A detection method and device for artificial tissues and organoids

Abstract

The present invention discloses a detection method and device for artificial tissues and organoids. The present invention uses a deep learning network for artificial tissue and organoid segmentation to achieve the segmentation of artificial tissues and organoids, and obtains three-dimensional images of artificial tissue and organoid regions; its labels are images after contrast enhancement; morphological indexes and ATP values of artificial tissues and organoids in each culture well of the corresponding culture plate on the last day of culture are calculated according to connected components; a covariance matrix is calculated for the matrix constructed after standardizing the morphological indexes and ATP values of artificial tissues and organoids; the covariance matrix is subjected to singular value decomposition to obtain a right singular matrix and a singular value matrix; principal components are extracted according to the variance contribution rate to obtain comprehensive indexes of artificial tissues and organoids. The comprehensive indexes of artificial tissues and organoids are used for the study of drug efficacy.

Description

Technical Field

[0001] The present invention belongs to the field of artificial tissue and organ detection, and particularly relates to a detection method and device for artificial tissues and organoids based on OCT. Background Art

[0002] Artificial tissues and organoids are 3D cell complexes formed by human-derived cells or stem cells driven by self-assembly, which can simulate the structure and function of human organs and can be used for optimized screening of drugs, pathological exploration, etc. During the cultivation of artificial tissues and organoids or when screening drugs for artificial tissues and organoids with specific physiological structures, the morphological indicators of artificial tissues or organoids are closely related to their functional states. The growth and development status of artificial tissues or organoids can be detected through high-resolution continuous morphological observation. However, due to the small characteristic morphological structures of artificial tissues and organoids and the insignificant differences from the surrounding media, as the cultivation time extends, it is impossible to observe the changes in their characteristic morphological structures with the naked eye. Therefore, it is necessary to compare and reflect the growth of artificial tissues and organoids through the segmentation and quantitative analysis of characteristic morphological structures to study the therapeutic effect of a certain drug on a disease.

[0003] Traditional detection methods are all invasive, such as histological staining, fluorescence microscopy, bright-field microscopy, etc., but they cannot evaluate the true state of cells and do not have the ability to image volume structures and quantify volume morphological characteristics. Because these methods detect artificial tissues or organoids based on planar information or approximate artificial tissues or organoids as spheres or ellipsoids for quantitative analysis. However, the growth of organoids is irregular, which will cause large errors. At the same time, the imaging penetration depth of these methods is limited, making it difficult to perform deep tissue imaging.

[0004] Optical Coherence Tomography (OCT) technology has the advantages of non-destructive, non-invasive, high resolution, and its imaging method is volume imaging, and the imaging depth can reach several millimeters. The present invention detects the growth process of artificial tissues and organoids by using OCT, combines the characteristic morphological indicators of artificial tissues and organoids, such as the overall volume, total surface area, average volume, average surface area, etc. of the characteristic structures of artificial tissues or organoids and the indicator adenosine triphosphate (ATP) representing the activity of artificial tissues or organoids, combines the morphological indicators with ATP through principal component analysis to obtain comprehensive indicators, and uses the comprehensive indicators of artificial tissues or organoids to study the drug efficacy. Summary of the Invention

[0005] The first object of the present invention is to propose a detection method for artificial tissues and organoids based on OCT in view of the deficiencies of the prior art. In order to use OCT to characterize the growth of artificial tissues and organoids, morphological indexes of the characteristic structures of artificial tissues and organoids, such as total volume, total surface area, average volume, average surface area, etc., and the index adenosine triphosphate (ATP) representing the activity of artificial tissues and organoids are combined to establish a comprehensive index for detecting artificial tissues and organoids, so as to realize the detection of the growth and extinction of artificial tissues and organoids by OCT over time.

[0006] A detection method for artificial tissues and organoids based on OCT, comprising the following steps:

[0007] Step (1): Use an OCT device to collect the original three-dimensional grayscale images of artificial tissues and organoids;

[0008] Open the system software of the OCT device, set the acquisition data format and save path, and select the 3D acquisition mode. Set the OCT field of view size, single pixel size, and scanning speed according to the diameter of the well of the artificial tissue and organoid culture plate. The well to be detected should be within the field of view, and the pixels are set according to the required image clarity and data size. The refractive index is adjusted according to the material of the actual acquisition object. Before collecting data, adjust the reference arm, sample arm, and light intensity adjustment button to achieve the purpose of high signal-to-noise ratio, the artificial tissue or organoid culture well located in the center of the field of view, and the clearest image display. At the same time, adjust the angle of the well plate to eliminate the influence of fringe noise. The placement position of the lens is referenced to the ground. When the lens is facing up, it is called reverse acquisition, and when the lens is facing down, it is called forward acquisition. Since the artificial tissue or organoid grows against the bottom of the well plate, reverse acquisition is used in the present invention to shorten the distance between the lens and the sample and make the image clearer. When starting data acquisition, use the preview function to observe the three-dimensional image of the acquired data under the acquisition scheme, and judge whether the acquisition scheme meets the expectations. If it meets the expectations, perform the acquisition; otherwise, adjust the acquisition parameters.

[0009] Step (2): Construct a labeled two-dimensional image or three-dimensional image data set, and divide the data set into a training set and a test set;

[0010] The specific process of constructing the labeled two-dimensional image is to split the original three-dimensional grayscale image of the artificial tissue or organoid into several original two-dimensional grayscale images, and then enhance the contrast of these two-dimensional grayscale images to obtain the corresponding two-dimensional grayscale images after contrast enhancement; the original two-dimensional grayscale images are labeled and used as the two-dimensional image data set, where the label is the two-dimensional grayscale image corresponding to the original two-dimensional grayscale image after contrast enhancement;

[0011] The process of constructing the tagged three-dimensional image specifically involves splitting the original three-dimensional grayscale image of the artificial tissue or organoid into several original two-dimensional grayscale images, and then performing contrast enhancement on these two-dimensional grayscale images to obtain the corresponding two-dimensional grayscale images after contrast enhancement; then combining the two-dimensional grayscale images after contrast enhancement to obtain a three-dimensional image after contrast enhancement; tagging the original three-dimensional grayscale image as a three-dimensional image dataset, where the tag is the three-dimensional image after contrast enhancement corresponding to the original three-dimensional grayscale image.

[0012] Preferably, the contrast enhancement of these two-dimensional grayscale images specifically involves processing these two-dimensional grayscale images using one or more of the attenuation coefficient (Depth-resolved model-based reconstruction of attenuation coefficients), logarithmic intensity variance algorithm (Logarithmic intensity variance (LIV)), late OCT correlation decay speed algorithm (Late OCT correlation decay speed (OCDSl)), inverse power law exponent α algorithm of the speckle fluctuation spectrum (The robustness of the power law exponent, α), motility amplitude M algorithm based on autocorrelation (motility amplitude, M), inter-frame complex correlation algorithm (complex-correlation algorithm), and traditional image preprocessing algorithms.

[0013] The traditional image preprocessing algorithms include denoising, image grayscale transformation, etc.

