A device for estimating the dynamics of a scattering object, a method for estimating the dynamics, a method for generating training data, a regression model, and a program.

Abstract

The type and characteristics of motion of scattering particles within the sample are estimated and quantified from the quantitative dynamic OCT signal. [Solution] The scattering dynamics estimation method (S1) includes a derivation step (S13) of deriving the logarithmic intensity variance of the measured dynamic OCT signal and the speed of the scattering body using a simulator that simulates the relationship between the movement of scattering bodies in biological tissue and the characteristic value of the dynamic OCT signal representing said movement, and an estimation step (S14) of applying the derived logarithmic intensity variance and speed to a regression model to estimate at least one of the speed of movement of the scattering body and the proportion of dynamic scattering bodies among the scattering bodies.

Description

[Technical Field]

[0001] The present invention relates to a device for estimating the dynamics of a scattering object, a method for estimating dynamics, a method for generating training data, a regression model, and a program. [Background technology]

[0002] In recent years, a technique called "OCT microscopy," which uses optical coherence tomography (OCT) to image in vitro / ex vivo samples, has been studied. However, OCT microscopy is generally a technique for morphological imaging and cannot image the function of biological tissues such as metabolism. In response to this, a signal analysis method called "dynamic OCT (DOCT)" has been proposed (see, for example, Non-Patent Document 1).

[0003] However, with DOCT, visualizing and observing rapid dynamics within biological tissues, such as necrosis and apoptosis (cell death), in three dimensions requires extremely long measurement times of several minutes or an ultra-high-speed OCT system. Therefore, visualizing and observing rapid dynamics within biological tissues is not practical using such techniques.

[0004] On the other hand, for example, Patent Document 1 discloses a method for calculating the time-varying characteristics of biological tissue based on two sets of signal values ​​acquired at different time intervals. This method performs statistical analysis of the time fluctuations of OCT signals, making it possible to evaluate samples with fast and slow dynamics using image acquisition with limited resources. [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2023-059461 [Non-patent literature]

[0006] [Non-Patent Document 1] El-Sadek et al., “Optical coherence tomography-based tissue dynamics imaging for longitudinal and drug response evaluation of tumor spheroids.” Biomedical Optics Express 11, 6231 (2020). [Overview of the project] [Problems that the invention aims to solve]

[0007] However, the conventional technologies described above do not directly provide the types of dynamics or dynamic parameters within biological tissues.

[0008] One aspect of the present invention aims to realize a technique for estimating and quantifying the type and characteristics of motion of scatterers within a measurement sample from quantitative dynamic OCT signals. [Means for solving the problem]

[0009] To solve the above problems, a method for estimating the dynamics of a scattering body according to one aspect of the present invention includes a derivation step of deriving the logarithmic intensity variance of the measured dynamic OCT signal and the speed of the scattering body using a simulator that simulates the relationship between the movement of a scattering body in biological tissue and the characteristic value of a dynamic OCT signal representing said movement, and an estimation step of applying the derived logarithmic intensity variance and the speed to a regression model to estimate at least one of the speed of movement of the scattering body and the proportion of dynamic scattering bodies among the scattering body.

[0010] To solve the above problems, a scattering dynamics estimation device according to one aspect of the present invention includes: a derivation unit that uses a simulator that simulates the relationship between the movement of scattering objects in biological tissue and the characteristic value of a dynamic OCT signal representing said movement to derive the logarithmic intensity variance of the measured dynamic OCT signal and the speed of the scattering objects; and an estimation unit that applies the derived logarithmic intensity variance and the speed to a regression model to estimate at least one of the speed of movement of the scattering objects and the proportion of dynamic scattering objects among the scattering objects.

[0011] To solve the above problems, a regression model according to one aspect of the present invention is a regression model trained to output at least one of the speed of motion of the scattering body and the proportion of dynamic scattering bodies among the scattering bodies, when the density of the scattering body, the type of motion of the scattering body, the log intensity variance and speed obtained from the dynamic OCT signal of the scattering body, the speed of motion of the scattering body, and the proportion of dynamic scattering bodies among the scattering body are input, using multiple sets of training data, each set consisting of six data points: density of scattering body, type of motion of the scattering body, log intensity variance and speed.

[0012] To solve the above problems, a method for generating training data for a regression model according to one aspect of the present invention is a method for generating training data for training the above regression model, comprising the steps of setting the density of scattering bodies in biological tissue, the type of motion, the speed of the motion, and the proportion of dynamic scattering bodies among the scattering bodies, and obtaining the logarithmic intensity variance and rapidity obtained from the dynamic OCT signals of the set scattering bodies by simulation.

