WO2024203936A1 - 解析装置、予測システム、及び予測方法 - Google Patents

解析装置、予測システム、及び予測方法 Download PDF

Info

Publication number
WO2024203936A1
WO2024203936A1 PCT/JP2024/011453 JP2024011453W WO2024203936A1 WO 2024203936 A1 WO2024203936 A1 WO 2024203936A1 JP 2024011453 W JP2024011453 W JP 2024011453W WO 2024203936 A1 WO2024203936 A1 WO 2024203936A1
Authority
WO
WIPO (PCT)
Prior art keywords
tissue
cell
learning
image
vascular network
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.)
Ceased
Application number
PCT/JP2024/011453
Other languages
English (en)
French (fr)
Japanese (ja)
Inventor
賢一 坂根
史朗 北野
寿美 上野
竜司 加藤
健二郎 田中
咲希 林
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tokai National Higher Education and Research System NUC
Toppan Holdings Inc
Original Assignee
Tokai National Higher Education and Research System NUC
Toppan Holdings Inc
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tokai National Higher Education and Research System NUC, Toppan Holdings Inc filed Critical Tokai National Higher Education and Research System NUC
Priority to JP2025510767A priority Critical patent/JPWO2024203936A1/ja
Publication of WO2024203936A1 publication Critical patent/WO2024203936A1/ja
Priority to US19/335,634 priority patent/US20260017788A1/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G06V20/695—Preprocessing, e.g. image segmentation
    • C—CHEMISTRY; METALLURGY
    • C12—BIOCHEMISTRY; BEER; SPIRITS; WINE; VINEGAR; MICROBIOLOGY; ENZYMOLOGY; MUTATION OR GENETIC ENGINEERING
    • C12M—APPARATUS FOR ENZYMOLOGY OR MICROBIOLOGY; APPARATUS FOR CULTURING MICROORGANISMS FOR PRODUCING BIOMASS, FOR GROWING CELLS OR FOR OBTAINING FERMENTATION OR METABOLIC PRODUCTS, i.e. BIOREACTORS OR FERMENTERS
    • C12M1/00—Apparatus for enzymology or microbiology
    • C12M1/34—Measuring or testing with condition measuring or sensing means, e.g. colony counters
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/50—Depth or shape recovery
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/72—Data preparation, e.g. statistical preprocessing of image or video features
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G06V20/698—Matching; Classification
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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
    • G06T2207/30024—Cell structures in vitro; Tissue sections in vitro
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G06T2207/30101—Blood vessel; Artery; Vein; Vascular
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/03—Recognition of patterns in medical or anatomical images

