WO2020004101A1 - 表示制御装置、表示制御方法及び表示制御プログラム - Google Patents
表示制御装置、表示制御方法及び表示制御プログラム Download PDFInfo
- Publication number
- WO2020004101A1 WO2020004101A1 PCT/JP2019/023833 JP2019023833W WO2020004101A1 WO 2020004101 A1 WO2020004101 A1 WO 2020004101A1 JP 2019023833 W JP2019023833 W JP 2019023833W WO 2020004101 A1 WO2020004101 A1 WO 2020004101A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- certainty factor
- display control
- graph
- gate
- image
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—Two-dimensional [2D] image generation
- G06T11/20—Drawing from basic elements
- G06T11/26—Drawing of charts or graphs
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2431—Multiple classes
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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
- G06N3/09—Supervised learning
-
- 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
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- 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
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
Definitions
- the present invention relates to a display control device, a display control method, and a display control program.
- the object of the present invention is to provide a display control device, a display control method, and a display control program capable of presenting the class determination performance by an artificial neural network in a format that can be easily interpreted by humans.
- One embodiment of the present invention is A certainty factor obtaining unit that obtains the certainty factor calculated by the certainty factor calculation unit that calculates the certainty factor of the class determination result of the input image, A display control unit configured to display a distribution of the certainty factor obtained by the certainty factor obtaining unit for each of the input images, wherein at least one of the display axes of the graph is a certainty factor axis indicating the certainty factor.
- a certainty factor obtaining unit that obtains the certainty factor calculated by the certainty factor calculation unit that calculates the certainty factor of the class determination result of the input image
- a display control unit configured to display a distribution of the certainty factor obtained by the certainty factor obtaining unit for each of the input images, wherein at least one of the display axes of the graph is a certainty factor axis indicating the certainty factor.
- One embodiment of the present invention is Obtain the confidence of the class judgment result of the input image, The distribution of the certainty factor for each input image, at least one of the display axes of the graph is displayed as a certainty factor axis indicating the certainty factor, This is a display control method.
- One embodiment of the present invention is In the computer provided in the display control device, Obtaining a certainty factor of the class determination result of the input image; A step of displaying the distribution of the input image of the certainty factor on at least one axis among display axes of a graph as a certainty factor axis indicating the certainty factor, Is a display control program for executing the program.
- the class determination performance by the artificial neural network can be presented in a format that is easy for humans to interpret.
- FIG. 2 is a diagram illustrating an example of a functional configuration of a class determination system according to the present embodiment.
- FIG. 3 is a diagram illustrating an example of a configuration of a class determination unit according to the present embodiment.
- FIG. 6 is a diagram illustrating an example of a calculation result of a certainty factor by the class determination system according to the present embodiment.
- FIG. 4 is a diagram illustrating an example of the operation of the class determination system according to the present embodiment. The figure which shows an example of the certainty factor graph image of this embodiment.
- FIG. 4 is a diagram illustrating an example of a graph displayed by a display control unit according to the embodiment.
- FIG. 4 is a diagram illustrating an example of gating by the display control device according to the present embodiment.
- FIG. 9 is a diagram illustrating a modification of gating by the display control device of the present embodiment.
- FIG. 9 is a diagram illustrating another modification of gating by the display control device of the present embodiment.
- the figure which shows an example of the structure of the neural network of this embodiment. Diagram showing an example of the relationship between the cell image and the certainty factor for each class The figure which shows an example of the histogram of a certainty factor, and a gating result The figure which shows an example of the result of having quantified the classification accuracy of the platelet aggregate.
- FIG. 1 shows an example of a functional configuration of a class determination system 1 of the present embodiment.
- the class determination system 1 is provided in an imaging cytometer (Imaging Cytometer) (not shown) which is a form of a flow cytometer (FCM) (not shown).
- the class determination system 1 acquires a cell image, and determines the type of a cell captured in the acquired cell image.
- the class CL to be determined by the class determination system 1 is “cell type”.
- the class determination system 1 performs the class determination using the cell type as the class CL, obtains the certainty factor PR for each class CL (that is, the result of estimating the cell type) as the determination result, and outputs the certainty factor. Displayed in the unit 230. Further, a gate GT is set on the certainty output unit 230, and it is determined whether or not the acquired image is included in the set gate GT (gate determination).
- the cell sorter having the cell sorting function outputs the gate determination result GR to a sorting mechanism (not shown). The sorting mechanism of the cell sorter sorts cells based on the gate determination result GR output from the class determination system 1.
- the class determination system 1 is described as being connected to a cell sorter of an imaging cytometry device and determining the type of a cell, but is not limited thereto.
- the class determination system 1 is not limited to the one used in the exemplified imaging cytometry.
- other cell observation applications such as a general flow cytometer, a mass cytometer (Mass Cytometer), and a microscope, and other than cells It may be used for classifying images.
- the class determination system 1 includes a display control device 10 and a class determination device 20.
- the display control device 10 and the class determination device 20 may be realized by one computer device or may be realized by separate computer devices.
- the class determination system 1 includes an operation detection unit 30 and a display unit 40.
- the operation detection unit 30 is an operation device such as a keyboard, a mouse, and a touch panel of a computer device, and detects an operation by an operator.
- the display unit 40 is, for example, a display device such as a liquid crystal display, and displays an image.
- the class determination device 20 includes an image acquisition unit 210, a class determination unit 220, and a certainty factor output unit 230.
- the image acquisition unit 210 acquires a cell image captured by an imaging unit (not shown) of the cell sorter as an input image IP.
- the image acquisition unit 210 outputs the acquired input image IP to the class determination unit 220.
- the class determination unit 220 determines the type of the cell captured in the input image IP as the class CL.
- the class determination unit 220 is implemented as a learned model by a convolutional neural network (CNN / hereinafter, also referred to as a neural network CNN).
- CNN convolutional neural network
- the example where the class determination unit 220 is realized by the neural network CNN is an example.
- the class determination unit 220 includes, for example, a deep neural network (deep @ neural @ network; DNN), a probabilistic neural network (Probabilistic @ neural @ network; PNN), a forward-propagation neural network (feedforward neural network; FFNN), and a recurrent neural network ( Recurrent (neural networks) (RNNs), Autoassociator, Deep Belief Network (DBNs), Radial basis function (RBF) network, Boltzmann machine, Boltzmann machine, etc.