[0013] Each aspect of the present invention may be implemented by a computer, in which case a control program for the dynamics estimation device that implements the dynamics estimation device by operating the computer as each part (software element) of the dynamics estimation device, and a computer-readable recording medium on which the program is recorded, also fall within the scope of the present invention.

Advantages of the Invention

[0014] According to one aspect of the present invention, it is possible to estimate and quantify the type and characteristics of the movement of scatterers in a measurement sample from a quantitative dynamic OCT signal.

Brief Description of the Drawings

[0015] [Figure 1] It is a block configuration diagram of a dynamics estimation device for estimating the dynamics of scatterers in a biological tissue according to Embodiment 1 of the present invention. [Figure 2] It is a flowchart of a dynamics estimation method for estimating the dynamics of scatterers in a biological tissue according to Embodiment 1 of the present invention. [Figure 3] It is a schematic diagram showing an outline of a method for training a machine model to generate a trained regression model. [Figure 4] It is a schematic diagram showing a method for generating a DOCT signal from an OCT signal. [Figure 5] It is a schematic diagram of a graph plotted with the LIV of each window on the vertical axis and the window time on the horizontal axis. [Figure 6] It is a schematic diagram when plotting the low-resolution LIV and the high-resolution LIV in the case where the scatterers have a single-direction flow. [Figure 7] It is a schematic diagram when plotting the low-resolution LIV and the high-resolution LIV in the case where the scatterers move in a random straight line. [Figure 8] It is an example of a three-dimensional plot showing the relationship between the rapidity of the DOCT signal and the speed (Sp) and DSR of the scatterers. [Figure 9] It is an example of a three-dimensional plot showing the relationship between the aLIV of the DOCT signal and the speed (Sp) and DSR of the scatterers. [Figure 10] It is a flowchart showing the flow of generation method S2 for generating training data for training a machine model using a simulator.

Modes for Carrying Out the Invention

[0016] [Embodiment 1] The following describes in detail one embodiment of the present invention. Figure 1 is a block diagram of a dynamics estimation device 1 for estimating the dynamics of scattering particles in biological tissue according to this embodiment. The dynamics estimation device 1 is a device that estimates the type and characteristics of the movement of scattering particles in biological tissue, i.e., the dynamics, by dynamically analyzing the OCT signal of biological tissue acquired using an optical coherence tomography (OCT) device. Therefore, the dynamics estimation device 1 may be used in combination with any optical coherence tomography device. Alternatively, it may be used independently of the optical coherence tomography device. In this case, the dynamics estimation device 1 may be configured to acquire an arbitrary OCT signal from an external source, for example via the Internet, analyze the OCT signal, and output the analysis results to an external source. Note that "OCT image" is an image of the intensity of each pixel of the "OCT signal," and in this embodiment, both descriptions have the same meaning.

[0017] As shown in Figure 1, the dynamics estimation device 1 comprises an acquisition unit 10, a derivation unit 11, an estimation unit 12, an input / output interface (input / output IF) 20, at least one processor 30, and at least one memory 40. The dynamics estimation device 1 may be connected to a DOCT simulator 50 (hereinafter also simply referred to as "simulator 50"), a regression model 60, a database 70, etc., in a communication-enabled manner.

[0018] The acquisition unit 10 acquires OCT signal data of the target biological tissue. For example, the acquisition unit 10 may acquire OCT signals from an external database 70 via the input / output IF 20. Alternatively, the acquisition unit 10 may acquire OCT signals stored in the memory 40, or it may acquire OCT signals transmitted from any optical coherence tomography device via the input / output IF 20.

[0019] The derivation unit 11 uses a simulator 50 that simulates the relationship between the movement of scatterers within biological tissue and the characteristic values ​​of the dynamic OCT signal (hereinafter referred to as "DOCT" or "DOCT signal") representing that movement to derive the logarithmic intensity variance (hereinafter also referred to as "LIV") of the measured DOCT signal and the swiftness of the scatterers. The DOCT signal is information obtained by analyzing the OCT signal. The DOCT signal obtained by analyzing the measured OCT signal is also referred to as the "measured DOCT signal." Details of the DOCT signal and LIV will be described later.

[0020] The estimation unit 12 applies the derived log-intensity variance and speed to the regression model 60 to estimate at least one of the following: the speed of motion of the scatterers and the proportion of dynamic scatterers among the scatterers (DSR, Dynamic Scatterer Ratio) (i.e., only the speed of motion of the scatterers, only the proportion of dynamic scatterers among the scatterers, or both). It is known that scatterers can consist of moving dynamic scatterers and stationary static scatterers. The DSR is the ratio of the number of dynamic scatterers to the total number of scatterers. The regression model 60 will be described later.

