Method for rapidly measuring absorption coefficient of underlying biological tissue based on light field camera

Abstract

The invention relates to a method for rapidly measuring the absorption coefficient of a bottom biological tissue based on a light field camera, and belongs to the technical field of optical diagnosis of biological tissues. The problems of complex detection process and low reconstruction efficiency are solved. Comprising the steps of setting double-layer media with different optical parameters, and obtaining a reflected radiation signal by using a time domain radiation transfer equation. A BP neural network is determined and initialized, and the network is updated through an intelligent algorithm. And inputting a reflected signal for forward propagation, and outputting a predicted optical parameter. And calculating an error, updating a weight through back propagation, and iterating to a precision requirement. A short pulse light source, a spectroscope and an attenuator are used for generating pulse laser, the pulse laser vertically enters the surface of a medium, a light field camera collects reflected signals, and finally optical parameters are output through a trained BP neural network. The light field camera is introduced to obtain radiation light field signals, accurate capture of multi-dimensional light field information is achieved, and rapid and accurate reconstruction of the absorption coefficient of the bottom biological tissue is achieved in combination with a light signal analysis means.

Description

Technical Field

[0001] The present invention relates to a method for rapidly measuring the absorption coefficient of underlying biological tissues based on a light field camera, and belongs to the technical field of optical diagnosis of biological tissues. Background Art

[0002] When light propagates in biological tissues, various interactions such as absorption, scattering, and reflection occur. Different tissue components and structures have different absorption and scattering characteristics for light, resulting in different optical parameter distributions of tissues. As a typical non-uniform participating medium, the structure of biological tissues is composed of scattering layers with different thicknesses and optical properties. Among them, the reconstruction of optical parameters of biological tissues based on the double-layer medium model has important value in optical imaging and spectral detection. However, at present, the double-layer medium model faces many problems. The absorption coefficient sensitivity of the underlying medium is poor, the inverse problem reconstruction is ill-posed, and the reconstruction of optical parameters is difficult. However, the absorption coefficient of the underlying medium contains key information of different components and is crucial for the diagnosis of biological tissues.

[0003] Most traditional research on the reconstruction of biological tissue parameters uses contact fiber optic devices to measure optical signals, providing limited measurement information. And usually, it is necessary to measure optical signals repeatedly for many times, and analyze them in combination with complex reconstruction algorithms. The parameter reconstruction efficiency is low, and the reconstruction accuracy is limited. For the double-layer medium model, as the thickness of the top layer medium increases, the reconstruction difficulty of the absorption coefficient of the bottom layer medium further increases, further increasing the difficulty of optical parameter reconstruction. Usually, it is necessary to select multiple detection sites for repeated measurements many times to obtain sufficient detection optical signals to realize the parameter reconstruction of the bottom layer medium. The measurement process is complex and the reconstruction efficiency is low. Usually, the multi-layer structure of the BP neural network is used to enable it to fit non-linear functions and realize the efficient reconstruction of optical parameters.

[0004] The commonly used weights of the BP neural network perform gradient descent along the local direction and are prone to falling into local extrema; when the network structure is complex and the number of hidden layers is set to be large, the situation of gradient disappearance is likely to occur, resulting in a slow convergence speed.

[0005] Therefore, there is an urgent need to propose a method for rapidly measuring the absorption coefficient of underlying biological tissues based on a light field camera to solve the above technical problems. Summary of the Invention

[0006] The present invention is to solve the problems of complex detection process, strong ill-posedness of inverse problems, and low reconstruction efficiency formed by common technologies. A brief overview of the present invention is given below to provide a basic understanding of certain aspects of the present invention. It should be understood that this overview is not an exhaustive overview of the present invention. It is not intended to identify the key or important parts of the present invention, nor is it intended to limit the scope of the present invention.

