JP6819445B2 - Information processing equipment, control methods, and programs - Google Patents

Description

The present invention relates to image analysis of pathological images.

As one of the methods for diagnosing humans and animals, pathological diagnosis using pathological images is performed. A pathological image is an image obtained by photographing a tissue in the body of a human or an animal with a camera.

Then, a technique for detecting a lesion or the like by image analysis of pathological image data (hereinafter, pathological image data) has been developed. For example, Patent Document 1 discloses a technique for evaluating the degree of deformity of a gland duct based on the distribution of nuclei in the gland duct and outputting the evaluation result.

Japanese Unexamined Patent Publication No. 2010-281636

The degree of glandular dysplasia is not the only index that can be used for pathological diagnosis. The present inventor has found a new index that can be used for pathological diagnosis. One of the objects of the present invention is to provide a technique for obtaining a new index that can be used for pathological diagnosis from pathological image data.

The information processing apparatus of the present invention is composed of 1) a first layer composed of normal gland ducts and a tumor gland duct from a tissue region included in the pathological image data by image processing the pathological image data. It has a detection means for detecting the second layer, and 2) a calculation means for calculating an index value based on the thickness or area of the first layer and the second layer.

The control method of the present invention is executed by a computer. The control method is as follows: 1) By image processing the pathological image data, a first layer composed of normal gland ducts and a second layer composed of tumor gland ducts are selected from the tissue region included in the pathological image data. It has a detection step for detecting and 2) a calculation step for calculating an index value based on the thickness or area of the first layer and the second layer.

The program of the present invention causes a computer to execute each step of the control method of the present invention.

According to the present invention, there is provided a technique for obtaining a new index that can be used for pathological diagnosis from pathological image data.

It is a figure which conceptually illustrates the operation of the information processing apparatus of Embodiment 1. It is a block diagram which illustrates the functional structure of the information processing apparatus of Embodiment 1. It is a figure which illustrates the computer for realizing the information processing apparatus. It is a flowchart which illustrates the flow of the process executed by the information processing apparatus of Embodiment 1. It is a flowchart which illustrates the flow of the process of detecting the 1st layer and the 2nd layer from a tissue area. It is a figure which illustrates the effect of the expansion process. It is a figure which conceptually illustrates the pathological image data after the determination whether or not it is a duct region. It is a figure which illustrates the peripheral area of a closed area. It is a figure which conceptually illustrates the method of detecting the pixel of the duct region of a tumor duct farthest from the surface of a tissue. It is a figure which illustrates the 1st layer. It is a figure which illustrates the 2nd layer. It is a figure which illustrates the functional structure of the information processing apparatus of Embodiment 2.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In all drawings, similar components are designated by the same reference numerals, and description thereof will be omitted as appropriate. Further, in each block diagram, unless otherwise specified, each block represents a functional unit configuration rather than a hardware unit configuration.

[Embodiment 1]
FIG. 1 is a diagram conceptually illustrating the operation of the information processing apparatus 2000 of the first embodiment. Note that FIG. 1 merely shows an example of the operation of the information processing apparatus 2000 in order to facilitate understanding, and does not limit the functions of the information processing apparatus 2000 at all.

The information processing device 2000 performs image processing on the pathological image data 10. The pathological image data 10 is image data obtained by photographing a tissue in the body of a person or other animal (hereinafter referred to as a patient) to be diagnosed with a camera. More specifically, for example, pathological image data 10 is obtained by collecting tissue from the patient's body, magnifying the tissue with a microscope, and imaging the magnified tissue with a camera.

The information processing apparatus 2000 is composed of an image region (hereinafter, tissue region) representing a tissue included in the pathological image data 10, a layer composed of normal gland ducts (hereinafter, first layer), and a tumor duct. Layer (hereinafter referred to as the second layer) is detected. For example, in the pathological image data 10 of FIG. 1, the first layer 22 and the second layer 24 are detected from the tissue region 20. Here, the tumor duct is defined as, for example, "a duct in which tumor cells proliferate outside the proliferative part located in the glandular neck". Since it is difficult to distinguish whether a duct is a tumor or a non-tumor in the proliferating part, it is possible to treat a duct containing a large amount of tumor as a tumor duct instead of treating all the ducts containing a tumor as a tumor duct. Good. Normal ducts are ducts other than tumor ducts.

The information processing apparatus 2000 calculates an index value based on the detected first layer 22 and second layer 24. For example, the information processing apparatus 2000 calculates an index value based on the thickness of the first layer 22 and the second layer 24, or the detected area of the first layer 22 and the second layer 24.