[0021] The input / output IF20 is an interface for sending and receiving data to and from the outside. Communication between the input / output IF20 and the outside may be performed, for example, via the internet. The input / output IF20 may be equipped with a short-range communication device such as Wi-Fi® or Bluetooth® that can connect wirelessly to an internet connection point. Alternatively, it may be a wired connection interface such as a USB connector.

[0022] The processor 30 can be configured using at least one general-purpose processor such as an MPU (Micro Processing Unit) or CPU (Central Processing Unit). The processor 30 may also include a dedicated processor composed of an ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array), or PLD (Programmable Logic Device).

[0023] The memory 40 may include multiple types of memory, such as ROM (Read Only Memory) and RAM (Random Access Memory). Furthermore, the memory 40 may include internal or external memory such as an HDD (Hard Disk Drive) or SSD (Solid State Drive). As an example, the processor 30 realizes the functions of the acquisition unit 10, the derivation unit 11, and the estimation unit 12 by loading various control programs recorded in the ROM of the memory 40 into the RAM and executing them.

[0024] The analysis method for DOCT signals according to this embodiment will now be described. First, logarithmic intensity variance (LIV) is the variance of the time fluctuations of the OCT signal within a predetermined time domain. More specifically, LIV is the time variance of the logarithm of the signal intensity of the OCT signal value. Since the time variance of the logarithm of the OCT signal intensity is calculated logarithmically, LIV is not affected by the signal intensity and is only affected by the time-varying component of the signal. The unit of LIV is, for example, dB. 2 That is the case.

[0025] The LIV analysis method will be explained with reference to the diagram. Figure 4 is a schematic diagram showing a method for generating a DOCT signal from an OCT signal. First, as shown at the top of Figure 4, multiple (e.g., 32) OCT images are acquired in a time sequence. Numerous subsets are created from the acquired multiple OCT images by changing the window time along the time series. The LIV (time dispersion) for each pixel of these subsets is calculated using the following equation (1). The horizontal axis is time, and the vertical axis is the intensity of the OCT signal (I dB Plotting (x, z, ti) yields a graph like the one shown at the bottom of Figure 4.

[0026]

number

[0027] Next, as shown in Figure 5, if we create a graph plotting the LIV of each window on the vertical axis and the window time on the horizontal axis, we obtain a curve graph that gradually approaches the saturation value. This curve can be approximated by the linear saturation function f(x) = a·exp(-τ / b), where a is the saturation log intensity variance (dB). 2 This is also called Authentic LIV (aLIV). In the following dynamics analysis, this saturated log intensity variance (aLIV) may be used instead of the log intensity variance (LIV). b is the time constant (s), and its reciprocal (1 / b) is called Swiftness. a and b can be determined by fitting the obtained curve graph with f(x). This curve graph has the characteristic that it saturates quickly when the movement of the scattering object being measured is large, but takes longer to saturate when the movement is slow.

[0028] Furthermore, the OCT correlation decay speed (OCDS) is the slope of the uncorrelated curve within a specific time delay range. The slope of the uncorrelated curve is obtained by linear regression of the uncorrelated curve. OCDS represents the speed of movement of the tissue being measured. A steep slope of the uncorrelated curve means that the correlation decreases rapidly, i.e., the tissue is moving quickly.

[0029] The inventors mathematically modeled the movement of various organs and other biological tissues, and further mathematically modeled how such movements generate OCT signals. Using these mathematical models, they created a simulator 50 that estimates the movement of biological tissue and the dynamic OCT signals it generates. Simulator 50 outputs the LIV of the generated dynamic OCT signal and the speed of the scatterers when the speed of movement of scatterers within the biological tissue, the proportion of dynamic scatterers among the scatterers, the type of movement of the scatterers, and the density of the scatterers are set as input. By using this simulator, it is possible to estimate the movement of biological tissue from the obtained OCT signals, and conversely, it is also possible to estimate the OCT signals generated from the assumed movement of biological tissue.

[0030] This section explains how to model the movement of biological tissues in simulations. First, we list the target tissues of the biological system. For example, proteins, amino acids, intracellular organelles, and cells. Next, we list the types of movement that primarily govern these tissues. For example, active movement, passive movement, cyclosis (cytoplasmic streaming), cell movement, and flow. These movements are categorized into three types: (1) random linear movement, (2) unidirectional flow, and (3) diffusive movement (and combinations thereof). (1) is the movement of each scattering object moving linearly in a random direction at the same speed, (2) is the movement of all scattering objects moving linearly in the same direction at the same speed, and (3) is movement due to diffusion.