Definitions

  • the present disclosure relates to an analysis device, a prediction system, and a prediction method.
  • This application claims priority based on Japanese Patent Application No. 2023-047986, filed in Japan on March 24, 2023, the contents of which are incorporated herein by reference.
  • Patent Document 1 discloses a technique for evaluating the effects of anticancer drugs, in which cells are cultured to create a three-dimensional cell tissue in which the cancer cells coexist with interstitium, such as endothelial cells and fibroblasts. Cancer cells are seeded into the three-dimensional cell tissue, which is then cultured and an anticancer drug is administered. This makes it possible to evaluate the effects of an anticancer drug on cancer cells present in a three-dimensional cell tissue that is closer to a living organism than cells grown on a flat plate.
  • the present disclosure has been made in light of these circumstances, and aims to provide an analysis device, a prediction system, and a prediction method that can predict the characteristics of a cellular tissue produced from a cell at an early stage after the start of cell culture.
  • the analysis device of the first aspect of the present disclosure includes a first acquisition unit that acquires cell characteristic data indicating characteristics of a learning cell, which is a cell that can become a cellular tissue having a vascular network structure, based on a cell image of the learning cell, a second acquisition unit that acquires tissue characteristic data indicating characteristics of the learning tissue, which is a cellular tissue produced by culturing the learning cell, and a generation unit that generates correspondence data that associates the cell characteristic data with the tissue characteristic data.
  • the prediction system of the second aspect of the present disclosure includes the analysis device of the first aspect described above, a trained model generation unit that generates a trained model that predicts the characteristics of a cellular tissue produced from a target cell based on a target cell image in which the target cell, which is the subject of estimation, is captured by having a trained model learn the correspondence between the cell and the cellular tissue using the correspondence data generated by the analysis device as a training data set, a third acquisition unit that acquires target cell characteristic data based on the target cell image, and a prediction unit that predicts the characteristics of a cellular tissue produced by culturing the target cell using the trained model generated by the trained model generation unit.
  • the prediction method of the third aspect of the present disclosure is a prediction method in which a computer makes predictions, in which a first acquisition unit acquires cell characteristic data indicating characteristics of a learning cell, which is a cell that can become a cellular tissue having a vascular network structure, based on a cell image obtained by capturing the learning cell, a second acquisition unit acquires tissue characteristic data indicating characteristics of a learning tissue, which is a cellular tissue created by culturing the learning cell, a generation unit generates correspondence data that associates the cell characteristic data with the tissue characteristic data, a trained model generation unit uses the correspondence data as a training data set to cause a learning model to learn the correspondence between cells and cellular tissues, thereby generating a trained model that predicts characteristics of a cellular tissue created from a target cell, which is an estimation target, based on a target cell image obtained by capturing the target cell, a third acquisition unit acquires target cell characteristic data based on the target cell image, and a prediction unit predicts characteristics of a cellular tissue created by
  • FIG. 1 is a block diagram showing an example of the configuration of a prediction system according to an embodiment.
  • FIG. 1 is a diagram for explaining a learning dataset according to an embodiment.
  • FIG. 2 is a diagram showing an example of an organization of an embodiment.
  • FIG. 2 is a diagram for explaining a process performed by a prediction system according to an embodiment.
  • FIG. 1 is a diagram showing an example of a learning cell image according to an embodiment.
  • FIG. 13 is a diagram showing an example of a learning tissue image according to an embodiment.
  • FIG. 13 is a diagram showing an example of a learning tissue image according to an embodiment.
  • FIG. 13 is a diagram showing an example of a learning tissue image according to an embodiment.
  • FIG. 13 is a diagram showing an example of quality evaluation of a learning structure image according to an embodiment.
  • FIG. 13 is a diagram showing an example of quality evaluation of a learning tissue image according to an embodiment.
  • FIG. 13 is a diagram showing an example of quality evaluation of a learning tissue image according to an embodiment.
  • FIG. 13 is a diagram showing an example of quality evaluation of a learning structure image according to an embodiment.
  • FIG. 13 is a diagram showing an example of quality evaluation of a learning structure image according to an embodiment.
  • FIG. 2 is a sequence diagram showing a flow of processing performed by the prediction system of the embodiment.
  • the tissue characteristic data may be a measurement result of measuring the shape of the tissue without using an image, or a result of an expert evaluating the quality of the tissue by visual inspection.
  • the tissue characteristic data By acquiring tissue characteristic data based on a quantitative index, it becomes possible to manage the quality of the tissue, which has a large variation and is difficult to manage in the first place.
  • an example of predicting the quality of the tissue using a prediction model will be described as an example, but is not limited to this.
  • the prediction system 1 is a system for predicting the properties of a cellular tissue produced from cells.
  • a cellular tissue having a vascular network structure is the target will be described as an example, but is not limited to this.
  • the prediction system 1 can be applied to predict the properties of any cellular tissue produced from cells.
  • the cellular tissue may be referred to simply as "tissue.”
  • FIG. 1 is a block diagram showing an example of the configuration of a prediction system 1 according to an embodiment.
  • the prediction system 1 includes, for example, an analysis device 10, a learning device 20, a prediction device 30, and an imaging terminal 40 (imaging terminals 40-1 to 40-3).
  • the analysis device 10 is a computer, and can be, for example, a server device, a cloud, a PC (Personal Computer), etc.
  • the analysis device 10 generates a learning dataset.
  • the learning dataset is data that is used to train a learning model, and is information that is a set in which input and output are associated with each other.
  • the input here is the information that is input to the learning model when it is trained.
  • the output is information that indicates the correct answer that the learning model should output for the input.
  • the learning model adjusts its internal parameters so that the correct output is output for the input.
  • a learning model that can output an accurate prediction value for an untrained input is adopted as a trained model.
  • the training dataset generated by the analysis device 10 is information that associates cell characteristic data as input with tissue characteristic data as output.
  • the training dataset is an example of "corresponding data.”
  • Cell characteristic data is information that indicates the characteristics of a cell. Details of the cell characteristic data will be described later.
  • Tissue characteristic data is information that indicates the characteristics of a tissue. Details of the tissue characteristic data will be described later.
  • the learning device 20 is a computer, and may be, for example, a server device, a cloud, a PC (Personal Computer), etc.
  • the learning device 20 generates a learned model.
  • the learned model is a model that has been trained to predict output (tissue characteristic data) for unlearned input (cell characteristic data).
  • the learning device 20 has the learning model learn the correspondence between the input (cell characteristic data) and the output (tissue characteristic data) by having the learning model learn the learning dataset generated by the analysis device 10. In this way, the learned model is generated.
  • the prediction device 30 is a computer, and can be, for example, a server device, a cloud, a PC (Personal Computer), etc.
  • the prediction device 30 predicts the output (tissue characteristic data) for the input (cell characteristic data) using the trained model generated by the learning device 20. This makes it possible to predict the characteristics of a tissue produced from a cell at an early stage after starting cell culture.
  • the imaging terminal 40 is a computer that captures images, and may be, for example, a digital camera or a communication device including a camera module, such as a mobile phone, smartphone, or tablet terminal.
