CN114494217A - Method and device for detecting artificial tissues and organoids - Google Patents
Method and device for detecting artificial tissues and organoids Download PDFInfo
- Publication number
- CN114494217A CN114494217A CN202210110925.6A CN202210110925A CN114494217A CN 114494217 A CN114494217 A CN 114494217A CN 202210110925 A CN202210110925 A CN 202210110925A CN 114494217 A CN114494217 A CN 114494217A
- Authority
- CN
- China
- Prior art keywords
- organoid
- artificial tissue
- matrix
- artificial
- dimensional
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 210000002220 organoid Anatomy 0.000 title claims abstract description 286
- 210000001519 tissue Anatomy 0.000 title claims abstract description 266
- 238000000034 method Methods 0.000 title claims abstract description 37
- 239000011159 matrix material Substances 0.000 claims abstract description 128
- 230000011218 segmentation Effects 0.000 claims abstract description 34
- 238000013135 deep learning Methods 0.000 claims abstract description 27
- 230000000877 morphologic effect Effects 0.000 claims abstract description 24
- 239000003814 drug Substances 0.000 claims abstract description 21
- 230000000694 effects Effects 0.000 claims abstract description 21
- 238000004364 calculation method Methods 0.000 claims abstract description 15
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 13
- ZKHQWZAMYRWXGA-UHFFFAOYSA-N Adenosine triphosphate Natural products C1=NC=2C(N)=NC=NC=2N1C1OC(COP(O)(=O)OP(O)(=O)OP(O)(O)=O)C(O)C1O ZKHQWZAMYRWXGA-UHFFFAOYSA-N 0.000 claims description 33
- 238000012014 optical coherence tomography Methods 0.000 claims description 32
- 238000004422 calculation algorithm Methods 0.000 claims description 29
- ZKHQWZAMYRWXGA-KQYNXXCUSA-J ATP(4-) Chemical compound C1=NC=2C(N)=NC=NC=2N1[C@@H]1O[C@H](COP([O-])(=O)OP([O-])(=O)OP([O-])([O-])=O)[C@@H](O)[C@H]1O ZKHQWZAMYRWXGA-KQYNXXCUSA-J 0.000 claims description 26
- 239000013598 vector Substances 0.000 claims description 19
- 238000002372 labelling Methods 0.000 claims description 15
- 238000012360 testing method Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 10
- 229940079593 drug Drugs 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 4
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 3
- 230000033001 locomotion Effects 0.000 claims description 3
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 230000035945 sensitivity Effects 0.000 claims description 2
- 238000001228 spectrum Methods 0.000 claims description 2
- 238000011160 research Methods 0.000 abstract description 5
- GHASVSINZRGABV-UHFFFAOYSA-N Fluorouracil Chemical compound FC1=CNC(=O)NC1=O GHASVSINZRGABV-UHFFFAOYSA-N 0.000 description 6
- 229960002949 fluorouracil Drugs 0.000 description 6
- 238000003384 imaging method Methods 0.000 description 6
- 238000012163 sequencing technique Methods 0.000 description 6
- 208000005016 Intestinal Neoplasms Diseases 0.000 description 4
- 201000002313 intestinal cancer Diseases 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 210000004027 cell Anatomy 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 210000000056 organ Anatomy 0.000 description 3
- YXTKHLHCVFUPPT-YYFJYKOTSA-N (2s)-2-[[4-[(2-amino-5-formyl-4-oxo-1,6,7,8-tetrahydropteridin-6-yl)methylamino]benzoyl]amino]pentanedioic acid;(1r,2r)-1,2-dimethanidylcyclohexane;5-fluoro-1h-pyrimidine-2,4-dione;oxalic acid;platinum(2+) Chemical compound [Pt+2].OC(=O)C(O)=O.[CH2-][C@@H]1CCCC[C@H]1[CH2-].FC1=CNC(=O)NC1=O.C1NC=2NC(N)=NC(=O)C=2N(C=O)C1CNC1=CC=C(C(=O)N[C@@H](CCC(O)=O)C(O)=O)C=C1 YXTKHLHCVFUPPT-YYFJYKOTSA-N 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- KVUAALJSMIVURS-ZEDZUCNESA-L calcium folinate Chemical compound [Ca+2].C1NC=2NC(N)=NC(=O)C=2N(C=O)C1CNC1=CC=C(C(=O)N[C@@H](CCC([O-])=O)C([O-])=O)C=C1 KVUAALJSMIVURS-ZEDZUCNESA-L 0.000 description 2
- 235000008207 calcium folinate Nutrition 0.000 description 2
- 239000011687 calcium folinate Substances 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 108010082117 matrigel Proteins 0.000 description 2
- 238000000691 measurement method Methods 0.000 description 2
- DWAFYCQODLXJNR-BNTLRKBRSA-L oxaliplatin Chemical compound O1C(=O)C(=O)O[Pt]11N[C@@H]2CCCC[C@H]2N1 DWAFYCQODLXJNR-BNTLRKBRSA-L 0.000 description 2
- 229960001756 oxaliplatin Drugs 0.000 description 2
- 239000011148 porous material Substances 0.000 description 2
- 238000004445 quantitative analysis Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000004904 shortening Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 206010059866 Drug resistance Diseases 0.000 description 1
- 108010076504 Protein Sorting Signals Proteins 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- OIRDTQYFTABQOQ-KQYNXXCUSA-N adenosine Chemical compound C1=NC=2C(N)=NC=NC=2N1[C@@H]1O[C@H](CO)[C@@H](O)[C@H]1O OIRDTQYFTABQOQ-KQYNXXCUSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- OIRDTQYFTABQOQ-UHFFFAOYSA-N ara-adenosine Natural products Nc1ncnc2n(cnc12)C1OC(CO)C(O)C1O OIRDTQYFTABQOQ-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000000339 bright-field microscopy Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000007877 drug screening Methods 0.000 description 1
- 238000000799 fluorescence microscopy Methods 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000001338 self-assembly Methods 0.000 description 1
- 238000010186 staining Methods 0.000 description 1
- 210000000130 stem cell Anatomy 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000011426 transformation method Methods 0.000 description 1
- 235000011178 triphosphate Nutrition 0.000 description 1
- 239000001226 triphosphate Substances 0.000 description 1
- UNXRWKVEANCORM-UHFFFAOYSA-N triphosphoric acid Chemical compound OP(O)(=O)OP(O)(=O)OP(O)(O)=O UNXRWKVEANCORM-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G06T5/70—
-
- G06T5/90—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10101—Optical tomography; Optical coherence tomography [OCT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30028—Colon; Small intestine
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30096—Tumor; Lesion
Abstract
The invention discloses a method and a device for detecting artificial tissues and organoids. The invention realizes the segmentation of the artificial tissues and the organoid by utilizing an artificial tissue and organoid segmentation deep learning network to obtain three-dimensional images of the artificial tissues and the organoid regions; the label adopts an image after contrast enhancement; calculating the morphological indexes and ATP values of the artificial tissues and the organoids in each culture hole of the corresponding culture plate from the last day of culture according to the connected domain; carrying out standardization treatment on morphological indexes of artificial tissues and organoids and ATP values to construct a matrix calculation covariance matrix; performing singular value decomposition on the covariance matrix to obtain a right singular matrix and a singular value matrix; and extracting the main components according to the variance contribution rate to obtain the comprehensive indexes of the artificial tissues and the organoids. The comprehensive indexes of the artificial tissues and the organoids are utilized to research the curative effect of the medicine.
Description
Technical Field
The invention belongs to the field of artificial tissue and organ detection, and particularly relates to an artificial tissue and organ detection method and device based on OCT.