[0014] Preferably, to save the acquisition and reconstruction time of OCT contrast-enhanced images and meet the requirements of deep learning network training convergence at the same time, the original two-dimensional grayscale images can also be data-augmented and then contrast-enhanced.

[0015] More preferably, the data augmentation is one or more of the methods of combining distortion augmentation and non-distortion augmentation for the image, image rotation method, and image scale transformation method.

[0016] Step (3): Construct a deep learning network for artificial tissue or organoid segmentation, and train it using the training set; finally, use the test set to test and verify the trained deep learning network for artificial tissue or organoid segmentation.

[0017] The deep learning network for artificial tissue or organoid segmentation uses EG-Net, ResNet50v2, VGG19, Xception, and DenseNet121 CNNs. Its input is a two-dimensional grayscale image, and its output is a two-dimensional grayscale image with enhanced contrast. Finally, all the two-dimensional grayscale images with enhanced contrast corresponding to the original three-dimensional grayscale image are combined;

[0018] The deep learning network for artificial tissue or organoid segmentation uses Resnet-3D. Its input is the original three-dimensional grayscale image, and its output is a three-dimensional image with enhanced contrast;

[0019] Step (4): Use the tested and verified deep learning network for artificial tissue or organoid segmentation to achieve artificial tissue or organoid segmentation, and obtain a three-dimensional image of the artificial tissue or organoid region;

[0020] Step (5): Quantify the three-dimensional image of the artificial tissue or organoid region obtained in step (4):

[0021] Calculate the morphological indexes and ATP values of the artificial tissues or organoids in each culture well of the culture plate corresponding to the last day of culture according to the connected components. The morphological indexes are: total volume, total surface area, total sphericity, number, average volume, average surface area, and average sphericity. Standardize the morphological indexes and ATP values of the artificial tissues or organoids to eliminate the influence of dimensions. Calculate the covariance for the standardized matrix to obtain the covariance matrix. Perform singular value decomposition on the covariance matrix to obtain the right singular matrix and the singular value matrix sorted in descending order of singular values. Extract the principal components according to the variance contribution rate, extract the principal components with a cumulative variance contribution rate greater than 85%, use the column vectors of the right singular matrix as the coefficients of a certain principal component expression, and the ratio of this singular value to the total singular value as the coefficient of the comprehensive index, and finally obtain the comprehensive index of the artificial tissue or organoid.

[0022] Specifically:

[0023] 5-1 Based on the three-dimensional image of the artificial tissue or organoid region segmented in step (4), obtain the number of pixels Volume_number occupied by the volume of a single artificial tissue or organoid and the number of pixels Surface_number occupied by the surface area in the three-dimensional image; then, according to the three-dimensional pixel size and two-dimensional pixel size, and combining formulas (1)-(2), respectively obtain the volume Organoid_volume and surface area Organoid_surface of a single artificial tissue or organoid;

[0024] Organoid_volume = Volume_number * (PixelSize_x * PixelSize_y * PixelSize_z) Formula (1)

[0025] Organoid_surface = surface_number * (PixelSize_x * PixelSize_y) Equation (2)

[0026] PixelSize_x = PixelSize_y = PixelSize_z

[0027] where PixelSize_x, PixelSize_y, and PixelSize_z respectively represent the sizes of the pixel points in the original three-dimensional grayscale image in the x-axis, y-axis, and z-axis directions;

[0028] The number of pixels Volume_number occupied by the single artificial tissue or organoid volume, and the number of pixels Surface_number occupied by the surface area are obtained by calculating the three-dimensional image of the artificial tissue or organoid region based on the three-dimensional connected domain, that is, by using the bwlabeln function in MATLAB to label the connected regions.

[0029] 5-2 According to the above single artificial tissue or organoid volume Organoid_volume and surface area Organoid_surface, combined with Equation (3), the sphericity of the single artificial tissue or organoid is obtained:

[0030]

[0031] 5-3 Calculate the total number of artificial tissues or organoids in a single culture well based on the three-dimensional connected domain of the three-dimensional image of the artificial tissue or organoid region:

[0032] Organoid_number = Connected_domains_number Equation (4)

[0033] where Connected_domains_number represents the number of connected domains in the three-dimensional image;

[0034] Preferably, the equivalent diameter of the connected domain is greater than or equal to 32 microns;

[0035] Calculate the total volume of the artificial tissue or organoid in a single culture well:

[0036]

[0037] where Organoid_volume(i) represents the volume of the i-th artificial tissue or organoid in a single culture well;

[0038] Calculate the total surface area of the artificial tissue or organoid in a single culture well:

[0039]

[0040] where Organoid_surface(i) represents the surface area of the i-th artificial tissue or organoid in a single culture well;

[0041] Calculate the total sphericity of the artificial tissues or organoids in a single culture well:

[0042]

[0043] where Organoid_spherecity(i) represents the sphericity of the i-th artificial tissue or organoid in a single culture well;

[0044] 5-4 Calculate the average volume of a single artificial tissue or organoid in a certain culture well based on the total volume of the artificial tissues or organoids in a single culture well:

[0045] Organoid_average_volume = Organoid_sum_volume / Organoid_number Equation (8)

[0046] Calculate the average surface area of a single artificial tissue or organoid in a certain culture well based on the total surface area of the artificial tissues or organoids in a single culture well:

[0047] Organoid_average_surface = Organoid_sum_surface / Organoid_number Equation (9)

[0048] Calculate the average sphericity of a single artificial tissue or organoid in a certain culture well based on the total sphericity of the artificial tissues or organoids in a single culture well:

[0049] Organoid_average_spherecity

[0050] = Organoid_sum_spherecity / Organoid_number

[0051] Equation (10)

[0052] 5-5 For each culture well of the culture plate corresponding to the last day of culture, obtain the adenosine triphosphate value Organoid_ATP using the chemiluminescence measurement method;

[0053] 5-6 Obtain the standardized artificial tissue or organoid indicators in each culture well, including total volume, total surface area, total sphericity, total number, average volume, average surface area, average sphericity, ATP;

[0054]

[0055] Among them, x1 represents the total volume of the standardized artificial tissue or organoid in the j-th culture well; Organoid_sum_volume(j) represents the total volume of the artificial tissue or organoid in the j-th culture well; m represents the number of culture wells.

[0056]

[0057] Among them, x2 represents the total surface area of the standardized artificial tissue or organoid in the j-th culture well; Organoid_sum_surface(j) represents the total surface area of the artificial tissue or organoid in the j-th culture well;

[0058]

[0059] Among them, x3 represents the total sphericity of the standardized artificial tissue or organoid in the j-th culture well; Organoid_sum_spherecity(j) represents the total sphericity of the artificial tissue or organoid in the j-th culture well;

[0060]

[0061] Among them, x4 represents the total number of the standardized artificial tissue or organoid in the j-th culture well; Organoid_number(j) represents the total number of the artificial tissue or organoid in the j-th culture well;

[0062]

[0063] Among them, x5 represents the average volume of the standardized artificial tissue or organoid in the j-th culture well; Organoid_average_volume(j) represents the average volume of the artificial tissue or organoid in the j-th culture well;

[0064]

[0065] Among them, x6 represents the average surface area of the standardized artificial tissue or organoid in the j-th culture well; Organoid_average_surface(j) represents the average surface area of the artificial tissue or organoid in the j-th culture well;

[0066]

[0067] Among them, x7 represents the average sphericity of the standardized artificial tissue or organoid in the j-th culture well; Organoid_average_spherecity(j) represents the average sphericity of the artificial tissue or organoid in the j-th culture well;

[0068]

[0069] where x8 is the standardized ATP value of the artificial tissue or organoid in the j-th culture well; Organoid_ATP(j) represents the ATP value of the artificial tissue or organoid in the i-th culture well;

[0070] The reason for standardization is that the dimensions of each parameter are inconsistent, in order to eliminate the influence of dimensions.