  • the imaging terminal 40 captures an image of the imaging target and obtains the captured image.
  • a microscope may be used to capture an enlarged image of the imaging target.
  • the imaging target is a cell or tissue, it is desirable to capture an image that is enlarged to an extent that the characteristics of the cell or tissue can be analyzed.
  • the subjects to be imaged here are learning cells, learning tissues, and target cells.
  • Learning cells are cells that can become a specific tissue, for example a cellular tissue having a vascular network-like structure, and are cells used in the learning dataset.
  • Learning tissues are tissues created from learning cells.
  • Target cells are prediction targets, and are cells that predict the properties of tissues created from the target cells.
  • FIG. 1 shows an example in which the imaging terminal 40-1 images the learning cells, the imaging terminal 40-2 images the learning tissue, and the imaging terminal 40-3 images the target cells, but this is not limited to the example.
  • One imaging terminal 40 may image all or a combination of the learning cells, learning tissue, and target cells.
  • the analysis device 10 includes, for example, a cell feature extraction unit 11, a tissue feature extraction unit 12, a quality evaluation unit 13, a learning dataset generation unit 14, and a learning dataset storage unit 15.
  • the cell feature extraction unit 11 acquires cell feature data indicating the features of the learning cell based on the learning cell image GC.
  • the learning cell image GC is an image of a learning cell.
  • the learning cell is a cell that can become a specific tissue, for example a cellular tissue having a vascular network-like structure.
  • the learning cell image GC is used as input for a learning dataset.
  • the cell feature extraction unit 11 acquires the learning cell image GC captured by the imaging terminal 40-1, and acquires cell feature data by performing image processing or the like on the acquired learning cell image GC.
  • the cell feature extraction unit 11 outputs the acquired cell feature data to the learning dataset generation unit 14.
  • the tissue characteristic extraction unit 12 acquires tissue characteristic data indicating the characteristics of the learning tissue based on the training tissue image GF.
  • the training tissue image GF is an image of the training tissue.
  • the training tissue is a cellular tissue produced by culturing training cells.
  • the training tissue image GF is used as an output of a training dataset.
  • the tissue characteristic extraction unit 12 acquires the training tissue image GF captured by the imaging terminal 40-2, and acquires tissue characteristic data by performing image processing, etc. on the acquired training tissue image GF.
  • the tissue characteristic extraction unit 12 outputs the acquired tissue characteristic data to the quality evaluation unit 13.
  • the quality evaluation unit 13 evaluates the quality of the tissue.
  • the quality evaluation unit 13 acquires tissue characteristic data from the tissue characteristic extraction unit 12, and based on the acquired tissue characteristic data, judges that, for example, a tissue in which the amount of vascular network formation is equal to or greater than a threshold is of good quality.
  • the quality evaluation unit 13 may acquire the result of a human evaluating the quality of the tissue by visual inspection, etc.
  • the quality evaluation unit 13 may acquire the result of evaluating the quality of the tissue based on the learning tissue image GF using a machine learning technique.
  • the quality evaluation unit 13 may judge the quality of the tissue based on a binary value of good or bad quality, or may judge the quality of the tissue based on multiple levels such as good quality, slightly good quality, and bad quality.
  • the training dataset generation unit 14 generates a training dataset.
  • the training dataset generation unit 14 generates a training dataset by matching cell characteristic data as input with tissue characteristic data as output.
  • the tissue characteristic data as output is data related to tissues created from cells from which cell characteristic data was obtained as input.
  • the training dataset generation unit 14 stores the generated training dataset in the training dataset storage unit 15.
  • the learning dataset storage unit 15 stores the learning dataset.
  • the learning dataset storage unit 15 is composed of a storage medium such as a HDD (Hard Disk Drive), flash memory, EEPROM (Electrically Erasable Programmable Read Only Memory), RAM (Random Access read/write Memory), ROM (Read Only Memory), or a combination of these.
  • a storage medium such as a HDD (Hard Disk Drive), flash memory, EEPROM (Electrically Erasable Programmable Read Only Memory), RAM (Random Access read/write Memory), ROM (Read Only Memory), or a combination of these.
  • the learning device 20 includes, for example, a trained model generation unit 21 and a trained model storage unit 22.
  • the trained model generation unit 21 generates a trained model.
  • the trained model generation unit 21 generates a trained model by having the trained model learn the correspondence between cells and tissues using a training dataset.
  • the trained model generation unit 21 stores information indicating the generated trained model, such as parameter setting values within the model, in the trained model storage unit 22.
  • the trained model storage unit 22 stores information indicating the trained model generated by the trained model generation unit 21.
  • the trained model storage unit 22 is configured with a storage medium such as a HDD, a flash memory, an EEPROM, a RAM, a ROM, or a combination of these.
  • models that are applied to existing machine learning such as deep learning using CNN (Convolutional Neural Network), DCNN (Deep CNN), decision trees, hierarchical Bayes, and SVM (Support Vector Machine), may be used.
  • CNN Convolutional Neural Network
  • DCNN Deep CNN
  • decision trees hierarchical Bayes
  • SVM Small Vector Machine
  • the prediction device 30 includes, for example, a cell feature extraction unit 31, a prediction unit 32, a vascular network prediction model storage unit 33, and a prediction result storage unit .
  • the cell feature extraction unit 31 acquires cell feature data indicating the features of the target cell based on the target cell image TC.
  • the target cell image TC is an image of the target cell.
  • the target cell is a cell for which the quality of the tissue to be produced from that cell is predicted.
  • the cell feature extraction unit 31 acquires the target cell image TC captured by the imaging terminal 40-3, and acquires cell feature data by performing image processing, etc. on the acquired target cell image TC.
  • the cell feature extraction unit 31 outputs the acquired cell feature data to the prediction unit 32.
  • the prediction unit 32 inputs the cell characteristic data of the target cell into the vascular network prediction model, and determines the output obtained as the quality of the tissue to be produced from the target cell.
  • the vascular network prediction model is a model that predicts the quality of the tissue to be produced from the cell based on the cell characteristic data, and is a trained model stored in the trained model storage unit 22.
  • the prediction unit 32 acquires information of the trained model to be used in the vascular network prediction model by referring to the trained model storage unit 22, and stores it in the vascular network prediction model storage unit 33.
  • the prediction unit 32 constructs the vascular network prediction model by referring to the vascular network prediction model storage unit 33.
  • the prediction unit 32 inputs the cell characteristic data acquired from the cell characteristic extraction unit 31 into the vascular network prediction model, and acquires the output value output from the vascular network prediction model.
  • the vascular network prediction model outputs information indicating the quality of tissue produced from cells and the degree to which that quality is predicted to be achieved, such as information indicating that there is a 70% probability that good quality tissue will be produced and a 30% probability that poor quality tissue will be produced.
  • the prediction unit 32 sets the output value output from the vascular network prediction model as the prediction result.
  • the prediction unit 32 stores the prediction result in the prediction result storage unit 34.
  • the vascular network prediction model storage unit 33 stores information indicating the vascular network prediction model.
  • the prediction result storage unit 34 stores the prediction result predicted by the prediction unit 32.
  • the vascular network prediction model storage unit 33 and the prediction result storage unit 34 are configured by storage media such as HDD, flash memory, EEPROM, RAM, ROM, etc., or a combination of these.
  • the learning dataset (cell characteristic data and tissue characteristic data) of this embodiment will now be described with reference to Figures 2 to 8.