- the network exemplified here is also referred to as an artificial neural network in the following description.
- FIG. 2 illustrates an example of a configuration of the class determination unit 220 according to the present embodiment.
- any of various cells such as platelet aggregates PA, platelets SP, and white blood cells LK is imaged.
- platelet aggregate PA is associated with class CLa
- platelet SP is associated with CLb
- white blood cell LK is associated with class CLc.
- the leukocytes LK include granulocytes GL (neutrophils, eosinophils and basophils), lymphocytes LC, monocytes MC and the like.
- the lymphocytes LC are associated with the class CLd
- the T lymphocytes LC among the lymphocytes LC are associated with the class CLd1
- the B lymphocytes LC are associated with the class CLd2.
- the granulocyte GL is associated with the class CLe.
- the class determination unit 220 determines which of the classes CL the type of cell captured in the input image IP corresponds to, and determines the certainty factor PR for each class CL. Is output.
- the class determination unit 220 outputs a certainty factor (Probability) for the class CL.
- the certainty factor PR will be described.
- the certainty factor PR is associated with each of a plurality of types of classes CL at a ratio of 1: 1.
- the class CLa has the certainty PRa
- the class CLb has the certainty PRb
- the class CLc has the certainty PRc.
- the certainty factor PR of a certain class CL indicates that the larger the value is, the higher the likelihood that the input signal (for example, the input image IP) to the neural network CNN corresponds to the class CL is. This indicates that the likelihood that the input signal falls into the class CL is low.
- the confidence PR of the class CLa (the platelet aggregate PA in the above example) is high when the input image IP is an image of the platelet aggregate PA, and the input image IP is an image other than the platelet aggregate PA. In some cases, it is calculated lower.
- the certainty factor PR may be normalized to a certain value range (for example, 0 to 1). In this case, the certainty factor PR of a certain class CL is such that the closer the value is to 1, the higher the likelihood that the input signal falls in the class CL, and the closer the value is to 0, the more likely the input signal falls in the class CL. Indicates low degree.
- the sum total obtained by adding the certainty factor PR of each class CL for each class CL is a certain value (for example, 1). For example, when an input image IP is classified into the above-described three types of classes CL, the sum of the certainty factors PRa to PRc corresponding to these classes CL is 1.
- a vector value corresponding to the number of types of the class CL is set as an output value.
- the output value from the final layer can be the certainty factor PR.
- the certainty factor PR may be any value as long as the likelihood that a certain input signal corresponds to a certain class CL can be expressed by comparison between the classes CL.
- FIG. 3 shows an example of a calculation result of the certainty factor PR by the class determination system 1 of the present embodiment.
- the class determination unit 220 calculates the certainty factor PR of each class CL for each input image IP.
- the class determination unit 220 is convinced of the likelihood that the input image IP1 corresponds to the class CLa for the first input image IP1.
- the degree PRa the likelihood that the input image IP1 corresponds to the class CLb is calculated as the certainty degree PRb
- the likelihood that the input image IP1 corresponds to the class CLc is calculated as the certainty degree PRc.
- the class determination unit 220 calculates the confidence PRa as 95%, the confidence PRb as 4%, and the confidence PRc as 1%.
- the confidence output unit 230 outputs the confidence PR calculated by the class determination unit 220 to the display control device 10.
- the display control device 10 includes a certainty factor acquisition unit 110, a display control unit 120, a set value acquisition unit 130, and a gate determination unit 140.
- the certainty obtaining unit 110 obtains the certainty PR calculated by the class determining unit 220 (the certainty calculating unit) that calculates the certainty PR of the class determination result RCL of the input image IP.
- the display control unit 120 generates an image of the graph G based on the distribution of the certainty factor PR obtained by the certainty factor obtaining unit 110 for each input image IP.
- the image of the graph G generated by the display control unit 120 includes the certainty axis PAX as the display axis DAX of the graph G.
- the display axis DAX refers to an axis of the graph G displayed on the display unit 40.
- the certainty axis PAX is an axis indicating the magnitude of the certainty PR among the axes of the graph G.
- the display control unit 120 displays at least one of the display axes DAX of the graph G as the confidence axis PAX.
- the class determination system 1 displays the X axis as the certainty axis PAX among these two axes.
- the display control unit 120 may display both of these two axes (that is, the X axis and the Y axis) as the certainty axis PAX.
- the graph G does not necessarily have to have two axes, and may have a number of axes corresponding to the number of types of the class CL determined by the class determination unit 220.
- the display control unit 120 may display a three-axis graph G.
- the class determination system 1 may use all of the three axes of the graph G as the confidence axes PAX. That is, the display control unit 120 sets the distribution of the certainty factor PR acquired by the certainty factor acquisition unit 110 for each input image IP to at least one of the display axes DAX of the graph G as the certainty factor axis PAX indicating the certainty factor PR.
- the image of the graph G generated by the display control unit 120 is also referred to as a certainty graph image PG.
- the setting value acquisition unit 130 acquires a gate setting value GV for the certainty factor PR displayed on the graph G.
- the gate determining unit 140 determines whether or not the input image IP is inside the gate setting value GV based on the gate setting value GV acquired by the setting value acquiring unit 130.
- FIG. 4 shows an example of the operation of the class determination system 1 of the present embodiment.
- the image acquisition unit 210 acquires an image of a cell (or tissue) captured from a flow cytometry device (not shown) as an input image IP (Step S210).
- the image acquisition unit 210 outputs the acquired input image IP to the class determination unit 220.
- the class determination unit 220 calculates a certainty factor PR for each determination target class CL for the input image IP acquired in step S210 (step S220).
- the certainty factor output unit 230 outputs the certainty factor PR calculated in step S220 to the display control device 10 (step S230).
- the certainty factor acquisition unit 110 acquires the certainty factor PR calculated in step S230 (step S110).
- the display control unit 120 generates an image of the graph G based on the certainty factor PR acquired in step S110 (step S120).
- the display control unit 120 outputs the certainty factor graph image PG generated in step S120 to the display unit 40 (step S130).
- the certainty graph image PG generated in step S120 is displayed on the display unit 40.
- the setting value acquisition unit 130 acquires the gate setting value GV (Step S140).