[0031] This model was used to simulate how different movements of scattering bodies in biological tissue generate different OCT signals. The parameters input to the simulator are, as mentioned above, (1) the speed of scattering body movement, (2) the proportion of dynamic scattering bodies among the scattering bodies, (3) the type of scattering body movement, and (4) the density of scattering bodies. When these parameters are input to the simulator, the LIV of the DOCT signal and the speed of the scattering bodies are output.

[0032] Analysis of the simulation results revealed that a (Authentic LIV) corresponds to the proportion of scattering particles moving dynamically in the measured region, and Swiftness corresponds to the speed at which individual scattering particles (target tissue) are moving. Based on these findings, estimation method S1 for estimating the dynamics of scattering particles in biological tissue will be described. Figure 2 is a flowchart showing the flow of the dynamics estimation method S1 for scattering particles in biological tissue according to this embodiment. At least a part of the dynamics estimation method S1 can be performed using the dynamics estimation device 1.

[0033] As shown in Figure 2, the dynamics estimation method S1 includes steps S11 to S14. It is not necessary to perform steps S11 to S14 consecutively. For example, steps S11 and S12 may be performed in advance. Then, steps S13 and S14 may be performed after acquiring the actual OCT images.

[0034] Step S11 is the step of creating a simulator that outputs the log-intensity variance (LIV) and Swiftness of the scatterers of the dynamic OCT signal, given the speed of movement of scatterers in the biological tissue, the proportion of dynamic scatterers among the scatterers, the type of movement of the scatterers, and the density of the scatterers.

[0035] Step S12 is the step of creating a regression model that estimates the relationship between the speed of motion of the scatters, the proportion of dynamic scatters among the scatters, the type of motion of the scatters, and the density of the scatters, and LIV and speed.

[0036] Step S13 is a derivation step in which the log-intensity variance (LIV) of the measured dynamic OCT signal and the speed of the scatterers are derived using the above-mentioned simulator, that is, a simulator that simulates the relationship between the movement of scatterers in biological tissue and the characteristic values ​​of the dynamic OCT signal representing that movement. Derivation step S13 may be performed by the derivation unit 11 of the dynamics estimation device 1.

[0037] Step S14 is an estimation step in which the derived LIV and speed are applied to a regression model to estimate at least one of the speed of motion of the scatterers (Sp) and the proportion of the scatterers that are dynamic scatterers (DSR). Estimation step S14 may be performed by the estimation unit 12 of the dynamics estimation device 1.

[0038] Furthermore, by repeating the above simulation, numerous datasets of the four parameters mentioned above, along with their corresponding LIV and velocity, can be obtained. Using these datasets, a regression model can be created to estimate the relationship between the speed of scattering motion (nm / s), the proportion of dynamic scattering among the scattering bodies, the type of scattering motion, the density of the scattering bodies, LIV, and velocity. Using this regression model, the remaining four parameters can be estimated from the two data points of LIV and velocity. Two approaches are possible for the regression model: one is an analytical regression model obtained by a regression method based on an analytical model, and the other is a mechanical regression model obtained by deep learning.

[0039] First, let's explain the analytical regression model (analytical regression method). Generally, it is not possible to analytically determine four parameters from two data points. Therefore, we generate an analytical regression model using the following method. First, the type of motion of the scatterer is determined by the following method. In this embodiment, the types of motion of the scatterer to be determined are random linear motion and unidirectional flow, but are not limited to these. These two types of motion are determined by comparing the LIV of the measured dynamic OCT signal with the LIV of a low-resolution dynamic OCT signal obtained by reducing the resolution of the original OCT signal, and determining whether the LIV is resolution-dependent.

[0040] The inventors discovered that LIV depends on the resolution of the OCT image only when the scattering material flows in a unidirectional direction. Therefore, they generate a low-resolution dynamic OCT signal by holographic processing the acquired original OCT signal to reduce its resolution, and then determine whether LIV is resolution-dependent. Specifically, they generate a low-resolution OCT signal using equation (2) below, where x in equation (2) represents a vector.