  • FIG. 2 is a diagram for explaining the training dataset of the embodiment.
  • the process for producing a cell tissue includes a number of steps, for example, steps 1 to 4, and it takes about two weeks to go through all the steps.
  • the first step is a step of performing pre-culture.
  • Pre-culture is a step of growing cells on a plate before producing a three-dimensional cell tissue.
  • cells are cultured on a plate.
  • the second process is a process of forming the cells cultured on the plate into a three-dimensional cellular tissue.
  • the cells grown in the first process are collected and recovered, and a cell aggregate is formed on the substrate.
  • the cells grown in the first process may be dispensed to form a plurality of three-dimensional cellular tissues.
  • the third process is a process of culturing a three-dimensional cellular tissue to form a vascular network.
  • a vascular network is formed as a cellular tissue by culturing the cell aggregate formed on the substrate in the second process. In this way, a cellular tissue is produced.
  • the fourth step is a step of evaluating the cellular tissue. In the fourth step, the vascular network as the cellular tissue is visualized by staining or the like, and the quality of the vascular network is evaluated by visual inspection or the like.
  • the image of the cells cultured in the first process is the learning cell image GC.
  • the image of the cell tissue, for example, a stained vascular network, in the fourth process is the learning tissue image GF.
  • the cell characteristic data is a learning cell image GC.
  • the tissue characteristic data is a learning tissue image GF.
  • the learning dataset DST is information that associates the learning cell image GC with the learning tissue image GF.
  • FIG. 3 is a diagram showing an example of an organization in this embodiment.
  • Figure 3 shows schematic images of cells and tissues corresponding to several specific passages from the early passage state to the late passage state.
  • cells in the early passage state have a regular cell shape, whereas cells in the late passage state are enlarged and lose their shape. In this way, the shape (appearance) of the cells changes with repeated passages.
  • FIG. 4 is a diagram for explaining the training dataset of the embodiment.
  • a vascular network prediction model is generated through each phase of experiment, analysis, and prediction.
  • a tissue is actually produced from cells, and a learning cell image GC and a learning tissue image GF are acquired during the production process.
  • cell characteristic data is obtained from the learning cell image GC
  • tissue characteristic data is obtained from the learning tissue image GF
  • the cell characteristic data and the tissue characteristic data are matched to generate a learning dataset DST.
  • a vascular network prediction model generated by performing machine learning using the learning dataset DST is used to predict the quality of the tissue (vascular network) to be created from the cell from which the target cell image TC was obtained, based on the target cell image TC.
  • the cell features extracted from the learning cell image GC may be used as cell feature data.
  • the cell feature extraction unit 11 of the analysis device 10 acquires the cell features from the learning cell image GC.
  • the cell feature extraction unit 11 acquires, for example, statistics such as the size, brightness, or shape of the cells captured in the learning cell image GC as cell feature data.
  • the cell feature extraction unit 11 identifies the area in the image where the cell is captured, for example, by performing image processing on the learning cell image GC to extract the cellular portion. For example, the cell feature extraction unit 11 generates a learning cell image GC1# (see FIG. 5) from the learning cell image GC1 (see FIG. 5) by identifying the area in the image where the cell is captured.
  • the cell feature extraction unit 11 measures indices indicating the size of each cell contained in an image, such as the cell's area, perimeter, length (length in the height direction (longitudinal direction)), width (width in the horizontal direction (direction perpendicular to the longitudinal direction)), and length/width ratio, based on the identified area (area in which the cell is imaged), and calculates statistics such as the average value of the measured area and other indices as cell feature data.
  • the cell feature extraction unit 11 extracts, based on the cell contours, indices indicating the brightness of each cell in an image, such as the average brightness, standard deviation, and coefficient of variation in the area surrounded by the contours, as cell feature data.
  • the cell feature extraction unit 11 also calculates a representative value indicating the brightness of each cell (e.g., the average brightness), and calculates statistics indicating the brightness of the cell, such as the average value, as cell feature data, based on the indices for these cells.
  • the cell feature extraction unit 11 may treat the number of days that the cells imaged in the learning tissue image GF have been cultured as cell feature data.
  • the cell feature extraction unit 11 may acquire cell features along the timeline of cell culture as cell feature data. For example, the cell feature extraction unit 11 acquires cell feature data from a cell image on the first day of culture, and acquires cell feature data from the cell image on the second day of culture for a cell for which cell feature data has been acquired on the first day of culture.
  • the cell feature extraction unit 11 may treat the changes in features acquired on the first and second days of culture as cell feature data.
  • tissue characteristics extracted from the training tissue image GF may be used as tissue characteristic data.
  • the tissue characteristic extraction unit 12 of the analysis device 10 acquires the tissue characteristics from the training tissue image GF.
  • the tissue characteristic extraction unit 12 obtains, for example, the quality of the tissue captured in the learning tissue image GF (shown as "Grade” in Figure 4) as tissue characteristic data.
  • the tissue feature extraction unit 12 acquires, as tissue feature data, the quality evaluated by a user (human) who visually inspects the tissue captured in the learning tissue image GF.
  • the tissue feature extraction unit 12 may acquire, as tissue feature data, the result of judging the quality based on the tissue feature data, for example, the area of the vascular network.
  • the organization feature extraction unit 12 may determine whether the quality is good or bad using a machine learning technique. A method in which the organization feature extraction unit 12 determines whether the quality is good or bad using a machine learning technique will be described in detail later.
  • the quality assessment unit 13 may determine whether the quality of the above-mentioned tissue is good or bad (for example, using a machine learning technique).
  • the tissue characteristic extraction unit 12 also acquires statistics such as the size or shape of the tissue as tissue characteristic data. If the cellular tissue has a vascular network-like structure, the tissue characteristic data may be information indicating the characteristics of the vascular network. In this case, for example, the tissue characteristic data is statistics such as the quality of the vascular network captured in the tissue image, the density of the vascular network, or the shape.
  • the tissue feature extraction unit 12 performs image processing, such as thinning processing, on the training tissue image GF to extract tissues, and identifies the area in the image where the vascular network is captured. For example, the tissue feature extraction unit 12 generates training tissue images GF1# to GF4# (see FIG. 7) from the training tissue images GF1 to GF4 (see FIG. 6) by identifying the area in the image where the tissues are captured.
  • image processing such as thinning processing
  • the tissue feature extraction unit 12 divides the image into areas where tissue is captured (vascular network-like structure areas) and areas where tissue is not captured (non-vascular network-like structure areas) based on the training tissue image GF that identifies the areas where tissue is captured. For example, the tissue feature extraction unit 12 divides the training tissue image GF into multiple areas. If the outline of the container that contains the tissue during the culture process is circular when viewed from the imaging direction, for example, the tissue feature extraction unit 12 divides the tissue image equally into sectors having the same central angle centered at the center of the image. The tissue feature extraction unit 12 divides each of the equally divided small areas into vascular network-like structure areas and non-vascular network-like structure areas.