- the gate setting value GV may be given manually by a keyboard, a mouse, a touch panel, or the like, or may be automatically calculated based on the distribution state of the certainty factor graph image PG.
- the setting value acquisition unit 130 outputs the acquired gate setting value GV to the display control unit 120.
- the gate setting value GV is a value for setting a gate GT for the certainty factor graph image PG displayed in step S130.
- the display control unit 120 displays the gate image GP based on the acquired gate setting value GV on the certainty factor graph image PG.
- the display control unit 120 outputs the certainty graph image PG generated in step S130 and the gate setting value GV for the certainty graph image PG to the gate determination unit 140.
- the gate determination unit 140 determines whether the input image IP acquired in step S210 is inside the gate setting value GV created in step S140, and generates a gate determination result GR as the determination result. (Step S150).
- whether the input image IP is inside the gate set value GV is determined by plotting the certainty factor PR of the input image IP within the range of the certainty factor PR indicated by the gate set value GV on the certainty factor graph image PG. Or not.
- the gate determination unit 140 determines the input image IP based on the certainty graph image PG output from the display control unit 120 in step S140 and the gate setting value GV for the certainty graph image PG. Is within the gate set value GV.
- the gate determination unit 140 outputs the generated gate determination result GR to a cell sorter (not shown).
- the class determination system 1 can be programmed to selectively sort cells included (or not included) in the gate set value GV. Further, the class determination system 1 can selectively display data (images, numerical data, plots, etc.) of cells (or cells not included) included in the gate setting value GV based on the gate determination result GR.
- the display control unit 120 can display data such as an input image IP and a graph G corresponding to a desired cell using the gate determination result GR by the gate determination unit 140.
- the display control unit 120 may calculate and output a statistical value such as the number or ratio of the input images IP included in the gate GT based on the gate determination result GR, or may display the statistical value on the graph G. It is possible.
- the display control unit 120 determines the position of the data point for the data point corresponding to the input image included in the gate GT based on the gate determination result GR of whether or not the individual input image IP is included in the gate GT. It is also possible to change the color and shape of the point indicated.
- FIG. 5 shows an example of the certainty factor graph image PG of the present embodiment.
- the display control unit 120 displays the distribution of the certainty factor PR for each input image IP with at least one of the display axes DAX of the graph G as the certainty factor axis PAX.
- the display control unit 120 displays the graph G with the X axis as the confidence axis PAX and the Y axis as the appearance frequency of the confidence PR among the display axes DAX of the graph G. .
- the appearance frequency of the certainty factor PR on the Y-axis is a value obtained by counting the certainty factor PR of a certain class CL for each input image IP when the class determination unit 220 determines the class CL for each of the plurality of input images IP. It is.
- the appearance frequency of the certainty factor PR is also referred to as a frequency F of the input image IP at a certain certainty factor PR.
- the certainty graph image PG is a histogram in which the X axis indicates the certainty axis PAX and the Y axis indicates the frequency F.
- the class determination unit 220 determines whether or not the input image IP is an image of the platelet aggregate PA.
- the class determination unit 220 calculates the certainty factor PRa that the cell (or tissue) captured in the input image IP is the platelet aggregate PA. For example, when the image of the platelet aggregate PA is given as the input image IP, the class determination unit 220 sets the confidence PR (that is, the confidence PRa) that the input image IP is the image of the platelet aggregate PA to 95. Calculate as%. That is, in this case, the class determination unit 220 determines that the given input image IP has a high likelihood of being an image of the platelet aggregate PA.
- the class determination unit 220 calculates the certainty factor PRa as 4%. That is, in this case, the class determination unit 220 determines that the given input image IP has a low likelihood of being an image of the platelet aggregate PA.
- the class determination unit 220 calculates a certainty factor PRa for each of the plurality of input images IP, and plots the distribution of the calculated certainty factors PRa on the X axis as the certainty axis PAX and the Y axis as the axis indicating the frequency F. Is displayed on the display unit 40 as a certainty graph image PG.
- FIGS. 5A to 5C show differences in the class determination performance of the class determination unit 220.
- FIG. When the class determination performance is relatively high, the certainty factor PR is distributed as shown in FIG. That is, when the class determination performance is relatively high, the distribution of the certainty factor PR is divided into a “distribution DS1A” region where the certainty factor PR is relatively low and an “distribution DS1B” region where the certainty factor PR is relatively high. Divide clearly. In other words, when the certainty factor graph image PG shown in FIG. 5A is displayed on the display unit 40, the operator can interpret that the class determination performance is relatively high. When the class determination performance is relatively low, the certainty factor PR is distributed as shown in FIG.
- the distribution of the certainty factor PR is divided into a “distribution DS1C” region where the certainty factor PR is relatively low and an “distribution DS1D” region where the certainty factor PR is relatively high. Although divided, these two areas are closer to each other than in the case shown in FIG. That is, when the class determination performance is relatively low, it is difficult to clearly separate an area having a relatively low confidence PR from an area having a relatively high confidence PR. In other words, when the certainty factor graph image PG shown in FIG. 5B is displayed on the display unit 40, the operator can interpret that the class determination performance is relatively low. If the class determination performance is even lower, the certainty factor PR is distributed as shown in FIG.
- the distribution of the certainty factor PR is only the area of the distribution DS1E, and it becomes impossible to discriminate between the area having the high certainty factor PR and the area having the low certainty factor PR.
- the certainty factor graph image PG shown in FIG. 5C is displayed on the display unit 40, the operator determines that the class determination performance is lower than the state shown in FIG. Can be interpreted as The “class determination performance” is also referred to as “class CL classification accuracy”.
- the determination performance of the class determination unit 220 is in a format that is easy for the operator (ie, human) to interpret. That is, according to the class determination system 1 of the present embodiment, the performance of determining the class CL by the neural network CNN can be presented in a format that is easy for the operator to interpret.
- the neural network CNN determines the class CL of the input image IP.
- the learning state of the neural network CNN is appropriate (that is, the class determination performance of the neural network CNN)
- an input signal for example, an input image IP
- RCL output by the neural network CNN indicates “platelet aggregate PA”
- the class determination result RCL output from the neural network CNN is determined only as two choices of correct or incorrect, and the degree of validity of the class determination result RCL, that is, the class determination of the neural network CNN, The performance could not be determined.