As described above, according to the information processing apparatus 2000 of the present embodiment, the first layer 22 composed of normal gland ducts and the second layer 24 composed of tumor ducts are detected from the pathological image data 10. Index values based on the detected first layer 22 and second layer 24 are calculated. As described above, according to the information processing apparatus 2000 of the present embodiment, a new index that can be used for pathological diagnosis is provided.

For example, the above index value can be used as an index for quantitatively expressing how deep the tumor has invaded from the tissue surface (depth of infiltration). The depth of infiltration is an important factor in assessing the risk of metastasis / recurrence and selecting a treatment method, and it is clinically significant to express this quantitatively.

Hereinafter, the present embodiment will be described in more detail.

<Example of functional configuration>
FIG. 2 is a block diagram illustrating the functional configuration of the information processing apparatus 2000. The information processing device 2000 has a detection unit 2020 and a calculation unit 2040. The detection unit 2020 performs image processing on the pathological image data 10 to obtain a first layer 22 composed of normal gland ducts and a second layer composed of tumor ducts from the tissue region 20 included in the pathological image data 10. Layer 24 is detected. The calculation unit 2040 calculates the index value based on the detected first layer 22 and second layer 24.

<Example of hardware configuration of information processing device 2000>
Each functional component of the information processing apparatus 2000 may be realized by hardware (eg, a hard-wired electronic circuit) that realizes each functional component, or a combination of hardware and software (eg, example). It may be realized by a combination of an electronic circuit and a program that controls it). Hereinafter, a case where each functional component of the information processing apparatus 2000 is realized by a combination of hardware and software will be further described.

FIG. 3 is a diagram illustrating a calculator 1000 for realizing the information processing apparatus 2000. The calculator 1000 is any kind of calculator. For example, the computer 1000 is a personal computer (PC), a server machine, a tablet terminal, a smartphone, or the like. The computer 1000 may be a dedicated computer designed to realize the information processing device 2000, or may be a general-purpose computer.

The computer 1000 includes a bus 1020, a processor 1040, a memory 1060, a storage device 1080, an input / output interface 1100, and a network interface 1120. The bus 1020 is a data transmission line for the processor 1040, the memory 1060, the storage device 1080, the input / output interface 1100, and the network interface 1120 to transmit and receive data to and from each other. The processor 1040 is an arithmetic processing unit such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The memory 1060 is a main storage device realized by RAM (Random Access Memory) or the like. The storage device 1080 is an auxiliary storage device realized by a hard disk, an SSD (Solid State Drive), a ROM, a memory card, or the like. However, the storage device 1080 may be realized by the same hardware as the hardware used for realizing the main storage device, such as RAM.

The input / output interface 1100 is an interface for connecting the computer 1000 and the input / output device. For example, a keyboard, mouse, display device, or the like is connected to the input / output interface 1100.

The network interface 1120 is an interface for connecting to a communication network such as WAN (Wide Area Network) or LAN (Local Area Network).

The storage device 1080 stores a program module that realizes each function of the information processing device 2000. The processor 1040 reads each of these program modules into the memory 1060 and executes them to realize each function corresponding to the program module.

<Processing flow>
FIG. 4 is a flowchart illustrating a flow of processing executed by the information processing apparatus 2000 of the first embodiment. The information processing device 2000 acquires the pathological image data 10 (S102). The detection unit 2020 performs image processing on the pathological image data 10 to obtain a first layer 22 composed of normal gland ducts and a second layer composed of tumor ducts from the tissue region 20 included in the pathological image data 10. Layer 24 is detected (S104). The calculation unit 2040 calculates an index value based on the detected first layer 22 and second layer 24 (S106).

<Acquisition of pathological image data 10: S102>
The information processing apparatus 2000 acquires the pathological image data 10 to be image-analyzed (S102). The method by which the information processing apparatus 2000 acquires the pathological image data 10 is arbitrary. For example, the information processing device 2000 acquires the pathological image data 10 by accessing the storage device in which the pathological image data 10 is stored. The storage device in which the pathological image data 10 is stored may be provided inside the camera that generates the pathological image data 10, or may be provided outside the camera. Further, for example, the information processing apparatus 2000 may acquire the pathological image data 10 by receiving the pathological image data 10 transmitted from the camera.

<Detection of first layer 22 and second layer 24: S104>
The detection unit 2020 detects the first layer 22 and the second layer 24 from the tissue region 20 included in the pathological image data 10 (S104). The detection of the first layer 22 and the second layer 24 is performed, for example, in the flow shown in FIG. FIG. 5 is a flowchart illustrating a flow of processing for detecting the first layer 22 and the second layer 24 from the tissue area 20.