Background
The artificial tissue and organoid are 3D cell complex formed by self-assembly driven by human source cells or stem cells, can simulate the structure and function of human organs, can be used for optimizing and screening medicaments, pathological exploration and the like, and can detect the growth and development states of the artificial tissue or organoid by high-resolution continuous morphological observation, wherein the morphological indexes of the artificial tissue or organoid are closely related to the functional state of the artificial tissue or organoid in the culture process of the artificial tissue and organoid or during drug screening of the artificial tissue and organoid with a specific physiological structure. However, since the characteristic morphological structures of the artificial tissues and the organoids are small and have no obvious difference with the surrounding medium, and the change of the characteristic morphological structures cannot be observed by naked eyes along with the extension of the culture time, the growth conditions of the artificial tissues and the organoids need to be comparatively embodied by the segmentation and quantitative analysis of the characteristic morphological structures so as to research the treatment effect of certain drugs on diseases.
The traditional detection methods are invasive, such as staining histology, fluorescence microscopy, bright field microscopy and the like, but cannot evaluate the true state of cells, have no volumetric structure imaging capability, and cannot quantify the volumetric morphological characteristics, because the methods detect the artificial tissues or organoids based on plane information or perform quantitative analysis by approximating the artificial tissues or organoids to spheres or ellipsoids, but the organoids grow irregularly, which causes great errors. Meanwhile, the imaging penetration depth of the method is limited, and deep tissue imaging is difficult to perform.
The Optical Coherence Tomography (OCT) technique has the advantages of no loss, non-invasion, high resolution, and volumetric imaging, with imaging depths up to several millimeters. The invention detects the growth process of the artificial tissue and the organoid by using OCT, combines the characteristic morphological indexes of the artificial tissue and the organoid, such as the indexes of the characteristic structure of the artificial tissue or the total volume, the total surface area, the average volume, the average surface area and the like of the organoid and the index of Adenosine Triphosphate (ATP) representing the activity of the artificial tissue or the organoid by main component analysis to obtain a comprehensive index, and utilizes the comprehensive index of the artificial tissue or the organoid to research the curative effect of the medicament.
Disclosure of Invention
In order to use OCT to represent the growth condition of the artificial tissue and the organoid, morphological indexes of characteristic structures of the artificial tissue and the organoid, such as total volume, total surface area, average volume, average surface area and the like, and an index Adenosine Triphosphate (ATP) representing the activity of the artificial tissue and the organoid are combined to establish a comprehensive index for detecting the artificial tissue and the organoid, so that the OCT can detect the growth and the death of the artificial tissue and the organoid along with time.
An artificial tissue and organoid detection method based on OCT, comprising the following steps:
step (1): acquiring original three-dimensional gray scale images of the artificial tissues and the organoids by using an OCT (optical coherence tomography) device;
opening system software of the OCT equipment, setting a data acquisition format and a storage path, selecting a 3D acquisition mode, setting the size of an OCT view field, the size of a single pixel and the scanning speed according to the diameters of holes of artificial tissues and organoid culture holes, wherein the detected hole is positioned in the view field range, and the pixel is set according to the required image definition and data size. The refractive index is adjusted according to the material of the actual acquisition object. Before data acquisition, the reference arm, the sample arm and the light intensity adjusting button are adjusted to achieve the purposes of high signal-to-noise ratio, positioning of the artificial tissue or organoid culture hole in the center of a visual field and clearest image display, and meanwhile, the angle of the pore plate is adjusted to eliminate the influence of stripe noise. The position of the lens is taken as reference, the lens is upward to be called reverse sampling, and the lens is downward to be called forward sampling. When data acquisition is started, a preview function is used for observing a three-dimensional image of acquired data under an acquisition scheme so as to judge whether the acquisition scheme is in accordance with expectation, if so, acquisition is carried out, and otherwise, acquisition parameters are adjusted.
Step (2): constructing a labeled two-dimensional image or three-dimensional image data set, and dividing the data set into a training set and a testing set;
the construction process of the labeled two-dimensional image is specifically to split an original three-dimensional gray image of an artificial tissue or an organoid into a plurality of original two-dimensional gray images, and then to perform contrast enhancement on the two-dimensional gray images to obtain corresponding contrast-enhanced two-dimensional gray images; labeling the original two-dimensional gray image to be used as a two-dimensional image data set, wherein the label is the two-dimensional gray image after the contrast corresponding to the original two-dimensional gray image is enhanced;
the construction process of the labeled three-dimensional image is specifically that an original three-dimensional gray image of an artificial tissue or an organoid is split into a plurality of original two-dimensional gray images, and then contrast enhancement is carried out on the two-dimensional gray images to obtain corresponding contrast-enhanced two-dimensional gray images; combining the two-dimensional gray level images with enhanced contrast to obtain a three-dimensional image with enhanced contrast; labeling the original three-dimensional gray level image to be used as a three-dimensional image data set, wherein the label is a three-dimensional image with the contrast ratio corresponding to the original three-dimensional gray level image enhanced;
preferably, The contrast enhancement of The two-dimensional grayscale images is specifically to process The two-dimensional grayscale images by one or more of a Depth-resolved model-based reconstruction of The intensity coefficients, a Logarithmic intensity variance algorithm (LIV), a delayed correlation attenuation speed algorithm (Late OCT compensation speckle (OCDSl)), an inverse power law exponent alpha algorithm of The fluctuation spectrum (alpha), a motion amplitude M algorithm based on autocorrelation (M), a complex correlation algorithm (complex-correlation algorithm), and a conventional image preprocessing algorithm;
the traditional image preprocessing algorithm comprises denoising, image gray level transformation and the like.
Preferably, in order to save acquisition and reconstruction time of an OCT contrast enhancement image and meet the requirement of deep learning network training convergence, the original two-dimensional gray image can be subjected to contrast enhancement after data enhancement;
more preferably, the data enhancement is one or more of a method combining distortion enhancement and non-distortion enhancement, an image rotation method and an image scale transformation method;
and (3): constructing an artificial tissue or organoid segmentation deep learning network, and training by using a training set; finally, testing and verifying the trained artificial tissue or organoid segmentation deep learning network by using a test set;
the artificial tissue or organoid segmentation deep learning network adopts EG-Net, ResNet50v2, VGG19, Xception and DenseNet121CNN, the input of the artificial tissue or organoid segmentation deep learning network is a two-dimensional gray image, and the output of the artificial tissue or organoid segmentation deep learning network is a two-dimensional gray image after contrast enhancement; finally, combining all the contrast-enhanced two-dimensional gray images corresponding to the original three-dimensional gray image;
the artificial tissue or organoid segmentation deep learning network adopts Resnet-3D, the input of the Resnet-3D is an original three-dimensional gray image, and the output of the Resnet-3D is a three-dimensional image with enhanced contrast;
and (4): the artificial tissue or the organoid is segmented by utilizing the artificial tissue or the organoid segmentation deep learning network after test and verification to obtain a three-dimensional image of the artificial tissue or the organoid region;
and (5): quantifying the three-dimensional image of the artificial tissue or organoid region obtained in the step (4):
calculating the morphological index and ATP value of the artificial tissue or organoid in each culture hole of the corresponding culture plate in the last day of culture according to the connected domain, wherein the morphological index is as follows: total volume, total surface area, total sphericity, number, average volume, average surface area, average sphericity. The morphological index of the artificial tissue or the organoid is standardized with the ATP value, and the influence of the dimension is eliminated. And calculating covariance of the normalized matrix to obtain a covariance matrix. And carrying out singular value decomposition on the covariance matrix to obtain a right singular matrix and a singular value matrix which are sequenced from large to small according to singular values. And extracting principal components according to the variance contribution rate, extracting the principal components with the accumulative variance contribution rate more than 85%, taking the column vector of the right singular matrix as a coefficient of a certain principal component expression, and taking the ratio of the singular value to the total singular value as a coefficient of the comprehensive index to finally obtain the comprehensive index of the artificial tissue or the organoid.
