Estimator training device, estimator training method, and estimator training program

Abstract

To estimate whether a human skin lesion is in a predetermined state.SOLUTION: An estimator training device comprises: an image acquisition unit that acquires an image of a human skin lesion captured in multiple bands; a color information acquisition unit that calculates absorbance in a region corresponding to the skin lesion in the image; and an estimator training unit that trains an estimator for estimating whether the skin lesion is in a predetermined state from the absorbance, based on the relationship between the absorbance and information indicating whether the skin lesion is in the predetermined state.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to an estimator learning device, an estimator learning method, and an estimator learning program. [Background technology]

[0002] Diagnosis of skin removed by surgery or other procedures is made based on the specialized knowledge of dermatologists and the diagnosis of pathologists using pathological specimens. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2021 / 015249 [Patent Document 2] Patent Publication No. 2021-4890 Summary of the Invention [Problem to be solved by the invention]

[0004] Some skin lesions are difficult to diagnose as benign (benign tumors) or malignant (malignant tumors). In response to this, the application of multispectral cameras to pathology specimens is being considered as a new method of providing diagnostic information. Examples of benign tumors include moles (cellular nevi), atheromas, lipomas, dermoid cysts, calcifying epithelioma, foreign body granulomas, parotid gland tumors, submandibular gland tumors, and accessory breast tumors. Examples of malignant tumors include basal cell carcinoma, squamous cell carcinoma, Bowen's disease, Paget's disease, extramammary Paget's disease, malignant melanoma, and soft tissue malignant tumors.

[0005] A dermatologist removes a skin lesion from a person's skin, and by photographing the lesion with a multispectral camera, images of the skin lesion in multiple bands can be obtained. It is desirable to estimate the benignity or malignancy of the skin lesion using the images in multiple bands.

[0006] An object of the present invention is to provide a technique capable of estimating whether or not the skin is in a specified condition. [Means for solving the problem]

[0007] In order to solve the above problems, the following measures are adopted.

[0008] That is, the first aspect is an image acquisition unit for acquiring images of a human skin lesion using multiple bands; a color information acquisition unit that calculates absorbance in the region of the skin lesion of the image; an estimator learning unit configured to learn an estimator that estimates whether the skin lesion is in a predetermined state from the absorbance based on a relationship between the absorbance and information on whether the skin lesion is in a predetermined state; This is an estimator learning device.

[0009] The disclosed aspects may be realized by a program being executed by an information processing device. That is, the disclosed configuration may be specified as a program for causing an information processing device to execute the processes executed by each of the means in the above-mentioned aspects, or a computer-readable recording medium on which the program is recorded. The disclosed configuration may also be specified by a method in which an information processing device executes the processes executed by each of the above-mentioned means. The disclosed configuration may also be specified as a system including an information processing device that executes the processes executed by each of the above-mentioned means. Effect of the Invention

[0010] According to the present invention, it is possible to provide a technique capable of estimating whether or not the skin in an image is in a predetermined condition. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 is a diagram illustrating an example of a configuration of an estimator learning system according to an embodiment. [Diagram 2] FIG. 2 is a diagram illustrating an example of the configuration of a benign / malignant presumption system according to the embodiment. [Diagram 3] FIG. 3 is a diagram illustrating an example of a hardware configuration of an information processing device. [Figure 4] FIG. 4 is a diagram illustrating an example of an operational flow when the estimator learning device learns a cancer estimator. [Diagram 5] FIG. 5 is a diagram showing an example of an operational flow when the inference device infers benign / malignant. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, an embodiment will be described with reference to the drawings. The configurations of the embodiments are merely examples, and the configurations of the invention are not limited to the specific configurations of the disclosed embodiments. In implementing the invention, a specific configuration according to the embodiment may be appropriately adopted.

[0013] [Embodiment] (Configuration example) 1 is a diagram showing an example of the configuration of an estimator learning system of this embodiment. The estimator learning system 10 of this embodiment includes an estimator learning device 100 and a multispectral camera 200. The estimator learning device 100 includes an image acquisition unit 102, a color information acquisition unit 104, a pigment amount calculation unit 106, an absorbance estimation unit 108, an estimator learning unit 110, and a storage unit 112. The estimator learning device 100 and the multispectral camera 200 are connected to each other directly or via a network or the like so as to be able to communicate with each other.

