WO2005010206A1 - 染色体状態の評価方法および評価システム - Google Patents
染色体状態の評価方法および評価システム Download PDFInfo
- Publication number
- WO2005010206A1 WO2005010206A1 PCT/JP2004/010852 JP2004010852W WO2005010206A1 WO 2005010206 A1 WO2005010206 A1 WO 2005010206A1 JP 2004010852 W JP2004010852 W JP 2004010852W WO 2005010206 A1 WO2005010206 A1 WO 2005010206A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- chromosome
- quantifying
- cell
- state
- territory
- 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
Classifications
-
- 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
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
-
- 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
- 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
Definitions
- the present invention relates to a method and system for evaluating the condition of cells and chromosomes contained in cells based on the arrangement of chromosome territories in the interphase of the cell cycle.
- Non-Patent Document 1 Non-Patent Document 2.
- Non-Patent Document 1 T. Cremer, C. Cremer, CHROMOSOME TERRITORI ES NUCLEAR ARCHITECTURE AND VENE REGULATION IN MAMMALIAN CELLS, NATURE REVIEWS
- Nonpatent literature 2 Hideyuki Tanabe, "Chromosome territory: Recent research on intranuclear arrangement and nuclear higher order structure of chromosomes in interphase nuclei", Environ. Mutagen Res., 25, 11 1 22, 2003
- Patent Document 1 Japanese Patent Application Laid-Open No. 2001-92980
- Non-Patent Document 2 describes the idea that chromosome territories arranged in close proximity to each other have a high probability of causing translocations as compared to those arranged in a distance from each other. From the above, it is expected to evaluate cell changes by examining the spatial arrangement of chromosome territories. However, conventionally, techniques for quantitatively analyzing the spatial arrangement of chromosome territories and evaluating cell changes have been provided.
- the inventors of the present application evaluate the state of cells by examining the state of arrangement of chromosome territories in the interphase of the cell cycle, and based on the evaluation results, it is used as an index for the diagnosis of diseases such as cancer.
- the following invention was devised for the purpose.
- a method of evaluating the state of a cell based on image information obtained by photographing a cell containing a chromosome territory which comprises the steps of: extracting a chromosome territory from an image;
- An evaluation method comprising the steps of: quantifying the state of arrangement after standardization of the state of arrangement; and evaluating the state of cells based on the state of arrangement of the quantified chromosome territory.
- the arrangement state can include at least one of the relative position of the chromosome territory within the nucleus of the cell, the orientation of the principal axis of the chromosome territory, and the spread of the chromosome territory.
- normalization is a process for quantifying the shape of chromosomal regions in a plurality of cells with a unified index. By performing such processing, quantification can be performed with a unified index for a plurality of cells.
- to evaluate the state of a cell means to evaluate the chromosomal state contained in the cell, change in the cell, anticipation of morphological abnormality of the cell, chromosomal translocation, and the cell.
- evaluate the possibility of suffering from a disease such as cancer based on change since the shape of the chromosome territory can be quantified, it can be expected to early detect the presence or absence of the morphological abnormality of the cell or the disease based on the quantified result.
- the image obtained by photographing the cell can be a plurality of images obtained by imaging the nucleus of the cell in a nuclear cross section, and the step of extracting the chromosome territory extracts the chromosome territory from the plurality of images, respectively. It is possible to synthesize and extract a chromosome territory image in three dimensions.
- the nucleus of the cell in the step of extracting the chromosome territory, the nucleus of the cell can be extracted and the chromosome territory included in the nucleus can be extracted, and the step of quantifying the arrangement state is Transforming the coordinates of the chromosome territory based on the step of standardizing the shape of the nucleus, and the shape change of the nucleus before normalization and the nucleus after normalization, , Can be included.
- the change in shape of the core may include both the change in force and the change in size.
- the step of transforming the coordinates of chromosome territory is
- Coordinate conversion can be performed based on the relative relationship between the nucleus and the chromosome territory before normalization, and the relative relationship between the nucleus and the chromosome territory after normalization.
- the evaluation method of the present invention may further include the step of statistically processing the results of quantifying the arrangement state of a plurality of cells, and in the step of evaluating the cell state, quantifying the arrangement state of one cell. It is possible to assess the state of one cell S based on the results of transformation and statistical processing.
- the step of performing statistical processing can include the step of calculating a standard value of the quantification result of the disposition state based on the result of quantifying the disposition state of the plurality of cells.
- the state of one cell can be evaluated based on the result of quantifying the state of arrangement of one cell and a standard value.
- the step of calculating the standard value can include the step of calculating an average value of the results of quantifying the arrangement of chromosomes of a plurality of cells.
- a method of evaluating the state of a cell based on image information obtained by photographing a cell containing a chromosome territory which comprises the steps of: extracting a plurality of chromosome territories from an image; The steps of quantifying the arrangement state of the chromosome territory, statistically processing the result of quantifying the arrangement state of a plurality of cells, the result of quantifying the arrangement state of one cell, and the result of the statistical processing And assessing the state of one cell based on the method.
- the degree of abnormality of the one cell can be evaluated by comparing the result of quantifying the arrangement state of the chromosome territory of one cell in this manner with the result of statistical processing. Further, according to the evaluation method of the present invention, since the shape of the chromosome territory can be quantified, it can be expected to early detect the presence or absence of the morphological abnormality of the cell or the disease based on the quantified result.
- the arrangement state can include a distance between a plurality of chromosome territories, or at least one of the arrangement directions of a plurality of chromosome territories.
- the quantifying step is performed by measuring the distance between the plurality of chromosome territories.
- a step of calculating separation may be included.
- the step of calculating the distance can include the steps of calculating the centers of gravity of each of a plurality of chromosome territories, and calculating the distance between the centers of gravity.
- the step of calculating the distance can be performed as follows.
- the step of calculating the distance can include the step of calculating the distance between the chromosome territories using the Mahalanobis distance.
- the step of calculating the distance can include the step of calculating the distance between the boundaries of the chromosome territory.
- the shortest distance between the boundaries can be used as the distance between the boundaries of the chromosome territory.
- the step of calculating the distance includes the steps of approximating the shapes of a plurality of chromosome territories with a normal distribution, the steps of calculating the average value of the normal distribution of each chromosome territory, and the average value. Calculating the distance between the two.
- the step of calculating the distance includes the steps of detecting a skeleton line of each of a plurality of chromosome territories, calculating the center of gravity of the skeleton line, and calculating the distance between the centers of gravity. And b.
- the quantifying step can include the step of quantifying the arrangement direction of the plurality of chromosome territories.
- the step of quantifying the arrangement direction can include a step of detecting principal axes of the plurality of chromosome territories, and a step of calculating an angle of the principal axes. .
- the step of detecting the principal axis includes the steps of approximating the shapes of a plurality of chromosome territories with a normal distribution, and based on the shapes of the approximated chromosome territories.
- the steps of calculating the mean value and the covariance matrix, and eigendecomposing the covariance matrix may be included.
- the cell is A method of evaluating the state of the cells, the steps of extracting the cell nucleus and chromosome territory from the image, quantifying the arrangement state of the nucleus and chromosome territory, and quantifying the arrangement states of a plurality of cells. Including the steps of statistically processing the converted results, evaluating the status of one cell based on the result of quantifying the disposition of one cell, and the result of the statistical processing. Evaluation methods are provided.
- the quantifying step can include the step of calculating the distance between the reference point in the nucleus and the chromosome territory.
- the reference point in the nucleus can be, for example, the nuclear wall or the center of gravity.
- the step of quantifying the arrangement state of the chromosome territory further includes the step of standardizing the arrangement state of the chromosome territory, and the arrangement state of the chromosome territory is standardized. It can be quantified.