  • the tissue feature extraction unit 12 calculates an index indicating the size of the vascular network-like structure region in each of the equally divided small regions, for example, the area of the vascular network-like structure region in the small region, and calculates statistics of the calculated area, for example, the average area, variance, and the ratio of the number of small regions whose difference from the average is equal to or greater than the variance to the total number of divided small regions, as tissue feature data.
  • the tissue feature extraction unit 12 may also extract the total length of the vascular network as tissue as tissue feature data. For example, the tissue feature extraction unit 12 divides the area of the vascular network by performing thinning processing, and determines the number of pixels having pixel values corresponding to the vascular network in the thinned portion divided as the vascular network as the total length of the vascular network.
  • the tissue feature extraction unit 12 may also extract the number of endpoints and branch points of the vascular network as tissue as tissue feature data. For example, the tissue feature extraction unit 12 divides the area of the vascular network by performing a thinning process. The tissue feature extraction unit 12 acquires the endpoints and branch points of the thinned portion divided as the vascular network by performing pattern recognition or the like. The tissue feature extraction unit 12 extracts the acquired number of endpoints, number of branch points, and ratio of the number of endpoints to the number of branch points as tissue feature data.
  • tissue feature extraction unit 12 may also calculate, as tissue feature data, the statistics of the area of the vascular network structure region in each of the equally divided small regions in the learning tissue image GF before the thinning process.
  • the tissue feature extraction unit 12 may also calculate statistics of the area of non-vascular network structure regions as tissue feature data. For example, the tissue feature extraction unit 12 uses the learning tissue image GF before thinning processing to calculate statistics of the area of non-vascular network structure regions in each of the equally divided small regions.
  • the tissue feature extraction unit 12 calculates an index indicating the size of the non-vascular network structure area in each of the equally divided small regions, for example, the area of the non-vascular network structure area in the small region, and calculates statistics of the calculated area, for example, the average area, variance, and the ratio of the number of small regions whose difference from the average is equal to or greater than the variance to the total number of divided small regions, as tissue feature data.
  • the tissue characteristic extraction unit 12 may calculate, as tissue characteristic data, the average value and variance of the area of each non-vascular network structure region, and the ratio of the number of non-vascular network structure regions whose difference from the average value is equal to or greater than the variance value to the total number of non-vascular network structure regions present in the small region.
  • the tissue characteristic extraction unit 12 may calculate, as tissue characteristic data, the statistics of the statistics for each small region, for example, the average value, variance, and coefficient of variation of the first variance value as the statistics of the variance value (first variance value) of the areas of multiple non-vascular network structure regions present in each small region, and the ratio of the number of small regions whose difference from the average value is equal to or greater than the variance value to the total number of small regions.
  • FIG. 5 is a diagram showing an example of a learning cell image GC in an embodiment.
  • FIG. 5 shows an example of a learning cell image GC1 before image processing is applied, and a learning cell image GC1# after image processing is applied to the learning cell image GC1.
  • the image processing here is a region identification process that identifies a region in the image where a single cell or a cell group is captured.
  • the region identification process is performed using, for example, a contour extraction filter (hereinafter referred to as a cell image analysis filter) that extracts a shape similar to a single cell or a cell group.
  • a contour extraction filter hereinafter referred to as a cell image analysis filter
  • the size and shape of cells vary depending on the type of cell.
  • multiple filters with different internal parameter settings are prepared in the prediction system 1.
  • cell images are prepared by capturing images of multiple types of cultured cells with different origins. It is desirable to prepare a number of cell images that satisfy the desired prediction accuracy, generalization performance, etc. in the prediction system 1 to be constructed.
  • a cell image analysis filter is applied to each cell image, and it is determined, for example by visual inspection, whether or not individual cells or cell clusters have been correctly extracted from the cell image.
  • a cell image analysis filter for which the probability of correctly extracting individual cells or cell clusters is below a threshold is selected as an appropriate cell image analysis filter.
  • the threshold value here may be set arbitrarily in the prediction system 1 according to the desired performance.
  • FIGS. 6 and 7 are diagrams showing examples of training tissue images GF in an embodiment.
  • FIG. 6 shows training tissue images GF1-GF4 before image processing is applied.
  • FIG. 7 shows training tissue images GF1#-GF4# after image processing is applied to the training tissue images GF1-GF4 in FIG. 6 in an embodiment.
  • the image processing here is a process for identifying an area in the image where a vascular network is captured. For example, a process for extracting the mesh of the vascular network (hereinafter referred to as a vascular network analysis filter) is executed using a thinning process or the like.
  • a three-dimensional cellular tissue containing cancer cells can also be used, which is obtained by seeding and culturing cancer cells in a three-dimensional cellular tissue.
  • the prediction system 1 it is desirable to select an appropriate vascular network analysis filter for both three-dimensional cellular tissue not containing cancer cells and three-dimensional cellular tissue containing cancer cells.
  • the efficacy can also be evaluated, for example, using the area size of the cancer cells as an index. From this perspective, it is possible to evaluate the efficacy of a drug even if the formation status of the vascular network in the three-dimensional cellular tissue or the quality of the vascular network has not been evaluated.
  • tissue characteristic data may be extracted.
  • tissue characteristic data in a three-dimensional cellular tissue may be obtained using, for example, the degree to which it is easy to recognize the proliferation or death of cancer cells as an index, rather than the presence or absence of a vascular network.
  • the top row shows a tissue image of a three-dimensional cellular tissue containing cancer cells (simply referred to as "tissue containing cancer cells” in Figure 6).
  • the bottom row shows a tissue image of a three-dimensional cellular tissue not containing cancer cells (simply referred to as "tissue not containing cancer cells” in Figure 6).
  • Learning tissue images GF1 and GF2 show images of three-dimensional cellular tissue containing cancer cells.
  • Learning tissue images GF3 and GF4 show images of three-dimensional cellular tissue not containing cancer cells.
  • learning tissue image GF1# shows an image in which a vascular network analysis filter has been applied to learning tissue image GF1.
  • Learning tissue image GF2# shows an image in which a vascular network analysis filter has been applied to learning tissue image GF2.
  • Learning tissue image GF3# shows an image in which a vascular network analysis filter has been applied to learning tissue image GF3.
  • Learning tissue image GF4# shows an image in which a vascular network analysis filter has been applied to learning tissue image GF4.
  • the prediction system 1 selects a vascular network analysis filter that appropriately extracts the meshwork of a vascular network, regardless of whether or not it contains cancer cells. Alternatively, if the results of extracting the meshwork of a vascular network using a vascular network analysis filter significantly differ depending on whether the tissue contains cancer cells or does not contain cancer cells, that vascular network analysis filter is not adopted, or a different vascular network analysis filter is adopted for tissues that contain cancer cells and tissues that do not contain cancer cells.
  • Fig. 8 to Fig. 11 are diagrams showing an example of the quality evaluation of a learning tissue image GF according to an embodiment.
  • the quality evaluation of the learning structure image GF in the embodiment may be performed by the quality evaluation unit 13 .
  • tissue images are associated with the results of their quality assessment.
  • the quality is classified into one of three grades: Grade A, Grade B, and Grade C.
  • FIG. 8 shows the results of judging the quality of a vascular network from an image of the vascular network (vascular network image) using a machine learning technique.