- the class determination system 1 of the present embodiment causes the neural network CNN to output the certainty factor PR used for the class determination from the neural network CNN, and displays the certainty factor PR as the display axis DAX of the graph G.
- the certainty factor PR is a value calculated by the neural network CNN as an index indicating the likelihood that a certain input signal corresponds to a certain class CL
- the degree of validity of the class determination by the neural network CNN is determined. Is shown.
- the distribution of the certainty factor PR can be an index indicating the class determination performance of the neural network CNN.
- the class determination system 1 of the present embodiment can present the class determination performance by the neural network CNN in a form that is easy for humans to interpret by showing the graph G with the distribution of the certainty factor PR as an axis.
- the gate determination unit 140 may evaluate the class determination performance.
- the cause of the case where the class determination performance is low is that the performance of the model of the neural network CNN that has performed the class determination is poor, or the image quality of the input image is low. Possible cause.
- it is effective to perform re-learning.
- the re-learning method For example, if it is determined from the distribution of the certainty factor of a certain parameter that the gate is not valid, it is effective to add a new image for learning to the class of the parameter and re-learn. It is. Further, it is effective to generate and add an image used for learning by performing image processing such as inversion and rotation on the existing image used for learning, and to perform re-learning.
- image processing such as inversion and rotation on the existing image used for learning
- FIG. 5C when it is not possible to discriminate between a region having a high degree of certainty and a region having a low degree of certainty, a plurality of cells having different shapes are included in the population classified into one class. May be apparently present.
- the gate determination unit 140 may determine whether re-learning is necessary or determine a re-learning method based on the evaluation result of the class determination performance.
- FIG. 6 shows an example of a graph G displayed by the display control unit 120 of the present embodiment.
- the display control unit 120 displays the certainty graph image PG with the certainty PRd of the class CLd as the X axis and the certainty PRE of the class CLe as the Y axis among the display axes DAX of the graph G.
- the class CLd corresponds to the lymphocyte LC
- the class CLe corresponds to the granulocyte GL.
- the distribution DS2A indicates an area where the class determination unit 220 has determined the class of the input image IP as an image of the lymphocyte LC.
- the distribution DS2B indicates an area where the class determination unit 220 has determined that the input image IP is a class image of the granulocyte GL.
- the distribution DS2C indicates an area in which the class determination unit 220 has determined that the input image IP is not an image of any of the lymphocyte LC and the granulocyte GL. That is, in the example shown in the drawing, the display control unit 120 causes a plurality of display axes DAX of the display axes DAX of the graph G to be displayed as the certainty degree axes PAX. As described above, by displaying the graph G with the plurality of display axes DAX being the confidence axes PAX, the class determination system 1 can present the class determination performance by the neural network CNN in a format that is easy for humans to interpret.
- the confidence axis PAX is described as an axis that directly indicates the confidence PR of a certain class CL, but is not limited thereto.
- the certainty axis PAX may be an axis indicating the certainty PR (ie, the combined certainty PRCB) obtained by combining the certainty PRs of the plurality of classes CL.
- the X-axis of the graph G may be a combined certainty factor PRCB obtained by adding the certainty factors PRa and PRb by a polynomial.
- the combination certainty factor PRCB is represented by the sum of a value obtained by multiplying the certainty factor PRa by a weight and a value obtained by multiplying the certainty factor PRb by a weight. That is, in this case, the display control unit 120 displays an axis indicating the combination certainty factor PRCB based on the combination of the certainty factors PR of the plurality of classes CL indicated by the class determination result RCL as the certainty factor axis PAX.
- the display control device 10 can display the gate image GP over the graph G by performing an operation of inputting the gate setting value GV to the operation detection unit 30 by the operator.
- gate refers to a threshold for extracting a specific region from a graph G such as a histogram.
- the term “gating” refers to an operation of extracting a specific region by performing a gate on the graph G. For example, this gating is effective when only information of a specific type of cell is to be extracted from a certain graph G. Further, for example, when the class CL to be gated extends over a plurality of classes CL, it is effective to use a combination certainty factor PRCB obtained by adding the plurality of certainty factors PR by a polynomial as described above. An example of this gating will be described with reference to FIG.
- FIG. 7 shows an example of gating by the display control device 10 of the present embodiment.
- FIG. 7A shows again the certainty factor graph image PG shown in FIG.
- the display control device 10 can display a certainty graph image PG (FIG. 7B) for the distribution DS2A among the certainty graph images PG.
- the operator performs an operation of inputting the gate setting value GV to the operation detection unit 30.
- the operation of inputting the gate setting value GV includes, for example, a numerical value input and a shape input such as a rectangle, a circle, or a polygon for the image of the graph G.
- the operation detection unit 30 When detecting the operation of inputting the gate setting value GV, the operation detection unit 30 outputs information indicating the detected operation to the display control device 10.
- the setting value acquisition unit 130 acquires the gate setting value GV from the information output by the operation detection unit 30.
- the display control unit 120 causes the graph G to display a gate image GP indicating the gate GT based on the gate setting value GV acquired by the setting value acquiring unit 130.
- the setting value acquisition unit 130 acquires the gate setting value GV1.
- the display control unit 120 causes the display unit 40 to display the gate image GP1 (that is, the polygonal gate image GP) so as to overlap the certainty factor graph image PG.
- the gate image GP1 shown in FIG. 7A surrounds the area of the distribution DS2A.
- the distribution DS2A indicates an area where the certainty factor PRd of the class CLd is relatively high (that is, an area where the input image IP has a high likelihood of being an image of the lymphocyte LC).
- the display control unit 120 determines that the area surrounded by the gate image GP is an area in which the certainty factor PRd of the class CLd is relatively high, the display control unit 120 generates a new image of the graph G and causes the display unit 40 to display the image. (FIG. 7 (B)).
- the graph G shown in FIG. 7B is a certainty graph image PG in which the X axis is a certainty factor PRd and the Y axis is a frequency F. That is, the display control unit 120 can display a new graph G based on the position of the gate image GP in the display area of the graph G.
- a gate can also be set for the graph G in FIG. That is, the operator performs an operation of inputting the gate setting value GV2 to the operation detection unit 30 in order to set a gate for the graph G shown in FIG. 7B.