The method comprises the following steps:
5-1, obtaining the pixel number Volume _ number occupied by the Volume of a single artificial tissue or organ-like part and the pixel number Surface _ number occupied by the Surface area in the three-dimensional image according to the three-dimensional image of the artificial tissue or organ-like part obtained by segmentation in the step (4); then according to the size of the three-dimensional pixel and the size of the two-dimensional pixel, combining the formulas (1) and (2) to respectively obtain the volume Organoid _ volume and the surface area Organoid _ surface of a single artificial tissue or Organoid;
Organoid_volume＝Volume_number*(PixelSize_x*PixelSize_y*PixelSize_z)
formula (1)
Organic _ surface _ number (PixelSize _ x PixelSize _ y) formula (2)
PixelSize_x＝PixelSize_y＝PixelSize_z
The pixel size of the original three-dimensional gray image is represented by PixelSize _ x, PixelSize _ y and PixelSize _ z;
the number of pixels Volume _ number occupied by the Volume of the single artificial tissue or the organoid and the number of pixels Surface _ number occupied by the Surface area are obtained by calculating a three-dimensional image of the artificial tissue or the organoid area according to a three-dimensional connected domain, namely, the connected area is marked by a bwleaeln function in MATLAB.
5-2 obtaining the sphericity of the single artificial tissue or Organoid according to the volume of the single artificial tissue or Organoid _ volume and the surface area of the artificial tissue or Organoid _ surface and combining the formula (3):
5-3, calculating the total number of the artificial tissues or the organoids in a single culture hole according to the three-dimensional images of the three-dimensional connected domain:
connected _ domains _ number (4)
The Connected _ domains _ number represents the number of Connected domains in the three-dimensional image;
preferably, the equivalent diameter of the connected domains is greater than or equal to 32 micrometers;
calculating the total volume of the artificial tissues or organoids of a single culture well:
wherein Organoid _ volume (i) represents the ith artificial tissue or Organoid volume within a single culture well;
calculate total surface area of artificial tissue or organoid for individual culture wells:
wherein Organoid surface (i) represents the surface area of the ith artificial tissue or Organoid within a single culture well;
calculating the total sphericity of the artificial tissues or organoids of a single culture well:
wherein Organoid _ sphere (i) indicates the ith artificial tissue or Organoid sphericity within a single culture well;
5-4 calculating the average volume of the single artificial tissue or organoid in a culture well according to the total volume of the artificial tissue or organoid in the single culture well:
Organoid_average_volume＝Organoid_sum_volume/Organoid_number
formula (8)
Calculating the average surface area of the individual artificial tissues or organoids in a culture well based on the total surface area of the individual artificial tissues or organoids in the individual culture well:
Organoid_average_surface＝Organoid_sum_surface/Organoid_number
formula (9)
Calculating the average sphericity of the single artificial tissue or organoid in a culture well according to the total sphericity of the artificial tissue or organoid in the single culture well:
Organoid_average_spherecity
＝Organoid_sum_spherecity/Organoid_number
formula (10)
5-5, obtaining adenosine triphosphate Organoid _ ATP by a chemiluminescence measurement method for each culture hole of the corresponding culture plate on the last day of culture;
5-6 obtaining standardized artificial tissue or organoid indexes in each culture hole, wherein the indexes comprise total volume, total surface area, total sphericity, total number, average volume, average surface area, average sphericity and ATP;
wherein x1Expressing the total normalized artificial tissue or organoid volume in the jth culture well; organic _ sum _ volume (j) represents the total volume of the artificial tissue or Organoid in the jth culture well; m represents the number of culture wells.
Wherein x2Represents the total surface area of the normalized artificial tissue or organoid in the jth well; organoid _ sum _ surface (j) indicates the summary of artificial tissues or organoids in the jth wellArea;
wherein x is3Expressing the total sphericity of the normalized artificial tissues or organoids in the jth culture well; organic _ sum _ specificity (j) represents the total sphericity of the artificial tissue or Organoid in the jth culture well;
wherein x is4Expressing the total number of standardized artificial tissues or organoids in the jth culture well; organoid _ number (j) represents the total number of artificial tissues or organoids in the jth culture well;
wherein x5Represents the mean volume of the normalized artificial tissue or organoid in the jth well; organic _ average _ volume (j) represents the average volume of artificial tissue or Organoid in the jth culture well;
wherein x6Represents the normalized average surface area of the artificial tissue or organoid in the jth culture well; organic _ average _ surface (j) represents the average surface area of the artificial tissue or Organoid in the jth culture well;
wherein x7Represents the average sphericity of the normalized artificial tissue or organoid in the jth culture well; organic _ average _ specificity (j) indicates the level of the artificial tissue or Organoid in the jth wellThe sphericity is equalized;
wherein x8(ii) normalized artificial tissue or organoid ATP values in the jth culture well; organoid _ ATP (j) indicates ATP value of artificial tissue or Organoid in the ith culture well;
the reason for normalization is that the dimensions of the respective parameters are not uniform in order to eliminate the influence of the dimensions.
5-7, performing principal component analysis according to the obtained standardized artificial tissue or organoid indexes; in particular to
5-7-1, forming a matrix A by the indexes of the standardized artificial tissues or organoids, namely:
A＝[x1，x2，x3，x4，x5，x6，x7，x8]m×8formula (19)
Where m is the number of culture wells, where m >8, i.e., the number of sets of data, and matrix A is a matrix of m rows and 8 columns.
5-7-2, calculating the matrix A to obtain a covariance matrix B:
where cov () represents the covariance function, the covariance matrix B is a matrix of 8 rows and 8 columns.
5-7-3 Singular Value Decomposition (SVD) is used to compute singular values and eigenvectors of the covariance matrix B:
B＝U∑VTformula (21)
The matrixes U and V respectively represent a left singular matrix and a right singular matrix which are orthogonal matrixes, column vectors of V and U are respectively base vectors of a row space and a column space of the covariance matrix B, and sigma represents a diagonal matrix containing singular values.
BTB＝V∑2VTFormula (22)
Right singular matrixV, using a right singular matrix of the SVD to reduce the dimension of the column number, wherein the first k eigenvectors of the PCA are the first k columns of the V; knowledge of the application of linear algebra to the formula (22) solves for V, BTThe matrix formed by the characteristic vectors of B is the V matrix in SVD, BTThe matrix formed by the characteristic values of B is sigma2And (4) matrix.
Due to BTB is a square matrix which can be subjected to characteristic decomposition; the obtained characteristic value and the characteristic vector need to satisfy the following conditions:
(BTB)vk＝λkvkformula (23)
Wherein λkIs a characteristic value, vkIs λkThe corresponding feature vector, k ═ 1, 2.., 7; sigma denotes a diagonal matrix containing singular values2Representing a diagonal matrix containing eigenvalues, i.e. singular values being the evolution of the eigenvalues
Eigenvalue matrix sigma2And sequencing according to the eigenvalue from large to small, sequencing the singular value matrix sigma from large to small according to the singular value, and simultaneously, changing the column of V along with the sequencing change.
The ordered sigma2The matrix, the sigma matrix and the V matrix are as follows:
equation (26) can also be expressed as:
wherein λkIs a characteristic value, λk＞λk+1(k ═ 1, 2,.., 7). Characteristic value lambda1Is sigma2Maximum eigenvalue in the matrix, its corresponding eigenvector v1＝[a11，a21，a31，a41，a51，a61，a71，a81]TThe first row of the V matrix is required to be arranged; characteristic value lambda8Is sigma2The smallest eigenvalue in the matrix, its corresponding eigenvector v8＝[a18，a28，a38，a48，a58，a68，a78，a88]TIt is arranged in the last column of the V matrix.