The detection unit 2020 extracts the tissue region 20 from the pathological image data 10 (S202). The detection unit 2020 extracts a region representing a gland duct (hereinafter, gland duct region) from the tissue region 20 (S204). The detection unit 2020 classifies each gland duct region extracted from the tissue region 20 into a duct region of a normal duct and a duct region of a tumor duct (S206). The detection unit 2020 detects the first layer 22 and the second layer 24 from the tissue region 20 based on the result of classification by S206 (S208).

Hereinafter, each step in the flowchart of FIG. 5 will be described in more detail.

<Extraction of tissue area 20: S202>
The detection unit 2020 extracts the tissue region 20 from the pathological image data 10 (S202). Here, it is assumed that the tissue included in the pathological image data 10 is stained with hematoxylin / eosin.

In the pathological image data 10, the image region other than the tissue region 20 is a portion that does not include the stained tissue and can be treated as a so-called background region. On the other hand, in the pathological image data 10, the tissue region 20 can be treated as a foreground region. Therefore, the detection unit 2020 separates the background of the pathological image data 10 and extracts the foreground region obtained as a result as the tissue region 20.

First, the detection unit 2020 performs color correction (for example, gamma correction) of the pathological image data 10 so that the variation in staining becomes small. Then, the detection unit 2020 divides the pathological image data 10 into a foreground region and a background region, and extracts the foreground region as the tissue region 20. As a concrete method of background separation, various existing methods can be used. It is not always necessary to perform color correction of the pathological image data 10.

Here, it is preferable that the detection unit 2020 performs an expansion process on the pathological image data 10 when performing background separation. FIG. 6 is a diagram illustrating the effect of the expansion treatment. In FIG. 6, when the background separation is performed without performing the expansion treatment, the recessed portion around the tissue surface is not included in the tissue region 20. On the other hand, when the background is separated after the expansion treatment, the recessed portion around the tissue surface is also included in the tissue region 20. Therefore, by performing the expansion treatment, it is possible to prevent the gland duct region existing in the recessed portion around the tissue surface from being excluded from the extraction target of the gland duct region described later. That is, the gland duct region on the tissue surface can also be extracted by the treatment described later.

<Extraction of gland duct region: S204>
The detection unit 2020 extracts the gland duct region from the tissue region 20 (S204). Therefore, the detection unit 2020 first extracts a line presumed to be the boundary line of the gland duct region (hereinafter referred to as a candidate boundary line) from the tissue region 20. The tissue region 20 includes one or more closed regions surrounded by the candidate boundaries extracted here. Therefore, the detection unit 2020 determines whether or not each of these one or more closed regions is a ductal region. Then, the detection unit 2020 extracts the closed region determined to be the gland duct region as the gland duct region.

<< Extraction of candidate boundaries >>
The detection unit 2020 performs edge detection for detecting an edge included in the tissue region 20, and extracts the detected edge as a candidate boundary line. Here, what is treated as an edge is an image region stained with hematoxylin. For example, the detection unit 2020 was stained with hematoxylin from the pathological image data 10 by converting (quadrupling) the pathological image data 10 into an image composed of four values of hematoxylin, eosin, white, and red. Get the image area. It should be noted that existing technology can be used as a specific method for edge detection.

At the time of edge detection, the detection unit 2020 may perform a process of emphasizing the edge by using an edge enhancement filter. As a result, for example, even if a part of the boundary line of the gland duct region is interrupted and captured, the interrupted portion can be connected, so that such a duct region can be used as a pathological image. It can be detected from the data 10.

The strength of edge enhancement (parameter of the edge enhancement filter) may be statically determined in advance, or may be dynamically determined by the detection unit 2020. In the latter case, the detection unit 2020 determines the strength of edge enhancement using the information obtained from the pathological image data 10. For example, the detection unit 2020 determines the strength of edge enhancement based on the area Sh of the hematoxylin-stained area and the area Se of the eosin-stained area. More specifically, the smaller the value of Sh / (Sh + Se), the stronger the strength of edge enhancement. By doing so, the pathological image data 10 is corrected so that the smaller the portion stained with hematoxylin in the pathological image data 10, the larger the portion (the edge is emphasized).

Here, it is preferable that the detection unit 2020 performs a process of removing the image region represented by the lymphocyte region (hereinafter, lymphocyte region) from the tissue region 20 before extracting the candidate boundary line. This is because lymphocytes are also stained with hematoxylin, so that the lymphocyte region may also be extracted as an edge by the edge detection process.

Various methods can be used for detecting the lymphocyte region from the pathological image data 10. For example, the detection unit 2020 detects a lymphocyte region from the pathological image data 10 by using machine learning. Specifically, a detector that detects a lymphocyte region from image data is created in advance by using teacher data that labels the lymphocyte region. Then, the detection unit 2020 removes the lymphocyte region obtained by inputting the pathological image data 10 into the detector from the pathological image data 10. It is preferable to use deep learning for the above machine learning.