- the gate setting value GV2 is a numerical value indicating the width of the certainty factor PRd.
- the setting value acquisition unit 130 acquires the gate setting value GV2 from the operation detection unit 30.
- the display control unit 120 causes the graph G to display a gate image GP2 indicating the gate GT based on the gate setting value GV2 acquired by the setting value acquiring unit 130.
- the gate image GP2 is an image surrounding the distribution DS2E in the example illustrated in FIG. 7B.
- the distribution DS2E is a region where the confidence PRd is relatively high (that is, a region where the likelihood that the input image IP is an image of the lymphocyte LC is high). That is, in the example shown in FIG. 7B, gating on the lymphocytes LC is performed.
- the display control device 10 can display the plurality of graphs G stepwise based on the operation detected by the operation detection unit 30.
- the display control device 10 can present in a format that is easy for humans to interpret by visually expressing the narrowing down by the so-called gate region using the image of the graph G.
- FIG. 8 shows a modification of gating by the display control device 10 of the present embodiment.
- the graph G before gating (FIG. 7A) is the certainty degree graph image PG.
- This modification is different from the above-described example in that the graph G before gating is not the certainty graph image PG.
- the graph G shown in FIG. 8A is a distribution diagram in which the parameter PM-A is on the X axis and the parameter PM-B is on the Y axis.
- the parameter PM-A is a parameter indicating a degree of a reaction of each cell with an antibody (eg, CD45-FITC).
- the parameter PM-B is a parameter indicating the degree of scattered light (for example, side scattered light) of each cell.
- the graph G shown in the figure shows the distribution of granulocytes GL, monocytes MC, and lymphocytes LC. In this example, gating is performed on the lymphocytes LC.
- the operator sets the range surrounding the area of the lymphocyte LC in the graph G as the gate setting value GV3.
- the operation detection unit 30 detects a setting operation of the gate setting value GV3 and outputs the operation to the display control device 10.
- the display control unit 120 displays a gate image GP3 corresponding to the gate setting value GV3. Further, the display control unit 120 generates an image of the graph G shown in FIG. 8B based on the gate setting value GV3, and causes the display unit 40 to display the image.
- the graph G shown in FIG. 8B is a certainty graph image PG in which the X axis is the certainty PRd1 and the Y axis is the certainty PRd2.
- the certainty factor PRd1 is, for example, the certainty factor PR for the class CLd1 (T lymphocyte LCT).
- the certainty factor PRd2 is, for example, the certainty factor PR for the class CLd2 (B lymphocyte LCB).
- any of a region having a relatively high likelihood of being a T lymphocyte LCT (distribution DS3A) and a region having a relatively high likelihood of being a B lymphocyte LCB (distribution DS3B) are shown.
- a region having a relatively low degree (distribution DS3C). Note that gating can be performed on the graph G by the gate setting value GV4 and the gate image GP4.
- FIG. 9 shows another modification of gating by the display control device 10 of the present embodiment. Also in this modified example, gating can be performed with the gate set value GV, as in the example of FIG. 7B described above. Note that this modification is different from the above-described example in that the gate set value GV can be changed according to the purpose of gating.
- screening is an example of the purpose of gating. In the screening, when a larger number of false positives may be included in order to avoid missing positives, the gate set value GV is set wider (for example, the gate set value GV23 shown in FIG. 9C). ). Further, when it is assumed that positives may be omitted in order to eliminate false positives as much as possible, the gate set value GV is set narrower (for example, the gate set value GV21 shown in FIG. 9A). ).
- the case of imaging cytometry has been described as the class determination system 1.
- the entire image is Individual cell images may be cut out, and the cut out images may be provided to the class determination system 1.
- a microscope such as a confocal microscope or a two-photon microscope, which captures a three-dimensional structure of an object.
- the microscope may be combined with a microscope that performs various multidimensional imaging, such as performing time-lapse imaging in the time axis direction instead of the depth direction (Z direction) or imaging by switching wavelengths. Even in a multidimensional image obtained by multidimensional imaging, an image obtained by cutting out individual cells can be provided to the class determination system 1.
- Method 1 An operation (simple addition, deconvolution, maximum value for each pixel, and the like) between images is performed on a plurality of captured images having different depth directions (Z directions), and the images are combined into a single image. Then, a certainty factor is obtained for the obtained single image, and the class is determined.
- Method 2 A plurality of captured images having different depth directions (Z directions) are made into a single voxel image including the Z direction, a certainty factor is obtained for the voxel images, and a class determination is performed.
- Methodhod 3) A plurality of captured images having different depth directions (Z directions) are separately input to the class determination system 1, and certainty factors are obtained for each Z position.
- a set of certainty factors can be obtained for one cell.
- a certainty factor is obtained for each Z position. Therefore, in the method 3, the image at the optimum Z position can be selected based on the distribution of the certainty factor.
- a multidimensional image in the depth direction (Z direction) is described, but the same applies to a case where the time or the wavelength is changed.
- the display control unit 120 may display the certainty factor obtained for the multidimensional image obtained by the multidimensional imaging by mapping the certainty factor to the original multidimensional image.
- the mapping refers to displaying the distribution of the certainty factor on the original multidimensional image using a method such as a contour plot, a heat map, a dot plot, or a bar graph.
- the display control unit 120 may map the cells of the gated class to the original multidimensional image and display the cells. For example, it is conceivable to represent the position of the cell of the gated class as a dot on the original multidimensional image, or to represent the area where the cell of the gated class exists with a curve or by changing the color tone. . According to the display method as described above, it becomes easier for a human to interpret the class determination result for a cell image having a time distribution or a spatial distribution.
- FIG. 10 shows an example of the configuration of the neural network CNN of the present embodiment.
- FIG. 11 shows an example of the relationship between the cell image and the certainty factor PR for each class CL.
- FIG. 12 shows an example of a histogram of the certainty factor PR and the gating result. In the example illustrated in FIG. 12, the histogram indicates the certainty factor PR classified as the white blood cell LK.
- an area where the certainty factor PR is equal to or less than a predetermined value is gated. That is, the leukocyte LK having low likelihood is gated.
- cells (or tissues) other than the leukocyte LK are extracted.
- a result of the gating cells (or tissues) containing the platelet aggregate PA are extracted.