[0007] Technical solution of the present invention:

[0008] A rapid measurement method for the absorption coefficient of underlying biological tissue based on a light field camera, comprising the following steps:

[0009] Step 1: Set up n double-layer media with different optical parameter distributions, substitute the optical parameter data set into the time-domain radiative transfer equation for solution, and correspondingly obtain the reflected radiation signals at the boundaries of the double-layer media under n different optical parameters where i = 1, 2, 3, …, n;

[0010] Step 2: Determine the BP neural network model, input the initial network weights and thresholds of the model into the intelligent algorithm, and then update them into the BP neural network model;

[0011] Step 3: Input the data set of the reflected radiation signals in Step 1 into the BP neural network model for forward propagation, and output the corresponding predicted optical parameter data set;

[0012] Step 4: Calculate the error between the predicted optical parameter data set obtained in Step 3 and the corresponding actual optical parameter data set, and perform backpropagation according to the calculated error to update the weights and thresholds of the BP neural network model;

[0013] Step 5: Repeat Step 3 to Step 4, update and iterate until the BP neural network model meets the prediction accuracy requirements. At this time, the training of the BP neural network model improved based on the intelligent algorithm is completed;

[0014] Step 6: Turn on the short-pulse continuous light source, decompose the white light into monochromatic lights of different wavelengths through a beam splitter and a variable neutral density attenuator, and use an interference filter composed of special filter sheets to obtain pulsed laser light;

[0015] Step 7: Vertically incident the pulsed laser light through an optical fiber from the surface of the participating medium, and the light field camera collects the reflected radiation signal M on the surface of the participating medium corresponding to the wavelength d,f , d represents the corresponding detection point position, and f represents the direction of the detected radiation light signal;

[0016] Step 8: Input the reflected radiation measurement signal M d,f into the BP neural network model trained in Step 5, and output the final optical parameters of the double-layer medium through the forward propagation and prediction of the neural network.

[0017] Preferably: Set up n double-layer media with different optical parameter distributions, and their optical parameters are respectively: where μ a1 represents the absorption coefficient of the top layer medium, μ a2 represents the absorption coefficient of the bottom layer medium, μs1 represents the scattering coefficient of the top - layer medium, μ s2 represents the scattering coefficient of the bottom - layer medium;

[0018] And substitute its optical parameters into the time - domain radiative transfer equation for calculation respectively, and finally obtain the reflected radiation signals at the boundaries of the medium to be measured under n groups of different optical parameters Take the reflected radiation signal obtained by solving the time - domain radiative transfer equation as a set of data X m , where X m corresponds to the optical parameter data Y m ;

[0019]

[0020] Pair the reflected radiation signal data X m with the corresponding optical parameter data Y m to form the data set of the BP neural network model, where m = 1, 2, … n, and the data set can be divided into 70% training data set and 30% test data set.

[0021] Preferably: The forward propagation in step three is a propagation path that receives external data through the input layer, then propagates from the input layer to the hidden layer, and finally from the hidden layer to the output layer;

[0022] The BP neural network model in step three consists of an input layer, six hidden layers, and an output layer. Specifically, an input layer is connected to an output layer through six hidden layers in sequence, and the six hidden layers are also connected to each other; there are 4 neurons in the input layer, 8 neurons in each hidden layer, and 4 neurons in the output layer.

[0023] Preferably: The back - propagation in step four uses the gradient descent method and gradient search technology with the goal of minimizing the mean square error between the actual output value and the desired output value.

[0024] Preferably: The convergence criterion of the BP neural network in step four is by setting the maximum number of iterations T max , when the number of iterations reaches T max , stop training;

[0025] Define the function of the BP neural network model improved based on the intelligent algorithm as BP(x), then where, X train is the training data set of the reflected radiation signal, X test is the test data set of the reflected radiation signal, and The predicted values of the optical parameters output for the training dataset and the test dataset respectively, and the prediction accuracy of the test model can be characterized by the following formula:

[0026]

[0027] In the formula: and are the squared residual coefficients of the training data and the test dataset respectively; and are the output average values of the training dataset and the test dataset respectively.