<< Judgment of whether or not it is a ductal region >>
The detection unit 2020 determines whether or not each of the closed regions surrounded by the candidate boundary line is a ductal region. FIG. 7 is a diagram conceptually illustrating the pathological image data 10 after the determination as to whether or not it is a ductal region is performed. In FIG. 7, the closed region 30 is a region surrounded by a dotted line frame, and the gland duct region 32 is a region surrounded by a solid line frame. In FIG. 7, a part of the closed region 30 included in the pathological image data 10 is detected as the ductal region 32.

The determination of whether or not the closed region 30 is the ductal region 32 is realized by using, for example, machine learning. For example, by using the image data labeled with the gland duct region as teacher data, a discriminator for identifying whether or not the closed region is the gland duct region is created. The classifier is realized as, for example, a support vector machine.

The discriminator discriminates whether or not the closed region 30 is a ductal region based on the feature amount obtained from the inside or the periphery of the closed region 30. For example, as this feature amount, 1) the ratio of the area of the lymphocyte region in the closed region 30, 2) the ratio of the area of the white region in the closed region 30, and 3) the ratio of the area of the stromal region in the closed region 30. 4) Percentage of cytoplasmic area within the closed region 30, 5) Percentage of gray area in the white region within the closed region 30, 6) Percentage of lymphocyte area in the peripheral region of the closed region 30, and 7 ) Any one or more of the ratio of the area of the cell nucleus region to the peripheral region of the closed region 30 can be used. The peripheral area will be described later.

Here, the white region in the closed region 30 is an image region composed of pixels having a brightness equal to or higher than a predetermined first threshold value among the pixels included in the closed region 30. For example, the detection unit 2020 uses the above-mentioned quaternized pathological image data 10 to extract a region composed of white pixels included in the quaternized closed region 30 as a white region. The first threshold value can be any value. The first threshold value may be set in advance in the detection unit 2020, or may be stored in a storage device accessible from the detection unit 2020.

The gray area in the closed area 30 is an image area composed of pixels having a brightness equal to or higher than the first threshold value and lower than the second threshold value among the pixels included in the closed area 30. That is, the gray region is an image region composed of pixels having relatively low brightness among the pixels constituting the white region. The second threshold value can be any value equal to or higher than the first threshold value. The second threshold value may be set in advance in the detection unit 2020, or may be stored in a storage device accessible from the detection unit 2020.

The interstitial region in the closed region 30 is a region obtained as follows, for example. First, the detection unit 2020 obtains an eosin-stained image region from the pathological image data 10 by quaternizing the pathological image data 10 into an image composed of four values of hematoxylin, eosin, white, and red. .. However, if an image obtained by digitizing the pathological image data 10 has already been generated, that image can be used. Then, the detection unit 2020 sets an image region having a hue value equal to or higher than a predetermined threshold value among the image regions stained with eosin and being within the closed region 30 as an interstitial region within the closed region 30. Get as.

The threshold value may be a predetermined value or a dynamically determined value. In the latter case, for example, the detection unit 2020 determines the threshold value as follows. First, the detection unit 2020 calculates the average color of the colors of a plurality of pixels (for example, 100 pixels) in the cell nucleus region. The average color is a color composed of an average value of R values, an average value of G values, and an average value of B values of a plurality of pixels in the cell nucleus region. The detection unit 2020 calculates the Euclidean distance in the three-dimensional coordinates composed of the R value, the G value, and the B value for the color of each pixel of the pathological image data 10 and the average color. Then, the Euclidean distance is normalized so as to be in the range of 0 or more and 255 or less. The image data composed of the values obtained by normalizing the Euclidean distance calculated for each pixel in this way is called the distance image data. The detection unit 2020 identifies an image region S composed of pixels smaller than a predetermined value for the distance image data.

Further, the detection unit 2020 identifies an image region R in which the pathological image data 10 is stained with eosin and whose hue is larger than the threshold value t. The detection unit 2020 repeatedly identifies the image area R while reducing the threshold value t so that the image areas R and S overlap. Then, the threshold value t when the image regions R and S overlap or the threshold value t immediately before the overlap is set as the threshold value for obtaining the interstitial region in the closed region 30. The initial value of the threshold value t is, for example, the maximum value of hue.

Here, the image region S obtained from the distance image data represents a roughly estimated duct region. Therefore, by the above method, an eosin region having a high hue and a region that does not overlap with the roughly estimated gland duct region can be obtained as an interstitial region.

The cytoplasmic region in the closed region 30 is a portion of the region in the closed region 30 excluding the interstitial region.