- a graph G is displayed using the confidence PR of the platelet aggregate PA as the X axis, and a region where the likelihood of the platelet aggregate PA is high is gated. In the vicinity of the gate, the number and ratio of cells (or tissues) included in the gate are displayed. Displaying the statistical value indicating the gating result in this manner is useful because the classification result can be quantitatively recognized.
- a group of input images IP having a high likelihood of being platelet aggregate PA is limited from all input images IP.
- a histogram of the certainty factor PR belonging to the platelet aggregate PA is drawn, and a group of the input images IP having a certainty factor value equal to or greater than a certain value is designated by a gate.
- FIG. 13 shows an example of the result of quantifying the classification accuracy of the platelet aggregate PA.
- a result of quantifying the classification accuracy of the platelet aggregate PA by the gate setting shown in FIG. 12 using a data set that has been artificially classified in advance is shown.
- FIG. 14 shows an example of the sorting performance by a cell sorter (when a stimulus by TRAP is given).
- FIG. 15 shows an example of the sorting performance by the cell sorter (when no stimulation by TRAP is given).
- a cell population to be sorted using a cell sorter (not shown) is designated by the gate setting shown in FIG. 12, and cell sorting is performed in accordance with the gating setting.
- FIG. 16 shows an example of a histogram display of the certainty factor PR classified as the platelet aggregate PA.
- the group is relatively biased to both ends of the histogram.
- more groups belong to the middle region of the histogram.
- FIG. 17 shows an example where the display format of the graph G is a scatter diagram.
- FIG. 17 is an example of a scatter diagram in which the horizontal axis (X axis) shows the certainty factor PR for the platelet aggregate PA and the vertical axis (Y axis) shows the certainty factor PR for the single platelet (platelet SP).
- X axis shows the certainty factor PR for the platelet aggregate PA
- Y axis shows the certainty factor PR for the single platelet (platelet SP).
- FIG. 18 is a diagram illustrating an example of a result of classifying the input image IP by the certainty factor PR.
- FIG. 18A shows a group of input images IP having a certainty factor greater than 0.999
- FIG. 18B shows a group of input images IP having a certainty factor between 0.99 and 0.999
- FIG. 9 shows a group of input images IP included between 0.98 and 0.99.
- the class determination unit 220 may be realized by existing means such as decision tree learning, support vector machine, clustering, Bayesian network, hidden Markov model, ensemble learning, and boosting.
- each of the above devices has a computer inside.
- the process of each process of each device described above is stored in a computer-readable recording medium in the form of a program, and the program is read and executed by a computer to perform the process.
- the computer-readable recording medium refers to a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, a semiconductor memory, or the like.
- the computer program may be distributed to a computer via a communication line, and the computer that has received the distribution may execute the program.
- the program may be for realizing a part of the functions described above. Furthermore, what can realize the above-mentioned function in combination with a program already recorded in the computer system, that is, a so-called difference file (difference program) may be used.
- difference file difference program
- SYMBOLS 1 ... Class determination system, 10 ... Display control device, 110 ... Confidence acquisition part, 120 ... Display control part, 130 ... Setting value acquisition part, 20 ... Class determination device, 210 ... Image acquisition part, 220 ... Class determination part, 230: confidence output unit, PR: confidence, IP: input image IP, PG: confidence graph image
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Priority Applications (7)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2019554569A JP6753622B2 (ja) | 2018-06-27 | 2019-06-17 | 表示制御装置、表示制御方法及び表示制御プログラム |
| CN201980041251.5A CN112313702B (zh) | 2018-06-27 | 2019-06-17 | 显示控制设备、显示控制方法以及显示控制程序 |
| US15/733,808 US12136148B2 (en) | 2018-06-27 | 2019-06-17 | Method, system, device, and program for displaying probabilistic classification results |
| AU2019293020A AU2019293020B2 (en) | 2018-06-27 | 2019-06-17 | Display control device, display control method, and display control program |
| CN202311687873.XA CN117496511A (zh) | 2018-06-27 | 2019-06-17 | 显示控制设备、显示控制方法以及显示控制程序 |
| CA3102749A CA3102749A1 (en) | 2018-06-27 | 2019-06-17 | Display control device, display control method, and display control program |
| EP19824977.3A EP3816935A4 (en) | 2018-06-27 | 2019-06-17 | DISPLAY CONTROL DEVICE, DISPLAY CONTROL METHOD AND DISPLAY CONTROL PROGRAM |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2018-122096 | 2018-06-27 | ||