Selecting principal component according to the singular value ratio greater than threshold value alpha (85% can be used) and called variance contribution rate, and the variance contribution rate b of first principal component1And a second principal component variance contribution rate b2Namely:
b1+b2alpha type > threshold (30)
The size and activity of the space occupied by the artificial tissue or organoid constituted by the first principal component F1:
F1＝-(a11*x1+a21*x2+a31*x3+a41*x4+a51*x5+a61*x6+a71*x7+a81*x8)
formula (31)
From the second principal component F2Phenotypic characteristics and quantities constituting artificial tissues or organoids:
F2＝-(a12*x1+a22*x2+a32*x3+a42*x4+a52*x5+a62*x6+a72*x7+a82*x8)
the formula (32) is the synthetic index organic _ F of the artificial tissue or Organoid composed of the size and activity of the space occupied by the artificial tissue or Organoid, and the phenotypic characteristics and quantity:
Organoid_F＝b1*F1+b2*F2formula (33)
It is a second object of the present invention to provide an OCT-based artificial tissue or organoid detection apparatus, comprising:
the image acquisition module is used for receiving the three-dimensional gray level image of the artificial tissue or the organoid acquired by the OCT equipment;
the artificial tissue or organoid segmentation module is used for segmenting the three-dimensional gray level image of the artificial tissue or organoid by utilizing an artificial tissue or organoid segmentation deep learning network to obtain a three-dimensional image of the artificial tissue or organoid region;
the first calculation module is used for calculating the morphological index and ATP value of the artificial tissue or organoid in each culture hole of the corresponding culture plate from the last day of culture;
the second calculation module is used for carrying out standardized processing on the morphological index of the artificial tissue or the organoid and the ATP value to construct a matrix calculation covariance matrix, and then carrying out singular value decomposition on the matrix calculation covariance matrix to obtain a right singular matrix and a singular value matrix; and extracting main components according to the variance contribution rate, and constructing to obtain the artificial tissue or organoid comprehensive index.
The third purpose of the invention is to provide a drug sensitivity test method, which utilizes the comprehensive index of the artificial tissue or the organoid obtained by the method to judge the curative effect condition of the artificial tissue or the organoid under different drug concentration.
By drawing a curve of comprehensive index change of artificial tissues or organoids along with time under the action of different drugs, the problems of the most effective single drug at which concentration and drug resistance at which concentration can be generated can be researched, and the problem of the best treatment effect of the combined mode of the combined drugs can also be researched.
The invention has the beneficial effects that:
1. in order to achieve the purpose of shortening the segmentation time of the artificial tissues or the organoids, the invention provides a method for constructing a training set to replace artificial labeling by using a contrast enhancement algorithm based on attenuation coefficients and the like and automatically segmenting by shortening the calculation time through a deep learning network.
2. The invention detects the growth process of the artificial tissue or the organoid by using OCT, calculates the covariance matrix of the matrix constructed after standardizing the parameters of the morphological index of the artificial tissue or the organoid and the adenine nucleoside triphosphate (ATP) which represents the activity of the artificial tissue or the organoid, and then carries out singular value decomposition to obtain the right singular matrix and the singular value matrix. And extracting the main component according to the variance contribution rate to finally obtain the synthetic index of the artificial tissue or the organoid. The comprehensive indexes of the artificial tissues or the organoids are utilized to research the curative effect of the medicine, so as to realize accurate treatment. For example, the treatment condition of 5-fluorouracil (5-FU) on artificial tissues or organoids under different concentrations is researched, and the optimal concentration is screened out to achieve the optimal curative effect; or researching the matching form of the combined medicine, such as FOLFOX (oxaliplatin, calcium folinate and fluorouracil) combined medicine, and searching a medicine combination scheme which is most suitable for the artificial tissue or organoid treatment by designing different medicine combination modes.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is an OCT cross-sectional view of an artificial tissue or organoid of intestinal cancer; (a) raw images acquired by OCT; (b) comparing the enhanced image by using an attenuation coefficient algorithm;
FIG. 3 is a graph of log intensity variance comparison; (a) raw images acquired by OCT; (b) contrast-enhanced images using number intensity variance;
FIG. 4 is a three-dimensional reconstruction of an artificial tissue or organoid of intestinal cancer; (a) raw images acquired by OCT; (b) manually marking the segmentation drawing; (c) the map is segmented using attenuation coefficients.
FIG. 5 is a schematic representation of a single pixel size of an artificial tissue or organoid; (a) a three-dimensional representation of a single artificial tissue or organoid; (b) a single pixel size schematic diagram under three-dimensional coordinates; (c) a schematic diagram of a single pixel size in two-dimensional coordinates x-o-y; (d) a schematic diagram of a single pixel size in two-dimensional coordinates x-o-z; (e) schematic of a single pixel size in two-dimensional coordinates y-o-z.
Detailed Description
The invention is further analyzed with reference to the following specific examples.
An OCT-based artificial tissue or organoid quantification method, see fig. 1, comprising the steps of:
step (1): acquiring an original three-dimensional gray scale image of an artificial tissue or an organoid by using an OCT (optical coherence tomography) device;
and opening the OCT system, and setting a file format and a stored path of the acquired data, wherein the file format is set according to the gray data to be acquired. The acquisition modes are selected in a software interface, the modes comprise 2D, 3D, Doppler and speckle, the embodiment selects the 3D mode, and the size of the field of view, the pixels and the scanning speed of the acquisition are set in the 3D mode. The detected hole is positioned in the field range, the pixel is set according to the required image definition and data size, and the scanning speed is defaulted to 48 kHz. Ascan and Bscan are both set to 1 without repeated collection, and the refractive index is set to 1. Before data acquisition, the reference arm, the sample arm and the light intensity adjusting button are adjusted to achieve the purposes of high signal-to-noise ratio, the artificial tissue or organoid culture hole positioned in the center of a visual field and clearest image display, and meanwhile, the angle of the pore plate is adjusted to be about 5 degrees to eliminate the influence of stripe noise. The artificial tissue or organoid is acquired by using the back-off method as shown in fig. 2(a), fig. 3(a) and fig. 4 (a).
Step (2): constructing a labeled two-dimensional image or three-dimensional image data set, and dividing the data set into a training set and a testing set;
the construction process of the labeled two-dimensional image is specifically to split an original three-dimensional gray image of an artificial tissue or an organoid into a plurality of original two-dimensional gray images, and then to perform contrast enhancement on the two-dimensional gray images to obtain corresponding contrast-enhanced two-dimensional gray images; labeling the original two-dimensional gray image to be used as a two-dimensional image data set, wherein the label is the two-dimensional gray image after the contrast corresponding to the original two-dimensional gray image is enhanced;
the construction process of the labeled three-dimensional image is specifically that an original three-dimensional gray image of an artificial tissue or an organoid is split into a plurality of original two-dimensional gray images, and then contrast enhancement is carried out on the two-dimensional gray images to obtain corresponding contrast-enhanced two-dimensional gray images; combining the two-dimensional gray level images with enhanced contrast to obtain a three-dimensional image with enhanced contrast; labeling the original three-dimensional gray level image to be used as a three-dimensional image data set, wherein the label is a three-dimensional image with the contrast ratio corresponding to the original three-dimensional gray level image enhanced;
the contrast enhancement of The two-dimensional gray scale images is specifically to perform one or more of Depth-resolved attenuation coefficient algorithm (Depth-resolved model-based correlation of attenuation coefficients), Logarithmic intensity variance algorithm (LIV), delay-correlated attenuation speed algorithm (Late OCT compensation speed (OCDSl)), inverse power law index alpha algorithm (The robust of The power law output, alpha), auto-correlated motion amplitude M algorithm (M), complex-correlation algorithm (complex-correlation algorithm), and conventional image preprocessing algorithm on The two-dimensional gray scale images;
an advantage of the depth-resolved attenuation coefficient algorithm is that the intensity values on each pixel of the OCT X-Z cross-section are converted into attenuation coefficients, which need not be segmented to determine. The log intensity variance algorithm is one of dynamic imaging algorithms, and can display some physiological information inside an artificial tissue or an organoid. The log intensity variance is essentially the time variance of the fast acquisition OCT signal sequence.
And calculating the attenuation coefficient of the acquired intestinal cancer artificial tissue or organoid data on an X-Z plane according to the two algorithms, and comparing the original image with an attenuation coefficient map shown in figure 2. The artificial tissue or organoid in the original image shown in fig. 2(a) is similar to the matrigel, and if the artificial labeling is directly performed in fig. 2(a), a great error is generated, while the problem is solved in fig. 2(b) and the influence of background noise is suppressed by calculating the average value.
