fluorescent light source simulation sample in the sample model.
(5) And combining the forward network A and the reverse network B, and taking the output layer of the trained forward network A as the input layer of the trained reverse network B to obtain the final multilayer perception network.
In (2) and (4), since the output result of each layer is theoretically a positive value, but a negative value may be generated in the training process, and the negative value belongs to an error or an error, the output result of each layer needs to be corrected by using a correction function.
In a preferred embodiment of the invention, negative values in the output results of the linear units of the hidden and output layers can be modified by:
where X represents the output result of the linear unit of the current layer and ReLu represents the correction function. According to the above formula, when the output result is a positive value, no influence is generated, and when the output result is a negative value or zero, the correction function returns the negative value to zero.
In the preferred embodiment of the present invention, in (2) and (4), the relationship between the current layer and the previous layer is as follows:
Xi＝Dropout0.4(ReLu(WiXi-1+bi))i≥2；
wherein XiNode value, W, representing the i-th layeriRepresents the weight of the ith layer, biIndicating the bias, Dropout, of the ith layer0.4Is a random function that indicates that the nodes of each layer have a 40% probability of being zeroed out.
In the preferred embodiment of the present invention, in (2) and (4), the multi-layer perceptual network is constraint-trained by the following formula:
ytruefor training known information on the distribution of the Cerenkov fluorescence signal in the sample, ypredCorresponding predicted Cerenkov fluorescence signal distribution information output for the network; in (4), ytrueFor training the known information of the three-dimensional distribution of the Cerenkov fluorescent light source in the sample, ypredAnd outputting the three-dimensional distribution information of the corresponding predicted Cherenkov fluorescent light source for the network.
And thirdly, obtaining a preliminary reconstruction result by using the existing reconstruction method, wherein a three-dimensional model of the organism is shown in figure 5, and a red part is the position of the Cherenkov fluorescent light source. The method comprises the steps of firstly collecting Cherenkov fluorescence signals on the surface of an organism, and then reconstructing to obtain the three-dimensional distribution information of the Cherenkov fluorescence light source in the organism based on the reconstruction method provided by Chendoarene and the like in the patent application document 'an excited fluorescence tomography reconstruction method', the patent application No. 201510041349.4, the application date 2015-10-27, wherein the primary reconstruction result is shown in figure 6.
And step four, mapping the preliminary reconstruction result obtained in the step three to the constructed gridded sample model, and then inputting the preliminary reconstruction result into the trained multilayer perception network to obtain an accurate reconstruction result, wherein the reconstruction result is shown in fig. 7.
As can be seen from comparing fig. 6 and 7, in fig. 6, the difference between the reconstruction result obtained by applying the conventional reconstruction method and the real light source is large, and the result is inaccurate, but it can be seen that the reconstruction result obtained by the method of the present invention is closer to the real light source in terms of position and shape, that is, the reconstruction accuracy is improved.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (10)
1. The multilayer perception network-based Cerenkov fluorescence tomography reconstruction method is characterized by comprising the following steps of:
firstly, generating a gridded sample model and generating a training sample;
secondly, constructing and training a multilayer perception network of the Cerenkov fluorescence tomography, wherein the multilayer perception network can be divided into a forward network A and a reverse network B;
thirdly, acquiring a Cherenkov fluorescence signal on the surface of the organism, and reconstructing to obtain the three-dimensional distribution information of a Cherenkov fluorescence light source in the organism;
and fourthly, mapping the obtained preliminary reconstruction result to the constructed gridded sample model, and inputting the preliminary reconstruction result to the multilayer perception network to obtain an accurate reconstruction result.
2. The multilayer perceptron network-based Cerenkov fluorescence tomography reconstruction method of claim 1, in which the first step of generating training samples comprises:
(1) constructing a sample model with the size of 5 multiplied by 5mm3The constructed sample model is meshed by utilizing a finite element theory, and then a meshed sample is obtainedA model;
(2) constructing a simulation sample of the single-Cherenkov fluorescent light source, arranging a single spherical Cherenkov fluorescent light source in the gridded sample model in the step (1), wherein the radius of the light source is 0.1mm, and generating a simulation training sample of the single-Cherenkov fluorescent light source by using a Monte Carlo simulation MOSE platform;
(3) and (3) expanding the sample to obtain a multi-Cherenkov fluorescent light source simulation sample, and expanding the sample by using a sample combination method on the basis of obtaining the single-Cherenkov fluorescent light source simulation sample set in the step (2), so as to obtain the multi-Cherenkov fluorescent light source simulation training sample.
3. The multilayer perception network-based Cerenkov fluorescence tomography reconstruction method according to claim 1, wherein the second step of constructing the multilayer perception network for Cerenkov fluorescence tomography and training comprises:
(1) constructing a forward network A, wherein the forward network A comprises 1 input layer, 4 hidden layers and 1 output layer, the number of nodes of the input layer and the number of nodes of the hidden layers are the same as the number of nodes of a sample model grid, and the number of nodes of the output layer is the same as the number of nodes on the surface of the sample model grid;
(2) training a forward network A by using the obtained multiple-Cherenkov fluorescent light source sample, inputting distribution data of the Cherenkov fluorescent light source in the sample model, which is the multiple-Cherenkov fluorescent light source simulation sample, into the network, and outputting the distribution data of the predicted multiple-Cherenkov fluorescent light source simulation sample on the surface of the sample model by the network;
(3) constructing a reverse network B, wherein the reverse network B comprises 1 input layer, 4 hidden layers and 1 output layer, the number of nodes of the input layer is the same as that of nodes on the surface of a sample model grid, and the number of nodes of the output layer and that of nodes of the hidden layers are the same as that of nodes of the sample model grid;
(4) training a reverse network B by using the obtained multiple-Cherenkov fluorescent light source sample, inputting the distribution data of the multiple-Cherenkov fluorescent light source simulation sample on the surface of the sample model by the network, and outputting the predicted distribution data of the Cherenkov fluorescent light source of the multiple-Cherenkov fluorescent light source simulation sample in the sample model by the network;
(5) and combining the forward network A and the reverse network B, and taking the output layer of the trained forward network A as the input layer of the trained reverse network B to obtain the final multilayer perception network.
4. The multilayer perceptron-based Cerenkov fluorescence tomography reconstruction method of claim 3, characterized in that the output result of each layer in (2) and (4) is corrected by a correction function; the negative values in the output results of the linear units of the hidden and output layers are modified by:
wherein X represents the output result of the linear unit of the current layer, and ReLu represents a correction function; and when the output result is a negative value or zero, the correction function returns the negative value to zero.
5. The method for reconstructing Cerenkov fluorescence tomography based on multilayer perceptron network as claimed in claim 3, wherein in (2) and (4), the relationship between the current layer and the previous layer is as follows:
Xi＝Dropout0.4(ReLu(WiXi-1+bi)) i≥2；
wherein XiNode value, W, representing the i-th layeriRepresents the weight of the ith layer, biIndicating the bias, Dropout, of the ith layer0.4Is a random function that indicates that the nodes of each layer have a 40% probability of being zeroed out.
6. The multilayer perception network-based Cerenkov fluorescence tomography reconstruction method as claimed in claim 3, wherein in the steps (2) and (4), the multilayer perception network is constraint-trained according to the following formula:
wherein | · | purple2Denotes the second order norm, minypredRepresenting y satisfying a minimum second order normpred(ii) a In (2), ytrueFor training known information on the distribution of the Cerenkov fluorescence signal in the sample, ypredCorresponding predicted Cerenkov fluorescence signal distribution information output for the network; in (4), ytrueFor training the known information of the three-dimensional distribution of the Cerenkov fluorescent light source in the sample, ypredAnd outputting the three-dimensional distribution information of the corresponding predicted Cherenkov fluorescent light source for the network.
7. The multilayer perception network-based Cerenkov fluorescence tomography reconstruction method according to claim 1, wherein the third step of acquiring Cerenkov fluorescence signals from the surface of the organism and reconstructing to obtain three-dimensional distribution information of Cerenkov fluorescence light sources inside the organism specifically comprises:
(1) collecting a Cerenkov fluorescence signal on the surface of a biological body;
(2) and (4) carrying out reconstruction by using a reconstruction method to obtain a preliminary distribution result of the Cerenkov fluorescent light source in the organism.
8. The multilayer perception network-based Cerenkov fluorescence tomography reconstruction method according to claim 1, wherein the fourth step maps the obtained preliminary reconstruction result to the constructed gridded sample model; inputting the data into the trained multilayer perception network, and obtaining an accurate reconstruction result specifically comprises the following steps:
(1) mapping the primary reconstruction result to a sample model;
(2) and inputting the data mapped into the sample model into the combined multilayer perception network to obtain an accurate distribution result of the Cerenkov fluorescent light source in the organism.
9. An application of the multilayer perception network-based Cerenkov fluorescence tomography reconstruction method in optical molecular image processing according to any one of claims 1 to 8.
10. An information data processing terminal applying the multilayer perception network-based Cerenkov fluorescence tomography reconstruction method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010873011.6A CN112137581A (en) | 2020-08-26 | 2020-08-26 | Cerenkov fluorescence tomography reconstruction method based on multilayer perception network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010873011.6A CN112137581A (en) | 2020-08-26 | 2020-08-26 | Cerenkov fluorescence tomography reconstruction method based on multilayer perception network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112137581A true CN112137581A (en) | 2020-12-29 |
Family
ID=73888236
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010873011.6A Pending CN112137581A (en) | 2020-08-26 | 2020-08-26 | Cerenkov fluorescence tomography reconstruction method based on multilayer perception network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112137581A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180144465A1 (en) * | 2016-11-23 | 2018-05-24 | General Electric Company | Deep learning medical systems and methods for medical procedures |
| CN108451508A (en) * | 2018-04-28 | 2018-08-28 | 中国科学院自动化研究所 | Archebiosis fluorescent three-dimensional imaging method based on multi-layer perception (MLP) |
| CN108814550A (en) * | 2018-04-16 | 2018-11-16 | 北京工业大学 | A kind of near infrared spectrum tomography rebuilding method neural network based |
| CN109191564A (en) * | 2018-07-27 | 2019-01-11 | 中国科学院自动化研究所 | Exciting tomography fluorescence imaging three-dimensional rebuilding method based on deep learning |
-
2020
- 2020-08-26 CN CN202010873011.6A patent/CN112137581A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180144465A1 (en) * | 2016-11-23 | 2018-05-24 | General Electric Company | Deep learning medical systems and methods for medical procedures |
| CN108814550A (en) * | 2018-04-16 | 2018-11-16 | 北京工业大学 | A kind of near infrared spectrum tomography rebuilding method neural network based |
| CN108451508A (en) * | 2018-04-28 | 2018-08-28 | 中国科学院自动化研究所 | Archebiosis fluorescent three-dimensional imaging method based on multi-layer perception (MLP) |
| CN109191564A (en) * | 2018-07-27 | 2019-01-11 | 中国科学院自动化研究所 | Exciting tomography fluorescence imaging three-dimensional rebuilding method based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| XIN CAO等: "A Novel Stacked Denoising Autoencoder-Based Reconstruction Framework for Cerenkov Luminescence Tomography", 《IEEE ACCESS》 * |
| ZEYU ZHANG等: ""A novel Cerenkov luminescence tomography approach using multilayer fully connected neural network", 《PHYSICS IN MEDICINE & BIOLOGY》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109978778B (en) | Convolutional neural network medical CT image denoising method based on residual learning | |
| CN109191564B (en) | Depth learning-based three-dimensional reconstruction method for fluorescence tomography | |
| CN108451508B (en) | Biological autofluorescence three-dimensional imaging method based on multilayer perceptron | |
| CN111915733B (en) | LeNet network-based three-dimensional cone-beam X-ray luminescence tomography method | |
| CN107392977B (en) | Single-view Cerenkov luminescence tomography reconstruction method | |
| CN113409466B (en) | Excitation fluorescence tomography method based on GCN residual error connection network | |
| Cong et al. | A Born‐type approximation method for bioluminescence tomography | |
| CN109166103B (en) | Excitation fluorescence tomography method based on multilayer perception network | |
| CN111508076B (en) | Elastic network excitation fluorescence fault reconstruction system based on self-adaptive parameter search | |
| WO2017004851A1 (en) | Bioluminescence tomography reconstruction algorithm based on multi-task bayesian compressed sensing method | |
| Wang et al. | Fluorescence molecular tomography reconstruction of small targets using stacked auto-encoder neural networks | |
| Wang et al. | Optimization of reconstruction accuracy of anomaly position based on stacked auto-encoder neural networks | |
| CN110974166A (en) | Optical tomography method and system based on K-nearest neighbor local connection network | |
| Qin et al. | Galerkin-based meshless methods for photon transport in the biological tissue | |
| CN107713995B (en) | Bioluminescence tomography light source reconstruction method based on penalty algorithm | |
| Kaplan et al. | Monte-Carlo-based inversion scheme for 3D quantitative photoacoustic tomography | |
| Yu et al. | Bioluminescence tomography based on one-dimensional convolutional neural networks | |
| CN112137581A (en) | Cerenkov fluorescence tomography reconstruction method based on multilayer perception network | |
| CN113951835B (en) | Three-dimensional fluorescence microscopic imaging method based on optical fault reconstruction strategy | |
| CN113288188B (en) | Cone beam X-ray luminescence tomography method based on grouping attention residual error network | |
| CN115868923A (en) | Fluorescence molecule tomography method and system based on expanded cyclic neural network | |
| CN107374588B (en) | Multi-light-source fluorescent molecular tomography reconstruction method based on synchronous clustering | |
| CN112089434A (en) | Multispectral bioluminescence tomography method and system | |
| CN112037300A (en) | Optical reconstruction method and device based on alternative direction multiplier network | |
| Zhang et al. | Multi-branch attention prior based parameterized generative adversarial network for fast and accurate limited-projection reconstruction in fluorescence molecular tomography |
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: 20201229 |
|
| RJ01 | Rejection of invention patent application after publication |