Method for correcting fundus hyper-spectral image on basis of reference region, and image processing apparatus

Abstract

This method for correcting a fundus hyper-spectral image on the basis of a reference region comprises the steps of: receiving, by an image processing apparatus, an input of a fundus hyper-spectral image of a specific subject; determining, by the image processing apparatus, a reference region in the fundus hyper-spectral image; and correcting, by the image processing apparatus, a color value of the fundus hyper-spectral image on the basis of the reference region.

Description

Method for correcting a fundus hyperspectral image based on a reference area and an image processing device

[0001] The technique described below relates to a technique for standardizing hyperspectral images of the fundus.

[0002] Optical imaging technology is being utilized for various pathological assessments and surgical procedures. Meanwhile, hyperspectral imaging is a technology that generates images by collecting information from a sample's electromagnetic spectrum (measuring the reflection, emission, and absorption of electromagnetic radiation). Hyperspectral imaging constructs two-dimensional image information according to the spectral bands of electromagnetic waves into a hyperspectral cube, which is used to identify the condition, composition, characteristics, and variations of a subject.

[0003] Hyperspectral imaging is being used to diagnose ophthalmic diseases. For example, hyperspectral imaging of the fundus can be used to diagnose ophthalmic diseases based on absorption spectra or oxygen saturation in retinal blood vessels. Furthermore, recent research has shown that hyperspectral imaging of the fundus is useful in diagnosing Alzheimer's disease.

[0004] Hyperspectral imaging of the fundus collects information generated when externally irradiated light passes through the cornea and lens and is reflected from the fundus. The cornea and lens act as optical filters, with structural characteristics that vary from subject to subject. Therefore, hyperspectral images of the fundus may exhibit different characteristics for each subject, making analysis difficult.

[0005] The technique described below is intended to provide a technique for normalizing or standardizing hyperspectral images of the fundus.

[0006] A method for correcting a fundus hyperspectral image based on a reference region includes a step of an image processing device receiving a fundus hyperspectral image of a specific subject, a step of the image processing device determining a reference region in the fundus hyperspectral image, and a step of the image processing device correcting a color value of the fundus hyperspectral image based on the reference region.

[0007] An image processing device for standardizing a fundus hyperspectral image includes an input device for receiving a fundus hyperspectral image of a specific subject, a storage device for storing a program for correcting the color of the fundus hyperspectral image, and a calculation device for determining a reference area in the fundus hyperspectral image using the program, comparing the color value of the reference area with a previously established reference value, and correcting the fundus hyperspectral image.

[0008] The technology described below uniformly corrects hyperspectral images based on arterial and / or venous regions in fundus hyperspectral images to improve diagnostic accuracy for a subject.

[0009] Figure 1 is an example of a system for standardizing fundus hyperspectral images.

[0010] Figure 2 is an example of a process for standardizing fundus hyperspectral images.

[0011] Figure 3 is another example of the process of standardizing fundus hyperspectral images.

[0012] Figure 4 is an example of an application using standardized fundus hyperspectral images.

[0013] Figure 5 is an example of an image processing device that standardizes fundus hyperspectral images.

[0014] The technology described below is susceptible to various modifications and embodiments. Specific embodiments are illustrated and described in detail in the drawings. However, this does not limit the technology described below to specific embodiments, and it should be understood that all modifications, equivalents, and alternatives fall within the spirit and scope of the technology described below.

[0015] Terms such as first, second, A, and B may be used to describe various components, but these components are not limited by these terms and are used solely to distinguish one component from another. For example, without departing from the scope of the technology described below, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component. The term "and / or" includes a combination of a plurality of related described items or any of a plurality of related described items.

[0016] As used herein, the singular expressions should be understood to include the plural expressions unless the context clearly dictates otherwise, and the term "comprises" and the like should be understood to mean the presence of a described feature, number, step, operation, component, part, or combination thereof, but not to exclude the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0017] Before going into a detailed description of the drawings, it should be made clear that the division of components in this specification is merely a division based on the main function of each component. In other words, two or more components described below may be combined into a single component, or a single component may be further subdivided into two or more components with more detailed functions. In addition to its own main function, each component described below may additionally perform some or all of the functions of other components, and of course, some of the main functions of each component may be exclusively performed by other components.

[0018] Additionally, in performing a method or method of operation, each process constituting the method may occur in a different order than the stated order, unless the context clearly indicates a specific order. That is, each process may occur in the same order as the stated order, may be performed substantially simultaneously, or may be performed in the opposite order.