| JP2018122096 | 2018-06-27 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2020004101A1 true WO2020004101A1 (ja) | 2020-01-02 |
Family
ID=68986491
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2019/023833 Ceased WO2020004101A1 (ja) | 2018-06-27 | 2019-06-17 | 表示制御装置、表示制御方法及び表示制御プログラム |
Country Status (7)
| Country | Link |
|---|---|
| US (1) | US12136148B2 (cg-RX-API-DMAC7.html) |
| EP (1) | EP3816935A4 (cg-RX-API-DMAC7.html) |
| JP (2) | JP6753622B2 (cg-RX-API-DMAC7.html) |
| CN (2) | CN117496511A (cg-RX-API-DMAC7.html) |
| AU (1) | AU2019293020B2 (cg-RX-API-DMAC7.html) |
| CA (1) | CA3102749A1 (cg-RX-API-DMAC7.html) |
| WO (1) | WO2020004101A1 (cg-RX-API-DMAC7.html) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPWO2021193673A1 (cg-RX-API-DMAC7.html) * | 2020-03-25 | 2021-09-30 | ||
| CN113536884A (zh) * | 2021-03-29 | 2021-10-22 | 内蒙古卫数数据科技有限公司 | 一种基于机器视觉的虎红平板结果识别方法及系统 |
| JP2022049767A (ja) * | 2020-09-17 | 2022-03-30 | 株式会社スペースリー | 情報処理装置、情報処理方法、情報処理システム、情報処理プログラム |
| JP2023073140A (ja) * | 2021-11-15 | 2023-05-25 | 富士電機株式会社 | 管理装置、検査システム、管理方法、及びプログラム |
| JP2023542047A (ja) * | 2020-11-19 | 2023-10-04 | ソニーグループ株式会社 | 画像ベースの教師なしセルクラスタリング及びソーティングのためのフレームワーク |
| JP2023550927A (ja) * | 2020-11-19 | 2023-12-06 | ソニーグループ株式会社 | 細胞画像分類のための効率的かつロバストな高速ニューラルネットワーク |
| WO2024014102A1 (ja) * | 2022-07-12 | 2024-01-18 | 株式会社レゾナック | 欠陥分析装置、欠陥分析方法及びプログラム |
| JP7901948B1 (ja) * | 2026-06-08 | 2026-08-07 | 株式会社Cybo | 細胞画像解析結果の表示方法、表示装置、及びプログラム |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020004101A1 (ja) * | 2018-06-27 | 2020-01-02 | 株式会社Cybo | 表示制御装置、表示制御方法及び表示制御プログラム |
| WO2025263343A1 (ja) * | 2024-06-18 | 2025-12-26 | 株式会社Cybo | 顕微積層画像のデータ解析装置およびデータ解析方法 |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010076882A1 (ja) * | 2008-12-29 | 2010-07-08 | 株式会社日立ハイテクノロジーズ | 画像分類基準更新方法、プログラムおよび画像分類装置 |
| WO2013147114A1 (ja) | 2012-03-30 | 2013-10-03 | 公益財団法人神奈川科学技術アカデミー | イメージングセルソーター |
| WO2017072854A1 (ja) * | 2015-10-27 | 2017-05-04 | 株式会社日立製作所 | 監視装置、監視システムおよび監視方法 |
| WO2017081829A1 (ja) * | 2015-11-13 | 2017-05-18 | 富士通株式会社 | 行動検知装置、行動検知方法及び行動検知プログラム |
Family Cites Families (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH08293025A (ja) * | 1995-04-20 | 1996-11-05 | Olympus Optical Co Ltd | 画像分類装置 |
| US6947586B2 (en) * | 2000-04-24 | 2005-09-20 | International Remote Imaging Systems, Inc. | Multi-neural net imaging apparatus and method |
| JP2006318405A (ja) * | 2005-05-16 | 2006-11-24 | Mitsubishi Electric Building Techno Service Co Ltd | 設備機器の情報提供装置 |
| WO2007139929A2 (en) * | 2006-05-25 | 2007-12-06 | The Burnham Institute For Medical Research | Methods for culture and production of single cell populations of human embryonic stem cells |
| CN101187985B (zh) * | 2006-11-17 | 2012-02-01 | 东软集团股份有限公司 | 确定对象分类器分类边界的方法及装置 |
| US20080214944A1 (en) * | 2007-02-09 | 2008-09-04 | Morris Margaret E | System, apparatus and method for mobile real-time feedback based on changes in the heart to enhance cognitive behavioral therapy for anger or stress reduction |
| JP5237596B2 (ja) * | 2007-08-10 | 2013-07-17 | オリンパス株式会社 | 判別装置 |
| JP6070420B2 (ja) * | 2013-05-31 | 2017-02-01 | 富士ゼロックス株式会社 | 画像処理装置及びプログラム |
| US9489373B2 (en) * | 2013-07-12 | 2016-11-08 | Microsoft Technology Licensing, Llc | Interactive segment extraction in computer-human interactive learning |
| US9886669B2 (en) * | 2014-02-26 | 2018-02-06 | Microsoft Technology Licensing, Llc | Interactive visualization of machine-learning performance |
| JP2015176588A (ja) * | 2014-03-18 | 2015-10-05 | 株式会社東芝 | 表示装置、画像表示システムおよび情報処理方法 |
| CN106156161A (zh) * | 2015-04-15 | 2016-11-23 | 富士通株式会社 | 模型融合方法、模型融合设备和分类方法 |
| JPWO2017154706A1 (ja) * | 2016-03-09 | 2018-12-27 | 株式会社ニコン | 検出装置、情報処理装置、検出方法、検出プログラム、及び検出システム |
| CN109863384B (zh) | 2016-06-10 | 2022-10-28 | 加利福尼亚大学董事会 | 基于图像的细胞分选系统和方法 |
| US10460256B2 (en) * | 2016-08-09 | 2019-10-29 | Microsot Technology Licensing, LLC | Interactive performance visualization of multi-class classifier |
| US10945864B2 (en) * | 2016-08-17 | 2021-03-16 | Teledyne Scientific & Imaging, Llc | System and method for noninvasive identification of cognitive and behavioral goals |
| KR102469620B1 (ko) * | 2016-08-22 | 2022-11-21 | 아이리스 인터내셔널 인크. | 생물학적 입자의 분류 시스템 및 방법 |
| US11200483B2 (en) * | 2016-08-30 | 2021-12-14 | Lunit Inc. | Machine learning method and apparatus based on weakly supervised learning |
| JP7054787B2 (ja) * | 2016-12-22 | 2022-04-15 | パナソニックIpマネジメント株式会社 | 制御方法、情報端末、及びプログラム |
| CN106650823A (zh) * | 2016-12-30 | 2017-05-10 | 湖南文理学院 | 一种基于概率极限学习机集成的泡沫镍表面缺陷分类方法 |
| CN108062551A (zh) * | 2017-06-28 | 2018-05-22 | 浙江大学 | 一种基于邻接矩阵的图特征提取系统、图分类系统和方法 |
| WO2020004101A1 (ja) * | 2018-06-27 | 2020-01-02 | 株式会社Cybo | 表示制御装置、表示制御方法及び表示制御プログラム |
-
2019
- 2019-06-17 WO PCT/JP2019/023833 patent/WO2020004101A1/ja not_active Ceased
- 2019-06-17 CN CN202311687873.XA patent/CN117496511A/zh active Pending
- 2019-06-17 JP JP2019554569A patent/JP6753622B2/ja active Active
- 2019-06-17 CA CA3102749A patent/CA3102749A1/en active Pending
- 2019-06-17 EP EP19824977.3A patent/EP3816935A4/en active Pending
- 2019-06-17 CN CN201980041251.5A patent/CN112313702B/zh active Active
- 2019-06-17 AU AU2019293020A patent/AU2019293020B2/en active Active
- 2019-06-17 US US15/733,808 patent/US12136148B2/en active Active
-
2020
- 2020-08-17 JP JP2020137325A patent/JP7361395B2/ja active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010076882A1 (ja) * | 2008-12-29 | 2010-07-08 | 株式会社日立ハイテクノロジーズ | 画像分類基準更新方法、プログラムおよび画像分類装置 |
| WO2013147114A1 (ja) | 2012-03-30 | 2013-10-03 | 公益財団法人神奈川科学技術アカデミー | イメージングセルソーター |
| WO2017072854A1 (ja) * | 2015-10-27 | 2017-05-04 | 株式会社日立製作所 | 監視装置、監視システムおよび監視方法 |
| WO2017081829A1 (ja) * | 2015-11-13 | 2017-05-18 | 富士通株式会社 | 行動検知装置、行動検知方法及び行動検知プログラム |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021193673A1 (ja) * | 2020-03-25 | 2021-09-30 | 合同会社H.U.グループ中央研究所 | ゲート領域推定プログラム、ゲート領域推定方法、及びゲート領域推定装置 |
| JPWO2021193673A1 (cg-RX-API-DMAC7.html) * | 2020-03-25 | 2021-09-30 | ||
| JP2022049767A (ja) * | 2020-09-17 | 2022-03-30 | 株式会社スペースリー | 情報処理装置、情報処理方法、情報処理システム、情報処理プログラム |
| JP2023550927A (ja) * | 2020-11-19 | 2023-12-06 | ソニーグループ株式会社 | 細胞画像分類のための効率的かつロバストな高速ニューラルネットワーク |
| JP2023542047A (ja) * | 2020-11-19 | 2023-10-04 | ソニーグループ株式会社 | 画像ベースの教師なしセルクラスタリング及びソーティングのためのフレームワーク |
| US12056944B2 (en) | 2020-11-19 | 2024-08-06 | Sony Group Corporation | Efficient and robust high-speed neural networks for cell image classification |
| US12078597B2 (en) | 2020-11-19 | 2024-09-03 | Sony Group Corporation | Framework for image based unsupervised cell clustering and sorting |
| JP7596534B2 (ja) | 2020-11-19 | 2024-12-09 | ソニーグループ株式会社 | 細胞画像分類のための効率的かつロバストな高速ニューラルネットワーク |
| JP7700220B2 (ja) | 2020-11-19 | 2025-06-30 | ソニーグループ株式会社 | 画像ベースの教師なしセルクラスタリング及びソーティングのためのフレームワーク |
| CN113536884A (zh) * | 2021-03-29 | 2021-10-22 | 内蒙古卫数数据科技有限公司 | 一种基于机器视觉的虎红平板结果识别方法及系统 |
| JP2023073140A (ja) * | 2021-11-15 | 2023-05-25 | 富士電機株式会社 | 管理装置、検査システム、管理方法、及びプログラム |
| JP7803090B2 (ja) | 2021-11-15 | 2026-01-21 | 富士電機株式会社 | 管理装置、検査システム、管理方法、及びプログラム |
| WO2024014102A1 (ja) * | 2022-07-12 | 2024-01-18 | 株式会社レゾナック | 欠陥分析装置、欠陥分析方法及びプログラム |
| JP7901948B1 (ja) * | 2026-06-08 | 2026-08-07 | 株式会社Cybo | 細胞画像解析結果の表示方法、表示装置、及びプログラム |
Also Published As
| Publication number | Publication date |
|---|---|
| JP7361395B2 (ja) | 2023-10-16 |
| EP3816935A4 (en) | 2022-03-23 |
| AU2019293020B2 (en) | 2024-01-18 |
| EP3816935A1 (en) | 2021-05-05 |
| AU2019293020A1 (en) | 2021-01-07 |
| US12136148B2 (en) | 2024-11-05 |
| CA3102749A1 (en) | 2020-01-02 |
| JPWO2020004101A1 (ja) | 2020-07-02 |
| JP6753622B2 (ja) | 2020-09-09 |
| JP2021002354A (ja) | 2021-01-07 |
| CN112313702A (zh) | 2021-02-02 |
| CN117496511A (zh) | 2024-02-02 |
| US20210232869A1 (en) | 2021-07-29 |
| CN112313702B (zh) | 2023-12-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7361395B2 (ja) | 表示制御装置、表示制御方法及び表示制御プログラム | |
| Vaickus et al. | Automating the Paris System for urine cytopathology—a hybrid deep‐learning and morphometric approach | |
| US12270750B2 (en) | Adaptive sorting for particle analyzers | |
| JP5425814B2 (ja) | サポートベクタマシンを用いてフローサイトメトリーデータを分析するための方法及びシステム | |
| US11748981B2 (en) | Deep learning method for predicting patient response to a therapy | |
| US11788946B2 (en) | Optimized sorting gates | |
| Mohapatra et al. | Lymphocyte image segmentation using functional link neural architecture for acute leukemia detection | |
| EP4097631B1 (en) | Systematic characterization of objects in a biological sample | |
| Röhrl et al. | Towards interpretable classification of leukocytes based on deep learning | |
| CN107430587A (zh) | 自动化流式细胞术分析方法及系统 | |
| JP2022046085A (ja) | 重要度解析装置、方法及びプログラム | |
| EP1301894B1 (en) | Multi-neural net imaging apparatus and method | |
| JP2025536473A (ja) | 標本画像の異常の検出 | |
| HK40077070B (en) | Systematic characterization of objects in a biological sample | |
| HK40077070A (en) | Systematic characterization of objects in a biological sample | |
| WO2025173014A1 (en) | A method for selecting ovarian cancer patients for parp inhibitor treatment | |
| JP2025144331A (ja) | 検出システム、検出装置、検出方法及びプログラム | |
| WO2014112567A1 (ja) | 細胞群分類装置及び、細胞群分類方法 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| ENP | Entry into the national phase |
Ref document number: 2019554569 Country of ref document: JP Kind code of ref document: A |
|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19824977 Country of ref document: EP Kind code of ref document: A1 |
|
| ENP | Entry into the national phase |
Ref document number: 3102749 Country of ref document: CA |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| ENP | Entry into the national phase |
Ref document number: 2019293020 Country of ref document: AU Date of ref document: 20190617 Kind code of ref document: A |
|
| ENP | Entry into the national phase |
Ref document number: 2019824977 Country of ref document: EP Effective date: 20210127 |