The peripheral region of the closed region 30 is defined as, for example, a region having a predetermined width surrounding the closed region 30. FIG. 8 is a diagram illustrating a peripheral region of the closed region 30. In FIG. 8, the peripheral region 40 is a region having a width d surrounding the closed region 30.

The method for identifying the lymphocyte region is as described above. The cell nucleus region in the peripheral region 40 is a region in which the lymphocyte region is removed from the peripheral region 40.

<Classification of gland duct region: S206>
The detection unit 2020 classifies the extracted duct region 32 into a normal duct region and a tumor duct region (S206). Therefore, the detection unit 2020 determines whether each cell nucleus contained in the peripheral region 40 of one ductal region 32 is a cell nucleus of a normal cell or a cell nucleus of a tumor cell. Then, the detection unit 2020 sets the glandular region 32 as a normal glandular region based on the number of cell nuclei of normal cells and the number of cell nuclei of tumor cells contained in the peripheral region 40 of a certain glandular region 32. Identify which is the ductal region of the tumor.

Various methods can be used to distinguish between the cell nuclei of normal cells and the cell nuclei of tumor cells. For example, machine learning can be used for this identification. In this case, for example, by using the image data representing the cell nuclei of normal cells and the image data representing the cell nuclei of tumor cells as teacher data, a discriminator that discriminates between the cell nuclei of normal cells and the cell nuclei of tumor cells is prepared in advance. Configure it. Then, the detection unit 2020 inputs the peripheral region 40 of the gland duct region 32 to this discriminator, so that each cell nucleus contained in the peripheral region 40 is either a normal cell nucleus or a tumor cell nucleus. Obtain the identification result of whether or not there is. It is preferable to use deep learning for the above machine learning.

Here, the discriminator not only discriminates the cell nucleus into the cell nucleus of a normal cell and the cell nucleus of a tumor cell, but also 1) the cell nucleus of a lymphocyte, 2) the cell nucleus of a normal cell other than a lymphocyte, and 3). It may be identified in any of the cell nuclei of tumor cells. In this case, the discriminator can also be used to remove the lymphocyte region described above.

The method for identifying cell nuclei is not limited to the method using machine learning, and various existing methods can also be used.

As described above, the detection unit 2020 has a gland duct in which the glandular region 32 is normal, based on the number of cell nuclei of normal cells and the number of cell nuclei of tumor cells contained in the peripheral region 40 of the glandular region 32. Identify whether it is a region or a ductal region of the tumor. There are various specific methods. For example, the detection unit 2020 calculates the ratio of the number of cell nuclei of normal cells contained in the peripheral region 40 to the total number of cell nuclei contained in the peripheral region 40 of the ductal region 32. Then, the detection unit 2020 determines that the gland duct region 32 is a normal gland duct region when the calculated ratio is equal to or higher than a predetermined value. On the other hand, when the calculated ratio is less than a predetermined value, the detection unit 2020 determines that the duct region 32 is the duct region of the tumor. The number of lymphocyte cell nuclei may be excluded from the "total number of cell nuclei contained in the peripheral region 40 of the gland duct region 32".

There are various methods for determining the above-mentioned predetermined value. For example, the above predetermined value is manually set based on the experience of a doctor or the like. In addition, for example, the above-mentioned predetermined value is set by using machine learning. In the latter case, for example, image data labeled as to whether the image data of the gland duct region includes the normal duct region or the duct region of the tumor is used as the teacher data, and the image data of the duct region included in the image data is used. By counting the number of cell nuclei of normal cells detected from the peripheral region and the number of cell nuclei of tumor cells, the above-mentioned predetermined value for distinguishing between the normal duct region and the duct region of the tumor is determined. ..

<Detection of first layer 22 and second layer 24: S208>
The detection unit 2020 detects the first layer 22 and the second layer 24 from the tissue region 20 based on the result of classification by S206 (S208). As described above, the first layer 22 is an image region composed of a region of the normal gland duct. On the other hand, the second layer 24 is an image region composed of a region of the tumor gland duct.

<< Detection of first layer 22 >>
The detection unit 2020 identifies the boundary line between the region of the normal duct and the region of the tumor duct (hereinafter referred to as the first boundary line). The first boundary line is specified by, for example, the following method.

First, the detection unit 2020 determines the pixels of the duct region of the tumor duct, which is the farthest from the surface of the tissue (the boundary line between the tissue region 20 and the background region), for each row of the pixel matrix constituting the pathological image data 10. To detect. FIG. 9 is a diagram conceptually illustrating a method of detecting pixels in the ductal region of a tumor duct, which is farthest from the surface of the tissue. In FIG. 9, in the matrix of pixels constituting the pathological image data 10, three tumor ductal duct regions and one normal ductal duct region are present in the same row. Therefore, the detection unit 2020 sets the pixel at the lower end of the duct region of the tumor duct farthest from the tissue surface (third duct region from the top) to "the duct of the tumor duct farthest from the tissue surface". Detected as "pixels in the area".