[0071] 5-7 Perform principal component analysis based on the above-obtained standardized artificial tissue or organoid indicators; specifically

[0072] 5-7-1 Construct matrix A from the standardized artificial tissue or organoid indicators, that is:

[0073] A = [x1, x2, x3, x4, x5, x6, x7, x8] m×8 Equation (19)

[0074] where m is the number of culture wells, here m > 8, that is, the number of data groups, and matrix A is a matrix with m rows and 8 columns.

[0075] 5-7-2 Calculate the covariance matrix B for matrix A:

[0076]

[0077] where cov() represents the covariance function, and the covariance matrix B is a matrix with 8 rows and 8 columns.

[0078] 5-7-3 Use singular value decomposition (SVD) to calculate the singular values and eigenvectors of the covariance matrix B:

[0079] B = U∑V T Equation (21)

[0080] where matrices U and V represent the left and right singular matrices respectively, both of which are orthogonal matrices. The column vectors of V and U are the basis vectors of the row space and column space of the covariance matrix B respectively, and ∑ represents a diagonal matrix containing singular values.

[0081] B T B = V∑ 2 V T Equation (22)

[0082] The right singular matrix V uses the right singular matrix of SVD to reduce the dimensionality of the number of columns. The first k eigenvectors of PCA are the first k columns of V; solve V for formula (22) using linear algebra knowledge, that is B T The matrix composed of the eigenvectors of B is the V matrix in SVD, B T The matrix composed of the eigenvalues of B is ∑ 2Matrix

[0083] Since B T B is a square matrix and can be eigen-decomposed; the obtained eigenvalues and eigenvectors need to satisfy:

[0084] (B T B)v k =λ k v k Equation (23)

[0085] where λ k is the eigenvalue and v k is the eigenvector corresponding to λ k , k = 1, 2,..., 7; ∑ represents a diagonal matrix containing singular values, and ∑ 2 represents a diagonal matrix containing eigenvalues, that is, the singular value is the square root of the eigenvalue

[0086] The eigenvalue matrix ∑ 2 is sorted in descending order of eigenvalues, then the singular value matrix ∑ is sorted in descending order of singular values, and at the same time, the columns of V also change with the sorting.

[0087] The sorted ∑ 2 matrix, ∑ matrix and V matrix are specifically as follows:

[0088]

[0089]

[0090]

[0091] Formula (26) can also be expressed as:

[0092]

[0093] where λ k is the eigenvalue, λ k >λ k+1 (k = 1, 2,..., 7). The eigenvalue λ1 is the largest eigenvalue in the ∑ 2 matrix, and its corresponding eigenvector v1 = [a 11 , a 21 , a 31 , a 41 , a 51 , a 61 , a 71 , a 81 T needs to be arranged in the first column of the V matrix; the eigenvalue λ8 is the smallest eigenvalue in the ∑ 2 matrix, and its corresponding eigenvector v8 = [a​18 , a 28 , a 38 , a 48 , a 58 , a 68 , a 78 , a 88 T It needs to be arranged in the last column of the V matrix.

[0094] Select the principal components according to the singular value proportion greater than the threshold α (which can be 85%). The singular value proportion is called the variance contribution rate. The variance contribution rate b1 of the first principal component and the variance contribution rate b2 of the second principal component are as follows:

[0095]

[0096]

[0097] b1 + b2 > threshold α, Equation (30)

[0098] The size and activity of the artificial tissue or organoid are composed of the first principal component F1:

[0099] F1 = -(a 11 *x1 + a 21 *x2 + a 31 *x3 + a 41 *x4 + a 51 *x5 + a 61 *x6 + a 71 *x7 + a 81 *x8), Equation (31)

[0100] The phenotypic characteristics and quantity of the artificial tissue or organoid are composed of the second principal component F2:

[0101] F2 = -(a 12 *x1 + a 22 *x2 + a 32 *x3 + a 42 *x4 + a 52 *x5 + a 62 *x6 + a 72 *x7 + a 82 *x8), Equation (32) The comprehensive index Organoid_F of the artificial tissue or organoid is composed of the size and activity of the artificial tissue or organoid, and the phenotypic characteristics and quantity:

[0102] Organoid_F = b1 * F1 + b2 * F2, Equation (33)

[0103] ​The second object of the present invention is to provide an OCT-based artificial tissue or organoid detection device, comprising:

[0104] An image acquisition module for receiving three-dimensional grayscale images of artificial tissues or organoids collected by an OCT device;

[0105] An artificial tissue or organoid segmentation module for segmenting three-dimensional grayscale images of artificial tissues or organoids by using a deep learning network for artificial tissue or organoid segmentation to obtain three-dimensional images of artificial tissue or organoid regions;

[0106] A first calculation module for calculating the morphological indexes and ATP values of artificial tissues or organoids in each culture well of the culture plate corresponding to the last day of cultivation;

[0107] A second calculation module for calculating the covariance matrix of the matrix constructed after standardizing the morphological indexes and ATP values of artificial tissues or organoids, and then performing singular value decomposition on it to obtain a right singular matrix and a singular value matrix; extracting the principal components according to the variance contribution rate to construct a comprehensive index of artificial tissues or organoids.

[0108] The third object of the present invention is to provide a drug sensitivity test method for judging the efficacy of artificial tissues or organoids under different drug concentrations by using the comprehensive index of artificial tissues or organoids obtained by the above method.

[0109] By plotting the curve of the change of the comprehensive index of artificial tissues or organoids over time under the action of different drugs, it is possible to study the problem of at what concentration a single drug is most effective and at what concentration drug resistance will occur, and it is also possible to study the problem of which combination of combined drugs has the best therapeutic effect.

[0110] The beneficial effects of the present invention are:

[0111] 1. To achieve the purpose of shortening the segmentation time of artificial tissues or organoids, the present invention proposes a method of using a contrast enhancement algorithm based on attenuation coefficient to construct a training set instead of manual annotation and automatically performing segmentation through a deep learning network to shorten the calculation time.