Similarly, fig. 3, after calculation of the variance of the log intensity, also effectively distinguishes the artificial tissue or organoid from matrigel, with contrast enhancement effect. FIG. 3(b) shows a stronger contrast and the artificial tissue or organoid is more distinct than FIG. 3 (a).
And performing three-dimensional reconstruction on the image data with the attenuation coefficient and the enhanced contrast ratio, which is obtained by calculation, and comparing the result with an artificial labeling result to obtain a graph 4. As can be obtained from fig. 4(a) and (b), a high accuracy can be achieved already through the deep learning network, and meanwhile, the three-dimensional reconstructed images obtained by the artificial labeling and the attenuation coefficient construction training set in fig. 4(b) and 4(c) have strong similarity, so that the training set provided by the attenuation coefficient can be comparable to the artificial labeling. Statistics of the run times were performed using tic and toc in MATLAB functions, taking intestinal cancer artificial tissues or organoids as an example, and the table was drawn as follows.
Table 1: comparison of time consumption
| Method | Manual labeling | Attenuation coefficient calculation |
| Time consuming | 10min | 3.046566s |
The calculation time of the attenuation coefficient obtained from table 1 is much shorter than that of the manual labeling, and the consumption of manpower and material resources can be greatly reduced by the method.
And (3): constructing an artificial tissue or organoid segmentation deep learning network, and training by using a training set; finally, testing and verifying the trained artificial tissue or organoid segmentation deep learning network by using a test set;
the artificial tissue or organoid segmentation deep learning network adopts EG-Net, ResNet50v2, VGG19, Xception and DenseNet121CNN, the input of the artificial tissue or organoid segmentation deep learning network is a two-dimensional gray image, and the output of the artificial tissue or organoid segmentation deep learning network is a two-dimensional gray image after contrast enhancement; finally, combining all the contrast-enhanced two-dimensional gray images corresponding to the original three-dimensional gray image;
the artificial tissue or organoid segmentation deep learning network adopts Resnet-3D, the input of the Resnet-3D is an original three-dimensional gray image, and the output of the Resnet-3D is a three-dimensional image with enhanced contrast;
and (4): the artificial tissue or organoid segmentation is realized by using the artificial tissue or organoid segmentation deep learning network after test and verification, and a three-dimensional image of the artificial tissue or organoid region is obtained;
and (5): quantifying the three-dimensional image of the artificial tissue or organoid region obtained in the step (4):
calculating the morphological index and ATP value of the artificial tissue or organoid in each culture hole of the corresponding culture plate in the last day of culture according to the connected domain, wherein the morphological index is as follows: total volume, total surface area, total sphericity, number, average volume, average surface area, average sphericity. The morphological indexes of the artificial tissues or organoids are standardized with ATP values, and the influence of dimensions is eliminated. And calculating covariance of the normalized matrix to obtain a covariance matrix. And performing singular value decomposition on the covariance matrix to obtain a right singular matrix and a singular value matrix which are sequenced from large to small according to singular values. And extracting principal components according to the variance contribution rate, extracting the principal components with the accumulative variance contribution rate more than 85%, taking the column vector of the right singular matrix as the coefficient of a certain principal component expression, taking the ratio of the singular value to the total singular value as the coefficient of the comprehensive index, and finally obtaining the mathematical expression of the comprehensive index of the artificial tissue or the organoid.
The method comprises the following steps:
5-1, according to the three-dimensional image of the artificial tissue or organ-like region obtained by segmentation in the step (5), firstly establishing a three-dimensional space coordinate system of a single artificial tissue or organ-like region as shown in fig. 5(a), and obtaining the number of pixels Volume _ number occupied by the Volume of the single artificial tissue or organ-like region and the number of pixels Surface _ number occupied by the Surface area in the three-dimensional image; the artificial tissue or organoid volume and surface area are then calculated based on the pixel size. The pixel size is a parameter set in the original three-dimensional gray scale image of the artificial tissue or the organoid acquired by using the OCT apparatus in the step (1), and the size of a single pixel in a three-dimensional coordinate system in the volume of the artificial tissue or the organoid is as shown in fig. 5 (b). The size of a single pixel in a two-dimensional coordinate system in the surface area of the artificial tissue or organoid is shown in FIGS. 5(c), (d), and (e). Obtaining a single artificial tissue or Organoid volume and a surface area Organoid surface respectively by combining the formulas (1) to (2);
Organoid_volume＝Volume_number*(PixelSize_x*PixelSize_y*PixelSize_z)
formula (1)
Organoid_surface＝surface_number*(PixelSize_x*PixelSize_y)
Formula (2)
PixelSize_x＝PixelSize_y＝PixelSize_z＝5μn
The number of pixels Volume _ number occupied by the Volume of the single artificial tissue or the organoid and the number of pixels Surface _ number occupied by the Surface area are obtained by calculating a three-dimensional image of the artificial tissue or the organoid area according to a three-dimensional connected domain, namely, the connected area is marked by a bwleaeln function in MATLAB.
5-2 obtaining the sphericity of the single artificial tissue or Organoid according to the volume Organoid _ volume and the surface area Organoid _ surface of the single artificial tissue or Organoid and combining the formula (3):
5-3, calculating the total number of the artificial tissues or the organoids in a single culture hole according to the three-dimensional images of the three-dimensional connected domain:
connected _ domains _ number formula (4)
The Connected _ domains _ number represents the number of Connected domains in the three-dimensional image;
preferably, the equivalent diameter of the connected domains is greater than or equal to 32 micrometers;
calculating the total volume of the artificial tissues or organoids of a single culture well:
wherein Organoid volume (i) represents the ith artificial tissue or Organoid volume within a single culture well;
calculate total surface area of artificial tissue or organoid for individual culture wells:
wherein Organoid surface (i) represents the surface area of the ith artificial tissue or Organoid within a single culture well;
calculating the total sphericity of the artificial tissues or organoids of a single culture well:
wherein Organoid _ sphere (i) indicates the ith artificial tissue or Organoid sphericity within a single culture well;
5-4 calculating the average volume of the single artificial tissue or organoid in a culture well according to the total volume of the artificial tissue or organoid in the single culture well:
Organoid_average_volume＝Organoid_sum_volume/Organoid_number
formula (8)
Calculating the average surface area of the individual artificial tissues or organoids in a culture well based on the total surface area of the individual artificial tissues or organoids in the individual culture well:
organic _ average _ surface ═ organic _ surface/organic _ number formula (9)
Calculating the average sphericity of the single artificial tissue or organoid in a culture well according to the total sphericity of the artificial tissue or organoid in the single culture well:
organic _ average _ specificity ═ organic _ sum _ specificity/organic _ number formula (10)
5-5, obtaining adenosine triphosphate Organoid _ ATP by a chemiluminescence measurement method for each culture hole of the corresponding culture plate on the last day of culture;
5-6, obtaining standardized artificial tissue or organoid indexes in each culture hole, wherein the indexes comprise total volume, total surface area, total sphericity, total number, average volume, average surface area, average sphericity and ATP average value;
wherein x1Expressing the total normalized artificial tissue or organoid volume in the jth culture well; organic _ sum _ volume (j) represents the total volume of the artificial tissue or Organoid in the jth culture well; m represents the number of culture wells.
Wherein x2Represents the total surface area of the normalized artificial tissue or organoid in the jth well; organic _ sum _ surface (j) represents the total surface area of the artificial tissue or Organoid in the jth well;
wherein x3Expressing the total sphericity of the normalized artificial tissue or organoid in the jth culture well; organic _ sum _ specificity (j) represents the total sphericity of the artificial tissue or Organoid in the jth culture well;
wherein x4Expressing the total number of standardized artificial tissues or organoids in the jth culture well; organoid _ number (j) indicates the total number of artificial tissues or organoids in the jth well;
wherein x5Represents the mean volume of the normalized artificial tissue or organoid in the jth well; organic _ average _ volume (j) represents the average volume of artificial tissue or Organoid in the jth culture well;
wherein x6Represents the normalized average surface area of the artificial tissue or organoid in the jth culture well; organic _ average _ surface (j) represents the average surface area of the artificial tissue or Organoid in the jth culture well;
wherein x7Represents the average sphericity of the normalized artificial tissue or organoid in the jth culture well;
organoid _ average _ specificity (j) represents the average sphericity of the artificial tissue or Organoid in the jth culture well;
wherein x8(ii) normalized artificial tissue or organoid ATP values in the jth culture well; organoid _ ATP (j) indicates ATP value of artificial tissue or Organoid in the jth well;
the reason for normalization is that the dimensions of the respective parameters are not uniform in order to eliminate the influence of the dimensions.