The detection unit 2020 connects the pixels detected from each row by the above method to specify the first boundary line. For example, this connection is performed by linear interpolation.

Here, it is preferable that the detection unit 2020 performs a process of smoothing the generated first boundary line. For example, this smoothing can be performed by using various smoothing filters such as a Lowess filter.

The detection unit 2020 comprises an image region closed by the identified first boundary line and the outer frame of the tissue area 20 (the boundary line between the tissue area 20 and the background area), and is composed of a glandular region of the tumor gland duct. Detected as 1 layer 22. FIG. 10 is a diagram illustrating the first layer 22.

<< Detection of second layer 24 >>
In order to detect the second layer 24, the detection unit 2020 has a boundary line between the gland duct region of the normal duct and an image region other than the duct region included in the tissue region 20 (hereinafter, the second boundary line). ) Is generated. The second boundary line can be generated in the same manner as the first boundary line, for example, by the following method. Specifically, the detection unit 2020 detects the pixels in the glandular region of the normal gland duct farthest from the surface of the tissue for each row of the pixel matrix constituting the pathological image data 10. Next, the detection unit 2020 connects the pixels detected from each row to generate a second boundary line. It is preferable to smooth the second boundary line as well as the first boundary line.

The detection unit 2020 detects the image region closed by the outer frame of the first boundary line, the second boundary line, and the tissue area 20 as the second layer 24. FIG. 11 is a diagram illustrating the second layer 24.

<Calculation of index value: S104>
The calculation unit 2040 calculates an index value based on the first layer 22 and the second layer 24 detected by the detection unit 2020. Various index values can be adopted. For example, the calculation unit 2040 calculates an index value based on the area of the first layer 22 and the second layer 24 or the detected thickness of the first layer 22 and the second layer 24. In the former case, for example, the calculation unit 2040 calculates the ratio of the area of the first layer 22 to the area of the entire layer of the gland duct region (the area where the first layer 22 and the second layer 24 are combined) as an index value. For example, if the area of the first layer 22 is S1 and the area of the second layer 24 is S2, the index value is S1 / (S1 + S2). In addition, for example, the calculation unit 2040 may use the ratio of the area of the first layer 22 to the area of the second layer 24 (S1 / S2) as an index value.

On the other hand, when the index value is calculated from the thicknesses of the first layer 22 and the second layer 24, for example, the information processing apparatus 2000 uses the thickness of the first layer 22 as the index value with respect to the thickness of the entire layer of the gland duct region. calculate. For example, if the thickness of the first layer 22 is T1 and the thickness of the entire layer of the gland duct region is Ta, the index value is T1 / Ta. In addition, for example, the calculation unit 2040 may use the ratio (T1 / T2) of the thickness T1 of the first layer 22 to the thickness T2 of the second layer 24 as an index value.

Here, there are various methods for determining the thickness of the layer. For example, the calculation unit 2040 calculates the thickness of the first layer 22 for each column in the matrix of pixels constituting the pathological image data 10, and calculates the statistical value (mean value, maximum value, or maximum value) of the thickness in each column. The minimum value, etc.) is defined as the thickness of the first layer 22. The same applies to the thickness of the second layer 24 and the thickness of the entire layer of the ductal region.

The index value does not necessarily have to be expressed as a ratio or ratio. For example, the information processing apparatus 2000 may use the area or thickness of the first layer 22 and the first layer 22 as an index value.

[Embodiment 2]
FIG. 12 is a block diagram illustrating the information processing apparatus 2000 of the second embodiment. Except for the matters described below, the information processing apparatus 2000 of the second embodiment is the same as the information processing apparatus 2000 of the first embodiment.

The information processing device 2000 of the second embodiment has an output control unit 2060. Based on the index value calculated by the calculation unit 2040, the output device is made to perform a predetermined output.

The output control unit 2060 may always output the index value calculated by the calculation unit 2040, or may output the index value only when the calculated index value satisfies a predetermined condition. For example, the predetermined condition is a case where the index value is equal to or more than a predetermined threshold value, a case where the index value is equal to or less than a predetermined threshold value, and the like.

The relationship between the index value and the threshold value is determined by what kind of value the index value is calculated as. For example, when the area and thickness of the gland duct region layer (first layer 22) of the tumor duct region is used as an index value with respect to the area and thickness of the entire layer of the gland duct region, the tumor duct region 20 is the tumor duct. The index value tends to increase as the amount of is contained. Therefore, for example, when the condition that "the index value is equal to or higher than a predetermined threshold value" is satisfied, the output control unit 2060 outputs the index value.