[0112] 2. The present invention uses OCT to detect the growth process of artificial tissues or organoids, and constructs a covariance matrix by calculating the covariance matrix of the morphological indexes of artificial tissues or organoids and the index adenosine triphosphate (ATP) representing the activity of artificial tissues or organoids. After standardizing the above parameters, singular value decomposition is performed to obtain the right singular matrix and the singular value matrix. The principal components are extracted according to the variance contribution rate, and finally the comprehensive index of artificial tissues or organoids is obtained. The comprehensive index of artificial tissues or organoids is used to study the drug efficacy, and then precise treatment is realized. For example, the treatment of artificial tissues or organoids with different concentrations of 5-fluorouracil (5-FU) is studied to screen out the optimal concentration to achieve the best therapeutic effect; or the combination form of combined drugs is studied, such as the combination of FOLFOX (oxaliplatin, calcium folinate, fluorouracil). By designing different drug combination methods, the most suitable drug combination scheme for the treatment of this artificial tissue or organoid is found. BRIEF DESCRIPTION OF THE DRAWINGS

[0113] Figure 1 is the flowchart of the method of the present invention;

[0114] Figure 2 is the OCT cross-sectional view of the artificial tissue or organoid of colorectal cancer; (a) the original image collected by OCT; (b) the image after contrast enhancement using the attenuation coefficient algorithm;

[0115] Figure 3 is the logarithmic intensity variance comparison diagram; (a) the original image collected by OCT; (b) the image after contrast enhancement using the logarithmic intensity variance;

[0116] Figure 4 is the three-dimensional reconstruction diagram of the artificial tissue or organoid of colorectal cancer; (a) the original image collected by OCT; (b) the manually labeled segmentation diagram; (c) the segmentation diagram using the attenuation coefficient.

[0117] Figure 5 is the schematic diagram of the single pixel size of the artificial tissue or organoid; (a) the three-dimensional schematic diagram of a single artificial tissue or organoid; (b) the schematic diagram of the single pixel size in three-dimensional coordinates; (c) the schematic diagram of the single pixel size in two-dimensional coordinates x-o-y; (d) the schematic diagram of the single pixel size in two-dimensional coordinates x-o-z; (e) the schematic diagram of the single pixel size in two-dimensional coordinates y-o-z. DETAILED DESCRIPTION OF THE INVENTION

[0118] The following further analyzes the present invention in combination with specific embodiments.

[0119] A method for quantifying artificial tissues or organoids based on OCT, see Figure 1 , including the following steps:

[0120] Step (1): Use an OCT device to acquire the original three-dimensional grayscale image of artificial tissue or organoids;

[0121] Turn on the OCT system, set the file format for collecting data and the saving path. The file format is set according to the grayscale data to be collected. Select the acquisition mode in the software interface. The modes include 2D, 3D, Doppler, and speckle. In this embodiment, the 3D mode is selected. In the 3D mode, set the field of view size, pixels, and scanning speed for acquisition. The detected hole should be within the field of view. The pixels are set according to the required image clarity and data size. The scanning speed is defaulted to 48 kHz. Ascan and Bscan do not need to be collected repeatedly and are both set to 1. The refractive index is set to 1. Before collecting data, adjust the reference arm, sample arm, and light intensity adjustment buttons to achieve the purpose of high signal-to-noise ratio, the artificial tissue or organoid culture hole being in the center of the field of view, and the clearest image display. At the same time, adjust the angle of the well plate by about 5° to eliminate the influence of fringe noise. Use reverse acquisition to collect data on the artificial tissue or organoids as shown in Figure 2 (a), Figure 3 (a), Figure 4 (a).

[0122] Step (2): Construct a labeled two-dimensional image or three-dimensional image dataset, and divide the dataset into a training set and a test set;

[0123] The process of constructing the labeled two-dimensional image is specifically to split the original three-dimensional grayscale image of artificial tissue or organoids into several original two-dimensional grayscale images, and then enhance the contrast of these two-dimensional grayscale images to obtain the corresponding contrast-enhanced two-dimensional grayscale images; label the original two-dimensional grayscale images as the two-dimensional image dataset, where the label is the corresponding contrast-enhanced two-dimensional grayscale image of the original two-dimensional grayscale image;

[0124] The process of constructing the labeled three-dimensional image is specifically to split the original three-dimensional grayscale image of artificial tissue or organoids into several original two-dimensional grayscale images, and then enhance the contrast of these two-dimensional grayscale images to obtain the corresponding contrast-enhanced two-dimensional grayscale images; then combine the contrast-enhanced two-dimensional grayscale images to obtain a contrast-enhanced three-dimensional image; label the original three-dimensional grayscale image as the three-dimensional image dataset, where the label is the corresponding contrast-enhanced three-dimensional image of the original three-dimensional grayscale image;

[0125] The contrast enhancement of these two-dimensional grayscale images specifically involves processing these two-dimensional grayscale images using one or more of the depth-resolved model-based reconstruction of attenuation coefficients, logarithmic intensity variance (LIV), late OCT correlation decay speed (OCDSl), the robustness of the power law exponent, α, motility amplitude, M, complex-correlation algorithm, and traditional image preprocessing algorithms;

[0126] The advantage of the depth-resolved model-based reconstruction of attenuation coefficients is that it converts the intensity value at each pixel of the OCT X-Z section into an attenuation coefficient without the need to segment to determine these attenuation coefficients. The logarithmic intensity variance algorithm is one of the algorithms for dynamic imaging and can display some physiological information inside artificial tissues or organoids. Logarithmic intensity variance is essentially the temporal variance of quickly obtaining an OCT signal sequence.

[0127] According to the above two algorithms, the attenuation coefficients are calculated for the collected data of artificial tissues or organoids of colorectal cancer in the X-Z plane. The comparison between the original image and the attenuation coefficient image is shown in Figure 2 . Figure 2 (a) In the original image shown, the artificial tissue or organoid and the Matrigel are relatively similar. If manual annotation is directly performed in Figure 2 (a), it will also produce a large error. However, in Figure 2 (b), this problem is solved and the influence of background noise is suppressed through the calculation of the average value.

[0128] Similarly, after calculation by the logarithmic intensity variance, Figure 3 also effectively differentiates the artificial tissue or organoid and the Matrigel, with the effect of contrast enhancement. Figure 3 (b) has a stronger contrast compared to Figure 3 (a), and the artificial tissue or organoid is clearer.

[0129] The image data of the attenuation coefficient after contrast enhancement is reconstructed three-dimensionally and compared with the result of manual annotation to obtain Figure 4 . From Figure 4(a) and (b) show that through the deep learning network, high precision has been achieved. At the same time Figure 4 (b) and Figure 4 The 3D reconstruction maps obtained by constructing the training set with manual annotation and attenuation coefficient in (c) also have strong similarity. Therefore, the training set provided by the attenuation coefficient can be comparable to manual annotation. Using the tic and toc functions in MATLAB to count the running time, taking the artificial tissue or organoids of colorectal cancer as an example, the following table is drawn.

[0130] Table 1: Comparison of Time Consumption

[0131] Method Manual annotation Attenuation coefficient calculation Time-consuming 10 min 3.046566s

[0132] It can be seen from Table 1 that the calculation time of the attenuation coefficient is much less than that of manual annotation. By this method, the consumption of human and material resources can be greatly reduced.

[0133] Step (3): Construct a deep learning network for artificial tissue or organoid segmentation, and use the training set for training; finally, use the test set to test and verify the trained deep learning network for artificial tissue or organoid segmentation;

[0134] The deep learning network for artificial tissue or organoid segmentation adopts EG-Net, ResNet50v2, VGG19, Xception, and DenseNet121 CNNs. Its input is a two-dimensional grayscale image, and the output is a two-dimensional grayscale image after contrast enhancement; finally, all the two-dimensional grayscale images after contrast enhancement corresponding to the original three-dimensional grayscale image are combined;

[0135] The deep learning network for artificial tissue or organoid segmentation adopts Resnet-3D. Its input is the original three-dimensional grayscale image, and the output is a three-dimensional image after contrast enhancement;

[0136] Step (4): Use the deep learning network for artificial tissue or organoid segmentation after test verification to achieve artificial tissue or organoid segmentation, and obtain the three-dimensional image of the artificial tissue or organoid region;

[0137] Step (5): Quantify the three-dimensional image of the artificial tissue or organoid region obtained in step (4):

[0138] Calculate the morphological indexes and ATP values of artificial tissues or organoids in each culture well of the culture plate on the last day of culture according to the connected domain. The morphological indexes are: total volume, total surface area, total sphericity, number, average volume, average surface area, and average sphericity. Standardize the morphological indexes and ATP values of the artificial tissues or organoids to eliminate the influence of dimensions. Calculate the covariance of the standardized matrix to obtain the covariance matrix. Perform singular value decomposition on the covariance matrix to obtain the right singular matrix and the singular value matrix sorted in descending order of singular values. Extract the principal components according to the variance contribution rate, extract the principal components with a cumulative variance contribution rate greater than 85%, use the column vectors of the right singular matrix as the coefficients of a certain principal component expression, and the ratio of the singular value to the total singular value as the coefficient of the comprehensive index, and finally obtain the mathematical expression of the comprehensive index of the artificial tissue or organoid.

[0139] Specifically:

[0140] 5-1 Based on the three-dimensional image of the artificial tissue or organoid region segmented in step (5), first establish a three-dimensional space coordinate system for a single artificial tissue or organoid as shown in Figure 5 (a), and obtain the number of pixels Volume_number occupied by the volume of a single artificial tissue or organoid and the number of pixels Surface_number occupied by the surface area in the three-dimensional image; then calculate the volume and surface area of the artificial tissue or organoid according to the pixel size. The pixel size is the parameter set in the original three-dimensional gray image of the artificial tissue or organoid collected by the OCT device in step (1). The size of a single pixel in the three-dimensional coordinate system of the artificial tissue or organoid volume is as shown in Figure 5 (b). The size of a single pixel in the two-dimensional coordinate system of the artificial tissue or organoid surface area is as shown in Figure 5 (c), (d), (e). Combine formulas (1)-(2) to obtain the volume Organoid_volume and surface area Organoid_surface of a single artificial tissue or organoid respectively;

[0141] Organoid_volume = Volume_number * (PixelSize_x * PixelSize_y * PixelSize_z) Formula (1)

[0142] Organoid_surface = surface_number * (PixelSize_x * PixelSize_y) Formula (2)

[0143] PixelSize_x = PixelSize_y = PixelSize_z = 5μm

[0144] The number of pixels occupied by the volume of a single artificial tissue or organoid, Volume_number, and the number of pixels occupied by the surface area, Surface_number, are obtained by calculating the three-dimensional image of the artificial tissue or organoid region based on the three-dimensional connected domain, that is, by using the bwlabeln function in MATLAB to label the connected regions.

[0145] 5-2 According to the volume of a single artificial tissue or organoid, Organoid_volume, and the surface area, Organoid_surface, the sphericity of a single artificial tissue or organoid is obtained by combining formula (3):

[0146]

[0147] 5-3 Calculate the total number of artificial tissues or organoids in a single culture well based on the three-dimensional connected domain of the three-dimensional image of the artificial tissue or organoid region:

[0148] Organoid_number = Connected_domains_number Formula (4)

[0149] where Connected_domains_number represents the number of connected domains in the three-dimensional image;

[0150] Preferably, the equivalent diameter of the connected domain is greater than or equal to 32 microns;

[0151] Calculate the total volume of the artificial tissue or organoids in a single culture well:

[0152]

[0153] where Organoid_volume(i) represents the volume of the i-th artificial tissue or organoid in a single culture well;

[0154] Calculate the total surface area of the artificial tissue or organoids in a single culture well:

[0155]

[0156] where Organoid_surface(i) represents the surface area of the i-th artificial tissue or organoid in a single culture well;

[0157] Calculate the total sphericity of the artificial tissue or organoids in a single culture well:

[0158]

[0159] where Organoid_spherecity(i) represents the sphericity of the i-th artificial tissue or organoid in a single culture well;

[0160] 5-4 Calculate the average volume of a single artificial tissue or organoid in a culture well according to the total volume of artificial tissues or organoids in a single culture well:

[0161] Organoid_average_volume = Organoid_sum_volume / Organoid_number Equation (8)

[0162] Calculate the average surface area of a single artificial tissue or organoid in a culture well according to the total surface area of artificial tissues or organoids in a single culture well:

[0163] Organoid_average_surface = Organoid_sum_surface / Organoid_number Equation (9)

[0164] Calculate the average sphericity of a single artificial tissue or organoid in a culture well according to the total sphericity of artificial tissues or organoids in a single culture well:

[0165] Organoid_average_spherecity = Organoid_sum_spherecity / Organoid_number Equation (10)

[0166] 5-5 Obtain the adenosine triphosphate value Organoid_ATP for each culture well of the culture plate on the last day of culture using a chemiluminescence measurement method;

[0167] 5-6 Obtain the standardized artificial tissue or organoid indicators in each culture well, including total volume, total surface area, total sphericity, total number, average volume, average surface area, average sphericity, and average ATP;

[0168]

[0169] Among them, x1 represents the standardized total volume of artificial tissues or organoids in the j-th culture well; Organoid_sum_volume(j) represents the total volume of artificial tissues or organoids in the j-th culture well; m represents the number of culture wells.

[0170]

[0171] Among them, x2 represents the standardized total surface area of artificial tissues or organoids in the j-th culture well; Organoid_sum_surface(j) represents the total surface area of artificial tissues or organoids in the j-th culture well;

[0172]

[0173] Among them, x3 represents the total sphericity of the standardized artificial tissue or organoid in the j-th culture well; Organoid_sum_spherecity(j) represents the total sphericity of the artificial tissue or organoid in the j-th culture well;

[0174]

[0175] Among them, x4 represents the total number of the standardized artificial tissue or organoid in the j-th culture well; Organoid_number(j) represents the total number of the artificial tissue or organoid in the j-th culture well;

[0176]

[0177] Among them, x5 represents the average volume of the standardized artificial tissue or organoid in the j-th culture well; Organoid_average_volume(j) represents the average volume of the artificial tissue or organoid in the j-th culture well;

[0178]

[0179] Among them, x6 represents the average surface area of the standardized artificial tissue or organoid in the j-th culture well; Organoid_average_surface(j) represents the average surface area of the artificial tissue or organoid in the j-th culture well;

[0180]

[0181] Among them, x7 represents the average sphericity of the standardized artificial tissue or organoid in the j-th culture well;

[0182] Organoid_average_spherecity(j) represents the average sphericity of the artificial tissue or organoid in the j-th culture well;

[0183]

[0184] Among them, x8 is the ATP value of the standardized artificial tissue or organoid in the j-th culture well; Organoid_ATP(j) represents the ATP value of the artificial tissue or organoid in the j-th culture well;

[0185] The reason for standardization is that the dimensions of each parameter are inconsistent, in order to eliminate the influence of dimensions.