[0028] Preferably: The optical parameters of the double-layer medium obtained in Step VIII include the absorption coefficient of the top tissue, the scattering coefficient of the top tissue, the absorption coefficient of the bottom tissue, and the scattering coefficient of the bottom tissue.

[0029] Preferably: The expression of the time-domain radiative transfer equation in Step I is as follows:

[0030]

[0031] In the formula, c represents the speed of light in the medium, r represents the position, Ω represents the radiative transfer direction, t represents the time, represents the partial differential symbol, represents the Hamiltonian operator; I(r,Ω,t) is the radiative intensity along the direction of Ω at the position r and time t, μ a (r) and μ s (r) are the absorption coefficient and the scattering coefficient respectively, β(r) is the attenuation coefficient, and β(r) = μ a (r) + μ s (r), Φ w (Ω′,Ω) is the scattering phase function.

[0032] The present invention has the following beneficial effects:

[0033] The present invention introduces a light field camera for measurement, obtains rich radiative light field signals, realizes the accurate capture of multi-dimensional light field information, and further combines the light signal analysis means based on deep learning to realize the rapid and accurate reconstruction of the absorption coefficient of the underlying biological tissue;

[0034] The present invention introduces a neural network model to realize the reconstruction of the optical parameters of the double-layer medium. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a flowchart of a method for rapid measurement of the absorption coefficient of the underlying biological tissue based on a light field camera;

[0036] Figure 2 is a schematic structural diagram of a method for rapid measurement of the absorption coefficient of the underlying biological tissue based on a light field camera;

[0037] Figure 3 It is a sensitivity test diagram of the reflected signal in the top layer medium with different thicknesses with respect to the absorption coefficient of the bottom layer medium.

[0038] Figure 2 In the figure: 1 - short - pulse continuous light source, 2 - beam splitter, 3 - variable neutral density attenuator, 4 - interference filter, 5 - light - field camera, 6 - data acquisition and processing system, 7 - participating medium;

[0039] Figure 3 In the figure: the thickness of the top - layer medium in Figure a is 1 cm, and the thickness of the top - layer medium in Figure b is 2 cm. For the sensitivity of the reflected signal with respect to the absorption coefficient of the bottom - layer medium, when the thickness of the top - layer medium increases, the sensitivity of the optical parameters of the bottom - layer medium further decreases. The influence of the optical parameters of the bottom - layer medium on the reflected signal is more reflected in the tail optical signal. Therefore, it is necessary to use a light - field camera to measure the optical signal to obtain more - dimensional and more accurate tail - optical - signal information. Specific implementation manners

[0040] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be described below through specific embodiments shown in the accompanying drawings. However, it should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the descriptions of well - known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present invention.

[0041] Specific implementation manner one: Combining Figures 1-3 To illustrate this implementation manner, the fast measurement method of the absorption coefficient of the bottom - layer biological tissue based on a light - field camera in this implementation manner includes the following steps:

[0042] Step one: Simulate the reflected radiation signal

[0043] First, set n double - layer media with different optical - parameter distributions. These optical parameters usually include absorption coefficient, scattering coefficient, and refractive index, etc. Substitute these optical - parameter data sets into the time - domain radiation transfer equation and solve it. By solving the equation, the reflected radiation signals on the boundaries of the double - layer media under different optical - parameter conditions can be obtained, such as where i = 1, 2, 3, …, n;

[0044] Step two: Train the BP neural network model

[0045] Determine a BP (back - propagation) neural network model suitable for this problem. After initializing the weights and thresholds of the network, input them into an intelligent optimization algorithm (such as genetic algorithm or particle swarm optimization algorithm) to obtain the optimal network weights and thresholds, and then update these optimal weights and thresholds into the BP neural network;