[0014] The estimator learning device 100 acquires an image of a sample for each band captured by a multispectral camera 200. The estimator learning device 100 calculates color information (absorbance of each band) from the image of the sample, and calculates the amount of pigment from the obtained color information. The estimator learning device 100 estimates absorbance from the spectral absorption coefficient and the calculated amount of pigment. The estimator learning device 100 generates an estimator based on the absorbance based on the image, the estimated amount of pigment, and information on whether the skin lesion is benign or malignant. The sample is, for example, a biological material sample such as a skin lesion. Programs, data, tables, and the like used by the estimator learning device 100 are stored in a memory unit 112 (storage means) of the estimator learning device 100.

[0015] The image acquisition unit 102 acquires an image of the sample from the multispectral camera 200 .

[0016] The color information acquisition unit 104 calculates color information (the absorbance of each band) from the image of the sample.

[0017] The dye amount calculation unit 106 calculates the dye amount from the color information of the acquired multiband image.

[0018] The absorbance estimation unit 108 estimates the absorbance based on the spectral absorption coefficient and the amount of dye calculated by the dye amount calculation unit 106 .

[0019] The storage unit 112 stores programs, data, tables, and the like used by the estimator learning device 100 .

[0020] The estimator learning unit 110 calculates the absorbance (color information) and absorbance estimates for each band of the skin lesion in the image. The estimator is trained based on the absorbance estimated by the unit 108 and information on whether the skin lesion is benign or malignant. The estimator training unit 110 stores the trained estimator in a storage means. Whether the skin lesion is benign or malignant is an example of whether the skin lesion is in a predetermined state. An example of whether the skin lesion is in a predetermined state is whether the skin lesion is a disease or not.

[0021] The multispectral camera 200 is a camera capable of capturing an image of an object in each of a plurality of divided bands (frequency bands, wavelength bands). The multispectral camera 200 captures an image of a sample (skin lesion) in each band. The frequency (wavelength) of the object captured by the multispectral camera 200 is, for example, from the infrared region to the ultraviolet region. The multispectral camera 200 can capture images in, for example, 10 or more wavelength bands. The multispectral camera 200 can capture images in more wavelength bands than an RGB camera.

[0022] Fig. 2 is a diagram showing an example of the configuration of a benign / malignant estimation system of this embodiment. The benign / malignant estimation system 20 of this embodiment includes a multispectral camera 200 and an estimation device 300. The estimation device 300 includes an image acquisition unit 302, a color information acquisition unit 304, a pigment amount calculation unit 306, an absorbance estimation unit 308, an estimation unit 310, an estimator 312, and a storage unit 314. The multispectral camera 200 is similar to the multispectral camera 200 in Fig. 1. The estimation device 300 and the multispectral camera 200 are connected to each other so as to be able to communicate with each other directly or via a network or the like.

[0023] The estimation device 300 acquires an image of a sample (skin lesion) for each band from the multispectral camera 200. The estimation device 300 calculates color information (absorbance of each band) from the image, and calculates the amount of pigment from the obtained color information. The estimation device 300 estimates absorbance from the spectral absorption coefficient and the estimated amount of pigment. The estimation device 300 estimates whether the skin lesion in the image is benign or malignant using the absorbance based on the image, the estimated amount of pigment, and an estimator trained by the estimator training device 100 of FIG. 1. The sample is, for example, a biological material such as a skin lesion excised from human skin. Programs, data, tables, and the like used by the estimation device 300 are stored in a memory unit 314 (storage means) of the estimation device 300.

[0024] The image acquisition unit 302, color information acquisition unit 304, pigment amount calculation unit 306, and absorbance estimation unit 308 are similar to the image acquisition unit 102, color information acquisition unit 104, pigment amount calculation unit 106, and absorbance estimation unit 108 in FIG.

[0025] The estimation unit 310 estimates whether a skin lesion in an image is benign or malignant based on the absorbance of the image, using an estimator 312 trained by the estimator training device 100 of FIG.

[0026] The estimator 312 is an estimator constructed by the estimator learning device 100 of FIG.

[0027] The memory unit 314 stores programs, data, tables, and the like used by the estimation device 300 .