[0019] The technology described below relates to a technique for correcting hyperspectral images of the fundus. Hyperspectral images refer to hyperspectral images of the fundus acquired with a hyperspectral camera. Hereinafter, hyperspectral images of the fundus region are referred to as fundus hyperspectral images.

[0020] Below, we describe how an image processing device calibrates and standardizes fundus hyperspectral images. The image processing device can be implemented as a variety of devices capable of processing specific data. For example, the image processing device can be implemented as a PC, a networked server, a smart device, or a chipset embedded with a dedicated program.

[0021] Figure 1 is an example of a system (100) for standardizing fundus hyperspectral images. In Figure 1, the image processing device is illustrated as an example of a computer terminal (130) and a server (140).

[0022] The fundus hyperspectral camera (110) generates a fundus hyperspectral image of a specific subject. The fundus hyperspectral camera (110) irradiates light from outside and collects spectral information reflected from the fundus to generate a hyperspectral image. The fundus hyperspectral image is composed of a set of images acquired in various frequency bands. The fundus hyperspectral image generated by the fundus hyperspectral camera (110) can be stored in an Electronic Medical Record (EMR, 120) or a separate database (DB).

[0023] A computer terminal (130) can be used to uniformly standardize a fundus hyperspectral image. The computer terminal (130) can receive a fundus hyperspectral image from a fundus hyperspectral camera (110) or an EMR (120) via a wired or wireless network. In some cases, the computer terminal (130) may be a device physically connected to the fundus camera (110). The computer terminal (130) can color-correct the fundus hyperspectral image based on a reference region. The reference region may be at least one of an arterial region and a venous region. The process of standardizing the fundus hyperspectral image will be described later. User A can check the standardized fundus hyperspectral image on the computer terminal (130). Furthermore, User A can further analyze the standardized fundus hyperspectral image or use it as diagnostic data using the computer terminal (130).

[0024] The server (140) can receive a fundus hyperspectral image from a fundus hyperspectral camera (110) or an EMR (120). The server (140) can color-correct the fundus hyperspectral image based on a reference region. The reference region can be at least one of an arterial region and a venous region. The fundus hyperspectral image standardization process will be described later. The server (140) can transmit the standardized fundus hyperspectral image to a terminal of user A. User A can check the standardized fundus hyperspectral image. User A can use the user's terminal to further analyze the standardized fundus hyperspectral image or use it as diagnostic data.

[0025] The computer terminal (130) and / or server (140) may store the standardized fundus hyperspectral image in the EMR (120) or a separate DB.

[0026] Figure 2 is an example of a process (200) for standardizing a fundus hyperspectral image.

[0027] The image processing device receives an initial hyperspectral image of the fundus (210). The initial hyperspectral image of the fundus is an image collected from a specific subject to be analyzed.

[0028] The image processing device can preprocess the initial fundus hyperspectral image in a predetermined manner (215). The preprocessing may include adjusting the resolution or size of the initial image, extracting the eye region from the initial fundus hyperspectral image, etc.

[0029] The image processing device determines a reference area for standardizing the color of the fundus hyperspectral image (220). The reference area refers to a reference area used to correct the color or color temperature of the entire fundus hyperspectral image.

[0030] In fundus images, structures such as the optic nerve head and retina are influenced by race and age. However, blood vessels in fundus images have optical characteristics for both arterial and venous blood, independent of age and race. Therefore, reference regions can utilize arterial and / or venous regions identified in fundus hyperspectral images. Image processing devices can utilize arterial and venous regions in certain areas of fundus hyperspectral images.

[0031] The image processing device can correct the entire fundus hyperspectral image based on the selected reference area (230).

[0032] (1) The reference region may include only the arterial region. Arteries are less affected by disease than veins. Therefore, the image processing device can correct the fundus hyperspectral image based solely on the arterial region.

[0033] (2) The reference region may include an arterial region and a venous region. The image processing device may correct the fundus hyperspectral image based on the venous region as an auxiliary reference region together with the arterial region.

[0034] (3) Only the blood vessel region located in the optic disc region can be used as the reference region. It is preferable to use the blood vessel region located close to the central blood vessel as the reference region. Blood vessels close to the central blood vessel are large, making measurement easy, and are least affected by the background retinal pigment epithelium. Furthermore, if the subject has a specific disease, blood vessels may also be affected, so it is preferable to use the blood vessels located in the optic disc region before the blood vessels branch within the eye as the reference region. Therefore, (i) the image processing device can correct the fundus hyperspectral image using only the arterial region in the optic disc region as the reference region. In addition, (ii) the image processing device can correct the fundus hyperspectral image using the arterial region and the venous region in the optic disc region as the reference region.