[0046] Step 3: Forward propagation to calculate optical parameters

[0047] Input the dataset of the reflected radiation signal obtained in Step 1 into the updated BP neural network model for forward propagation. Through network calculation, the neural network will output the corresponding dataset of optical parameters, i.e., the predicted values of absorption coefficient, scattering coefficient, etc.;

[0048] Step 4: Error calculation and backpropagation

[0049] Calculate the error between the predicted optical parameters obtained in Step 3 and the corresponding actual optical parameters. Through the error backpropagation algorithm, update the weights and thresholds of the neural network to continuously optimize the neural network model and make it more accurate in predicting optical parameters;

[0050] Step 5: Iterative update until the accuracy requirement is met

[0051] Repeat the process of Step 3 to Step 4, update and adjust the weights and thresholds of the neural network until the prediction accuracy of the neural network model reaches the set requirement. At this time, the training process of the BP neural network model improved based on the intelligent algorithm is completed, and it can provide high-precision optical parameter prediction;

[0052] Step 6: Laser light source and signal acquisition

[0053] Turn on the short-pulse continuous light source 1. The white light is decomposed into monochromatic lights of different wavelengths by the beam splitter 2 and the variable neutral density attenuator 3, and then through the interference filter 4 composed of special filter slices, a pulsed laser at a specific wavelength is selected;

[0054] Step 7: Acquisition of reflected radiation signal

[0055] Vertically incident the pulsed laser through the optical fiber on the surface of the participating medium 7. The light field camera 5 is used to collect the reflected radiation signal on the surface of the participating medium at this wavelength. During this process, the position of the detection point is represented by d, and the direction of the detection signal is represented by f;

[0056] Step 8: Output optical parameters

[0057] Input the reflected radiation measurement signal obtained in Step 7 into the neural network model trained in Step 5. Through the forward propagation and prediction of the neural network, output the final optical parameters of the double-layer medium, especially the prediction results of absorption coefficient, scattering coefficient, etc.;

[0058] Preferably: The important component, the microlens array, in the light field camera is located at the focal length of the main lens of the light field camera. Its function is to image the light passing through the main lens again, so as to record the direction information of the light and obtain accurate optical signals during the measurement process.

[0059] Specific Embodiment 2: Combined with Figures 1-3 Describe this embodiment. The fast measurement method of the absorption coefficient of the underlying biological tissue based on the light field camera in this embodiment sets a double-layer medium system containing n groups of different optical parameter distributions, and its optical parameters are respectively: Among them, μ a1 represents the absorption coefficient of the top layer medium, μ a2 represents the absorption coefficient of the bottom layer medium, μ s1 represents the scattering coefficient of the top layer medium, μ s2 represents the scattering coefficient of the bottom layer medium;

[0060] These parameters are usually used to describe the optical properties of biological tissues. The top layer medium may be the epidermis or other biological tissues, and the bottom layer medium may represent deeper tissues or the bottom substrate. The optical parameters have different values in different media, and with the change of tissues, the scattering and absorption characteristics of light will also be different;

[0061] And substitute the above optical parameters into the time-domain radiative transfer equation for calculation, and finally obtain the reflected radiation signals at the boundaries of the medium to be measured under n groups of different optical parameters Take the reflected radiation signal obtained by solving the time-domain radiative transfer equation as a set of data X m , where X m corresponds to the optical parameter data Y m ;

[0062]

[0063] Pair the reflected radiation signal data X m with the corresponding optical parameter data Y m to form the data set of the BP neural network model, where m = 1, 2,... n, and the data set can be divided into 70% training data set and 30% test data set.

[0064] By simulating the reflected radiation signals under different optical parameter conditions through the time-domain radiative transfer equation and combining with the backpropagation neural network for training and prediction, the absorption coefficient of biological tissues can be measured efficiently and accurately.