  • the prediction system 1 may be configured to judge the quality of a vascular network from a vascular network image using a machine learning technique.
  • multiple images of a vascular network are prepared. Then, the user visually scores the state of formation of the vascular network in each vascular network image. Using the scored vascular network images, a clustering analysis is performed to classify the quality of the vascular network into groups such as Grade X, Grade Y, and Grade Z. A judgment model is generated to judge whether the quality of the vascular network is good (OK) or bad (NG) using the vascular network images used in the clustering analysis as training data. Any machine learning method may be used as a method for generating the judgment model. For example, the judgment model is a model that estimates the probability that the vascular network shown in an unlearned vascular network image is of Grade A quality.
  • the judgment model is a model that estimates the probability that the vascular network shown in an unlearned vascular network image is of Grade C quality.
  • the tissue feature extraction unit 12 may use the judgment model created in this way to judge the quality of a cellular tissue such as a vascular network.
  • the quality of the vascular network is determined to be Grade X.
  • the probability of being Grade C is greater than or equal to a threshold, the quality of the vascular network is determined to be Grade Z. The rest are classified as Grade Y.
  • FIG. 8 is a diagram showing an example of quality evaluation of a learning tissue image GF in an embodiment.
  • tissue images are associated with the results of their quality evaluation.
  • an example is shown in which the quality is classified into one of three grades: Grade A, Grade B, and Grade C.
  • the system may be configured to set a judgment threshold for judging whether the quality of the vascular network is good or bad based on the judgment result of the judgment model, i.e., whether it is Grade A, Grade B, or Grade C.
  • the quality of the vascular network for Grade A or Grade B vascular network images is determined to be good (OK).
  • the quality of the vascular network for Grade C vascular network images is determined to be poor (NG).
  • the quality of the vascular network of a Grade A vascular network image may be determined to be good (OK), and the quality of the vascular network of a Grade B or Grade C vascular network image may be determined to be poor (NG).
  • the quality of the vascular network of a Grade C vascular network image may be determined to be good (OK), and the quality of the vascular network of a Grade A vascular network image may be determined to be poor (NG).
  • More vascular networks do not necessarily mean better quality, and the quality of a tissue may be determined arbitrarily depending on the performance and quality desired for that tissue.
  • a judgment model may be used to evaluate the accuracy of prediction in the vascular network prediction model.
  • a vascular network image is prepared.
  • a cell image of pre-culture is prepared in the prepared vascular network image.
  • a judgment model is used to judge the quality of the vascular network shown in the vascular network image. Furthermore, a vascular network prediction model is used to judge the quality of the vascular network created by culturing the cells shown in the cell image. It is judged whether the judgment result using the judgment model matches the judgment result using the vascular network prediction model. If they match, it is judged that the vascular network prediction model made a correct prediction. On the other hand, if they do not match, it is judged that the vascular network prediction model made an incorrect prediction. For example, the ratio of the number of cell images for which the vascular network prediction model made a correct prediction to the total number of cell images used in the judgment is used as an index showing the accuracy of predictions in the vascular network prediction model.
  • FIG. 10 and 11 show the evaluation results of the prediction accuracy of the vascular network prediction model.
  • the quality of the vascular network of the vascular network image is judged as good (OK) or poor (NG) based on certain conditions.
  • the quality of the vascular network in the vascular network image is determined to be good (OK) or poor (NG) based on conditions different from those in FIG. 10 and 11, the judgment result in the judgment model is shown horizontally as "Truth.”
  • the judgment result in the vascular network prediction model is shown vertically as "Predict.”
  • Figs. 10 and 11 show an example in which 63 images are used for judgment.
  • the results of the judgment using the judgment model and the judgment using the vascular network prediction model agreed in that the quality of the vascular network was good (OK) for 19 images, and disagreed in that the quality of the vascular network was good (OK) for 4 images.
  • the judgment results using the judgment model and the judgment results using the vascular network prediction model agreed in that the vascular network quality was poor (NG) for 37 images, and disagreed in that the vascular network quality was poor (NG) for 3 images.
  • the judgment results using the judgment model and the judgment results using the vascular network prediction model agreed in that the vascular network quality was good (OK) for 43 images, and disagreed in that the vascular network quality was good (OK) for 4 images.
  • the probability that the quality of the vascular network is judged to be poor (NG), i.e., Grade C, in the judgment model is approximately 25% ((12+4)/(43+4+12+4) x 100). Also, in FIG. 11, the probability that the judgment model and the vascular network prediction model match when the quality of the vascular network is judged to be poor (NG) is 25%.
  • Fig. 12 shows the results of evaluating the prediction accuracy of the vascular network prediction model by reviewing the conditions for obtaining the evaluation results shown in Fig. 11. In the evaluation shown in Fig. 12, 14 images were used for the judgment.
  • the judgment results using the judgment model and the judgment results using the vascular network prediction model agreed in that the quality of the vascular network was good (OK) for five images, and disagreed in that the quality of the vascular network was good (OK) for two images.
  • the probability that the judgment results that the quality of the vascular network was good (OK) was consistent was approximately 71% ( ⁇ 5/(5+2) x 100).
  • the probability of a match in the judgment result that the quality of the vascular network is poor (hereinafter referred to as the NG hit rate) has been significantly improved in the evaluation results in Fig. 12 compared to the evaluation results shown in Fig. 11.
  • the NG hit rate in Fig. 11 was 25%
  • the NG hit rate in Fig. 12 has improved to 86%.
  • the quality of the vascular network is determined using a determination model that utilizes a machine learning technique, and the vascular network prediction model is trained using tissue characteristic data that includes the determination result of the quality of the vascular network, but the following first and second modified examples may also be adopted.
  • a deviation criterion (threshold) is used to judge the quality of the vascular network. Items expressed in numerical values are selected from tissue characteristic data that can be obtained by, for example, performing image processing on the tissue image for learning. The items of tissue characteristic data to be used may be appropriately selected depending on the performance of the target tissue, the structure of the vascular network, etc.
  • deviation criteria are set. These deviation criteria may also be adjusted as appropriate depending on the performance of the target tissue, the structure of the vascular network, etc.
  • the quality of the vascular network is judged using the MT method using data judged using the deviation criterion.
  • a group of data selected according to the purpose is defined as a unit space, and the Mahalanobis distance (MD) from the center of the unit space of the tissue characteristic data to be judged is obtained. If the MD is small, it can be judged that the configuration is close to the unit space, and if the MD is large, it can be judged that the configuration is far from the unit space.
  • the "group of data selected according to the purpose" corresponds to the data of the tissue characteristic data judged not to deviate from the deviation criterion in the first modified example described above. As described above, since the unit space is made up of a group of data selected according to the purpose, if the MD is small, it can be judged that the purpose is met, that is, the quality of the vascular network is good.
  • FIG. 13 is a sequence diagram showing the flow of processing performed by the prediction system 1 of the embodiment.
  • the analysis device 10 acquires image data of the learning cell image GC (step S10).