[0041]

number

[0042] Equation (2) shows that the resolution is reduced by multiplying the original image signal by a complex Gaussian function. The product of the first and second terms on the left side is the signal of the original image, and σ in the denominator of the exponential function is the value that determines the resolution. The complex Gaussian function for blurring, which is the third term, is convolved into this. This calculation makes σ larger (reduces the resolution) as shown on the right side. When the LIV (Low resolution) obtained by such processing is plotted on top of the LIV (High resolution) of the original image, a graph like the one shown in Figure 6 or Figure 7 is obtained. Figure 6 is the graph obtained when the scatterer is flowing in a unidirectional direction as described above, and Figure 7 is the graph obtained when the scatterer is moving in a random linear direction as described above. In the case of unidirectional flow (Figure 6), the LIV value for a relatively high resolution (called HLIV) is larger than the LIV value for a relatively low resolution (called LLIV). In other words, when the scatterer is flowing in a unidirectional direction, HLIV becomes larger than LLIV depending on the resolution. When the scatterer undergoes random linear motion, the LIV (Likely Inertial Volume) is independent of the resolution, and the LIV value does not change significantly with resolution. This difference can be used to determine the type of motion of the scatterer.

[0043] The criteria for distinguishing between random linear motion and unidirectional flow can be set by the user based on various information. Examples of criteria include: (1) when the value of ΔLIV = HLIV - HLIV is negative, or (2) when 0 ≤ ΔLIV ≤ 2.5 dB. 2and HLIV > 15 dB 2 If so, it may be determined as random linear movement. Conversely, (3) ΔLIV > 2.5 dB 2 or (4) 0 ≤ ΔLIV ≤ 2.5 dB 2 and HLIV ≤ 15 dB 2 If so, it may be determined as a single - direction flow.

[0044] As a result of this determination, if the scatterer is determined to have a single - direction flow, it can be analyzed by the prior art. On the other hand, if it is determined as random linear movement, the following processing is performed.

[0045] First, in advance, by simulation, for each of Swiftness (rapidity) and aLIV (or the saturation value of LIV), the relationships with DSR and the speed of the scatterer (S p ) are obtained. As described above, DSR is the numerical ratio of the dynamic scatterer to the scatterer. According to the inventors' findings, it was found that the relationship between rapidity, the speed of the scatterer (S p ) and DSR can be approximated by a three - dimensional plane formula as shown in FIG. 8. Also, the relationship between aLIV and the speed of the scatterer (S p ) has a linear (straight - line) relationship as shown in FIG. 9, and the relationship between aLIV and DSR has a first - order saturation function relationship as shown in FIG. 9.

[0046] The above relationship formula of Swiftness (rapidity), the speed of the scatterer (S p ) and DSR (denoted by "R") is expressed by the following formula (3), and the relationship formula of aLIV, the speed of the scatterer (S p ) and R is expressed by the following formula (4). a, b, c, a', b', c' are coefficients (constants), and can be obtained by fitting the simulation numerical values to formulas (3) and (4).

Equation

[0047] Swiftness and aLIV can be obtained from the measured data using the method described above. Then, substitute a, b, c, a', b', c', Swiftness, and aLIV into equations (3) and (4) above and S p We then try to find R. However, we found that this system of equations cannot be solved analytically. Therefore, we use equation (5) below, where R is a function of a, and t is a parameter. R = at …(5)

[0048] Substituting equation (5) into equation (3) and rearranging, we obtain the following equation (6).

number

[0049] Substituting equation (6) above into equation (4) and rearranging, we obtain equation (7) below.

number

[0050] We can find t from equation (7) above, but equation (7) cannot be solved analytically. Therefore, we find t by numerical regression (fitting). Numerical regression can be performed using, for example, a known regression algorithm. Once t is found, we substitute it into equations (6) and (5) to get Swiftness(S p ) and R(DSR) can be calculated.

[0051] Note that the shapes of the Swiftness and aLIV graphs shown in Figures 8 and 9 differ slightly depending on the scattering density. Therefore, it is also possible to apply the scattering density to the regression model to estimate at least one of the speed of scattering motion and the proportion of dynamic scattering among the scattering particles. In other words, the graphs of Swiftness and aLIV for different scattering density cases should be obtained in advance by simulation. Then, the above analysis should be performed using Figures 8 and 9 corresponding to the scattering density of the target biological tissue. This method allows for a more accurate estimation of Swiftness(S p ) and R(DSR) can be calculated.

[0052] The scattering density value can be set by the user depending on the target. For example, the scattering density can be a value estimated using predetermined measured values. For instance, the scattering density can be estimated using a known density estimator that estimates density from OCT images. Alternatively, the scattering density may be set using a trained machine model that estimates scattering density.

[0053] Alternatively, known values ​​that have been estimated or measured in previous studies and clinical trials for different types of biological tissue may be used for the scattering material density. Conventionally, scattering material density has been measured for different types of biological tissue. Although there is some variation in the measured values ​​depending on the subject, the most reasonable value should be used.