[0035] Meanwhile, (i) the image processing device can use the artery region and vein region set (selected) by the user through the interface device as reference regions. (ii) the image processing device can also determine the reference region through a general image processing process. the image processing device can determine the reference region by using the shape and color characteristics of the artery region and vein region. The characteristics of the artery region and vein region that serve as reference regions will be described later. (iii) Alternatively, the image processing device can detect the reference region using a segmentation model. For example, the segmentation model can be a semantic segmentation model such as U-net or FCN (fully convolutional network). In this case, the image processing device can input the initial fundus hyperspectral image into the learned segmentation model to distinguish the reference region.

[0036] Image processing devices can correct fundus hyperspectral images in a variety of ways.

[0037] The image processing device compares the universal reference value of a previously collected reference region with the color value of the reference region extracted from the current fundus hyperspectral image. The universal reference value may be a color value or an average of the color values ​​of the reference region collected from a specific population. The universal reference value may also be a value for at least one of an arterial region, a venous region, an arterial region of the optic nerve head region, and a venous region of the optic nerve head region. In this case, the population may be composed of healthy normal individuals or patients with a specific disease. Alternatively, the population may be composed of a specific race. The image processing device may correct the fundus hyperspectral image of a specific subject based on the difference in color value between the universal reference value and the reference region of the specific subject.

[0038] Color differences can be computed in a variety of ways. Image processing devices can determine color differences based on the entire RGB channel, or based on specific color channels. Alternatively, image processing devices can convert RGB images to YCbCr and determine color differences in the CbCr space.

[0039] Furthermore, the image processing device can correct the entire fundus hyperspectral image based on the difference in color between the arterial region and the venous region among the reference regions.

[0040] The colors of arterial and venous regions originally have a certain degree of difference. Therefore, the arterial and venous regions will also have a certain degree of difference in the fundus hyperspectral image. The image processing device can color-correct the entire fundus hyperspectral image so that the color difference between the arterial and venous regions in the initial fundus hyperspectral image of the current subject has a certain value (reference value). Here, the reference value may be a specific value or a specific range. In other words, the image processing device can correct the color difference between the arterial and venous regions to a certain value or a certain range. The image processing device can adjust the color difference between the arterial and venous regions to a certain value by changing the values ​​of specific color channel(s). The image processing device can correct the color using a correction table prepared in advance for each color channel or using a correction function.

[0041] The color difference between arteries and veins is due to the difference in the ratio of oxygenated hemoglobin (HbO2) to deoxygenated hemoglobin (Hb). As is well known, arteries have a higher ratio of oxygenated hemoglobin than veins. Furthermore, because arteries have thicker muscle walls, while veins have thinner muscle walls, veins appear thicker and have a deeper red color than arteries.

[0042] There is a difference in arterial and venous oxygen saturation in fundus hyperspectral images. Studies have shown that there is a difference between arterial and venous oxygen saturation (Bahram Khoobehi et al, Retinal Arterial and Venous Oxygen Saturation Is Altered in Diabetic Patients, Invest Ophthalmol Vis Sci. 2013;54:7103-7106).

[0043] The study measured arterial and venous oxygen saturation in the retina in normal subjects, diabetic patients without diabetic retinopathy, patients with mild to moderate diabetic retinopathy, and patients with severe diabetic retinopathy. In brief, (i) in normal subjects, the oxygen saturation of arterial blood was 92.3 ± 4.2% and that of venous blood was 57.2 ± 6.0%, with a difference of 35.2%; (ii) in diabetic patients without diabetic retinopathy, the oxygen saturation of arterial blood was 96.3 ± 8.6% and that of venous blood was 58.7 ± 7.5%, with a difference of 37.6%; (iii) in patients with mild-moderate diabetic retinopathy, the oxygen saturation of arterial blood was 97.7 ± 5.8% and that of venous blood was 61.1% ± 7.6%, with a difference of 36.6%; and (iv) in patients with severe diabetic retinopathy, the oxygen saturation of arterial blood was 102 ± 10.2% and that of venous blood was 66.8 ± 8.4%, with a difference of 35.2%. That is, regardless of ocular disease, the difference in oxygen saturation between arterial and venous blood in the retina remains relatively constant. Furthermore, arterial oxygen saturation ranges from 92 to 103.6%, while venous oxygen saturation ranges from 57.2 to 66.8%.