  • the analysis device 10 extracts cell characteristic data from the learning cell image GC using a cell analysis filter or the like (step S11). Meanwhile, the analysis device 10 acquires image data of the learning tissue image GF (step S12).
  • the analysis device 10 extracts tissue characteristic data from the learning tissue image GF using a vascular network analysis filter or the like (step S13).
  • the analysis device 10 also judges the quality of the tissue shown in the learning tissue image GF (step S14).
  • the analysis device 10 may judge the quality of the tissue using the tissue characteristic data, or may judge the quality of the tissue using a judgment model.
  • the analysis device 10 generates a training dataset by associating the cell characteristic data with the tissue characteristic data (step S15).
  • the analysis device 10 stores the generated training dataset in the training dataset storage unit 15 (step S16).
  • the learning device 20 acquires a learning dataset by, for example, referring to the learning dataset storage unit 15 (step S17).
  • the learning device 20 uses the learning dataset to cause the learning model to learn the correspondence between the cell characteristic data and the tissue characteristic data (step S18).
  • the learning device 20 determines whether or not a learning end condition is satisfied (step S19).
  • the learning end condition is, for example, that the accuracy of predictions made by the learning model is equal to or greater than a threshold, but is not limited to this, and the learning end condition may be set arbitrarily.
  • the learning device 20 stores the learned learning model as a trained model in the trained model storage unit 22 (step S20). On the other hand, if the learning termination condition is not met, the learning device 20 returns to step S17 and repeats the learning.
  • the prediction device 30 acquires a vascular network prediction model by, for example, referring to the learned model storage unit 22 (step S21).
  • the prediction device 30 acquires image data of the target cell image TC (step S22).
  • the prediction device 30 extracts cell characteristic data from the target cell image TC using a cell analysis filter or the like (step S23).
  • the prediction device 30 predicts the quality of a cell tissue produced by culturing the cells shown in the target cell image TC (step S24). For example, the prediction device 30 inputs the cell characteristic data extracted from the target cell image TC to the vascular network prediction model, thereby predicting the predicted value output from the vascular network prediction model as the quality of the cell tissue.
  • the prediction device 30 stores the prediction result in the prediction result storage unit 34.
  • the analysis device 10 of the embodiment includes a cell feature extraction unit 11 (first acquisition unit), a tissue feature extraction unit 12 (second acquisition unit), and a learning dataset generation unit 14 (generation unit).
  • the cell feature extraction unit 11 acquires cell feature data indicating the features of the cell (learning cell) based on the learning cell image GC (cell image).
  • the cell image is an image of a cell that can become a cellular tissue having a vascular network structure.
  • the tissue feature extraction unit 12 acquires tissue feature data indicating the features of the tissue (learning tissue) based on the learning tissue image GF (tissue image).
  • the tissue image is an image of a cellular tissue produced by culturing a cell (learning cell).
  • the learning dataset generation unit 14 generates a learning dataset (corresponding data).
  • the learning dataset is information that associates cell feature data with tissue feature data.
  • the analysis device 10 of the embodiment can generate information (training dataset) that associates cell characteristic data with tissue characteristic data. Therefore, by training the learning model with the training dataset, it is possible to generate a trained model that predicts tissue characteristics based on the cell characteristic data. In other words, it is possible to predict the characteristics of a cellular tissue produced from a cell at an early stage after starting cell culture.
  • the cell feature data is image data of the learning cell image GC (cell image).
  • the tissue feature data is image data of the learning tissue image GF (tissue image).
  • the image in the learning cell image GC (cell image) itself can be used as a feature of the cell.
  • the image in the learning tissue image GF (tissue image) itself can be used as a feature of the tissue. It is possible to generate a trained model that has learned the features of these images.
  • the cell characteristic data is data indicating at least one of the characteristics of the size, brightness, and shape of the cell captured in the learning cell image GC (cell image).
  • the analysis device 10 of the embodiment at least one of the size, brightness, and shape of the cell in the image can be regarded as a characteristic of the cell, and a trained model that has learned such cell characteristics can be generated.
  • the tissue characteristic data is data that indicates at least one of the characteristics of the size or shape of a vascular network-like structure region having a vascular network-like structure in the learning tissue image GF (tissue image).
  • tissue image tissue image
  • at least one of the size or shape of a vascular network-like structure region in the image can be regarded as a characteristic of the tissue, and it becomes possible to generate a trained model that has learned the characteristics of such tissue.
  • the tissue characteristic data is data indicating the characteristics of the size of non-vascular network structure areas that do not have a vascular network structure in the learning tissue image GF (tissue image).
  • the size of the non-vascular network structure areas in the image can be considered as a characteristic of the tissue, and a trained model that has learned the characteristics of such tissue can be generated.
  • the tissue characteristic data is data indicating the quality of the cellular tissue imaged in the learning tissue image GF (tissue image).
  • the quality of the vascular network can be regarded as a characteristic of the tissue, and a trained model that has learned the characteristics of such tissue can be generated.
  • the prediction system 1 of the embodiment includes an analysis device 10, a trained model generation unit 21, a cell feature extraction unit 31, and a prediction unit 32.
  • the trained model generation unit 21 generates a trained model.
  • the trained model is a model that predicts the features of a cellular tissue produced from a target cell based on a target cell image TC.
  • the trained model generation unit 21 uses a training data set to make the trained model learn the correspondence between cells and cellular tissues, thereby generating a trained model.
  • the cell feature extraction unit 31 acquires cell feature data of cells based on the target cell image TC.
  • the prediction unit 32 predicts the features of a cellular tissue produced by culturing a target cell using the trained model (vascular network prediction model).
  • the prediction system 1 of the embodiment is capable of predicting the features of a tissue produced from a cell based on the features of the cell. In other words, it is possible to predict the properties of a cellular tissue produced from a cell at an early stage after the start of cell culture.
  • the prediction system 1 and the analysis device 10 in the above-mentioned embodiment may be realized in whole or in part by a computer.
  • a program for realizing this function may be recorded in a computer-readable recording medium, and the program recorded in the recording medium may be read into the computer system and executed to realize the function.
  • computer system here includes hardware such as an OS and peripheral devices.
  • computer-readable recording medium refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, and storage devices such as hard disks built into the computer system.
  • the term "computer-readable recording medium” may include a medium that dynamically holds a program for a short period of time, such as a communication line when a program is transmitted via a network such as the Internet or a communication line such as a telephone line, and a medium that holds a program for a certain period of time, such as a volatile memory inside a computer system that is a server or client in such a case.
  • the above-mentioned program may be a program for realizing a part of the above-mentioned function, or may be a program that can realize the above-mentioned function in combination with a program already recorded in the computer system, or may be a program that is realized using a programmable logic device such as an FPGA.
  • Prediction system 10 For Analysis device 11
  • Quality evaluation unit 14 for Learning dataset generation unit 15: Learning dataset storage unit 20: Learning device 21: Learned model generation unit 22: Learned model storage unit 30: Prediction device 31: Cell feature extraction unit 32: Prediction unit 33: Vascular network prediction model storage unit 34: Prediction result storage unit