- the step of performing statistical processing can include the step of calculating a standard value of the quantification result of the disposition state based on the result of quantifying the disposition state of the plurality of cells.
- the state of one cell can be evaluated based on the result of quantification of the state of arrangement of one cell and a standard value.
- the step of calculating the standard value can include the step of calculating an average value of the results of quantifying the arrangement state of a plurality of cells.
- a method of evaluating the state of a cell based on image information obtained by photographing a cell containing a chromosome territory wherein the step of extracting the nucleus of the cell and the plurality of chromosome territories respectively from the image.
- the step of evaluating the cell condition is a step of quantifying the difference between the result of quantifying the arrangement of one cell and the result of statistical processing using a t-test. Can be included.
- the step of evaluating the cell condition is a step of quantifying the difference between the result of quantifying the arrangement of one cell and the result of statistical processing using a chi-square test. Can be included.
- a system for evaluating the state of a cell based on image information obtained by photographing a cell containing a chromosome territory the extraction processing unit extracting a chromosome territory from the image, and the extracted chromosome
- a quantification processing unit that quantifies the arrangement state of territory a storage unit that stores the result of quantifying the arrangement state, a statistical processing unit that statistically processes the result of quantifying the arrangement states of a plurality of cells, and
- An evaluation system is provided that includes an evaluation unit that evaluates the state of one cell based on the result of quantifying the state of arrangement of cells and the result of statistical processing.
- the quantification processing unit normalizes the arrangement state of the chromosome territory extracted from the image, standardizes the arrangement state of the chromosome territory, and quantifies the arrangement state. Can.
- the extraction processing unit can extract the nucleus of the cell and extract the chromosome territory contained in the nucleus, and the quantification processing unit normalizes the shape of the nucleus,
- the coordinates of the chromosome territory can be transformed based on the pre-standardized nucleus and the post-normalized nuclear shape change.
- the present invention is a program for causing a computer to evaluate the state of a cell based on image information obtained by photographing a cell containing a chromosome territory, and the computer extracts the chromosome territory from the image.
- the method is characterized in that it functions as a means for quantifying the arrangement state after standardization of the arrangement state of the chromosome territory, a means for evaluating the state of the cell based on the arrangement state of the quantified chromosome territory.
- a program is provided.
- a program characterized by functioning as, is provided.
- the image is composed of a plurality of pixels each having an attribute value
- the step of extracting the chromosome territory comprises the steps of classifying the image into a plurality of classes, and any of the classes. Extracting from the image as a region representing a chromosome territory.
- the step of classifying includes setting an initial value of an attribute parameter indicating the attribute value of each class and a mixing ratio of each class, and based on the attribute parameter and the mixing ratio, each of the plurality of pixels is classified into each class.
- the evaluation function indicating the goodness of estimation is calculated based on the step of calculating the class belonging probability classified into two classes, the mixed probability distribution defined by the attribute parameter and the mixing ratio, and the belonging probability, and the evaluation function Determining whether the predetermined condition is satisfied or not, updating the attribute parameter and the mixing ratio if the evaluation function does not satisfy the predetermined condition, and determining the attribute parameter when the evaluation function satisfies the predetermined condition Classifying the plurality of pixels into one of a plurality of classes based on the mixture ratio. And wherein the door.
- chromosome territory can be extracted with high accuracy from the image.
- accurately extracting the chromosome territory it is possible to reduce the variation in the quantification result of the chromosome territory, and to evaluate the cell state well.
- the step of calculating the class belonging probability includes the steps of decomposing a plurality of pixels into a plurality of subspaces according to their attribute values, and a ratio of pixels contained in the subspaces. Calculating the coarse-grained empirical probability distribution, and calculating the coarse-grained class belonging probability in which each subspace is classified into each class based on the attribute parameter and the mixture ratio.
- the step of calculating the evaluation function may include the step of calculating the coarse-grained mixture probability distribution by averaging the mixture probability distribution defined by the attribute parameter and the mixture ratio in the subspace, The evaluation function can be calculated based on the empirical probability distribution, the coarse-grained mixture probability distribution, and the coarse-grained class belonging probability. By performing such rough-graining, the processing time can be significantly shortened. The coarse-graining process will be described in detail below.
- the following are also effective as a method of extracting a chromosome territory according to the present invention.
- each pixel on the image is classified into a plurality of classes, and in an extraction method for extracting a desired region,
- a set of pixels for which a data space consisting of all the possible attribute values of each pixel of an image is divided into subspaces at a given resolution, and attribute values in each subspace, and the attributes of the pixels A first step of constructing a coarse-grained data space holding an average of the values and the number of the pixels;
- a second step of calculating a coarse-grained empirical probability distribution over the coarse-grained data space by dividing the number of pixels of each subspace by the total number of pixels included in the image;
- a fourth step of calculating a coarse-grained conditional probability distribution by calculating the conditional probability distribution when a class is specified from the class parameters and averaging the conditional probability distributions in each subspace;
- a fifth step of calculating a class belonging probability which is a probability that each pixel belongs to each class by multiplying the mixture ratio of the classes by the coarse-grained conditional probability distribution;
- a sixth step of updating the class parameters and the mixture ratio of the classes
- coarse-grained log likelihood can be used as the evaluation function.
- the class parameter and the mixture ratio of classes can be updated so as to increase the evaluation function.
- the third step by initializing the number of classes Can.
- the number of classes can be set to a given value in advance.
- the evaluation function satisfies a given condition
- the fourth, fifth, sixth, seventh and eighth steps can be repeated.
- the average value of the attribute values of the pixels included in each partial space is calculated, and the coarse-grained in each partial space is calculated using the average value.
- the number of pixels belonging to each class is calculated by multiplying the mixing ratio of classes when the evaluation function satisfies a given end condition by the total number of pixels contained in the image, and the class belonging probability is calculated.
- An extraction method characterized in that pixels belonging to each class are determined by selecting the pixels of the number of pixels from the highest order.
- an extraction method is characterized in that AIC is used as an evaluation function, and parameters are changed so as to reduce the evaluation function in the sixth step.
- An extraction method comprising: in the seventh step, using MDL (Minimum Description Length) as an evaluation function, and changing the parameter so as to reduce the evaluation function in the sixth step.
- MDL Minimum Description Length
- step B4 select the subspace with the highest coarse-grained experience probability
- a subspace track having a representative value whose distance from the representative value of the subspace with the highest coarse-grained experience probability is within the vicinity radius is selected as a neighborhood set (step B5),
- Step B6 Check whether the shortest distance between the representative value of the subspace contained in the class that has already been classified and the representative value of the subspace contained in the neighborhood set is larger than the neighborhood radius (Step B6),
- step B6 if the shortest distance between the representative value of the subspace included in the class that has already been classified and the representative value of the subspace included in the neighborhood set is larger than the neighborhood radius, the neighborhood set is regarded as a new class. And delete the neighborhood set from the classification target subspace (step B7),
- step B6 If, in step B6, the shortest distance between the representative value of the subspace included in the already classified class and the representative value of the subspace included in the neighborhood set is smaller than the neighborhood radius, the class for which the neighborhood set has already been classified. While deleting the neighborhood set from the classification target subspace (step B8),
- step B9 if the classification target subspace is not an empty set, step B4 and subsequent steps are repeated.
- step B9 if the classification target subspace is an empty set, it is checked whether the number of classes for which classification has been completed is a given number or more (step B10), In step BIO, if the number of classes for which classification has been completed is smaller than the given number, the neighborhood radius is reduced (step B11), and the steps from step B3 are repeated,
- step B10 if the classification target subspace is an empty set and the number of classified classes is larger than the given number, class parameters are calculated in each class, and this is used as the initial value of the class parameters. And the ratio of the number of subspaces contained in each class as the initial value of the mixing ratio of the class (step B12).
- Steps B7 and B8 if the shortest distance between the representative value of the partial space included in the class that has already been classified and the representative value of the partial space included in the neighborhood set is equal to the neighborhood diameter, Any processing may be performed.
- an extraction device that classifies each pixel on the image into a plurality of classes and extracts a desired region
- An input device for reading an image An input device for reading an image
- a set of pixels for which a data space consisting of all the possible attribute values of each pixel of an image is divided into subspaces at a given resolution, and attribute values in each subspace, and the attributes of the pixels An area coarse-graining device that holds the average of the values and the number of the pixels and forms the coarse-grained data space;
- a coarse-grained empirical probability distribution calculator that divides the number of pixels in each subspace by the total number of pixels included in the image to calculate a coarse-grained empirical distribution over the coarse-grained data space;
- the class probability that defines the attributes of each class is initialized, the mixed ratio of the classes is initialized, and the conditional probability distribution when the class is specified is calculated from the class parameters that specify the attributes of each class, and the class is specified.
- a coarse-grained conditional probability distribution device that calculates a coarse-grained conditional probability distribution by averaging the conditional probability distributions in A in each subspace;
- a class belonging probability calculation device for calculating a class belonging probability that is a probability that each pixel of an image belongs to each class from the coarse-grained conditional probability distribution;
- a parameter updating device for updating the class parameter and the mixture ratio of the class, and an evaluation function calculating device for calculating an evaluation function using the coarse-grained conditional probability distribution; Each pixel is checked based on the class parameter, the mixture ratio of classes, and the class membership probability when the evaluation function satisfies a given end condition and whether the evaluation function satisfies a given end condition or not.
- a region extraction device that determines the class to which it belongs and extracts the desired region;
- An extraction device characterized by having.
- the evaluation function calculation apparatus can use coarse-grained log likelihood, AIC, MDL or structural risk as an evaluation function.
- the parameter update device can update the class parameter and the mixture ratio of classes so as to increase the evaluation function.
- the parameter update device updates the class parameter and the mixture ratio of classes so as to reduce the evaluation function.
- An extraction device further comprising a restoration device.
- the attribute value of each pixel constituting the image is regarded as a random variable, and a desired region is extracted based on the probability distribution of the estimated pixel values.
- the attribute value for example, in the case of a monochrome image, the value of luminance, and in the case of a color image, the intensity of a color element of red (R), green (G) and blue (B) can be used.
- each pixel is classified into a plurality of groups having similar attributes based on the attribute value of each pixel.
- a set of pixels having similar attributes is called a class.
- Each class is characterized, for example, by the average value or variance of attribute values belonging to that class.
- k is the number of classes.
- the probability that the j-th pixel force X takes an attribute value is a mixture of C can be expressed by distribution
- J I ⁇ is the condition given that the data originates from the ith class
- W is the mixing ratio of each class
- X is represented by an integer value such as 0 to 255. Also, the image is
- X is a three-dimensional vector (X 1, X 2, X i jl) 2 j whose component is the value of the RGB color component
- the mixing ratio w represents the area ratio (in the case of a two-dimensional area) or the volume ratio (in the case of a three-dimensional area) of the regions belonging to different classes.
- a bright image area (class 1) represented by a feature with an average brightness of 200 and a standard deviation of brightness of 20, a dark area (a class 2 with average brightness of 50 and a standard deviation of luminance of 10)
- a monok image composed of two area forces.
- 70% of the area of the image is bright and 30% is dark.
- the class parameters ⁇ (200, 20)
- ⁇ (50, 10)
- the mixture distribution of this image is
- n is the number of pixels included in the image.
- maximum likelihood the number of pixels included in the image.
- the parameter can be estimated using the expectation value Q of the complete log likelihood represented by here,
- the parameters w and ⁇ are updated so as to increase Q, and Q is calculated again. Repeat this procedure until Q does not eventually increase.
- the region w is extracted using parameters w and ⁇ when Q does not increase.
- the value of the probability of belonging to the i-th class of each pixel is first examined. Then, pixels with probability values greater than or equal to a certain value are classified as belonging to that class. When all pixels have been classified into each class, select the class with the desired attribute from k classes, and extract the pixels belonging to that class from the image. Can be extracted in
- a coarse-grained probability distribution is introduced in order to speed up this Q maximization.
- the coarse-grained probability distribution is constructed by decomposing a space consisting of all possible values of data (hereinafter referred to as data space) into N subspaces that do not intersect each other, and assigning probabilities to each subspace.
- the coarse-grained conditional probability distribution in the j-th subspace is
- the coarse-grained full log likelihood given in equation (9) is only for the subspace I'm doing the sum. As shown later, by performing such coarse-graining, the amount of calculation is significantly reduced.
- Ability to reduce S can. For example, in the case of an image of 512 ⁇ 512 pixels, using equation (4), the force S that needs to be summed for 260,000 or more pieces of data, and the coarse-grained distribution of the present invention The sum can be reduced to about 1000, and the ability to perform fast estimation S can be achieved.
- the coarse-grained probability value in each subspace is approximated by the probability value in the average value of the data contained in the subspace.
- the integration (or summation) operation in can be omitted, and the amount of calculation can be further reduced.
- parameters estimated using the coarse-grained probability distribution can be used as initial values, and estimation can be performed again with the original resolution.
- the number of successive parameter updates may be much smaller than in the case of estimation with the original resolution from the beginning. Therefore, highly accurate estimation can be performed at high speed.
- the evaluation function it is possible to select an AIC, an MDL, and a structural 'risk', respectively, and a model that gives the lowest value S with the lowest value. If AIC, MDL, or Structural 'risk is used as the evaluation function, the value of the evaluation function will increase if an excessive number of parameters are used, so it is possible to estimate the optimum number of parameters. This makes it possible to estimate the appropriate number of regions in which the image is formed of several types of regions.
- the extraction method of the present invention is based on the class belonging probability calculated from the class parameter. Since region extraction is performed, it is possible to perform region extraction with high accuracy without being affected by errors due to coarse-graining.
- FIG. 1 is a flow chart showing the processing procedure of the method for evaluating the state of cells in the embodiment of the present invention.
- FIG. 2 is a block diagram showing an evaluation device according to the present embodiment that executes the processing procedure shown in FIG. 1.
- FIG. 3 is a block diagram showing an extraction processing unit shown in FIG.
- FIG. 4 is a flowchart showing the processing procedure of the extraction processing unit.
- FIG. 5 is a view showing a coarse-grained data space and a partial space when coarse-grained.
- FIG. 6 A flowchart showing a processing procedure for determining initial values of parameters based on coarse-grained empirical distribution.
- FIG. 7 is a diagram for specifically explaining the procedure described with reference to FIG.
- FIG. 8 is a block diagram showing another example of the extraction processing unit.
- FIG. 9 A flowchart showing the processing procedure of the extraction processing unit shown in FIG.
- FIG. 10 A flow chart showing the processing procedure for quantifying the arrangement state of chromosome territory.
- Fig. 11 is a schematic view showing how coordinates of chromosome territory are transformed.
- FIG. 12 shows a plurality of chromosome territories and chromosome territories transformed into coordinates in a standard sphere.
- FIG. 13 is a block diagram showing another example of an evaluation device.
- FIG. 14 is a diagram showing a chromosome territory extracted by the extraction processing unit in the present embodiment.
- FIG. 15 In the pre- and post-differentiated cell groups, the distance between the centers of gravity of chromosomes 12 and 16 at the nearest position was compared with the normalized relative radius to the nucleus radius.
- FIG. 16 A diagram showing the radial distribution of the 12th chromosome before and after cell differentiation before normalization of the nuclear shape.
- FIG. 17 A diagram showing the radial distribution of No. 12 chromosome after cell differentiation before cell differentiation after normalization of the nuclear shape.
- FIG. 1 is a flowchart showing the processing procedure of the method for evaluating the state of a cell in the embodiment of the present invention.
- an image of a chromosome territory is extracted from an image obtained by photographing cells containing the chromosome territory (S20).
- an image of a cell nucleus and a plurality of chromosome territories is extracted from the three-dimensional image.
- the arrangement state of the chromosome territory is quantified (S22).
- the results of quantifying the alignment state of the chromosome territory are preserved.
- the process of step 20 and step 22 is repeated for a plurality of samples, and the quantification results of the plurality of samples are accumulated (S24). Also, for a certain sample, the process of step 20 and step 22 is performed, and the quantified result is compared with the quantified result of a plurality of samples accumulated in step 24 to evaluate the cell condition in the sample. (S26).
- FIG. 2 is a block diagram showing an evaluation device according to the present embodiment that executes the processing procedure shown in FIG.
- the evaluation device 100 includes an image acquisition unit 102, an extraction processing unit 104, a quantification processing unit 106, an evaluation unit 110, an output unit 112, and a quantification result storage unit 114.
- the image acquisition unit 102 acquires a three-dimensional image of a cell containing a chromosome territory.
- the extraction processing unit 104 extracts an image of a chromosome territory from the three-dimensional image.
- the quantification processing unit 106 quantifies the arrangement state of the chromosome territory.
- the quantification processing unit 106 stores the quantification result in the quantification result storage unit 114.
- the evaluation unit 110 compares the quantification result of a certain sample with the quantification results of a plurality of samples stored in the quantification result storage unit 114 to evaluate the cell state of the sample.
- the output unit 112 outputs the extraction result of the image of the chromosome region by the extraction processing unit 104 and the evaluation result of the cell state by the evaluation unit 110.
- the output unit 112 can be configured by
- Three-dimensional images of cells containing chromosome territories can be obtained using 3D-FISH (Fluorescent in situ hybridization) method.
- the image acquisition unit 102 can be configured by an image scanner, a digital camera, or the like, and can be capable of reading an image obtained using the 3D-FISH method.
- the image acquisition unit 102 is connected to a computer connected to a biological microscope. It is possible to configure more, and it is also possible to input an image directly through the network.
- a labeled probe (a chromosome territory of interest and a DNA fragment complementary to the nucleus) is added and reacted at about 37 ° C., and the probe complementarily binds to the chromosome territory of interest and the nucleus.
- the probe is attached with a dye that develops color according to the type of chromosome territory of interest and the nucleus for each nucleus. In this way, only the dye can be specifically developed by irradiating a wavelength at which a specific dye develops color.
- the sample prepared as described above is irradiated with a specific wavelength using a confocal heart laser microscope to obtain an image including a target chromosome territory.
- an image containing each chromosome territory and an image containing a nucleus By superimposing an image containing each chromosome territory and an image containing a nucleus, an image containing a plurality of chromosome territories can be obtained.
- chromosome territory or “nucleus”
- nucleus the chromosome territory image and the image of the chromosome territory on the image which is not the nucleus itself itself are shown.
- the extraction processing unit 104 shown in FIG. 2 performs image processing for extracting a chromosome territory from the image obtained as described above.
- the extraction processing unit 104 extracts an image of a chromosome territory and a nucleus from an image by classifying an image including a plurality of pixels into a plurality of classes and associating a plurality of classes with a chromosome territory and a nucleus respectively.
- the extraction processing unit 104 assumes the existence of a plurality of luminance sources having different luminances in each pixel, and based on the probability that each luminance source belongs to that pixel and the mixing ratio of the luminance sources in the image, Do classification.
- Patent Document 1 Such a method is described, for example, in Japanese Patent Application Laid-Open No. 2001-92980 (hereinafter referred to as Patent Document 1).
- Patent Document 1 the expectation value of the region belonging probability to which each pixel in the image belongs to each region is calculated from the image data, and the region belonging probability and the region parameter force are evaluated based on the mixed probability distribution obtained.
- the concept of coarse graining is further introduced to this method, and the optimum parameters are obtained at high speed.
- FIG. 3 is a block diagram showing the extraction processing unit 104 shown in FIG. Figure 4 shows the extraction processing unit
- 10 is a flowchart showing a processing procedure of 104.
- the extraction processing unit 104 in the embodiment of the present invention is an input device 1 for reading image data, a data space coarse-graining device 2, a parameter updating device 6, and a coarse-grained experience probability distribution calculating device 3 And a coarse-grained conditional probability distribution calculation unit 4, a class belonging probability calculation unit 5, an evaluation function calculation unit 7, an area extraction unit 8, and an output unit 9.
- the input device 1 inputs a color image of cells obtained by the 3D-FISH method as described above from the image acquisition unit 102 (step Al).
- the input device 1 sends the read data to the data space coarse-graining device 2.
- the data space coarse-graining device 2 decomposes the data space into N subspaces that do not intersect each other to form a coarse-grained data space (step A2).
- the data space means a set of all possible attribute values of each pixel of the image.
- the attributes of each pixel can be represented by intensity, which is typically a one-dimensional space represented by an integer value of 0 by 255.
- a three-dimensional space is usually represented by integer values of 0 to 255 for each of R, G and B color elements.
- the data space coarse-graining device 2 can divide each of RGB values into 8 and make an 8 ⁇ 8 ⁇ 8 cube into one partial space.
- the resolution does not have to be the same on the three-dimensional axis. For example, separate R axis by h, G axis by h, B axis by h
- h x h x h means to separate each RGB value.
- the data space coarse-graining device 2 holds a set of pixels (numbers of pixels) taking values in each subspace, an average of attribute values of those pixels, and the number of pixels.
- FIG. 5 shows a coarse-grained data space 11 and a portion when coarse-grained at a resolution h
- this pixel takes a value in the subspace designated by the index (16, 31, 2). This is expressed below as X is included in this subspace.
- a coarse-grained empirical probability distribution calculating device calculates the number of pixels included in each partial space and the number of pixels. Send to 3.
- the coarse-grained experience probability distribution calculating device 3 calculates a coarse-grained experience probability distribution (step A3).
- the coarse-grained empirical probability distribution refers to a probability distribution calculated by dividing the observed data included in each subspace of the coarse-grained data space by the total number of pixels.
- the probability distribution calculator 3 sends the coarse-grained empirical probability distribution to the coarse-grained conditional probability distribution calculator 4.
- the coarse-grained conditional probability distribution calculating device 4 initializes class parameters (step A4). The procedure for determining the initial value of the parameter will be described later.
- the coarse-grained conditional probability distribution calculator 4 calculates the coarse-grained conditional probability distribution (step A5). Specifically, the coarse-grained conditional probability distribution is calculated as follows. Here, the conditional probability power under the condition that the j-th pixel value originates from the i-th class Multidimensional normal distribution such as
- X is a three-dimensional vector representing the value of each color of RGB
- ⁇ is i
- ⁇ ⁇ is the covariance matrix of the first class
- I ⁇ I, 1 respectively represent the determinant of the matrix ⁇ , the inverse matrix.
- (X ⁇ ) ⁇ represents transposition.
- conditional probability under the condition that it was generated from the j-th pixel value power class is the one-dimensional normal distribution as follows:
- X is a scalar value indicating the value of luminance
- ⁇ is a scalar value indicating the average luminance of the i-th class
- ⁇ is the standard deviation of the luminance of the i-th class.
- the coarse-grained conditional probability distribution calculation device 4 calculates this conditional probability by the equation given by the equation (6).
- the coarse-grained conditional probability distribution calculation device 4 can also approximate the coarse-grained conditional probability distribution according to equation (13). By this method, it is not necessary to perform the operation given by Eq. (6) each time the parameter is updated, and the amount of calculation can be significantly reduced.
- the coarse-grained conditional probability distribution calculation device 4 can also determine the initial value of the parameter based on the coarse-grained empirical probability distribution. Each subspace is roughly classified into classes, average values and variance values are obtained in each class obtained as a result of classification, and these values are used as initial values for parameter estimation.
- FIG. 6 is a flowchart showing a processing procedure for determining initial values of parameters based on the coarse-grained empirical distribution. The procedure for setting the initial value of the parameter will be described below with reference to FIG.
- step B1 the neighborhood radius and the number of classes to be classified are set.
- the near radius is a reference value to roughly classify all subspaces within that range as belonging to the same class. For example, in the case of color images, pixels of similar colors are close to RGB values, so it is natural to classify them into the same class. Because it is considered to be.
- the initial value of this neighborhood radius is set to a sufficiently large value, such as 50.
- the number of classes to be classified uses the given value as it is.
- step B2 a representative value of each partial space is set for each partial space.
- each subspace for example, a median of the subspace or the like can be used. In the following, the distance between these representative values is taken as the distance of the subspace.
- a set of subspaces to be classified is set (step B3).
- ⁇ a set consisting of the entire subspace including data.
- the number of classified classes i is 1, and the initial value of classified class C is an empty set.
- step B4 the subspace with the highest probability of coarse-grained experience is selected (step B4). Write this subspace A.
- the distance between A and the subspace belonging to ⁇ is selected (step B4).
- step B5 The separation is examined, all subspaces within the neighborhood radius r are selected, and this is set as the neighborhood set (step B5). Below we write the neighborhood set as B.
- Step B6 Find the shortest distance to the subspace where i s is to be examined, and check if it is greater than the neighborhood radius r (Step B6). If this shortest distance is greater than r, then neighborhood set B has already been classified
- step ⁇ 7 Since the class is complete, delete it from the classification set ⁇ . In Fig. 6, this deletion is indicated by the symbol " ⁇ -Q ⁇ Bs". After updating ⁇ , return to step ⁇ 4 (step ⁇ 7)
- the neighborhood set B is close to C in C and the attribute
- step B8 Delete the ⁇ force (step B8).
- step ⁇ 9 it is checked whether ⁇ is an empty set (step ⁇ 9), and if not an empty set, proceed to step ⁇ 4, and if ⁇ is an empty set, the number of classes for which classification has been completed is k
- step BIO step BIO
- step BIO step BIO
- step BIO step BIO
- step BIO step BIO
- step BIO step BIO
- step BIO step BIO
- step BIO multiply the neighborhood radius by a constant smaller than 1 and reduce the neighborhood radius.
- a value such as 0.9 can be used. Then repeat steps B3 and below.
- step B9 if ⁇ is an empty set and the number of classified classes is larger than the given number, the desired number of classifications have been completed, so classes within each class are completed.
- a parameter is calculated, which is used as an initial value of the class parameter, and the ratio of the number of partial spaces included in each class is used as an initial value of the mixing ratio of the class (step B12).
- FIG. 7 is a diagram for specifically explaining the procedure described with reference to FIG.
- the data space is one-dimensional and a total of 10 subspaces are set (Fig. 7 (a)).
- the horizontal axis represents the number of the subspace
- the vertical axis represents the coarse-grained experience probability distribution.
- step B1 for example, the number of classes is 2 and the neighborhood radius is r.
- step B2 a representative value of each subspace is set.
- step B3 since the initial value of the classification target set ⁇ is the entire subspace including data,
- ⁇ ⁇ , ⁇ , ⁇ , ⁇ , ⁇ , ⁇ , ⁇ , ⁇
- A, A, A, and A have a coarse-grained probability of 0, ie, included in these subspaces
- step B4 of the subspaces included in the classification target set, A7 having the highest coarse-grained empirical probability is selected as As (FIG. 7 (a)).
- step B5 select a subspace within the A radius radius r and let it be Bs. Since the subspaces within the neighborhood radius shown in Figure 7 (a) are A, A, A, and A,
- step B6 since there are no classes yet classified, this B is not changed.
- Adopt B as class C remove B from the classification set, and return to step B4.
- Figure 7 (b In the above, the bar graph indicating the height of the coarse-grained experience probability is shown in white, indicating that the partial space has been excluded from the grouping target force.
- step B 4 the one with the highest coarse-grained experience probability among the remaining classification target sets is A, so this is newly set to A (FIG. 7 (c)).
- step B5 from A
- step B6 classified classes, ie,
- A is examined, it includes A and A that are within the neighborhood radius r from A4. Therefore the present
- the classification target set is now empty, and classification of all subspaces is completed, but the number of classes classified is 1 and the desired number of classes is not 2 (step B10). Therefore, the neighborhood radius is reduced (step B11), and the steps after step B3 are repeated.
- This B is adopted as the first class C as it is, and the most of the remaining classification sets is
- Each subspace is classified into two desired classes.
- the mean or variance can be obtained in the classified class and used as an initial parameter of estimation to be performed later.
- the initial values of the parameters it is possible to reduce the amount of calculation after this and perform region extraction at high speed.
- setting the initial parameters appropriately is effective in preventing the local optimal solution from falling in the course of maximizing coarse-grained log likelihood.
- the coarse-grained conditional probability distribution calculation device 4 can obtain the coarse-grained conditional probability distribution with the parameters thus obtained as initial values.
- the coarse-grained conditional probability distribution calculating device 4 sends the obtained coarse-grained conditional probability distribution to the class belonging probability calculating device 5.
- the class belonging probability calculating device 5 calculates the class belonging probability using equation (11) (step A6).
- This class belonging probability represents the probability of belonging to the i th class of pixel powers contained in the j th subspace. Therefore, region extraction can be easily performed by calculating this class belonging probability for each pixel of the image and classifying each pixel into a class with high probability.
- the class belonging probability calculating unit 5 sends the calculated class belonging probability to the parameter updating unit 6.
- the parameter updating device 6 updates the parameters so as to maximize equation (9). Specifically, the parameters are updated as follows.
- I the average value of the data contained in the j-th subspace A, as defined in equation (14).
- the wi, ⁇ i update can be done similarly using the intensity value as X, and the update of the standard deviation ⁇ is
- the parameter update unit 6 sends the updated parameters to the evaluation function calculation unit 7.
- the evaluation function calculation device 7 calculates the coarse-grained log likelihood by using the equation (12) (step A8).
- m represents the number of all parameters. Since the AIC is proportional to the coarse-grained log likelihood multiplied by a minus value, the updating performed by the parameter change unit 6 changes the AIC in the decreasing direction. In addition, since a term proportional to the parameter is added, it is assumed that the estimation result using the model with few parameters is better if the coarse-grained log likelihood is the same. By using this evaluation function, it is possible to suppress the overpowering IJ-like fit to the data and to make a noise-resistant estimation.
- c, a and a are constants determined by the nature of the probability distribution
- the evaluation function calculation unit 7 determines whether the change of the evaluation function satisfies a given end condition or not If the termination condition is satisfied, the current parameter is sent to the region extraction unit 8. If the termination condition is not satisfied, the current parameter is sent to the coarse-grained conditional probability distribution calculation unit 4 (step A9).
- the end condition for example, the difference between the current evaluation function value and the previously calculated evaluation function value is divided by the current evaluation function value, and whether the absolute value of the value is less than or equal to 0. 0001, etc. Conditions can be used.
- the region extraction device 8 can also set the threshold automatically. In order to do this, the following process is performed. In order to extract the i-th region, first, the mixture ratio w of the estimated class is multiplied by the total number of pixels to obtain an estimated value of the number of pixels belonging to each class. Let this number be n.
- the value of the roughly classified class belonging probability in equation (11) becomes the threshold of the probability of belonging to the i-th region.
- the threshold automatically set in this manner, chromosome territory can be extracted with high accuracy from the image.
- the same effect can be obtained by the following.
- the probability of membership represented by equation (11) is calculated for all pixels. By this, the probability of belonging to each of the k classes is obtained for each pixel.
- the region of the chromosome territory should be displayed as a bright region on the image, so If the above-mentioned procedure is performed for a class having a large average brightness, it is possible to automatically extract a chromosome territory region.
- the region extraction device 8 sends the data of the extracted region to the output device 9 after the region extraction is completed.
- the data space coarse graining device 2 the coarse grained experience distribution calculation device 3, the coarse graining conditional probability distribution 4, the class belonging probability calculation device 5, the parameter update device 6, the evaluation function calculation device described above 7.
- the area extracting device 8 can be constructed using, for example, a computer such as a personal computer or a workstation or a super computer.
- the output unit 9 receives the area data from the area extraction unit 8 and outputs it to the quantification processing unit 106 in FIG. 2 (step All). Also, the output from the output device 9 is displayed by the output unit 112.
- FIG. 8 is a block diagram showing another example of the extraction processing unit 104.
- FIG. 9 is a flow chart showing the processing procedure of the extraction processing unit 104 shown in FIG.
- the same components as those described with reference to FIGS. 3 and 4 are denoted by the same reference numerals, and the description thereof will be appropriately omitted.
- the region restoration apparatus 10 checks whether the coarse-grained resolution is equal to the original resolution (step A12), and coarse-grained If so, restore the data to its original resolution (step A13). If not coarse-grained, it means that the estimation at the original resolution is completed, so the parameter estimated at the original resolution is sent to the region extraction unit 8. To return the data to the original resolution, set the coarse-grained resolution to the smallest unit of data attribute value (for example, 1), and repeat the same method as the method described in the first embodiment. Just do it. In this case, although estimation takes more time than the method described with reference to FIGS. 3 and 4, parameters can be estimated with higher accuracy, and as a result, more accurate region extraction can be performed. It becomes. In addition, since the parameters estimated using the coarse-grained probability distribution are already estimated in the vicinity of the optimal parameters, the optimal parameters can be estimated with fewer parameter updates than estimation from the high resolution from the beginning. It is possible to perform region extraction much faster.
- the smallest unit of data attribute value for example, 1
- the quantification processing unit 106 converts the coordinates of the chromosome territory extracted by the extraction processing unit 104 according to a certain standard. Process to standardize. By performing such processing, it is possible to quantify the arrangement state of the chromosome territory on a uniform basis for a plurality of samples. As a result, it is possible to perform statistical processing or the like on the quantification result of the arrangement state of the chromosome territory in a plurality of samples.
- FIG. 10 is a flow chart showing a processing procedure for quantifying the arrangement state of chromosome territory.
- FIG. 11 is a schematic view showing how coordinates of chromosome territory are transformed.
- the processing procedure of the coordinate conversion will be described with reference to FIGS.
- the quantification processing unit 106 normalizes the cell nucleus 120 as shown in FIG. 11 (a) (FIG. 10, S 100). Here, first, the center of gravity P and volume V of the cell nucleus 120 are calculated. Continue,
- a standard sphere 124 which has a volume V centering on the center of gravity P is calculated.
- the nucleus 120 of the cell contains the chromosome territory 122.
- the cell nucleus 120 can contain multiple chromosomal territories.
- the coordinates of the chromosome territory 122 are transformed according to the shape and size of the cell nucleus 120 and the standard sphere 124 (FIG. 10, S102). Specifically, as shown in FIG. 11 (c), a straight line 1 passing through the center of the standard sphere 124 (same as the center of gravity P) and each point p in the chromosome territory 122
- the quantification processing unit 106 quantifies the arrangement state of the chromosome territory (FIG. 10, S104).
- FIG. 12 shows a plurality of chromosome territories 126a and chromosome territories 126b coordinate-transformed into a standard sphere 124.
- the quantification processing unit 106 calculates the distance between the chromosome territory 126a and the chromosome territory 126b thus coordinate-transformed.
- the nucleus of a cell may take various shapes depending on the time, even for the same cell.
- the influence of the change in the shape of the cell nucleus due to such a difference in time is eliminated, and the arrangement state of the chromosome territory is appropriately quantified.
- the quantification processing unit 106 divides the distance between the chromosome territories by the radius of the standard sphere 124 and calculates the value of the quantification result. can do. In this way, it is possible to process quantification results of arrangement states of chromosome territories in a plurality of samples each having a different nuclear size with a unified index.
- the arrangement state of the chromosome territory can be quantified S by using the distance between multiple chromosome territories, the spread of the chromosome territory, the arrangement direction, the distance between the chromosome territory and the reference point in the nucleus, and the like.
- the distances between multiple chromosome territories can be calculated using various methods.
- the quantification processing unit 106 can calculate the center of gravity of each chromosome territory, and can calculate the distance between the centers of gravity.
- the center of gravity is the center of gravity of the area assumed to have the same specific gravity of each pixel constituting the chromosome territory.
- the center of gravity of each chromosome territory is, for example, the center of gravity weighted by the value of luminance.
- I is the intensity value of the pixel at location X. Also,
- the quantification processing unit 106 can also calculate the distance between the boundaries of a plurality of chromosome territories.
- the distance between the boundaries of chromosome territories is, for example, a plurality of chromosome territories The shortest distance of the surface of
- the quantification processing unit 106 approximates the shapes of a plurality of chromosome territories with a normal distribution, calculates the average value of the normal distribution of each of these chromosome territories, and calculates the distance between the average values. You can also In addition, after the shapes of a plurality of chromosome territories are respectively approximated by a normal distribution, the center of gravity of each chromosome territory may be calculated as described above to calculate the distance between the centers of gravity, or the distance between boundaries may be calculated.
- the quantification processing unit 106 can also detect a skeleton line of each of a plurality of chromosome territories, calculate the center of gravity of each skeleton line, and calculate the distance between the centers of gravity.
- the quantification processing unit 106 determines the distances between a plurality of chromosome territories, the distance of Mahalanobis, the distance of Bhattacharyya, the distance of Kullback-Leibler, and the symmetry. It can be calculated using the distance between the Kullback and the Librar and the absolute distance.
- the Mahalanobis distance can be calculated according to the following equation.
- the intraclass covariance matrix defined by, w, W indicate the mixing ratio of classes.
- the Batacharya distance can be calculated according to the following equation.
- the Kullback Librar distance can be calculated according to the following equation.
- the distance of the symmetric Kullback Libror can be calculated according to the following equation.
- the absolute distance can be calculated according to the following equation c
- a method of approximating the shape of the chromosome territory using a normal distribution can be used. This makes it possible to quantify the direction and size of the spread of chromosome territory.
- the value of luminance is regarded as the magnitude of probability of existence of chromosome territory, and it is approximated by mixed normal distribution.
- This processing can be performed in the same procedure as the extraction method of the chromosome territory by the extraction processing unit 104 described above. At this time, the coordinates of each pixel on the image are regarded as a two-dimensional probability vector, and the luminance value at that position is regarded as a probability value.
- the luminance value of all the pixels in the image is obtained, and the luminance value of each pixel is divided by that value to perform normalization. In this way, the shape of the chromosome territory is quantified.
- the chromosome territory exists to some extent in the nucleus of a cell, and the shape of this spread is considered to be related to the ease of interaction with chromosomes or with the outside through the nuclear membrane. Therefore, it is expected that the cell state can be evaluated with high accuracy by taking into consideration the shape of the spread of the chromosome territory when quantifying the arrangement state of the chromosome territory.
- the principal axis of each chromosome territory can be detected, and the result of quantifying the angle formed by those principal axes can be obtained.
- the principal axis of the chromosome territory can also be calculated, for example, using a normal distribution as described above. Thus, by approximating the shape of the chromosome territory by normal distribution, the main axis of the spread of each chromosome territory can be made clear.
- ⁇ , ⁇ ) is a two-dimensional normal distribution
- ⁇ is the mean value
- ⁇ is the covariance matrix
- W is the ratio of the size of the spread
- ⁇ is the center position of the spread
- ⁇ is the main axis of the spread.
- the covariance matrix is subjected to eigenvalue decomposition to obtain the principal axis of spread specifically. As described above, by quantifying the shape, the arrangement state of chromosome territories can be quantified.
- the parallel interaction of the main axes of the spread of two chromosome territories has a larger influence on the area of interaction with each other than in the case of perpendicularity. It is considered easy. Therefore, when quantifying the arrangement state of the chromosome territory, it is expected that the cell state can be evaluated more accurately by considering the shape and arrangement direction of the spread of the chromosome territory.
- the reference point in the nucleus can be the nuclear membrane or the center of gravity. Further, in this case, the nuclear membrane can be the surface of a standard sphere calculated by the above-described method.
- the presence of chromosomes at the outer periphery in the nucleus of cells is more genetically active than that at the center. Therefore, it is expected that the genetic activity of the cell can be evaluated by quantifying the distance between the chromosome territory and the nuclear membrane of the cell and the distance between the chromosome territory and the nuclear membrane of the cell.
- the quantification processing unit 106 can quantify the arrangement state of the chromosome territory using one or more of the above methods.
- the state of cells can be evaluated in more detail by using the results of quantification by a plurality of methods.
- Evaluation of cell condition can be performed by various methods By statistically processing the quantification results obtained for a plurality of samples and comparing the results with the statistical processing, the cells of a certain sample can be evaluated. Can assess the state of
- the statistical processing can use various methods.
- the statistical processing unit 108 can calculate a standard value as a result of quantifying the arrangement state of the chromosome territory of a plurality of samples.
- the standard value it is possible to use the average value of the results of quantifying the state of arrangement.
- the evaluation device 100 can include a statistical processing unit 108 and a statistical result storage unit 116 in addition to the configuration shown in FIG. 2 as shown in FIG.
- the statistical processing unit 108 calculates, for example, the standard value described above, and stores the standard value in the statistical result storage unit 116.
- the evaluation unit 110 can evaluate the state of the cells of the sample by comparing the quantification result of a certain sample with a standard value.
- the statistical processing can also be processing of correlating and storing the result of quantifying the arrangement state of the chromosome territory of each sample and the actual diagnosis result of the provider of the sample.
- quantification data of chromosomal territorial arrangement of cells collected from patients suffering from a disease such as cancer and quantification data of chromosomal territorial arrangements of cells collected from non-diseased persons.
- the evaluation unit 110 evaluates whether or not the subject is afflicted by determining whether the quantification data of the arrangement state of the chromosome territory of a certain subject's force-collected cells is close to which quantification data. be able to.
- the state of the cells is evaluated using the result of quantifying the arrangement state of the chromosome territory, and the presence or absence of a disease is determined based on the result. Because changes in the alignment state of chromosome territories occur earlier than morphological changes occur in cells, early detection of disease can be expected.
- the statistical processing unit 108 can construct a two-eural network based on the quantification results obtained from a plurality of samples.
- a plurality of results quantified by various methods such as distances between multiple chromosome territories, spread of chromosome territory, orientation, and distance between chromosome territory and reference point in nucleus are input. It can be a parameter.
- the evaluation unit 110 can evaluate the state of the cell by the output of the neural network.
- the evaluation unit 110 can quantify the difference between the result of quantifying the arrangement state of one sample and the result of statistical processing using a t-test or a chi-square test.
- Adipocytes (cells 25 and 13) were used to obtain three-dimensional color images (554 x 576).
- color images were obtained by 3D-FISH using probes that complementarily bind to chromosomes 16 and 12, and the nucleus, respectively.
- Such a color image is input by the image acquisition unit 102.
- the extraction processing unit 104 extracts the chromosome territory and the nucleus from the image using the coarse-grained area extraction method described above with the coarse-grained resolution being 8 and the number of classes being 3.
- the classes were assigned to chromosome territory of chromosome 12, chromosome territory of chromosome 16 chromosome and nucleus respectively.
- FIG. 14 shows the extracted image.
- the quantification processing unit 106 calculated the distance between the centers of gravity of the chromosome territory of the 16th chromosome and the chromosome territory of the 12th chromosome.
- Chromosomes 16 and 12 include homologous chromosomes, among which the shortest distance between chromosomes 16 and 12 was adopted. This is because interactions such as translocation occur at the point where the distance of the chromosome territory is closest.
- a standard sphere of cell nuclei extracted from image data was calculated, and a chromosome territory of chromosome 12 and a chromosome territory of chromosome 16 were defined.
- the chromosome territory had a substantially flat shape, conversion in the z direction was not performed, and two-dimensional conversion was performed. The same processing procedure can be applied to the case of performing three-dimensional conversion.
- FIG. 15 shows an example in which the distance between the centers of gravity of chromosome territories is statistically compared between cell groups.
- the vertical axis in FIG. 15 is a relative value obtained by normalizing the distance between the centers of gravity of the chromosomes 12 and 16 at the closest position to the standardized nucleus radius.
- a shows relative distance (0.41) between chromosomes 12-16 territory in cell group before cell differentiation (before cell maturation)
- b shows after cell differentiation (after cell maturation)
- the relative distance (0. 33) between the 12-16 chromosome territories in the cell group is shown. This revealed that the dye territory at the closest position approached significantly (p ⁇ 0. 05) after cell differentiation. This result is obtained for the first time by applying the standardized method described in the embodiment to a group of cells having different sizes and shapes.
- FIG. 16 shows a radial distribution in the nucleus at the position of chromosome 12.
- Figure 16 (a) shows the radial distribution before cell differentiation (before cell maturation)
- Figure 16 (b) shows the radial distribution after cell differentiation (after cell maturation).
- FIG. The horizontal axis represents the distance ( ⁇ m) from the center of gravity of the nucleus
- the vertical axis represents the relative frequency. Since the core shape is different in normal measurement, the distance from the center is expressed here using actual measurement values. From this figure, it is not clear that the chromosomal arrangement changes before and after cell differentiation.
- Fig. 17 shows the radial distribution of the 12th chromosome after the standardization processing described in the present embodiment.
- FIG. 17 shows the radial distribution of the 12th chromosome after the standardization processing described in the present embodiment.
- FIG. 17 (a) shows the radial distribution before cell differentiation
- FIG. 17 (b) shows the radial distribution after cell differentiation.
- the horizontal axis represents relative position distance (%) from the center of gravity of the nucleus
- the vertical axis represents relative frequency.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Probability & Statistics with Applications (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Multimedia (AREA)
- Investigating Or Analysing Biological Materials (AREA)
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
Abstract
Description
Claims
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US10/565,669 US7840357B2 (en) | 2003-07-29 | 2004-07-29 | Method of evaluating chromosome state and evaluation system |
| JP2005512081A JP4609322B2 (ja) | 2003-07-29 | 2004-07-29 | 染色体状態の評価方法および評価システム |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2003-282122 | 2003-07-29 | ||
| JP2003282122 | 2003-07-29 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2005010206A1 true WO2005010206A1 (ja) | 2005-02-03 |
Family
ID=34100990
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2004/010852 Ceased WO2005010206A1 (ja) | 2003-07-29 | 2004-07-29 | 染色体状態の評価方法および評価システム |
Country Status (3)
| Country | Link |
|---|---|
| US (1) | US7840357B2 (ja) |
| JP (1) | JP4609322B2 (ja) |
| WO (1) | WO2005010206A1 (ja) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2012103077A (ja) * | 2010-11-09 | 2012-05-31 | Olympus Corp | 遺伝子異常細胞の解析方法 |
| JP2016522720A (ja) * | 2013-09-26 | 2016-08-04 | コリア インスティテュート オブ コンストラクション テクノロジー | 複合媒質からなる試片に対するx線ct画像の最小単位に存在する各純粋媒質の体積比の測定方法 |
| CN115791122A (zh) * | 2022-11-17 | 2023-03-14 | 中国直升机设计研究所 | 一种无轴承旋翼中央件疲劳试验设计方法 |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2109856B1 (en) * | 2007-02-05 | 2019-01-16 | Siemens Healthcare Diagnostics Inc. | System and method for cell analysis in microscopy |
| JPWO2010098105A1 (ja) * | 2009-02-26 | 2012-08-30 | 国立大学法人名古屋大学 | 培養状態評価装置、培養状態評価方法、インキュベータおよびプログラム |
| US8103672B2 (en) * | 2009-05-20 | 2012-01-24 | Detectent, Inc. | Apparatus, system, and method for determining a partial class membership of a data record in a class |
| JP2011022131A (ja) * | 2009-06-18 | 2011-02-03 | Olympus Corp | 医療診断支援装置、画像処理方法、画像処理プログラム、およびバーチャル顕微鏡システム |
| JP6824845B2 (ja) * | 2017-08-09 | 2021-02-03 | キヤノン株式会社 | 画像処理システム、装置、方法およびプログラム |
| TWI669616B (zh) * | 2018-08-13 | 2019-08-21 | 行政院原子能委員會核能研究所 | 圓柱狀和圓餅狀數位模型之模擬切割方法 |
| CN112037185B (zh) * | 2020-08-21 | 2023-09-05 | 湖南自兴智慧医疗科技有限公司 | 染色体分裂相图像筛选方法、装置及终端设备 |
| CN120992568B (zh) * | 2025-08-06 | 2026-03-27 | 奥辰生物(云南)有限公司 | 一种人工智能的脐带干细胞质量检测数据分析方法及系统 |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH05172810A (ja) * | 1991-06-10 | 1993-07-13 | Kagaku Gijutsucho Hoshasen Igaku Sogo Kenkyusho | 染色体異常検出装置 |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3534009B2 (ja) * | 1999-09-24 | 2004-06-07 | 日本電気株式会社 | 輪郭抽出方法及び装置 |
| JP3649328B2 (ja) * | 2002-04-10 | 2005-05-18 | 日本電気株式会社 | 画像領域抽出方法および装置 |
-
2004
- 2004-07-29 US US10/565,669 patent/US7840357B2/en not_active Expired - Fee Related
- 2004-07-29 JP JP2005512081A patent/JP4609322B2/ja not_active Expired - Lifetime
- 2004-07-29 WO PCT/JP2004/010852 patent/WO2005010206A1/ja not_active Ceased
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH05172810A (ja) * | 1991-06-10 | 1993-07-13 | Kagaku Gijutsucho Hoshasen Igaku Sogo Kenkyusho | 染色体異常検出装置 |
Non-Patent Citations (2)
| Title |
|---|
| PARADA L.A. ET AL.: "Chromosome positioning in the interphase nucleus", TRENDS IN CELL BIOLOGY, vol. 12, no. 9, September 2002 (2002-09-01), pages 425 - 432, XP002982245 * |
| TANABE H.: "Chromosome territory: recent advancement of studies on interphase chromosome positioning in relation to nuclear architecture", ENVIRON. MUTAGEN. RES., vol. 25, no. 1, April 2003 (2003-04-01), pages 11 - 22, XP002982246 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2012103077A (ja) * | 2010-11-09 | 2012-05-31 | Olympus Corp | 遺伝子異常細胞の解析方法 |
| JP2016522720A (ja) * | 2013-09-26 | 2016-08-04 | コリア インスティテュート オブ コンストラクション テクノロジー | 複合媒質からなる試片に対するx線ct画像の最小単位に存在する各純粋媒質の体積比の測定方法 |
| CN115791122A (zh) * | 2022-11-17 | 2023-03-14 | 中国直升机设计研究所 | 一种无轴承旋翼中央件疲劳试验设计方法 |
Also Published As
| Publication number | Publication date |
|---|---|
| JPWO2005010206A1 (ja) | 2007-09-27 |
| US20060246458A1 (en) | 2006-11-02 |
| US7840357B2 (en) | 2010-11-23 |
| JP4609322B2 (ja) | 2011-01-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP4496943B2 (ja) | 病理診断支援装置、病理診断支援プログラム、病理診断支援装置の作動方法、及び病理診断支援システム | |
| EP2391887B1 (en) | Image-based risk score-a prognostic predictor of survival and outcome from digital histopathology | |
| Huang et al. | Time-efficient sparse analysis of histopathological whole slide images | |
| RU2385494C1 (ru) | Способ распознавания изображения текстуры клеток | |
| KR20230012524A (ko) | 디지털 병리 영상에 대한 공간 특징 분석 | |
| JP7550459B2 (ja) | マルチパラメータ細胞・細胞内撮像データから細胞表現型多様性を特徴付けるシステム及び方法 | |
| US20240087122A1 (en) | Detecting tertiary lymphoid structures in digital pathology images | |
| CN116188423A (zh) | 基于病理切片高光谱图像的超像素稀疏解混检测方法 | |
| Song et al. | Hybrid deep autoencoder with curvature gaussian for detection of various types of cells in bone marrow trephine biopsy images | |
| US12272071B2 (en) | Scalable and high precision context-guided segmentation of histological structures including ducts/glands and lumen, cluster of ducts/glands, and individual nuclei in whole slide images of tissue samples from spatial multi-parameter cellular and sub-cellular imaging platforms | |
| CN110838094A (zh) | 病理切片染色风格转换方法和电子设备 | |
| CN114445356A (zh) | 基于多分辨率的全视野病理切片图像肿瘤快速定位方法 | |
| WO2005010206A1 (ja) | 染色体状態の評価方法および評価システム | |
| WO2023226217A1 (zh) | 微卫星不稳定预测系统及其构建方法、终端设备及介质 | |
| CN101390129A (zh) | 用于分析对象聚群的方法和设备 | |
| Sáez et al. | Neuromuscular disease classification system | |
| CN119888351B (zh) | 癌症治疗疗效预测方法、装置、设备及存储介质 | |
| Pati et al. | Deep positive-unlabeled learning for region of interest localization in breast tissue images | |
| CN117635551A (zh) | 宫颈细胞样本年龄及健康程度预测方法 | |
| Zanotelli et al. | A flexible image segmentation pipeline for heterogeneous multiplexed tissue images based on pixel classification | |
| CN115713505A (zh) | 基于聚焦交叉熵的多元特征融合的脑膜瘤分级方法及系统 | |
| JP6329651B1 (ja) | 画像処理装置及び画像処理方法 | |
| CN121687461B (zh) | 基于多模态光学影像融合的皮肤病辅助诊断方法及系统 | |
| EP4687143A1 (en) | Breast cancer classification | |
| Song et al. | A circumscribing active contour model for delineation of nuclei and membranes of megakaryocytes in bone marrow trephine biopsy images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW |
|
| AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): BW GH GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG |
|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
| WWE | Wipo information: entry into national phase |
Ref document number: 2005512081 Country of ref document: JP |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 2006246458 Country of ref document: US Ref document number: 10565669 Country of ref document: US |
|
| 122 | Ep: pct application non-entry in european phase | ||
| WWP | Wipo information: published in national office |
Ref document number: 10565669 Country of ref document: US |