[0044] Previous research findings that the arteries and veins of the retina each have a consistent oxygen saturation distribution, and that the difference in oxygen saturation between arteries and veins remains constant across subjects, lead to the conclusion that the color difference between arteries and veins in fundus hyperspectral images is also consistent. Therefore, an image processing device can correct fundus hyperspectral images based on the color differences between arterial and venous regions.

[0045] Figure 3 is another example of the process of standardizing fundus hyperspectral images.

[0046] The image processing device receives an initial hyperspectral image (310). The initial hyperspectral image is an image collected from a specific subject to be analyzed. One image has an x×y size and a wavelength of λ. The hyperspectral image is composed of multiple images generated from different spectra. Therefore, the initial hyperspectral image corresponds to a set of hyperspectral images.

[0047] The image processing device can preprocess the initial fundus hyperspectral image in a predetermined manner (315). The preprocessing may include adjusting the resolution or size of the initial image, extracting the eye region from the initial fundus hyperspectral image, etc.

[0048] The image processing device selects reference images from among a plurality of hyperspectral images generated in different spectral bands (320). The reference images are composed of hyperspectral images generated in different spectral bands (e.g., 900 nm, 600 nm). The reference images may be composed of images generated in at least two spectral bands.

[0049] The image processing device can determine a reference area from each of the selected reference images (330). The reference area is as described above in FIG. 2.

[0050] The image processing device can correct the entire fundus hyperspectral image based on the selected reference area (340).

[0051] Assume that an image processing device selects reference images of two spectral bands (600 nm and 900 nm) as reference images. (i) The image processing device determines a reference region in the 600 nm reference image. The image processing device can compare the color value of the 600 nm reference region of a specific subject with a universal reference value for the 600 nm reference region, and correct the 600 nm fundus hyperspectral image of the specific subject so that the fundus hyperspectral image of the specific subject becomes identical to the reference value (or a value close to the reference range). Furthermore, the image processing device can correct the fundus hyperspectral image of the specific subject by the amount by which the 600 nm fundus hyperspectral image is corrected, using fundus hyperspectral images of other spectral bands. (ii) The image processing device determines a reference region in the 900 nm reference image. The image processing device can compare the universal reference value for the 900 nm reference region with the color value of the 600 nm reference region of a specific subject, and correct the 900 nm fundus hyperspectral image of the specific subject so that the fundus hyperspectral image of the specific subject becomes the same as the reference value (or a value close to the reference range). Furthermore, the image processing device can correct the fundus hyperspectral image of another spectral band by the amount of the corrected value of the 900 nm fundus hyperspectral image. (iii) The image processing device can compare the reference region and the universal reference value of the corresponding spectral band for each of the selected reference images. The image processing device can calculate the difference by averaging the difference between the color value of the reference region of the specific subject and the universal reference value in each of a plurality of reference images. The image processing device can correct all fundus hyperspectral images of the specific subject based on the difference. (iv) The image processing device can calculate the color value difference of the arterial region and the venous region, which are reference regions, for each of the selected reference images. The image processing device can calculate the average value of the difference between the arterial and venous regions in each of multiple reference images. The image processing device can then correct all hyperspectral fundus images so that the difference remains constant based on the average value of the difference.

[0052] Figure 4 illustrates an example of an application utilizing standardized fundus hyperspectral images. Figure 4 illustrates an example of building a learning model or artificial intelligence (AI) model that analyzes standardized fundus hyperspectral images. The AI ​​model may be a model that analyzes a patient's fundus hyperspectral images to predict or diagnose ocular lesions.

[0053] Figure 4(A) illustrates an example of a process for training an AI model. An image processing device receives initial hyperspectral images generated by a fundus examination device. The image processing device may be a separate learning device. The image processing device may receive the initial hyperspectral images from a separate database. The image processing device standardizes the fundus hyperspectral images of a population through the aforementioned process. The image processing device may store the standardized hyperspectral images in a learning database. The image processing device trains an AI model using the standardized hyperspectral images. The image processing device inputs the standardized hyperspectral images into the AI ​​model, and updates the parameters of the AI ​​model by comparing the values ​​output by the AI ​​model with previously known label values, repeating the process. Through this, the image processing device establishes an AI model that predicts ocular lesions.