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Medical Informatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Chemical & Material Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Quality & Reliability (AREA)
  • Radiology & Medical Imaging (AREA)
  • Wood Science & Technology (AREA)
  • Zoology (AREA)
  • Biotechnology (AREA)
  • Organic Chemistry (AREA)
  • Geometry (AREA)
  • Microbiology (AREA)
  • Analytical Chemistry (AREA)
  • Sustainable Development (AREA)
  • Biochemistry (AREA)
  • General Engineering & Computer Science (AREA)
  • Genetics & Genomics (AREA)
  • Medicinal Chemistry (AREA)
  • Image Analysis (AREA)
  • Investigating Or Analysing Biological Materials (AREA)
PCT/JP2024/011453 2023-03-24 2024-03-22 解析装置、予測システム、及び予測方法 Ceased WO2024203936A1 (ja)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP2025510767A JPWO2024203936A1 (https=) 2023-03-24 2024-03-22
US19/335,634 US20260017788A1 (en) 2023-03-24 2025-09-22 Analysis device, prediction system, and prediction method

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2023047986 2023-03-24
JP2023-047986 2023-03-24

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US19/335,634 Continuation US20260017788A1 (en) 2023-03-24 2025-09-22 Analysis device, prediction system, and prediction method

Publications (1)

Publication Number Publication Date
WO2024203936A1 true WO2024203936A1 (ja) 2024-10-03

Family

ID=92905128

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2024/011453 Ceased WO2024203936A1 (ja) 2023-03-24 2024-03-22 解析装置、予測システム、及び予測方法

Country Status (3)

Country Link
US (1) US20260017788A1 (https=)
JP (1) JPWO2024203936A1 (https=)
WO (1) WO2024203936A1 (https=)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017183675A1 (ja) * 2016-04-19 2017-10-26 凸版印刷株式会社 薬剤耐性を有する細胞を作製する方法、抗ガン剤をスクリーニングするための方法、及び抗ガン剤スクリーニング用キット
WO2018101004A1 (ja) * 2016-12-01 2018-06-07 富士フイルム株式会社 細胞画像評価装置および細胞画像評価制御プログラム
WO2019039452A1 (ja) * 2017-08-21 2019-02-28 凸版印刷株式会社 抗がん効果の評価方法、及びがん免疫療法の奏効性予測方法
US20200105413A1 (en) * 2018-09-29 2020-04-02 Roche Molecular Systems, Inc. Multimodal machine learning based clinical predictor

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017183675A1 (ja) * 2016-04-19 2017-10-26 凸版印刷株式会社 薬剤耐性を有する細胞を作製する方法、抗ガン剤をスクリーニングするための方法、及び抗ガン剤スクリーニング用キット
WO2018101004A1 (ja) * 2016-12-01 2018-06-07 富士フイルム株式会社 細胞画像評価装置および細胞画像評価制御プログラム
WO2019039452A1 (ja) * 2017-08-21 2019-02-28 凸版印刷株式会社 抗がん効果の評価方法、及びがん免疫療法の奏効性予測方法
US20200105413A1 (en) * 2018-09-29 2020-04-02 Roche Molecular Systems, Inc. Multimodal machine learning based clinical predictor

Also Published As

Publication number Publication date
JPWO2024203936A1 (https=) 2024-10-03
US20260017788A1 (en) 2026-01-15

Similar Documents

Publication Publication Date Title
JP7583041B2 (ja) 組織画像分類用のマルチインスタンス学習器
JP7270058B2 (ja) 予測的組織パターン特定のためのマルチプルインスタンスラーナ
CN114846507B (zh) 用于使用人工智能（ai）模型进行非侵入性基因检测的方法和系统
Kan Machine learning applications in cell image analysis
US8831327B2 (en) Systems and methods for tissue classification using attributes of a biomarker enhanced tissue network (BETN)
JP2022528961A (ja) 胚を選択する方法及びシステム
Villaruz Deep convolutional neural network feature extraction for berry trees classification
CN108596046A (zh) 一种基于深度学习的细胞检测计数方法及系统
JP7001060B2 (ja) 情報処理装置、情報処理方法及び情報処理システム
US12406187B2 (en) Methods and systems for embryo classification using morpho-kinetic signatures
CN109145871A (zh) 心理行为识别方法、装置与存储介质
EP4616381A1 (en) Methods and systems for characterizing morphodynamic profiles of objects
WO2024203936A1 (ja) 解析装置、予測システム、及び予測方法
JP2020166711A (ja) 計数装置、計数方法、計数プログラムおよび記録媒体
Theagarajan et al. DeephESC: An automated system for generating and classification of human embryonic stem cells
CN114119446A (zh) 图像处理方法及装置、医学图像处理方法及装置
Wang et al. OC_Finder: A deep learning-based software for osteoclast segmentation, counting, and classification
Abisha et al. Impact of image pre-processing in image quality and segmentation for banana leaf disease prediction
Ravindra Analyzing the benthic cover of Crustose Coralline Algae using Mask-R CNN
Akter et al. BrainNet-7: A CNN model for diagnosing brain tumors from MRI images based on an ablation study
Asole Automated classification of retinopathy of prematurity in newborns
Caldas The Inference of Selective Sweep Parameters from their Genomic Footprint
Vasconcellos Caldas The Inference of Selective Sweep Parameters from their Genomic Footprint
Sugumar et al. Faux Reality Detector
Acharya et al. Enhanced Histopathological Image Classification through the fusion of Thepade Sorted Block Truncation Code and Otsu Binarization features

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 24780054

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2025510767

Country of ref document: JP

Kind code of ref document: A

WWE Wipo information: entry into national phase

Ref document number: 2025510767

Country of ref document: JP

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 24780054

Country of ref document: EP

Kind code of ref document: A1