[0028] The estimator learning device 100 that learns the estimator and the estimation device 300 that performs benign / malicious estimation may be integrated to operate as a single device. That is, a single device connected to the multispectral camera 200 may have the functions of the estimator learning device 100 in FIG. 1 and the functions of the estimation device 300 in FIG.

[0029] FIG. 3 is a diagram illustrating an example of the hardware configuration of an information processing device. The information processing device 90 illustrated in FIG. 3 has the configuration of a general computer. The estimator learning device 100, the estimation device 300, and a device in which the estimator learning device 100 and the estimation device 300 are integrated are each configured as shown in FIG. This is realized by using an information processing device 90. The information processing device 90 in Fig. 3 has a processor 91, a memory 92, a storage unit 93, an input unit 94, an output unit 95, and a communication control unit 96. These are connected to each other by a bus. The memory 92 and the storage unit 93 are computer-readable recording media. The hardware configuration of the information processing device is not limited to the example shown in Fig. 3, and components may be omitted, replaced, or added as appropriate.

[0030] The information processing device 90 has a processor 91 which loads a program stored in a recording medium into a working area of ​​a memory 92 and executes it, and each component, etc. is controlled through the execution of the program, thereby realizing functions consistent with a specified purpose.

[0031] The processor 91 is, for example, a CPU (Central Processing Unit) or a DSP (Digital Signal Processor).

[0032] The memory 92 includes, for example, a random access memory (RAM) and a read only memory (ROM). The memory 92 is also called a main storage device.

[0033] The storage unit 93 is, for example, an EPROM (Erasable Programmable ROM) or a hard disk. The storage unit 93 is a removable media. The removable media may include a USB (Universal Serial Bus) memory or a disk recording medium such as a CD (Compact Disc) or a DVD (Digital Versatile Disc). The storage unit 93 is also called a secondary storage device.

[0034] The storage unit 93 stores various programs, various data, and various tables used in the information processing device 90 in a readable and writable recording medium. An operating system (OS), various programs, various tables, etc. are stored in the storage unit 93. The information stored in the storage unit 93 may be stored in the memory 92. Moreover, the information stored in the memory 92 may be stored in the storage unit 93.

[0035] The operating system is software that mediates between software and hardware, manages memory space, manages files, and manages processes and tasks. The operating system includes a communication interface. The communication interface is a program that exchanges data with other external devices connected via the communication control unit 96. The external devices include, for example, other information processing devices, external storage devices, etc.

[0036] The input unit 94 includes a keyboard, a pointing device, a wireless remote control, a touch panel, etc. The input unit 94 can also include an input device for video or images such as a camera, and an input device for audio such as a microphone.

[0037] The output unit 95 is a liquid crystal display (LCD), an electroluminescence (EL) panel, The output unit 95 may include a display device such as a display panel, a CRT (Cathode Ray Tube) display, a PDP (Plasma Display Panel), etc., and an output device such as a printer, etc. The output unit 95 may also include an audio output device such as a speaker.

[0038] The communication control unit 96 connects to other devices and controls communication between the information processing device 90 and the other devices. The communication control unit 96 is, for example, a LAN (Local Area Network) interface board, a wireless communication circuit for wireless communication, or a communication circuit for wired communication. The LAN interface board and the wireless communication circuit are connected to a network such as the Internet.

[0039] The information processing device 90 is a device in which a processor executes a program stored in an auxiliary storage unit and a program to be executed in a main storage unit. The information processing device controls peripheral devices and the like through the execution of the program. This allows the information processing device to realize functions that meet a predetermined purpose. The main memory unit and the auxiliary memory unit are recording media that can be read by the information processing device.

[0040] (Example of operation) Estimator Learning FIG. 4 is a diagram illustrating an example of an operational flow when the estimator learning device learns a cancer estimator.

[0041] In S101, the image acquisition unit 102 acquires an image of a sample (skin lesion) captured by the multispectral camera 200 from the multispectral camera 200. The image acquisition unit 102 stores the acquired image in a storage unit. The sample is, for example, a biological material such as a skin lesion excised from a person for treatment or the like. Skin lesions have various colors. The colored parts and colors of the skin lesion depend on the substances present in the skin lesion. The multispectral camera 200 captures images in multiple bands (frequency bands). Examples of substances present in the skin lesion include melanin, oxyhemoglobin, deoxyhemoglobin, β-carotene, and bilirubin. Melanin, oxyhemoglobin, deoxyhemoglobin, β-carotene, and bilirubin are examples of substances that may be present in the skin lesion. Light that is not absorbed or polarized by these substances enters the multispectral camera 200 and is captured. By using the multispectral camera 200, more detailed information on the transmitted light that has passed through the sample can be obtained. It is known whether a substance contained in an image captured by the multispectral camera 200 is benign or malignant. Each image is stored in a storage unit in association with whether it is benign or malignant.