[0065] Specific Embodiment 3: Combined with Figures 1-3 Describe this embodiment. The fast measurement method of the absorption coefficient of the underlying biological tissue based on the light field camera in this embodiment, the forward propagation is through the input layer receiving external data, then the propagation from the input layer to the hidden layer, and finally the propagation path from the hidden layer to the output layer;

[0066] Among them, the BP neural network model consists of an input layer, six hidden layers, and an output layer. Specifically, an input layer is connected to an output layer through six hidden layers in sequence, and the six hidden layers are also connected to each other. The input layer mainly consists of 4 neurons, each hidden layer mainly consists of 8 neurons, and the output layer mainly consists of 4 neurons.

[0067] The forward propagation is the process of information flow in the neural network, and the specific steps are as follows:

[0068] The input layer receives the reflected radiation signal as external data. This reflected signal is collected by the light field camera and contains the reflected light intensity information. The input layer transmits this data to the first hidden layer;

[0069] The input data is transmitted from the input layer to the first hidden layer. Each neuron performs a weighted sum with the input data and then undergoes a non-linear process through an activation function to generate the output of the hidden layer;

[0070] The output of the first hidden layer is transmitted to the second hidden layer and undergoes a weighted sum and non-linear transformation in this layer. Similarly, this process is sequentially transmitted between the subsequent hidden layers until the sixth hidden layer;

[0071] The output of the sixth hidden layer is finally transmitted to the output layer and generates the final prediction result after passing through the activation function, that is, the optical parameters (absorption coefficient, scattering coefficient, etc.) corresponding to the 4 neurons in the output layer;

[0072] The 4 neurons in the output layer give the finally predicted optical parameters. This prediction result is the absorption coefficient and scattering coefficient of the underlying biological tissue deduced by the neural network through the reflected radiation signal.

[0073] Specific Embodiment 4: Combined Figures 1-3 To illustrate this embodiment, in the backpropagation described in Step 4 of the fast measurement method for the absorption coefficient of the underlying biological tissue based on the light field camera in this embodiment, the gradient descent method and gradient search technology are used, with the goal of minimizing the mean square error between the actual output value and the expected output value.

[0074] Backpropagation is an effective algorithm for training multi-layer feedforward neural networks. Its core idea is to calculate the gradient of each weight and update the weights of the network according to these gradients to minimize the error between the actual output and the expected output. This process is achieved by means of forward propagation, error calculation, backpropagation of the error, and weight update, thereby gradually optimizing the performance of the neural network.

[0075] The specific application process of backpropagation is as follows:

[0076] The input data is obtained through a light field camera, usually the reflected data of an image or a signal. Before entering the neural network, the input data will undergo normalization processing to ensure consistent data ranges, thereby accelerating the training process;

[0077] In each training iteration, the input data generates a predicted output through the forward propagation of the neural network. Subsequently, by calculating the difference between the loss function and the expected output value, the backpropagation algorithm updates the weights of each layer according to the gradient to reduce the error;

[0078] By adopting appropriate gradient search techniques, the network can converge to the optimal solution more quickly, reducing the training time and improving the accuracy of absorption coefficient prediction. Especially when facing complex non-linear relationships, the gradient descent method and gradient search techniques can better find the global optimal solution or local optimal solution and avoid falling into local minima.

[0079] Specific Embodiment 5: Combining Figures 1-3 To illustrate this embodiment, for the fast measurement method of the absorption coefficient of the underlying biological tissue based on a light field camera in this embodiment, the convergence criterion of the BP neural network in Step 4 is by setting the maximum number of iterations T max When the number of iterations reaches T max stop training;

[0080] Define the BP neural network model function improved based on the intelligent algorithm as BP(x), then where Xtrain is the training data set of the reflected radiation signal, and Xtest is the test data set of the reflected radiation signal, and are the predicted values of the optical parameters output by the training data set and the test data set respectively. The prediction accuracy of the test model can be characterized by the following formula:

[0081]

[0082]

[0083] In the formula: and are the squared residual coefficients of the training data and the test data set respectively; and are the output average values of the training data set and the test data set respectively.