The predetermined threshold value may be set in advance in the output control unit 2060, or may be stored in a storage device accessible from the output control unit 2060.

There are various modes in which the output control unit 2060 outputs the index value. For example, the output control unit 2060 displays an index value on the display screen of the display device. In this case, the output control unit 2060 controls the display device. In addition, for example, the output control unit 2060 outputs an index value by voice from the speaker. In this case, the output control unit 2060 controls the speaker. In addition, for example, the output control unit 2060 stores the index value in the storage device. In this case, the output control unit 2060 controls the storage device.

The index value may be output as it is as so-called raw data, or may be processed and output. In the latter case, for example, the output control unit 2060 collectively outputs a plurality of calculated index values in the form of a table or a graph. For example, assume that the calculation unit 2040 calculates an index value using each of the plurality of pathological image data 10 obtained from the same patient. In this case, the output control unit 2060 creates a table or a graph with each index value calculated using the plurality of pathological image data 10. In addition, for example, the output control unit 2060 creates a table or a graph with one or more index values calculated in the past for the same patient.

In addition, for example, the output control unit 2060 may be configured to output a predetermined message (for example, a warning message) when the index value satisfies the predetermined condition described above.

<Hardware configuration>
The hardware configuration of the computer that realizes the information processing apparatus 2000 of the second embodiment is represented by, for example, FIG. 3 as in the first embodiment. However, the storage device 1080 of the computer 1000 that realizes the information processing device 2000 of the present embodiment further stores a program module that realizes the function of the information processing device 2000 of the present embodiment. Further, various hardware (display device, speaker, etc.) controlled by the output control unit 2060 may be connected to the input / output interface 1100 of the computer 1000 that realizes the information processing device 2000 of the present embodiment.

Although the embodiments of the present invention have been described above with reference to the drawings, these are examples of the present invention, and a combination of the above embodiments or various configurations other than the above can be adopted.

Some or all of the above embodiments may also be described, but not limited to:
1. 1. A detection means for detecting a first layer composed of normal ducts and a second layer composed of tumor ducts from a tissue region included in the pathological image data by image processing the pathological image data. ,
An information processing device having a calculation means for calculating an index value based on the thickness or area of the first layer and the second layer.
2. 2. The detection means
A plurality of ductal regions are extracted from the tissue region, and it is determined whether each of the extracted ductal regions is a normal duct region or a tumor duct region.
1. The boundary between the first layer and the second layer is calculated using the position of the region of each normal duct and the location of the region of each tumor duct. The information processing device described in.
3. 3. 2. The process of determining whether the gland duct region is a normal duct region or a tumor duct region is performed by using deep learning. The information processing device described in.
4. The tissue region is an image region of a tissue stained with hematoxylin.
The detection means extracts a closed region surrounded by a hematoxylin-stained image region from the tissue region, and determines whether or not the closed region is the gland duct region based on the characteristics of the closed region. By doing so, the glandular canal region is extracted. Or 3. The information processing device described in.
5. The characteristics of the closed region are the ratio of the lymphocyte region in the closed region, the ratio of the white region in the closed region, the ratio of the gray region in the white region, the ratio of the interstitial region in the closed region, and the cytoplasm in the closed region. 4. Includes any one or more of the proportion of the region, the proportion of the lymphocyte region around the closed region, and the proportion of the cell nucleus region around the closed region. The information processing device described in.
6. The detection means removes the lymphocyte region from the tissue region and extracts the closed region from the tissue region from which the lymphocyte region has been removed. Or 5. The information processing device described in.
7. The treatment for removing the lymphocyte region is performed using deep learning. The information processing device described in.
8. The tissue region is an image region of a tissue stained with hematoxylin and eosin.
The detection means
Based on the ratio of the hematoxylin-stained image region to the eosin-stained image region, the hematoxylin-stained image region is enhanced.
3. Extract the closed region from the pathological image data after the enhancement treatment is performed. To 7. The information processing device according to any one.
9. 1. It has an output control means that outputs a predetermined value when the index value calculated by the calculation means satisfies a predetermined condition. ~ 8. The information processing device according to any one.