[0042] In S102, the color information acquisition unit 104 obtains the absorbance of each pixel (position) in the skin lesion in the image of each band from the extracted skin lesion area of ​​the image of each band. It is assumed that the skin lesion area in the image is specified in advance. The pixels in the skin lesion used here do not need to be all pixels included in the skin lesion area, but may be a predetermined number of pixels. Absorbance is an amount that indicates the degree to which the intensity of light is weakened when light passes through a medium. The absorbance of each pixel depends on the brightness of each pixel. The color information acquisition unit 104 obtains the absorbance of each pixel from the brightness of each pixel in the image for each band. The absorbance of each band depends on the amount of pigment of the substance included in the sample and the spectral absorption coefficient of the substance, etc.

[0043] In S103, the pigment amount calculation unit 106 calculates the pigment amount of the substance contained in the sample for each pixel based on the absorbance calculated in S102 and the known spectral absorption coefficients of the substances that may be present in the skin lesion. Here, the substances that may be present in the skin lesion are melanin, oxyhemoglobin, deoxyhemoglobin, β-carotene, and bilirubin. The spectral absorption coefficient is an amount that indicates how much light a medium absorbs when light is incident on the medium. The spectral absorption coefficient of a substance depends on the wavelength of the incident light. If the absorbance vector is a, the spectral absorption coefficient matrix is ​​X, the pigment amount is C, and the difference vector is e, the following is expressed according to Lambert-Beer's law. The absorbance vector a is a vector representation of the absorbance for each band. The spectral absorption coefficient matrix X is a matrix representation of the absorption coefficient for each substance and band contained in the sample. The spectral absorption coefficient matrix X is stored in advance in a storage means. The difference vector e is a vector (residual component) that cannot express the absorbance vector a using a combination of the spectral absorption coefficient matrix X and the dye amount C. The dye amount calculation unit 106 calculates the estimated dye amount C′ that minimizes the difference vector e by least squares estimation.

number

number

[0044] In S104, the absorbance estimation unit 108 can obtain an estimated absorbance vector a' by multiplying the absorption spectroscopic coefficient matrix X by the estimated dye amount C' obtained in S103 as follows: The absorbance vector a' is an absorbance vector (estimated absorbance) estimated by the substance's spectral absorption coefficient matrix X and the estimated dye amount C' of the substance present in the sample.

number

[0045]

number

[0046] Furthermore, the estimator learning device 100 can learn estimators for various substances by learning the estimator using images of skin lesions containing various substances. The estimator learning device 100 may also learn the estimator using images of skin lesions captured by a camera other than the multispectral camera 200, such as an RGB camera. The estimator learning device 100 learns the estimator using images captured by the multispectral camera 200 and images captured by an RGB camera, etc., so that the estimator is learned with higher accuracy.

[0047] <Presumption of benign or malignant> 5 is a diagram showing an example of an operation flow when the estimation device estimates whether a skin lesion is benign or malignant, in which the estimation device 300 estimates whether a skin lesion is benign or malignant by using an image of the skin lesion to be estimated and the above-mentioned estimator.

[0048] In S201, the image acquisition unit 302 acquires an image of a sample (skin lesion) captured by the multispectral camera 200 from the multispectral camera 200. The image acquisition unit 302 stores the acquired image in a storage means. The sample is a skin lesion excised from a person who is to be presumed to be benign or malignant. The colored portion and color of the skin lesion depend on the substance present in the skin lesion. The multispectral camera 200 captures images in a plurality of bands (frequency bands). The substance present in the skin lesion is, for example, melanin, oxyhemoglobin, deoxyhemoglobin, β-carotene, and bilirubin. By using the multispectral camera 200, more detailed information of the transmitted light that has passed through the sample can be obtained.

[0049] S202 to S204 are the same as S102 to S104 in FIG.