[0054] Next, we will explain machine learning regression models (mechanical regression models) obtained through deep learning. A machine learning regression model is a trained regression model that estimates the relationships between six parameters: the speed of motion of a scatterer (Sp), the proportion of dynamic scatterers among the scatterers (DSR), the type of motion of the scatterer, the density of the scatterer, LIV, and speed. The type of machine model used for training is not limited. Any machine model can be used, such as a CNN model or a deep learning model with attention or transformer functions.

[0055] This regression model may be one that, given the density of the scatterer, the type of motion of the scatterer, the LIV, and the speed, is trained to output the most likely combination of Sp and DSR, or Sp only, or DSR only. Alternatively, it may be a regression model that, given Sp, DSR, the type of motion of the scatterer, and the density of the scatterer, outputs the most likely combination of LIV and speed.

[0056] For example, using multiple sets of training data, each set consisting of six data points—the density of scatterers in biological tissue, the type of movement of the scatterers, the log-scale intensity variance and speed obtained from the dynamic OCT signal of the scatterers, the speed of movement of the scatterers, and the proportion of dynamic scatterers among the scatterers—a regression model can be generated that, given the input of the density of scatterers, the type of movement of the scatterers, the log-scale intensity variance, and speed, outputs at least one of the speed of movement of the scatterers and the proportion of dynamic scatterers among the scatterers.

[0057] Figure 3 is a schematic diagram illustrating a method for training a machine model to generate a trained regression model. This training method involves training a pre-trained machine model M using a training dataset D to generate a trained regression model LM. The training dataset D includes training data D1, D2...Dn. Training data D1 includes six data points: data 1A of the speed of motion of scatterers in biological tissue, data 1B of the DSR of scatterers, data 1C of the type of motion of scatterers, data 1D of the density of scatterers, data 1E of the LIV obtained from the DOCT signal of biological tissue, and data 1F of the speed obtained from the DOCT signal. The training dataset D contains n sets of such training data.

[0058] When training a model, it is arbitrary what inputs to input and what outputs to train it to produce. For example, given data 1C, 1D, 1E, and 1F from the training data, the model may be trained to output the set of data 1A and 1B. By setting the inputs and outputs in this way and training the model, a desired trained machine model can be generated.

[0059] Alternatively, such a training dataset may be generated using a simulator 50. Figure 10 is a flowchart showing the flow of the generation method S2 for generating training data to train a machine model using a simulator. As shown in the figure, the training data generation method S2 includes steps S21 to S23.

[0060] Step S21 is a setting step in which the density of scatterers in the biological tissue, the type of motion, the speed of motion, and the proportion of dynamic scatterers among the scatterers are set. This setting step is performed by the user. Alternatively, the user may specify a numerical range for predetermined parameters, and a learning data generation device (not shown) may generate multiple parameter sets based on predetermined conditions.

[0061] Step S22 is an acquisition step in which the log-intensity variance and rapidity obtained from the dynamic OCT signal of the set scatterer are acquired by simulation. The acquisition step is performed by the user using a simulator. Alternatively, a learning data generation device (not shown) may input the parameter set set in step S21 into the simulator and acquire the data.

[0062] Step S23 is a generation step in which steps S21 and S22 are repeated to generate multiple datasets. Then, a machine model is trained using the generated datasets (not shown). The method for training the machine model using a large number of datasets is arbitrary. The step of training the machine model may be performed using a programmed learning device (not shown).

[0063] The above generation method S2 can generate the training data necessary to train a machine model. Then, by training the machine model with different input and output combinations, a trained machine model with the desired functionality can be generated.

[0064] According to the method S1 and dynamics estimation apparatus 1 of this embodiment, which describe the dynamics estimation of scattering bodies in biological tissue as described above, it is possible to estimate and quantify the type and characteristics of motion of scattering bodies in a measurement sample from quantitative dynamic OCT signals.

[0065] [Examples of implementation using software] The function of the dynamics estimation device 1 (hereinafter referred to as "the device") is a program that causes a computer to function as the device, and can be realized by a program that causes a computer to function as each control block of the device (particularly the acquisition unit 10, the derivation unit 11, and the estimation unit 12).

[0066] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., memory) as hardware for executing the program. By executing the program using this control device and storage device, the functions described in each of the embodiments are realized.

[0067] The above program may be recorded on one or more computer-readable recording media, not temporary ones. These recording media may or may not be provided by the above device. In the latter case, the program may be supplied to the above device via any wired or wireless transmission medium.