5-6, performing principal component analysis according to the obtained standardized artificial tissue or organoid indexes; in particular to
5-7-1, forming a matrix A by the indexes of the standardized artificial tissues or organoids, namely:
A＝[x1，x2，x3，x4，x5，x6，x7，x8]m×8formula (19)
Where m is the number of culture wells, where m is 21, and matrix a is a matrix of m rows and 8 columns.
5-7-2, calculating the matrix A to obtain a covariance matrix B:
where cov () represents the covariance function, the covariance matrix B is a matrix of 8 rows and 8 columns.
5-7-3 Singular Value Decomposition (SVD) is used to compute singular values and eigenvectors of the covariance matrix B:
B＝U∑VTformula (21)
The matrixes U and V respectively represent a left singular matrix and a right singular matrix which are orthogonal matrixes, column vectors of V and U are respectively base vectors of a row space and a column space of the covariance matrix B, and sigma represents a diagonal matrix containing singular values.
BTB＝V∑2VTFormula (22)
The right singular matrix V uses the right singular matrix of the SVD to reduce the dimension of the column number, and the first k eigenvectors of the PCA are the first k columns of the V; knowledge of the application of linear algebra to the formula (22) solves for V, BTThe matrix formed by the characteristic vectors of B is the V matrix in SVD, BTThe matrix formed by the characteristic values of B is sigma2And (4) matrix.
Due to BTB is a square matrix which can be subjected to characteristic decomposition; the obtained characteristic value and the characteristic vector need to satisfy the following conditions:
(BTB)vk＝λkvkformula (23)
Wherein λkIs a characteristic value, vkIs λkThe corresponding feature vector, k ═ 1, 2.., 7; sigma denotes a diagonal matrix containing singular values2Representing a diagonal matrix containing eigenvalues, i.e. singular values being the evolution of the eigenvalues
Eigenvalue matrix sigma2And sequencing according to the eigenvalue from large to small, sequencing the singular value matrix sigma from large to small according to the singular value, and simultaneously, changing the column of V along with the sequencing change.
The sorted sigma2The matrix, the sigma matrix and the V matrix are as follows:
a11＝-0.4520，a21＝-0.4285，a31＝-0.0175，a41＝0.0303，a51＝-0.4470，a61＝0.4518，a71＝-0.2502，a81＝-0.3800；
a12＝-0.0294，a22＝-0.1214，a32＝-0.6146，a42＝-0.6118，a52＝0.1408，a62＝0.1089，a72＝0.4266，a82＝0.1370；
equation (26) can also be expressed as:
wherein λkIs a characteristic value, λk＞λk+1(k ═ 1, 2,.., 7). Characteristic value lambda1Is sigma2Maximum eigenvalue in the matrix, its corresponding eigenvector v1＝[a11a21，a31，a41，a51，a61，a71，a81]TThe first row of the V matrix is required to be arranged; characteristic value lambda8Is sigma2The smallest eigenvalue in the matrix, its corresponding eigenvector v8＝[a18，a28，a38，a48，a58，a68，a78，a88]TIt needs to be arranged in the last column of the V matrix.
Selecting principal component according to the singular value ratio greater than threshold value alpha (85% can be used) and called variance contribution rate, and the variance contribution rate b of first principal component1And a second principal component variance contribution rate b2Namely:
b1+b2>threshold alpha type (30)
The size and activity of the space occupied by the artificial tissue or organoid constituted by the first principal component F1:
F1＝-(a11*x1+a21*x2+a31*x3+a41*x4+a51*x5+a61*x6+a71*x7+a81*x8)
＝0.4520*x1+0.4285*x2+0.0175*x3-0.0303*x4+0.4470*x5-0.4518*x6+0.2502*x7+0.3800*x8formula (31)
From the second principal component F2Phenotypic characteristics and quantities constituting artificial tissues or organoids:
F2＝-(a12*x1+a22*x2+a32*x3+a42*x4+a52*x5+a62*x6+a72*x7+a82*x8)
＝0.0294*x1+0.1214*x2+0.6146*x3+0.6118*x4-0.1408*x5-0.1089*x6+0.4266*x7-0.1370*x8formula (32)
Since there is a moderate positive load on the normalization variables x1, x2, x5, and x8 in the first principal component, F1 can constitute the size and activity of the space occupied by the artificial tissue or organoid. While there is a moderate positive loading on the normalization variables x3, x4, x7 in the second principal component, F2 can constitute an artificial tissue or organoid phenotypic trait and quantity.
The size and activity of the space occupied by the artificial tissue or Organoid, and the phenotypic characteristics and quantity form an artificial tissue or Organoid comprehensive index Organoid _ F:
Organoid_F＝b1*F1+b2*F2＝0.5760*F1+0.3216*F2formula (33)
The contribution rate of the first principal component is 57.60%, the contribution rate of the second principal component is 32.16%, and the cumulative contribution rate of the first two principal components reaches 89.7655%, that is, the 2 principal components can represent the information content of 8 indexes 89.7655%.
The research on the curative effect of the medicine is carried out through the obtained comprehensive indexes, and then the accurate treatment is realized. For example, the treatment condition of 5-fluorouracil (5-FU) on artificial tissues or organoids under different concentrations is researched, and the optimal concentration is screened out to achieve the optimal curative effect; or researching the matching form of the combined medicine, such as FOLFOX (oxaliplatin, calcium folinate and fluorouracil) combined medicine, and searching a medicine combination scheme which is most suitable for the artificial tissue or organoid treatment by designing different medicine combination modes.
Claims (8)
1. A method for the detection of artificial tissues and organoids comprising the steps of:
step (1): acquiring an original three-dimensional gray scale image of an artificial tissue or an organoid by using an OCT (optical coherence tomography) device;
step (2): constructing a two-dimensional image or three-dimensional image data set after labeling, and dividing the data set into a training set and a testing set;
the construction process of the two-dimensional image after labeling is specifically to split an original three-dimensional gray image of an artificial tissue or an organoid into a plurality of original two-dimensional gray images, and then to perform contrast enhancement on the two-dimensional gray images to obtain corresponding contrast-enhanced two-dimensional gray images; labeling the original two-dimensional gray image to be used as a two-dimensional image data set, wherein the label is the two-dimensional gray image after the contrast corresponding to the original two-dimensional gray image is enhanced;
the construction process of the labeled three-dimensional image is specifically that an original three-dimensional gray image of an artificial tissue or an organoid is split into a plurality of original two-dimensional gray images, and then contrast enhancement is carried out on the two-dimensional gray images to obtain corresponding contrast-enhanced two-dimensional gray images; combining the two-dimensional gray level images with enhanced contrast to obtain a three-dimensional image with enhanced contrast; labeling the original three-dimensional gray level image to be used as a three-dimensional image data set, wherein the label is a three-dimensional image with the contrast ratio corresponding to the original three-dimensional gray level image enhanced;
and (3): constructing an artificial tissue or organoid segmentation deep learning network, and training by using a training set; finally, testing and verifying the trained artificial tissue or organoid segmentation deep learning network by using a test set;
the artificial tissue or organoid segmentation deep learning network adopts EG-Net, ResNet50v2, VGG19, Xception or DenseNet121CNN, the input of the artificial tissue or organoid segmentation deep learning network is a two-dimensional gray image, and the output of the artificial tissue or organoid segmentation deep learning network is a two-dimensional gray image after contrast enhancement; finally, combining all the contrast-enhanced two-dimensional gray images corresponding to the original three-dimensional gray image;
the artificial tissue or organoid segmentation deep learning network adopts Resnet-3D, the input of the Resnet-3D is an original three-dimensional gray image, and the output of the Resnet-3D is a three-dimensional image after contrast enhancement;
and (4): the artificial tissue or the organoid is segmented by utilizing the artificial tissue or the organoid segmentation deep learning network after test and verification to obtain a three-dimensional image of the artificial tissue or the organoid region;
and (5): quantifying the three-dimensional image of the artificial tissue or organoid region obtained in the step (4):
calculating the morphological index and ATP value of the artificial tissue or organoid in each culture hole of the corresponding culture plate from the last day of culture according to the connected domain, wherein the morphological index is as follows: total volume, total surface area, total sphericity, number, average volume, average surface area, average sphericity; carrying out standardization processing on morphological indexes of the artificial tissues or the organoids and ATP values to construct a matrix calculation covariance matrix; performing singular value decomposition on the covariance matrix to obtain a right singular matrix and a singular value matrix which are sequenced from large to small according to singular values; and extracting principal components according to the variance contribution rate, taking the column vector of the right singular matrix as a coefficient of a certain principal component expression, taking the ratio of the singular value to the total singular value as a coefficient of a comprehensive index, and finally obtaining the comprehensive index of the artificial tissue or the organoid.