[0050] In S205, the estimation unit 310 estimates whether the lesion is benign or malignant using the estimator 312 with the absorbance vector a of the skin lesion in the image calculated in S202 and the absorbance vector a' (difference vector e) of the skin lesion in the image estimated in S204 as input data. Here, the estimation unit 310 estimates whether the lesion is benign or malignant by using the estimator 312 trained by the estimator learning device 100. The estimation unit 310 stores the estimation result of whether the lesion is benign or malignant in a storage means such as a storage device, or outputs it to a display means such as a display. The items of the input data (absorbance vector, difference vector, etc.) are the same as the items of the input data when the estimator 312 was trained.

[0051] In this way, the estimation device 300 can perform benign / malignant estimation using an image including a skin lesion that is to be the target of benign / malignant estimation and the estimator 312 trained by the estimator training device 100.

[0052] (others) Here, melanin, oxyhemoglobin, deoxyhemoglobin, β-carotene, and bilirubin are listed as substances that may be present in skin lesions. The estimated absorbance vector a' is calculated based on the spectral absorption coefficients of the substances that may be present in the skin lesions listed here. Therefore, the estimated absorbance vector a' is considered to lack information on substances other than the substances that may be present in the skin lesions here. On the other hand, the difference vector e is considered to include information on substances other than the substances that may be present in the skin lesions here. When substances other than the substances that may be present in the skin lesions here are involved in the benignity or malignancy of the skin lesion, a more accurate estimation can be made by using the difference vector e.

[0053] Here, the performance of the estimation device 300 (cancer estimator) trained by the estimator learning device 100 is shown using images of slices of non-cancerous tissue (benign tumor) of human skin and images of slices of cancerous tissue (malignant tumor) of human skin. The range of a specified lesion in each tissue slice is determined in advance by a pathologist as being benign or malignant. For comparison, the performance of a trained cancer estimator is also shown using images taken by an RGB camera instead of the multispectral camera 200, similarly to the estimator learning device 100. Here, the estimation accuracy is shown by the accuracy of the estimated pixel area. That is, the estimation accuracy indicates the proportion of the area of ​​pixels designated by a pathologist as benign or malignant that are correctly determined as benign or malignant by machine learning. With the RGB camera, an image of a band including red, an image of a band including green, and an image of a band including blue are obtained. When images taken by the RGB camera were used, the estimation accuracy of the cancer estimator was 63.7%.

[0054] In the estimator training device trained by combining estimated dye amounts from images from the multispectral camera 200 and images from the RGB camera, the highest estimation accuracy of 69.25% was obtained when two types of dye amounts and green and blue images from the RGB camera were used. By using images from the multispectral camera 200, the performance of the cancer estimator was generally improved compared to when only images from the RGB camera were used.

[0055] Furthermore, when the estimator training device 100 was trained by combining an image from the RGB camera with a wavelength of 588 nm of the difference vector e obtained from the image from the multispectral camera 200, the estimation accuracy of the cancer estimator for the area ratio was 74.21%. By using a combination of the normalized RGB camera image and a specific wavelength of the difference vector e, it is possible to significantly improve the performance of the estimator compared to an estimator using only the RGB camera image and an estimator using the RGB camera image and the estimated dye amount.

[0056] (Actions and Effects of the Embodiments) The estimator learning device 100 acquires an image including a skin lesion captured by a multispectral camera 200 for each band. By using the multispectral camera 200, images of many bands can be obtained. The estimator learning device 100 calculates the absorbance of each pixel (position) of the skin lesion in the image from the region of the skin lesion in each band image. The estimator learning device 100 calculates the pigment amount of a substance contained in the skin lesion of the sample for each pixel based on the absorbance and the known spectral absorption coefficient of the substance contained in the skin lesion. The estimator learning device 100 also estimates an absorbance vector based on the calculated pigment amount and a spectral absorption coefficient matrix X of the substance. The estimator learning device 100 calculates a difference vector, which is the difference between the absorbance vector obtained from the image and the estimated absorbance vector. The estimator learning device 100 analyzes the relationship between the absorbance vector a of the skin lesion in the image, the difference vector e which is the difference between the absorbance vector a′ of the estimated skin lesion in the image and the absorbance vector, and information on whether the skin lesion contained in the image is benign or malignant, by machine learning or the like, and learns an estimator. The estimator can estimate whether the skin lesion is benign or malignant based on the absorbance vector of the skin lesion in the image or the like. The estimator learning device 100 can build an estimator with higher performance by using images of skin lesions in more bands. The estimator learning device 100 can calculate the difference vector e based on the Lambert-Beer law. The estimator learning device 100 can build an estimator with higher performance by learning the estimator using the difference vector e.