[0068] Furthermore, some or all of the functions of each of the above control blocks can also be realized by logic circuits. For example, an integrated circuit in which logic circuits functioning as each of the above control blocks are formed is also included in the scope of the present invention. In addition, it is also possible to realize the functions of each of the above control blocks by, for example, a quantum computer.

[0069] The present invention is not limited to the embodiments described above, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0070] (summary) (Aspect 1) A method for estimating the dynamics of a scattering body, comprising: a derivation step of deriving the logarithmic intensity variance of the measured dynamic OCT signal and the speed of the scattering body using a simulator that simulates the relationship between the movement of a scattering body in a biological tissue and the characteristic value of a dynamic OCT signal representing said movement; and an estimation step of applying the derived logarithmic intensity variance and the speed to a regression model to estimate at least one of the speed of movement of the scattering body and the proportion of dynamic scattering bodies among the scattering bodies.

[0071] (Aspect 2) The scattering motion type of the scattering body includes at least two types: random linear motion and unidirectional flow, and the two types of motion are determined by comparing the logarithmic intensity variance of the measured dynamic OCT signal with the logarithmic intensity variance of a low-resolution dynamic OCT signal obtained by reducing the resolution of the original OCT signal, and determining whether or not the logarithmic intensity variance is resolution-dependent, as described in Embodiment 1.

[0072] (Aspect 3) The method for estimating the dynamics of a scattering body according to embodiment 1 or 2, wherein the estimation step further includes applying the density of the scattering body to the regression model to estimate at least one of the speed of motion of the scattering body and the proportion of the scattering body that is dynamic.

[0073] (Aspect 4) The method for estimating the dynamics of a scattering body according to any one of embodiments 1 to 3, wherein the regression model is a regression model that estimates the relationship between the speed of motion of the scattering body, the proportion of dynamic scattering bodies among the scattering body, the type of motion of the scattering body, the density of the scattering body, the log-intensity variance, and the speed, and is an analytical regression model obtained by a regression method based on an analytical model, or a machine learning regression model obtained by deep learning.

[0074] (Appendix 5) The method for estimating the dynamics of a scattering body according to any one of embodiments 1 to 4, wherein the simulator is a simulator that outputs the logarithmic intensity variance of the generated dynamic OCT signal and the speed of the scattering body when the speed of motion of the scattering body, the proportion of the scattering body that is a dynamic scattering body, the type of motion of the scattering body, and the density of the scattering body are set and input.

[0075] (Aspect 6) The method for estimating the dynamics of a scattering body according to any one of embodiments 1 to 5, wherein the saturated log-intensity variance value of a graph obtained when the vertical axis is plotted with the log-intensity variance and the horizontal axis with the window time is used as the log-intensity variance.

[0076] (Aspect 7) The method for estimating the dynamics of a scattering body according to embodiment 2, wherein the low-resolution dynamic OCT signal is a signal obtained by holographic processing of the original OCT signal.

[0077] (Pattern 8) A scattering body dynamics estimation device, comprising: a derivation unit that uses a simulator to simulate the relationship between the movement of scattering bodies in biological tissue and the characteristic values ​​of a dynamic OCT signal representing said movement to derive the logarithmic intensity variance of the measured dynamic OCT signal and the speed of the scattering body; and an estimation unit that applies the derived logarithmic intensity variance and the speed to a regression model to estimate at least one of the speed of movement of the scattering body and the proportion of dynamic scattering bodies among the scattering bodies.

[0078] (Aspect 9) A regression model trained to output at least one of the speed of motion of the scattering body and the proportion of dynamic scattering bodies among the scattering bodies, when the density of the scattering body, the type of motion of the scattering body, the log-intensity variance and speed obtained from the dynamic OCT signal of the scattering body, the speed of motion of the scattering body, and the proportion of dynamic scattering bodies among the scattering body are inputs, using multiple sets of training data, each set consisting of six data points: density of scattering body, type of motion of the scattering body, log-intensity variance and speed obtained from the dynamic OCT signal of the scattering body, the speed of motion of the scattering body, and the proportion of dynamic scattering bodies among the scattering body.

[0079] (Aspect 10) A method for generating training data for training a regression model described in Embodiment 9, comprising the steps of: setting the density of scattering bodies in a biological tissue, the type of motion, the speed of the motion, and the proportion of dynamic scattering bodies among the scattering bodies; and obtaining the logarithmic intensity variance and rapidity obtained from the dynamic OCT signals of the set scattering bodies by simulation.

[0080] (Aspect 11) An estimation program for causing a computer to function as a scattering dynamics estimation device according to embodiment 8, comprising the derivation unit and the estimation program for causing a computer to function as the estimation unit.

[0081] (Aspect 12) A computer-readable non-temporary recording medium that records the estimation program described in Embodiment 11. [Explanation of Symbols]

[0082] 1. Dynamics Estimation Device 10...Acquisition part 11...Derivation part 12... Estimation part 20. Input / Output Interface 30 Processors 40...memory 50... Simulator 60. Regression Model 70... Database

Claims

1. A derivation step is performed to derive the logarithmic intensity variance of the measured dynamic OCT signal and the speed of the scatterer, using a simulator that simulates the relationship between the movement of scatterers within biological tissue and the characteristic values ​​of the dynamic OCT signal representing that movement. An estimation step of applying the derived log-intensity variance and the speed to a regression model to estimate at least one of the speed of motion of the scatterer and the proportion of the scatterer that is dynamic, A method for estimating the dynamics of a scattering body, including [specific details omitted].

2. The method for estimating the dynamics of a scattering body according to claim 1, wherein the type of motion of the scattering body includes at least two types: random linear motion and unidirectional flow, and the two types of motion are determined by comparing the logarithmic intensity variance of the measured dynamic OCT signal with the logarithmic intensity variance of a low-resolution dynamic OCT signal obtained by reducing the resolution of the original OCT signal, and determining whether or not the logarithmic intensity variance is resolution-dependent.

3. The method for estimating the dynamics of a scattering body according to claim 1, wherein the estimation step further includes applying the density of the scattering body to the regression model to estimate at least one of the speed of motion of the scattering body and the proportion of the scattering body that is dynamic.

4. The method for estimating the dynamics of a scattering body according to any one of claims 1 to 3, wherein the regression model is a regression model that estimates the relationship between the speed of motion of the scattering body, the proportion of dynamic scattering bodies among the scattering body, the type of motion of the scattering body, the density of the scattering body, the log-intensity variance, and the speed, and is an analytical regression model obtained by a regression method based on an analytical model, or a machine learning regression model obtained by deep learning.

5. The method for estimating the dynamics of a scattering body according to any one of claims 1 to 3, wherein the simulator is a simulator that outputs the logarithmic intensity variance of the generated dynamic OCT signal and the speed of the scattering body when the speed of motion of the scattering body, the proportion of the scattering body that is a dynamic scattering body among the scattering body, the type of motion of the scattering body, and the density of the scattering body are set and input.

6. The method for estimating the dynamics of a scattering body according to any one of claims 1 to 3, wherein the saturated log-intensity variance of a graph obtained when the vertical axis is plotted with the log-intensity variance and the horizontal axis with the window time is used as the log-intensity variance.

7. The method for estimating the dynamics of a scattering body according to claim 2, wherein the low-resolution dynamic OCT signal is a signal obtained by holographic processing of the original OCT signal.

8. A derivation unit that uses a simulator to simulate the relationship between the movement of scatterers within biological tissue and the characteristic values ​​of the dynamic OCT signal representing that movement, to derive the logarithmic intensity variance of the measured dynamic OCT signal and the speed of the scatterers, An estimation unit that applies the derived log-intensity variance and the speed to a regression model to estimate at least one of the speed of motion of the scatterer and the proportion of the scatterer that is dynamic. A device for estimating the dynamics of scattering materials, including [specific component / feature].

9. A regression model trained to output at least one of the speed of motion of the scattering body and the proportion of dynamic scattering bodies among the scattering bodies, when the density of the scattering body, the type of motion of the scattering body, the logarithmic intensity variance and speed obtained from the dynamic OCT signal of the scattering body, the speed of motion of the scattering body, and the proportion of dynamic scattering bodies among the scattering body are input, using multiple sets of training data, each set consisting of six data points: density of scattering body within biological tissue, type of motion of the scattering body, logarithmic intensity variance and speed obtained from the dynamic OCT signal of the scattering body, the speed of motion of the scattering body, and the proportion of dynamic scattering bodies among the scattering body.

10. A method for generating training data for training the regression model described in claim 9, A step of setting the density of scattering particles within the biological tissue, the type of movement, the speed of the movement, and the proportion of dynamic scattering particles among the scattering particles, A step of obtaining the logarithmic intensity variance and rapidity obtained from the dynamic OCT signal of the set scattering body by simulation, A method for generating training data, including [the specified data].

11. An estimation program for causing a computer to function as a scatterer dynamics estimation device according to claim 8, comprising the derivation unit and the estimation program for causing a computer to function as the estimation unit.

12. A computer-readable non-temporary recording medium that stores the estimation program described in claim 11.

Images