[0054] Figure 4(B) illustrates an example of a process for predicting ocular lesions using a trained AI model. The image processing device receives an initial hyperspectral image of the fundus captured by a hyperspectral camera. The image processing device standardizes the initial hyperspectral image. The image processing device inputs the standardized hyperspectral image into the AI ​​model trained through Figure 4(A). The image processing device uses the AI ​​model to make inferences about ocular lesions. The image processing device can derive analysis results on the presence or absence of an ocular lesion based on the value (probability value) output to the AI ​​model.

[0055] Figure 5 illustrates an example of an image processing device (400) that standardizes fundus hyperspectral images. The image processing device (400) corresponds to the aforementioned image processing devices (130 and 140 of Figure 1). The image processing device (400) may be physically implemented in various forms. For example, the image processing device (400) may take the form of a computer device such as a PC, a network server, a data processing-dedicated chipset, etc.

[0056] The image processing device (400) may include a storage device (410), a memory (420), a computing device (430), an interface device (440), a communication device (450), and an output device (460).

[0057] The storage device (410) can store the initial fundus hyperspectral image generated by the fundus hyperspectral camera.

[0058] The storage device (410) can store a code or program that standardizes the initial fundus hyperspectral image.

[0059] The storage device (410) can store reference information for image correction (reference values ​​of reference areas by spectrum band, reference values ​​for color differences between arteries and veins by spectrum band, etc.).

[0060] The storage device (410) can store a program or learning model (segmentation model) that distinguishes a reference area for standardizing a fundus hyperspectral image.

[0061] The storage device (410) can store standardized fundus hyperspectral images.

[0062] The memory (420) can store data and information generated during the process of standardizing (correcting) the fundus hyperspectral image by the image processing device (400).

[0063] The interface device (440) is a device that receives certain commands and data from an external source. The interface device (440) can receive an initial hyperspectral image of a specific subject from a physically connected input device or an external storage device. The interface device (440) can receive setting information for a reference area from the initial hyperspectral image. The interface device (440) can also transmit a standardized hyperspectral image of the fundus to an external entity.

[0064] The communication device (450) refers to a configuration that receives and transmits certain information via a wired or wireless network. The communication device (450) can receive an initial hyperspectral image of a specific subject from an external entity. The communication device (450) can receive setting information for a reference region from the initial hyperspectral image. The communication device (450) can also transmit a standardized hyperspectral image of the fundus to an external entity, such as a user terminal.

[0065] The interface device (440) may be a device that transmits data transmitted from a communication device (450) into the image processing device (400).

[0066] The output device (460) is a device that outputs certain information. The output device (460) can output interfaces required for data processing, initial fundus hyperspectral images, standardized fundus hyperspectral images, diagnosis results, etc.

[0067] The computing device (430) can perform the process described below using code, program, or learning model stored in the storage device (410).

[0068] The computational device (430) can preprocess the initial hyperspectral image in a predetermined manner. The preprocessing may be any one of the following operations: resolution or size change, cropping of a predetermined area, noise removal, etc.

[0069] The computing device (430) can determine a reference region that serves as a reference for initial fundus hyperspectral image correction. The computing device (430) can determine the reference region based on information input by the user. Furthermore, the computing device (430) can automatically determine the reference region based on an image processing program. Furthermore, the computing device (430) can also distinguish the reference region using a learned segmentation model. The reference region is as described in FIG. 2 .

[0070] The computing device (430) can correct the hyperspectral image of a specific subject by comparing the color value of the reference region with a reference value (reference value) collected in advance from the population data. This process is as described in FIG. 2. For example, the computing device (430) can correct the hyperspectral image of a specific subject based on the difference between the color value of the reference region of the specific subject and the color value (reference value) of the reference region of the hyperspectral image of the fundus of the population constructed in advance.

[0071] The computing device (430) can collect reference images in different spectral bands, and compare the reference region of each reference image with a previously collected reference value (reference value) to correct the hyperspectral image of a specific subject. This process is as described in FIG. 3. The computing device can select a plurality of hyperspectral images in different spectral bands from a hyperspectral image set, calculate the difference between the color value of the reference region for each of the plurality of hyperspectral images and the color value of the reference region of the hyperspectral image of a pre-constructed population, and correct the hyperspectral image set of the fundus.

[0072] Furthermore, the computing device (430) may input a standardized fundus hyperspectral image into a pre-trained AI model to perform inference regarding ocular lesions.

[0073] The computing device (430) may be a device such as a processor, AP, or chip embedded with a program that processes data and performs certain operations.