[0057] The estimation device 300 estimates whether a skin lesion in an image is benign or malignant based on an image including a skin lesion captured for each band by the multispectral camera 200 and an estimator 312 trained by the estimator training device 100.

[0058] Computer-readable recording medium A program for causing a computer or other machine or device (hereinafter, referred to as a computer, etc.) to realize any of the above functions can be recorded on a recording medium readable by the computer, etc. Then, the computer, etc. can provide the function by reading and executing the program from the recording medium.

[0059] Here, a computer-readable recording medium refers to a recording medium that stores information such as data and programs by electrical, magnetic, optical, mechanical, or chemical action and can be read by a computer, etc. Such a recording medium may be provided with elements that constitute a computer, such as a CPU and memory, and the CPU may be made to execute the program.

[0060] Among such recording media, those that are removable from a computer etc. include, for example, flexible disks, magneto-optical disks, CD-ROMs, CD-R / Ws, DVDs, DATs, 8mm tapes, memory cards and the like.

[0061] Furthermore, examples of recording media fixed to a computer include a hard disk and a ROM.

[0062] (others) Although the embodiments of the present invention have been described above, these are merely examples, and the present invention is not limited to these. Various modifications based on the knowledge of those skilled in the art, such as combinations of each component, are possible without departing from the spirit of the claims. [Explanation of symbols]

[0063] 10: System 20: System 100: Estimator learning device 102: Image acquisition unit 104: Color information acquisition section 106: Pigment amount calculation unit 108: Absorbance estimation section 110: Estimator learning unit 112: Storage section 200: Multispectral camera 300: Estimation device 302: Image acquisition unit 304: Color information acquisition unit 306: Pigment amount calculation unit 308: Absorbance estimation section 310: Estimation section 312: Estimator 314: Storage section 90: Information processing device 91: Processor 92: Memory 93: Storage section 94: Input section 95: Output section 96: Communication control section

Claims

1. an image acquisition unit for acquiring images of a human skin lesion using multiple bands; a color information acquisition unit that calculates absorbance in the region of the skin lesion of the image; an estimator learning unit configured to learn an estimator that estimates whether the skin lesion is in a predetermined state from the absorbance based on a relationship between the absorbance and information on whether the skin lesion is in a predetermined state; Estimator learner.

2. a pigment amount calculation unit that estimates a pigment amount of the substance contained in the skin lesion area based on the absorbance of the skin lesion area calculated by the color information acquisition unit and a spectral absorption coefficient of the substance that may be contained in the skin lesion; an absorbance estimation unit that estimates an estimated absorbance, which is an estimated absorbance in the region of the skin lesion, based on the pigment amount and the spectral absorption coefficient; the estimator learning unit learns the estimator that estimates whether the skin lesion is in a predetermined state from the absorbance and the difference, based on a relationship between the absorbance and the difference between the absorbance and the estimated absorbance, and information on whether the skin lesion is in a predetermined state. The estimator learning device according to claim 1 .

3. The absorbance and the difference are each normalized. The estimator learning device according to claim 2 .

4. Whether the skin lesion is in the predetermined state or not is whether the skin lesion is benign or malignant. The estimator learning device according to claim 1 .

5. An estimator learning device, Images of human skin lesions were taken using multiple bands. calculating the absorbance in the region of the skin lesion of the image; learning an estimator that estimates whether the skin lesion is in a predetermined state from the absorbance based on a relationship between the absorbance and information on whether the skin lesion is in a predetermined state; The estimator training method performs the above.

6. The estimator learning device Images of human skin lesions were taken using multiple bands. calculating the absorbance in the region of the skin lesion of the image; learning an estimator that estimates whether the skin lesion is in a predetermined state from the absorbance based on a relationship between the absorbance and information on whether the skin lesion is in a predetermined state; An estimator learning program that executes the above.

Images