[0186] 5-6 According to the above-obtained standardized indicators of artificial tissue or organoid, principal component analysis is carried out; specifically

[0187] 5-7-1 A matrix A is composed of the standardized indicators of artificial tissue or organoid, that is:

[0188] A = [x1, x2, x3, x4, x5, x6, x7, x8] m×8 Equation (19)

[0189] Where m is the number of culture wells, here m = 21, and matrix A is a matrix with m rows and 8 columns.

[0190] 5-7-2 Calculate the covariance matrix B for matrix A:

[0191]

[0192] Where cov() represents the covariance function, and the covariance matrix B is a matrix with 8 rows and 8 columns.

[0193] 5-7-3 Use singular value decomposition (SVD) to calculate the singular values and eigenvectors of the covariance matrix B:

[0194] B = U∑V T Equation (21)

[0195] Where matrices U and V represent the left and right singular matrices respectively, both of which are orthogonal matrices. The column vectors of V and U are the basis vectors of the row space and column space of the covariance matrix B respectively, and ∑ represents a diagonal matrix containing the singular values.

[0196] B T B = V∑ 2 V T Equation (22)

[0197] The right singular matrix V uses the right singular matrix of SVD to reduce the dimensionality of the number of columns. The first k eigenvectors of PCA are the first k columns of V; solve V for formula (22) using linear algebra knowledge, that is, B T The matrix composed of the eigenvectors of B is the V matrix in SVD, B T The matrix composed of the eigenvalues of B is ∑ 2 matrix.

[0198] Since B T B is a square matrix, it can be eigen-decomposed; the obtained eigenvalues and eigenvectors need to satisfy:

[0199] (B T B)v k = λ k v k Equation (23)

[0200] Where λ k is the eigenvalue, v k is the eigenvector corresponding to λ k , k = 1, 2,..., 7; ∑ represents a diagonal matrix containing the singular values, ∑ 2represents a diagonal matrix containing eigenvalues, i.e., the singular values are the square roots of the eigenvalues

[0201] Eigenvalue matrix ∑ 2 Sorted by eigenvalues from largest to smallest, the singular value matrix ∑ is sorted by singular values from largest to smallest, and at the same time the columns of V also change with the sorting.

[0202] Sorted ∑ 2 The matrix, ∑ matrix, and V matrix are specifically as follows:

[0203]

[0204]

[0205]

[0206] a 11 = -0.4520, a 21 = -0.4285, a 31 = -0.0175, a 41 = 0.0303, a 51 = -0.4470, a 61 = 0.4518, a 71 = -0.2502, a 81 = -0.3800;

[0207] a 12 = -0.0294, a 22 = -0.1214, a 32 = -0.6146, a 42 = -0.6118, a 52 = 0.1408, a 62 = 0.1089, a 72 = 0.4266, a 82 = 0.1370;

[0208] Formula (26) can also be expressed as:

[0209]

[0210] where λ k is the eigenvalue, λ k > λ k+1 (k = 1, 2,..., 7). The eigenvalue λ1 is the largest eigenvalue in the ∑ 2 matrix, and its corresponding eigenvector v1 = [a 11 a 21 a 31 a 41 a 51, a 61 , a 71 , a 81 T needs to be arranged in the first column of the V matrix; the eigenvalue λ8 is the smallest eigenvalue in the ∑ 2 matrix, and its corresponding eigenvector v8 = [a 18 , a 28 , a 38 , a 48 , a 58 , a 68 , a 78 , a 88 T needs to be arranged in the last column of the V matrix.

[0211] Select the principal components according to the singular value ratio greater than the threshold α (which can be 85%). The singular value ratio is called the variance contribution rate. The variance contribution rate b1 of the first principal component and the variance contribution rate b2 of the second principal component are as follows:

[0212]

[0213]

[0214] b1 + b2 > threshold α Equation (30)

[0215] The size and activity of the artificial tissue or organoid occupied by the first principal component F1 are:

[0216] F1 = -(a 11 *x1 + a 21 *x2 + a 31 *x3 + a 41 *x4 + a 51 *x5 + a 61 *x6 + a 71 *x7 + a 81 *x8)

[0217] = 0.4520 * x1 + 0.4285 * x2 + 0.0175 * x3 - 0.0303 * x4 + 0.4470 * x5 - 0.4518 * x6 + 0.2502 * x7 + 0.3800 * x8 Equation (31)

[0218] The phenotypic characteristics and quantity of the artificial tissue or organoid formed by the second principal component F2 are:

[0219] F2 = -(a 12 *x1 + a 22 *x2 + a 32 *x3 + a 42 *x4 + a 52 *x5 + a 62 ​​*x6 + a 72 *x7 + a 82 *x8)

[0220] = 0.0294 * x1 + 0.1214 * x2 + 0.6146 * x3 + 0.6118 * x4 - 0.1408 * x5 - 0.1089 * x6 + 0.4266 * x7 - 0.1370 * x8 Equation (32)

[0221] There are medium positive loadings on the standardized variables x1, x2, x5, x8 in the first principal component. Therefore, F1 can constitute the size and activity of the space occupied by artificial tissues or organoids. And there are medium positive loadings on the standardized variables x3, x4, x7 in the second principal component. Therefore, F2 can constitute the phenotypic characteristics and quantity of artificial tissues or organoids.

[0222] The comprehensive index Organoid_F of artificial tissues or organoids is composed of the size and activity of the space occupied by artificial tissues or organoids, and the phenotypic characteristics and quantity:

[0223] Organoid_F = b1 * F1 + b2 * F2 = 0.5760 * F1 + 0.3216 * F2 Equation (33)

[0224] The contribution rate of the first principal component is 57.60%, and the contribution rate of the second principal component is 32.16%. The cumulative contribution rate of the first two principal components reaches 89.7655%, that is, these 2 principal components can represent 89.7655% of the information of 8 indicators.

[0225] The research on the drug efficacy is carried out through the comprehensive index obtained above, and then precision treatment is realized. For example, study the treatment of artificial tissues or organoids with different concentrations of 5-fluorouracil (5-FU), and screen out the optimal concentration to achieve the best therapeutic effect; or study the combination form of combined drugs, such as the combination of FOLFOX (oxaliplatin, calcium folinate, fluorouracil), and find the most suitable drug combination plan for the treatment of this artificial tissue or organoid by designing different drug combination methods.