10. A control method performed by a computer
A detection step of detecting a first layer composed of normal ducts and a second layer composed of tumor ducts from a tissue region included in the pathological image data by image processing the pathological image data. ,
A control method comprising a calculation step of calculating an index value based on the thickness or area of the first layer and the second layer.
11. In the detection step
A plurality of ductal regions are extracted from the tissue region, and it is determined whether each of the extracted ductal regions is a normal duct region or a tumor duct region.
10. The boundary between the first layer and the second layer is calculated using the position of the region of each normal duct and the location of the region of each tumor duct. The control method described in.
12. The process of determining whether the gland duct region is a normal duct region or a tumor duct region is performed by using deep learning. The control method described in.
13. The tissue region is an image region of a tissue stained with hematoxylin.
In the detection step, a closed region surrounded by a hematoxylin-stained image region is extracted from the tissue region, and whether or not the closed region is the glandular region is determined based on the characteristics of the closed region. By doing so, the glandular region is extracted. Or 12. The control method described in.
14. The characteristics of the closed region are the ratio of the lymphocyte region in the closed region, the ratio of the white region in the closed region, the ratio of the gray region in the white region, the ratio of the interstitial region in the closed region, and the cytoplasm in the closed region. 13. Includes one or more of the proportion of regions, the proportion of lymphocyte regions around the closed region, and the proportion of cell nucleus regions around the closed region. The control method described in.
15. 13. In the detection step, the lymphocyte region is removed from the tissue region and the closed region is extracted from the tissue region from which the lymphocyte region has been removed. Or 14. The control method described in.
16. The treatment for removing the lymphocyte region is performed using deep learning. The control method described in.
17. The tissue region is an image region of a tissue stained with hematoxylin and eosin.
In the detection step
Based on the ratio of the hematoxylin-stained image region to the eosin-stained image region, the hematoxylin-stained image region is enhanced.
The closed region is extracted from the pathological image data after the enhancement treatment is performed. To 16. The control method according to any one.
18. 10. It has an output control step that outputs a predetermined output when the index value calculated in the calculation step satisfies a predetermined condition. ~ 17. The control method according to any one.

19. 10. ~ 18. A program that causes a computer to execute each step of the control method described in any one of them.

10 Pathological image data 20 Tissue area 22 1st layer 24 2nd layer 30 Closed area 32 Glandular area 40 Peripheral area 1000 Computer 1020 Bus 1040 Processor 1060 Memory 1080 Storage device 1080 Storage device 1100 Input / output interface 1120 Network interface 2000 Information processing Device 2020 Detection unit 2040 Calculation unit 2060 Output control unit

Claims (11)

A detection means for detecting a first layer composed of normal ducts and a second layer composed of tumor ducts from a tissue region included in the pathological image data by image processing the pathological image data. ,
An information processing device having a calculation means for calculating an index value based on the thickness or area of the first layer and the second layer. The detection means
A plurality of ductal regions are extracted from the tissue region, and it is determined whether each of the extracted ductal regions is a normal duct region or a tumor duct region.
The information processing apparatus according to claim 1, wherein the boundary between the first layer and the second layer is calculated by using the position of the region of each normal duct and the position of the region of each tumor duct. The information processing apparatus according to claim 2, wherein the process of determining whether the gland duct region is a normal duct region or a tumor duct region is performed by using deep learning. The tissue region is an image region of a tissue stained with hematoxylin.
The detection means extracts a closed region surrounded by an image region stained with hematoxylin from the tissue region, and determines whether or not the closed region is the glandular region based on the characteristics of the closed region. The information processing apparatus according to claim 2 or 3, wherein the glandular tube region is extracted by the method. The characteristics of the closed region are the ratio of the lymphocyte region in the closed region, the ratio of the white region in the closed region, the ratio of the gray region in the white region, the ratio of the interstitial region in the closed region, and the cytoplasm in the closed region. The information processing apparatus according to claim 4, further comprising one or more of a ratio of regions, a ratio of lymphocyte regions around the closed region, and a ratio of cell nucleus regions around the closed region. The information processing apparatus according to claim 4 or 5, wherein the detection means removes a lymphocyte region from the tissue region and extracts the closed region from the tissue region from which the lymphocyte region has been removed. The information processing apparatus according to claim 6, wherein the process of removing the lymphocyte region is performed by using deep learning. The tissue region is an image region of a tissue stained with hematoxylin and eosin.
The detection means
Based on the ratio of the hematoxylin-stained image region to the eosin-stained image region, the hematoxylin-stained image region is enhanced.
The information processing apparatus according to any one of claims 4 to 7, wherein the closed region is extracted from the pathological image data after the enhancement treatment is performed. The information processing apparatus according to any one of claims 1 to 8, further comprising an output control means that outputs a predetermined value when the index value calculated by the calculation means satisfies a predetermined condition. A control method performed by a computer
A detection step of detecting a first layer composed of normal ducts and a second layer composed of tumor ducts from a tissue region included in the pathological image data by image processing the pathological image data. ,
A control method comprising a calculation step of calculating an index value based on the thickness or area of the first layer and the second layer. A program that causes a computer to execute each step of the control method according to claim 10.

Images