[0084] In practical applications, the following intelligent algorithms can be used to improve the BP neural network to improve its prediction accuracy and convergence speed;

[0085] Particle swarm optimization: Use the PSO algorithm to optimize the weights of the neural network, thereby accelerating network convergence and improving accuracy;

[0086] Genetic algorithm: Optimize the network structure and parameters through the genetic algorithm, enabling the neural network to achieve better performance in a shorter time;

[0087] Simulated annealing algorithm: Use the simulated annealing algorithm to jump out of the local optimal solution, search for the global optimal solution, and optimize the training process of the neural network.

[0088] Specific Embodiment Six: Combine Figures 1-3 To illustrate this embodiment, for the fast measurement method of the absorption coefficient of the underlying biological tissue based on the light field camera in this embodiment, the optical parameters of the double-layer medium obtained in Step 8 include the absorption coefficient of the top tissue, the scattering coefficient of the top tissue, the absorption coefficient of the bottom tissue, and the scattering coefficient of the bottom tissue.

[0089] By determining these parameters, the optical properties of biological tissues can be further understood, and important data can be provided for related medical imaging, optical diagnosis, or biophysical research.

[0090] Specific Embodiment Seven: Combine Figures 1-3 To illustrate this embodiment, for the fast measurement method of the absorption coefficient of the underlying biological tissue based on the light field camera in this embodiment, the expression of the time-domain radiative transfer equation in Step 1 is as follows:

[0091]

[0092] In the formula, c represents the speed of light in the medium, r represents the position, Ω represents the radiative transfer direction, t represents the time, represents the partial differential symbol, represents the Hamiltonian operator; I(r,Ω,t) is the radiative intensity at position r at time t along the direction Ω, μ a (r) and μ s (r) are the absorption coefficient and the scattering coefficient respectively, β(r) is the attenuation coefficient, and β(r) = μ a (r) + μ s (r), Φ w (Ω′,Ω) is the scattering phase function.

[0093] This equation describes the change of radiative intensity in the medium with time and spatial position, and takes into account absorption, scattering, and light propagation and interaction in different directions.

[0094] It should be noted that in the above embodiments, as long as the technical solutions are not contradictory, they can be arranged and combined. Those skilled in the art can exhaust all possibilities according to the mathematical knowledge of permutation and combination. Therefore, the present invention will no longer explain the technical solutions after permutation and combination one by one, but it should be understood that the technical solutions after permutation and combination have been disclosed by the present invention.

[0095] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A rapid measurement method of the absorption coefficient of underlying biological tissue based on a light field camera, characterized in that: The steps include: Step 1: Set up n double-layer media with different optical parameter distributions, substitute the optical parameter data set into the time-domain radiation transfer equation for solution, and obtain the reflected radiation signal of the double-layer medium boundary under n different optical parameters. Where i = 1, 2, 3, ..., n; Step 2: Determine the BP neural network model, input the model's initial network weights and thresholds into the intelligent algorithm, and then update them into the BP neural network model; Step 3: Input the data set of the reflected radiation signal in step 1 into the BP neural network model for forward propagation, and output the corresponding predicted optical parameter data set; Step 4: Calculate the error between the predicted optical parameter data set obtained in step 3 and the corresponding actual optical parameter data set, and perform back propagation based on the calculated error to update the weights and thresholds of the BP neural network model; Step 5: Repeat steps 3 to 4, and update and iterate until the BP neural network model reaches the prediction accuracy requirement. At this time, the training of the BP neural network model improved based on the intelligent algorithm is completed; Step 6: Turn on the short pulse continuous light source (1), decompose the white light into monochromatic light of different wavelengths through a beam splitter (2) and a variable neutral density attenuator (3), and use an interference filter (4) composed of special filters to obtain a pulsed laser; Step 7: The pulsed laser is incident vertically on the surface of the participating medium (7) through the optical fiber, and the light field camera (5) collects the reflected radiation signal M of the surface of the participating medium (7) of the corresponding wavelength d,f , d represents the corresponding detection point position, and f represents the direction of the detection radiation light signal; Step 8: Transform the reflected radiation measurement signal M d,f The data are input into the BP neural network model trained in step 5, and the final optical parameters of the double-layer medium are output through the forward propagation and prediction of the neural network.