Claims

1. A detection method for artificial tissues and organoids, characterized in that It includes the following steps: Step (1): Use an OCT device to collect the original three-dimensional grayscale images of artificial tissues or organoids; Step (2): Construct a labeled two-dimensional image or three-dimensional image dataset, and divide the dataset into a training set and a test set; The process of constructing the labeled two-dimensional image is specifically to split the original three-dimensional grayscale image of the artificial tissue or organoid into several original two-dimensional grayscale images, and then perform contrast enhancement on these two-dimensional grayscale images to obtain the corresponding contrast-enhanced two-dimensional grayscale images; The original two-dimensional grayscale images are labeled as the two-dimensional image dataset, where the label is the corresponding contrast-enhanced two-dimensional grayscale image of the original two-dimensional grayscale image; The process of constructing the labeled three-dimensional image is specifically to split the original three-dimensional grayscale image of the artificial tissue or organoid into several original two-dimensional grayscale images, and then perform contrast enhancement on these two-dimensional grayscale images to obtain the corresponding contrast-enhanced two-dimensional grayscale images; Then combine the contrast-enhanced two-dimensional grayscale images to obtain a contrast-enhanced three-dimensional image; The original three-dimensional grayscale image is labeled as the three-dimensional image dataset, where the label is the corresponding contrast-enhanced three-dimensional image of the original three-dimensional grayscale image; Step (3): Construct a deep learning network for artificial tissue or organoid segmentation, and use the training set for training; Finally, use the test set to test and verify the trained deep learning network for artificial tissue or organoid segmentation; The deep learning network for artificial tissue or organoid segmentation uses EG-Net, ResNet50v2, VGG19, Xception or DenseNet121 CNN, its input is a two-dimensional grayscale image, and the output is a contrast-enhanced two-dimensional grayscale image; Finally, combine all the contrast-enhanced two-dimensional grayscale images corresponding to the original three-dimensional grayscale image; The deep learning network for artificial tissue or organoid segmentation uses Resnet-3D, its input is the original three-dimensional grayscale image, and the output is a contrast-enhanced three-dimensional image; Step (4): Use the deep learning network for artificial tissue or organoid segmentation after test verification to achieve artificial tissue or organoid segmentation, and obtain a three-dimensional image of the artificial tissue or organoid region; Step (5): Quantify the three-dimensional image of the artificial tissue or organoid region obtained in step (4): Calculate the morphological indexes and ATP values of the artificial tissues or organoids in each culture well of the culture plate corresponding to the last day of culture according to the connected components. The morphological indexes are: total volume, total surface area, total sphericity, number, average volume, average surface area, average sphericity; Calculate the covariance matrix of the matrix constructed after standardizing the morphological indexes and ATP values of the artificial tissues or organoids; Perform singular value decomposition on the covariance matrix to obtain a right singular matrix and a singular value matrix sorted in descending order of singular values; Extract the main components according to the variance contribution rate, use the column vectors of the right singular matrix as the coefficients of a certain main component expression, and the proportion of the singular value in the total singular value as the coefficient of the comprehensive index, and finally obtain the comprehensive index of the artificial tissue or organoid; Among them, the principal components are extracted according to the variance contribution rate, the column vectors of the right singular matrix are used as the coefficients of a certain principal component expression, and the proportion of the singular value in the total singular value is used as the coefficient of the comprehensive index, and finally the comprehensive index of the artificial tissue or organoid is obtained. Specifically: The principal components are selected according to the proportion of the singular value greater than the threshold α. The proportion of the singular value is called the variance contribution rate. The variance contribution rate b1 of the first principal component and the variance contribution rate b2 of the second principal component are as follows: where σ k represents the k-th singular value; b1 + b2 > threshold α Equation (30) The size and activity of the space occupied by the artificial tissue or organoid are composed of the first principal component F1: F1 = -(a 11 * x1 + a 21 * x2 + a 31 * x3 + a 41 * x4 + a 51 * x5 + a 61 * x6 + a 71 * x7 + a 81 * x8) Equation (31) where x1, x2, x3, x4, x5, x6, x7, x8 respectively represent the total volume, total surface area, total sphericity, total number, average volume, average surface area, average sphericity, and ATP value of the standardized artificial tissue or organoids in the j-th culture well; the eigenvalue λ1 is the largest eigenvalue in the ∑ 2 matrix of singular value decomposition, and the corresponding eigenvector is v1 = [a 11, a 21, a 31, a 41, a 51, a 61, a 71, a 81 T ;​ The phenotypic characteristics and quantity of the artificial tissue or organoid are composed of the second principal component F2: F2 = -(a 12 * x1 + a 22 * x2 + a 32 * x3 + a 42 * x4 + a 52 * x5 + a 62 * x6 + a 72 * x7 + a 82 * x8) Equation (32) where the eigenvalue λ8 is the smallest eigenvalue in the Σ matrix of the singular value decomposition, and the corresponding eigenvector is v8 = [a a a a a a a] 2 18, a 28, a 38, a 48, a 58, a 68, a 78， a 88 T ;​​ Furthermore, the calculation of the comprehensive index Organoid_F of the artificial tissue or organoid is as follows: Organoid_F = b1 * F1 + b2 * F2 Equation (33).

2. The method according to claim 1, characterized in that The specific contrast enhancement of these two-dimensional grayscale images is to perform one or more of the attenuation coefficient algorithm, logarithmic intensity variance algorithm, delay-related attenuation speed algorithm, inverse power-law exponent α algorithm of the speckle fluctuation spectrum, motion amplitude M algorithm based on autocorrelation, inter-frame complex correlation algorithm, and image preprocessing algorithm on these two-dimensional grayscale images.

3. The method according to claim 1, characterized in that A data enhancement step is added before the contrast enhancement of the original two-dimensional grayscale images.

4. The method according to claim 3, wherein The data enhancement is one or more of the methods of combining distortion enhancement and non-distortion enhancement, image rotation method, and image scale transformation method for the image.

5. The method according to claim 1, wherein Step (5) is specifically: 5-1 According to the three-dimensional image of the artificial tissue or organoid region segmented in step (4), obtain the number of pixels Volume_number occupied by the volume of a single artificial tissue or organoid and the number of pixels Surface_number occupied by the surface area in the three-dimensional image; then according to the size of a single pixel in the three dimensions and the size of a single pixel in the two dimensions, combine formulas (1)-(2) to obtain the volume Organoid_volume and surface area Organoid_surface of a single artificial tissue or organoid respectively; Organoid_volume = Volume_number * (PixelSize_x * PixelSize_y * PixelSize_z) Equation (1) Organoid_surface = surface_number * (PixelSize_x * PixelSize_y) Equation (2) PixelSize_x = PixelSize_y = PixelSize_z Among them, PixelSize_x, PixelSize_y, and PixelSize_z respectively represent the sizes of the pixel points in the x-axis, y-axis, and z-axis directions of the original three-dimensional grayscale image; 5-2 According to the volume Organoid_volume and surface area Organoid_surface of the single artificial tissue or organoid above, combine formula (3) to obtain the sphericity of the single artificial tissue or organoid: 5-3 Calculate the total number of artificial tissues or organoids in a single culture well based on the three-dimensional connected domains for the three-dimensional image of the artificial tissue or organoid region: Organoid_number = Connected_domains_number Equation (4) where Connected_domains_number represents the number of connected domains in the three-dimensional image; Calculate the total volume of the artificial tissue or organoids in a single culture well: where Organoid_volume(i) represents the volume of the i-th artificial tissue or organoid in a single culture well; Calculate the total surface area of the artificial tissue or organoids in a single culture well: where Organoid_surface(i) represents the surface area of the i-th artificial tissue or organoid in a single culture well; Calculate the total sphericity of the artificial tissue or organoids in a single culture well: where Organoid_spherecity(i) represents the sphericity of the i-th artificial tissue or organoid in a single culture well; 5-4 Calculate the average volume of a single artificial tissue or organoid in a certain culture well based on the total volume of the artificial tissue or organoids in a single culture well: Organoid_average_volume = Organoid_sum_volume / Organoid_number Equation (8) Calculate the average surface area of a single artificial tissue or organoid in a certain culture well based on the total surface area of the artificial tissue or organoids in a single culture well: Organoid_average_surface = Organoid_sum_surface / Organoid_number Equation (9) Calculate the average sphericity of a single artificial tissue or organoid in a certain culture well based on the total sphericity of the artificial tissue or organoids in a single culture well: Organoid_average_spherecity = Organoid_sum_spherecity / Organoid_number Equation (10) 5-5 For each culture well of the culture plate corresponding to the last day of culture, obtain the adenosine triphosphate value Organoid_ATP using the chemiluminescence measurement method; 5-6 Obtain the standardized artificial tissue or organoid indicators in each culture well, including total volume, total surface area, total sphericity, total number, average volume, average surface area, average sphericity, ATP; where x1 represents the standardized total volume of the artificial tissue or organoids in the j-th culture well; Organoid_sum_volume(j) represents the total volume of the artificial tissue or organoids in the j-th culture well; m represents the number of culture wells; where x2 represents the standardized total surface area of the artificial tissue or organoids in the j-th culture well; Organoid_sum_surface(j) represents the total surface area of the artificial tissue or organoids in the j-th culture well; Where x3 represents the total normalized sphericity of artificial tissues or organoids in the j-th culture well; Organoid_sum_spherecity(j) represents the total sphericity of artificial tissues or organoids in the j-th culture well; Where x4 represents the total normalized number of artificial tissues or organoids in the j-th culture well; Organoid_number(j) represents the total number of artificial tissues or organoids in the j-th culture well; Where x5 represents the average normalized volume of artificial tissues or organoids in the j-th culture well; Organoid_average_volume(j) represents the average volume of artificial tissues or organoids in the j-th culture well; Where x6 represents the average normalized surface area of artificial tissues or organoids in the j-th culture well; Organoid_avergae_surface(j) represents the average surface area of artificial tissues or organoids in the j-th culture well; Where x7 represents the average normalized sphericity of artificial tissues or organoids in the j-th culture well; Organoid_avergae_spherecity(j) represents the average sphericity of artificial tissues or organoids in the j-th culture well; Where x8 represents the normalized ATP value of artificial tissues or organoids in the j-th culture well; Organoid_ATP(j) represents the ATP value of artificial tissues or organoids in the j-th culture well; 5-7 According to the above-obtained normalized indicators of artificial tissues or organoids, principal component analysis is performed; specifically: 5-7-1 A matrix A is composed of the normalized indicators of artificial tissues or organoids, that is: A = [x1, x2, x3, x4, x5, x6, x7, x8] m×8 Equation (19) Where m is the number of culture wells, here m>8, that is, the number of data groups, and the matrix A is an m×8 matrix; 5-7-2 Calculate the covariance matrix B for the matrix A: Where cov() represents the covariance function, and the covariance matrix B is an 8×8 matrix; 5-7-3 Use singular value decomposition (SVD) to calculate the singular values and eigenvectors of the covariance matrix B: B = U∑V T Equation (21) Where the matrices U and V represent the left and right singular matrices respectively, both of which are orthogonal matrices. The column vectors of V and U are the basis vectors of the row space and column space of the covariance matrix B respectively, and ∑ represents a diagonal matrix containing singular values; B T B = V∑ 2 V T Equation (22) The right singular matrix V uses the right singular matrix of SVD to reduce the dimensionality of the number of columns. The first k eigenvectors of PCA are the first k columns of V. Solve V for formula (22) using the knowledge of linear algebra, that is, B T The matrix composed of the eigenvectors of B is the V matrix in SVD, B T The matrix composed of the eigenvalues of B is ∑ 2 matrix; Since B T B is a square matrix and can be eigen-decomposed; the obtained eigenvalues and eigenvectors need to satisfy: (B T (B)v k = λ k v k Equation (23) where λ k is an eigenvalue, v k is the eigenvector corresponding to λ k , k = 1, 2, ..., 7; ∑ represents a diagonal matrix containing singular values, ∑ 2 represents a diagonal matrix containing eigenvalues, i.e., the singular value is the square root of the eigenvalue Eigenvalue matrix ∑ 2 Sorted by eigenvalues from largest to smallest, the singular value matrix ∑ is sorted by singular values from largest to smallest, and the columns of V also change with the sorting; Sorted ∑ 2 The matrix, ∑ matrix, and V matrix are specifically as follows: Formula (26) can also be expressed as: V = [v 1, v 2, v 3, v 4, v 5, v 6, v 7, v8] Equation (27) where λ k is an eigenvalue, λ k > λ k+1 ; The eigenvalue λ1 is the largest eigenvalue in the ∑ 2 matrix, and its corresponding eigenvector v1 = [a 11, a 21, a 31, a 41, a 51, a 61, a 71, a 81 T needs to be arranged in the first column of the V matrix; The eigenvalue λ8 is the smallest eigenvalue in the ∑ 2 matrix, and its corresponding eigenvector v8 = [a 18, a 28, a 38, a 48, a 58, a 68, a 78， a 88 T needs to be arranged in the last column of the V matrix;​​ Extract the principal components according to the variance contribution rate, use the column vectors of the right singular matrix as the coefficients of a certain principal component expression, and the ratio of the singular value to the total singular value is the coefficient of the comprehensive index, and finally obtain the comprehensive index of artificial tissues or organoids.

6. The method according to claim 5, characterized in that The equivalent diameter of the connected domain is greater than or equal to 32 microns.

7. An OCT-based artificial tissue or organoid detection device, characterized in that Including: An image acquisition module for receiving the three-dimensional gray-scale image of artificial tissues or organoids collected by the OCT device; An artificial tissue or organoid segmentation module for segmenting the three-dimensional gray-scale image of artificial tissues or organoids by using a deep learning network for artificial tissue or organoid segmentation to obtain a three-dimensional image of the artificial tissue or organoid region; A first calculation module for calculating the morphological indexes and ATP values of artificial tissues or organoids in each culture well of the culture plate corresponding to the last day of culture; A second calculation module, configured to calculate a covariance matrix for a matrix constructed by standardizing the morphological indexes of the artificial tissue or organoid and the ATP values, and then perform singular value decomposition on the covariance matrix to obtain a right singular matrix and a singular value matrix; extract the principal components according to the variance contribution rate to construct a comprehensive index of the artificial tissue or organoid.

8. A drug susceptibility testing method, characterized in that Judging the situation of the artificial tissue or organoid under different drug concentrations by using the comprehensive index of the artificial tissue or organoid obtained by the method according to any one of claims 1-6.

Images