2. The method according to claim 1, wherein the contrast enhancement of the two-dimensional gray scale images is performed by performing one or more of an attenuation coefficient algorithm, a logarithmic intensity variance algorithm, a delay correlation attenuation speed algorithm, an inverse power law index alpha algorithm of speckle fluctuation spectrum, an autocorrelation-based motion amplitude M algorithm, an interframe complex correlation algorithm, and an image preprocessing algorithm on the two-dimensional gray scale images.
3. The method of claim 1, wherein the original two-dimensional gray scale image is subjected to a data enhancement step prior to contrast enhancement.
4. The method of claim 3, wherein the data enhancement is one or more of a combination of warping and non-warping enhancement, image rotation, and image scaling.
5. The method according to claim 1, characterized in that step (5) is in particular:
5-1, obtaining the pixel number Volume _ number occupied by the Volume of a single artificial tissue or organ-like part and the pixel number Surface _ number occupied by the Surface area in the three-dimensional image according to the three-dimensional image of the artificial tissue or organ-like part obtained by segmentation in the step (4); then according to the size of the three-dimensional single pixel and the size of the two-dimensional single pixel, respectively obtaining a single artificial tissue or organ-like volume and a surface area organic surface by combining formulas (1) to (2);
Organoid_volume＝Volume_number*(PixelSize_x*PixelSize_y*PixelSize_z)
formula (1)
Organic _ surface _ number (user ixelSize _ x PixelSize _ y) formula (2)
PixelSize_x＝PixelSize_y＝PixelSize_z
The pixel size of the original three-dimensional gray image is represented by PixelSize _ x, PixelSize _ y and PixelSize _ z;
5-2 obtaining the sphericity of the single artificial tissue or Organoid according to the volume Organoid _ volume and the surface area Organoid _ surface of the single artificial tissue or Organoid and combining the formula (3):
5-3, calculating the total number of the artificial tissues or the organoids in a single culture hole according to the three-dimensional images of the three-dimensional connected domain:
connected _ domains _ number (4)
Wherein Connected domains number represents the number of Connected domains in the three-dimensional image;
calculating the total volume of the artificial tissues or organoids of a single culture well:
wherein Organoid _ volume (i) represents the ith artificial tissue or Organoid volume within a single culture well;
calculate total surface area of artificial tissue or organoid for individual culture wells:
wherein Organoid surface (i) represents the surface area of the ith artificial tissue or Organoid within a single culture well;
calculating the total sphericity of the artificial tissues or organoids of a single culture well:
wherein Organoid _ sphere (i) indicates the ith artificial tissue or Organoid sphericity within a single culture well;
5-4 calculating the average volume of the single artificial tissue or organoid in a culture well according to the total volume of the artificial tissue or organoid in the single culture well:
Organoid_average_volume＝Organoid_sum_volume/Organoid_number
formula (8)
Calculating the average surface area of the individual artificial tissues or organoids in a culture well based on the total surface area of the individual artificial tissues or organoids in the individual culture well:
Organoid_average_surface＝Organoid_sum_surface/Organoid_number
formula (9)
Calculating the average sphericity of the single artificial tissue or organoid in a culture well according to the total sphericity of the artificial tissue or organoid in the single culture well:
Organoid_average_spherecity＝Organoid_sum_spherecity/Organoid_number
formula (10)
5-5 obtaining adenosine triphosphate Organoid _ ATP by a chemiluminescence measuring method for each culture hole of the corresponding culture plate from the last day of culture;
5-6, acquiring standardized artificial tissue or organoid indexes in each culture hole, wherein the indexes comprise total volume, total surface area, total sphericity, total number, average volume, average surface area, average sphericity and ATP;
wherein x1Expressing the total normalized artificial tissue or organoid volume in the jth culture well; organoid _ sum _ volume (j) represents the total volume of artificial tissue or Organoid in the jth culture well; m represents the number of culture wells;
wherein x2Represents the total surface area of the normalized artificial tissue or organoid in the jth well; organic _ sum _ surface (j) represents the total surface area of the artificial tissue or Organoid in the jth well;
wherein x3Expressing the total sphericity of the normalized artificial tissue or organoid in the jth culture well; organic _ sum _ specificity (j) represents the total sphericity of the artificial tissue or Organoid in the jth culture well;
wherein x4Expressing the total number of standardized artificial tissues or organoids in the jth culture well; organoid _ number (j) represents the total number of artificial tissues or organoids in the jth culture well;
wherein x5Represents the mean volume of the normalized artificial tissue or organoid in the jth well; organic _ average _ volume (j) represents the average volume of artificial tissue or Organoid in the jth culture well;
wherein x6Represents the normalized average surface area of the artificial tissue or organoid in the jth culture well; organic _ average _ surface (j) represents the average surface area of the artificial tissue or Organoid in the jth culture well;
wherein x7Represents the average sphericity of the normalized artificial tissue or organoid in the jth culture well;
organic _ average _ specificity (j) represents the average sphericity of the artificial tissue or Organoid in the jth culture well;
wherein x is8(ii) normalized artificial tissue or organoid ATP values in the jth culture well; organoid _ ATP (j) indicates ATP value of artificial tissue or Organoid in the jth well;
5-7, performing principal component analysis according to the obtained standardized artificial tissue or organoid indexes; the method comprises the following steps:
5-7-1, forming a matrix A by the indexes of the standardized artificial tissues or organoids, namely:
A＝[x1，x2，x3，x4，x5，x6，x7，x8]m×8formula (19)
Wherein m is the number of culture wells, where m >8, i.e. the number of sets of data, and matrix A is a matrix of m rows and 8 columns; 5-7-2, calculating the matrix A to obtain a covariance matrix B:
where cov () represents the covariance function, covariance matrix B is a matrix of 8 rows and 8 columns;
5-7-3 Singular Value Decomposition (SVD) is used to compute singular values and eigenvectors of the covariance matrix B:
B＝U∑VTformula (21)
The matrixes U and V respectively represent a left singular matrix and a right singular matrix which are both orthogonal matrixes, column vectors of V and U are respectively base vectors of a row space and a column space of the covariance matrix B, and sigma represents a diagonal matrix containing singular values;
BTB＝V∑2VTformula (22)
The right singular matrix V uses the right singular matrix of the SVD to reduce the dimension of the column number, and the first k eigenvectors of the PCA are the first k columns of the V; knowledge of the application of linear algebra to the formula (22) solves for V, BTThe matrix formed by the characteristic vectors of B is the V matrix in SVD, BTMatrix composed of B eigenvaluesIs just2A matrix;
due to BTB is a square matrix which can be subjected to characteristic decomposition; the obtained characteristic value and the characteristic vector need to satisfy the following conditions:
(BTB)vk＝λkvkformula (23)
Wherein λkIs a characteristic value, vkIs λkThe corresponding feature vector, k ═ 1, 2.., 7; sigma denotes a diagonal matrix containing singular values2Representing a diagonal matrix containing eigenvalues, i.e. singular values being the evolution of the eigenvalues
Eigenvalue matrix sigma2Sorting according to the eigenvalues from large to small, then sorting the singular value matrix sigma according to the singular values from large to small, and changing the column of V along with the change of sorting;
the ordered sigma2The matrix, the sigma matrix and the V matrix are as follows:
equation (26) can also be expressed as:
V＝[v1，v2，v3，v4，v5，v6，v7，v8]formula (27)
Wherein λkIs a characteristic value, λk＞λk+1(ii) a Characteristic value lambda1Is sigma2The value of the largest feature in the matrix,its corresponding feature vector v1＝[a11，a21，a31，a41，a51，a61，a71，a81]TThe first row of the V matrix is required to be arranged; characteristic value lambda8Is sigma2The smallest eigenvalue in the matrix, its corresponding eigenvector v8＝[a18，a28，a38，a48，a58，a68，a78，a88]TThe data need to be arranged in the last column of the V matrix;
selecting principal components according to the ratio of the singular values to the singular values greater than a threshold value alpha, wherein the ratio of the singular values is called a variance contribution rate, and a variance contribution rate b of the first principal component1And a second principal component variance contribution rate b2Namely:
b1+b2>threshold alpha type (30)
The size and activity of the space occupied by the artificial tissue or organoid composed of the first principal component F1:
F1＝-(a11*x1+a21*x2+a31*x3+a41*x4+a51*x5+a61*x6+a71*x7+a81*x8)
formula (31)
From the second principal component F2Phenotypic characteristics and quantities constituting artificial tissues or organoids:
F2＝-(a12*x1+a22*x2+a32*x3+a42*x4+a52*x5+a62*x6+a72*x7+a82*x8)
formula (32)
The size and activity of the space occupied by the artificial tissue or Organoid, and the phenotypic characteristics and quantity form an artificial tissue or Organoid comprehensive index Organoid _ F:
Organoid_F＝b1*F1+b2*F2formula (33).