[0074] In addition, the fundus hyperspectral image processing method or the fundus hyperspectral image standardization method described above may be implemented as a program (or application) including an executable algorithm that can be executed on a computer. The program may be provided by being stored on a temporary or non-transitory computer-readable medium.

[0075] A non-transitory readable medium refers to a medium that stores data semi-permanently and can be read by a device, rather than a medium that stores data for a short period of time, such as a register, cache, or memory. Specifically, the various applications or programs described above may be stored and provided on a non-transitory readable medium, such as a CD, DVD, hard disk, Blu-ray disk, USB, memory card, ROM (read-only memory), PROM (programmable read only memory), EPROM (Erasable PROM, EPROM), EEPROM (Electrically EPROM), or flash memory.

[0076] Temporarily readable media refers to various types of RAM, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Synchronous DRAM (Synclink DRAM, SLDRAM), and Direct Rambus RAM (DRRAM).

[0077] The present embodiment and the drawings attached to the present specification only clearly illustrate a part of the technical idea included in the above-described technology, and it will be obvious that all modified examples and specific embodiments that can be easily inferred by a person skilled in the art within the scope of the technical idea included in the specification and drawings of the above-described technology are included in the scope of the rights of the above-described technology.

Claims

1. A step in which an image processing device receives a hyperspectral image of a specific subject's fundus; The step of the image processing device determining a reference area in the fundus hyperspectral image; and A method for correcting a fundus hyperspectral image based on a reference region, the method comprising a step of correcting the color value of the fundus hyperspectral image based on the reference region by the image processing device.

2. In paragraph 1, A method for correcting a fundus hyperspectral image based on a reference region, wherein the reference region is at least one of (i) an arterial region and (ii) an arterial region and a venous region.

3. In paragraph 2, The above reference area is a method for correcting a fundus hyperspectral image based on a reference area which is a vascular area located on the optic nerve head.

4. In paragraph 1, A method for correcting a fundus hyperspectral image based on a reference region in which the image processing device corrects the fundus hyperspectral image of the specific subject based on the difference between the color value of the reference region and the color value of the reference region of the fundus hyperspectral image of the pre-constructed population.

5. A step in which an image processing device receives a set of hyperspectral fundus images of a specific subject; A step in which the image processing device selects a plurality of hyperspectral images in different spectral bands from the fundus hyperspectral image set; A step in which the image processing device determines a reference area in each of the plurality of hyperspectral images; and A method for correcting a fundus hyperspectral image based on a reference region, the method comprising a step of correcting the color value of the fundus hyperspectral image based on the reference region of each of the plurality of hyperspectral images by the image processing device.

6. In paragraph 5, A method for correcting a fundus hyperspectral image based on a reference region, wherein the reference region is at least one of (i) an arterial region and (ii) an arterial region and a venous region located on the optic nerve head.

7. In paragraph 5, A method for correcting a fundus hyperspectral image based on a reference region in which the image processing device calculates the difference between the color value of the reference region and the color value of the reference region of the fundus hyperspectral image of a pre-constructed population for each of the plurality of hyperspectral images, and corrects the fundus hyperspectral image based on the average value of the difference values ​​of the plurality of hyperspectral images.

8. An input device for receiving a hyperspectral image of a specific subject; A storage device storing a program for color correction of a fundus hyperspectral image; and An image processing device for standardizing a fundus hyperspectral image, comprising a calculation device for determining a reference area in the fundus hyperspectral image using the above program and correcting the fundus hyperspectral image by comparing the color value of the reference area with a previously established reference value.

9. In paragraph 8, An image processing device for standardizing a fundus hyperspectral image, wherein the reference area is at least one of (i) an arterial area and (ii) an arterial area and a venous area located on the optic nerve head.

10. In paragraph 8, The above-mentioned calculation device is an image processing device that standardizes the fundus hyperspectral image by correcting the fundus hyperspectral image of the specific subject based on the difference between the color value of the reference area and the color value of the reference area of ​​the fundus hyperspectral image of the pre-constructed population.

11. In paragraph 8, The above hyperspectral image comprises multiple hyperspectral images generated in different spectral bands, The above-mentioned calculation device calculates the difference between the color value of the reference area and the color value of the reference area of ​​the fundus hyperspectral image of the pre-constructed population for each of the plurality of hyperspectral images, and standardizes the fundus hyperspectral image by correcting the fundus hyperspectral image based on the average value of the difference values ​​of the plurality of hyperspectral images.

Images