2. The method for rapid measurement of the absorption coefficient of underlying biological tissue based on a light field camera according to claim 1 is characterized in that: Set n double-layer media with different optical parameter distributions, and their optical parameters are: …, Among them, μ a1 represents the absorption coefficient of the top layer medium, μ a2 represents the absorption coefficient of the underlying medium, μ s1 represents the scattering coefficient of the top layer medium, μ s2 represents the scattering coefficient of the underlying medium; And its optical parameters Substitute them into the time domain radiation transfer equation for calculation, and finally obtain the reflected radiation signal of the boundary of the medium to be measured under n groups of different optical parameters. The reflected radiation signal obtained by solving the time domain radiation transfer equation is taken as a set of data X m , where X m and optical parameter data Y m correspond; The reflected radiation signal data X m The corresponding optical parameter data Y m Paired, the data set that constitutes the BP neural network model, where m = 1, 2, ... n, the data set can be divided into 70% training data set and 30% test data set.

3. The method for rapid measurement of the absorption coefficient of underlying biological tissue based on a light field camera according to claim 1, characterized in that: The forward propagation described in step 3 is the propagation path from the input layer to the hidden layer, and finally from the hidden layer to the output layer. The BP neural network model in step three is composed of an input layer, six hidden layers and an output layer. Specifically, an input layer is connected to an output layer through six hidden layers in sequence, and the six hidden layers are also connected to each other; the input layer consists of 4 neurons, each hidden layer consists of 8 neurons, and the output layer consists of 4 neurons.

4. The method for rapid measurement of the absorption coefficient of underlying biological tissue based on a light field camera according to claim 1, characterized in that: The back propagation described in step 4 is an algorithm for training a multi-layer feedforward network based on error back propagation.

5. The method for rapid measurement of the absorption coefficient of underlying biological tissue based on a light field camera according to claim 1, characterized in that: The convergence criterion of the BP neural network in step 4 is to set the maximum number of iterations T max , when the number of iterations reaches T max Stop training when Define the BP neural network model function based on intelligent algorithm improvement as BP(x), then Among them, X train is the reflected radiation signal training data set, X test is the reflected radiation signal test data set, and These are the predicted values ​​of optical parameters output for the training data set and the test data set respectively. The prediction accuracy of the test model can be characterized by the following formula: Where: and are the squared residual coefficients of the training data and the test data set, respectively; and are the average outputs of the training dataset and the test dataset, respectively.

6. The method for rapid measurement of the absorption coefficient of underlying biological tissue based on a light field camera according to claim 1, characterized in that: The optical parameters of the double-layer medium obtained in step eight include the absorption coefficient of the top tissue, the scattering coefficient of the top tissue, the absorption coefficient of the bottom tissue and the scattering coefficient of the bottom tissue.

7. The method for rapid measurement of the absorption coefficient of underlying biological tissue based on a light field camera according to claim 1, characterized in that: The expression of the time domain radiation transfer equation in step 1 is as follows: In the formula, c represents the speed of light in the medium, r represents the position, Ω represents the direction of radiation transmission, and t represents the time. represents the partial differential symbol, represents the Hamiltonian operator; I(r,Ω,t) is the radiation intensity along the Ω direction at position r at time t, μ a (r) and μ s (r) are the absorption coefficient and scattering coefficient respectively, β(r) is the attenuation coefficient, and β(r)=μ a (r)+μ s (r), Φ w (Ω′,Ω) is the scattering phase function.

Images