6. The method of claim 5, wherein said interconnected domains have an equivalent diameter of 32 microns or greater.
7. An OCT-based artificial tissue or organoid detection device, characterized by comprising:
the image acquisition module is used for receiving the three-dimensional gray level image of the artificial tissue or the organoid acquired by the OCT equipment;
the artificial tissue or organoid segmentation module is used for segmenting the three-dimensional gray level image of the artificial tissue or organoid by utilizing an artificial tissue or organoid segmentation deep learning network to obtain a three-dimensional image of the artificial tissue or organoid region;
the first calculation module is used for calculating the morphological index and ATP value of the artificial tissue or organoid in each culture hole of the corresponding culture plate from the last day of culture;
the second calculation module is used for carrying out standardized processing on the morphological index of the artificial tissue or the organoid and the ATP value to construct a matrix calculation covariance matrix, and then carrying out singular value decomposition on the matrix calculation covariance matrix to obtain a right singular matrix and a singular value matrix; and extracting main components according to the variance contribution rate, and constructing to obtain the artificial tissue or organoid comprehensive index.
8. A method for testing drug sensitivity, which is characterized in that the comprehensive index of the artificial tissue or organoid obtained by the method of any one of claims 1 to 6 is used for judging the curative effect of the artificial tissue or organoid under different drug concentration.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210110925.6A CN114494217A (en) | 2022-01-29 | 2022-01-29 | Method and device for detecting artificial tissues and organoids |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210110925.6A CN114494217A (en) | 2022-01-29 | 2022-01-29 | Method and device for detecting artificial tissues and organoids |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114494217A true CN114494217A (en) | 2022-05-13 |
Family
ID=81478989
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210110925.6A Pending CN114494217A (en) | 2022-01-29 | 2022-01-29 | Method and device for detecting artificial tissues and organoids |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114494217A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115082529A (en) * | 2022-06-30 | 2022-09-20 | 华东师范大学 | System and method for collecting and analyzing multi-dimensional information of gross tissue |
| CN116012838A (en) * | 2022-12-30 | 2023-04-25 | 创芯国际生物科技(广州)有限公司 | Artificial intelligence-based organoid activity recognition method and system |
| CN117274702A (en) * | 2023-09-27 | 2023-12-22 | 湖南景为电子科技有限公司 | Automatic classification method and system for cracks of mobile phone tempered glass film based on machine vision |
-
2022
- 2022-01-29 CN CN202210110925.6A patent/CN114494217A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115082529A (en) * | 2022-06-30 | 2022-09-20 | 华东师范大学 | System and method for collecting and analyzing multi-dimensional information of gross tissue |
| CN116012838A (en) * | 2022-12-30 | 2023-04-25 | 创芯国际生物科技(广州)有限公司 | Artificial intelligence-based organoid activity recognition method and system |
| CN116012838B (en) * | 2022-12-30 | 2023-11-07 | 创芯国际生物科技(广州)有限公司 | Artificial intelligence-based organoid activity recognition method and system |
| CN117274702A (en) * | 2023-09-27 | 2023-12-22 | 湖南景为电子科技有限公司 | Automatic classification method and system for cracks of mobile phone tempered glass film based on machine vision |
| CN117274702B (en) * | 2023-09-27 | 2024-03-29 | 湖南景为电子科技有限公司 | Automatic classification method and system for cracks of mobile phone tempered glass film based on machine vision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114494217A (en) | Method and device for detecting artificial tissues and organoids | |
| US20220230310A1 (en) | Three-dimensional object segmentation of medical images localized with object detection | |
| Zhang et al. | Photoacoustic image classification and segmentation of breast cancer: a feasibility study | |
| US7283652B2 (en) | Method and system for measuring disease relevant tissue changes | |
| US8059879B2 (en) | Brain function analysis apparatus and method | |
| CN110503635B (en) | Hand bone X-ray film bone age assessment method based on heterogeneous data fusion network | |
| Li et al. | Wavelet-based segmentation of renal compartments in DCE-MRI of human kidney: initial results in patients and healthy volunteers | |
| CN107615066A (en) | For biological cell and the interference system of biologic artifact and method including sperm | |
| CN109410167A (en) | A kind of analysis method and Related product of 3D galactophore image | |
| Hao et al. | Breast cancer histopathological images classification based on deep semantic features and gray level co-occurrence matrix | |
| CN101342075A (en) | Multi-optical spectrum autofluorescence dislocation imaging reconstruction method based on single view | |
| CN105842642A (en) | Fractional anisotropy microstructure characteristic extraction method based on kurtosis tensor and apparatus thereof | |
| CN112419320B (en) | Cross-modal heart segmentation method based on SAM and multi-layer UDA | |
| Gong et al. | Pixel level segmentation of early-stage in-bag rice root for its architecture analysis | |
| CN104217437A (en) | Lesion area partitioning method for prostate KVCT image | |
| CN111784704A (en) | MRI coxitis disease segmentation and classification automatic quantitative grading sequential method | |
| CN101947103A (en) | Optical bioluminescence tomography method | |
| Zhang et al. | Multispectral heterogeneity detection based on frame accumulation and deep learning | |
| Cao et al. | 3D convolutional neural networks fusion model for lung nodule detection onclinical CT scans | |
| Ivanovici et al. | Fractal dimension and lacunarity of psoriatic lesions-a colour approach | |
| Wei et al. | A machine vision system for automated non-invasive assessment of cell viability via dark field microscopy, wavelet feature selection and classification | |
| CN113299372B (en) | Photoacoustic pathological image processing method, storage medium and terminal device | |
| CN107076820A (en) | Method for assessing and improving the quality of data in fine-structure distribution data | |
| CN115101150A (en) | Specimen collection method for clinical tumor operation in general surgery department | |
| CN115312189A (en) | Construction method of breast cancer neoadjuvant chemotherapy curative effect prediction model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |