JP2021100572A - Ophthalmologic image processing apparatus, oct apparatus, ophthalmologic image processing program, and mathematical model construction method - Google Patents

Abstract

SOLUTION: A control part of an ophthalmologic processing apparatus acquires an ophthalmologic image or an addition average image photographed by an ophthalmologic imaging apparatus as a base image. The control part acquires an objective image with quality higher than that of the base image by inputting the base image in a mathematical model trained by a machine learning algorithm. The mathematical model is trained by training data for inputting and training data for outputting. Data based on L pieces of training ophthalmologic images out of a plurality of the training ophthalmologic images photographing a same site of tissues is used for the training data for inputting. Data of the addition average image obtained by performing addition averaging on H (H>L) pieces of the training ophthalmologic images is used for the training data for outputting.EFFECT: An ophthalmologic image of high quality is appropriately acquired.SELECTED DRAWING: Figure 3

Description

The present disclosure discloses an ophthalmic image processing device for processing an ophthalmic image of an eye to be inspected, an OCT device, an ophthalmic image processing program executed in the ophthalmic image processing device, and a mathematical model construction for constructing a mathematical model used by the ophthalmic image processing device. Regarding the method.

Conventionally, various techniques for acquiring high-quality images have been proposed. For example, the ophthalmologic imaging apparatus described in Patent Document 1 acquires a high-quality tomographic image by adding and averaging a plurality of tomographic images and suppressing speckle noise.

Japanese Unexamined Patent Publication No. 2015-9108

In the technique of adding and averaging a plurality of images, it is necessary to repeatedly shoot a large number of the same parts so as to improve the accuracy of suppressing the influence of noise. Therefore, there is a problem that the shooting time increases in proportion to the number of images to be added.

A typical object of the present disclosure is to provide an ophthalmic image processing apparatus, an OCT apparatus, an ophthalmic image processing program, and a method for constructing a mathematical model for appropriately acquiring high-quality ophthalmic images.

The ophthalmic image processing apparatus provided by the typical embodiment in the present disclosure is an ophthalmic image processing apparatus that processes an ophthalmic image which is an image of a tissue of an eye to be inspected, and a control unit of the ophthalmic image processing apparatus is an ophthalmic image. An ophthalmic image taken by the imaging device or an added average image obtained by adding and averaging a plurality of ophthalmic images obtained by photographing the same part of the tissue by the ophthalmic imaging device was acquired as a base image and trained by a machine learning algorithm. By inputting the basic image into the mathematical model, a target image having a higher image quality than the basic image is acquired, and the mathematical model is L (L) out of a plurality of training ophthalmic images obtained by photographing the same part of the tissue. The data based on the training ophthalmic image of L ≧ 1) is used as the input training data, and the data of the added average image obtained by adding and averaging the H sheets (H> L) of the training ophthalmic images is used as the output training data. Has been done.

The OCT apparatus provided by a typical embodiment in the present disclosure captures an ophthalmic image of the tissue by processing an OCT signal by the reference light and the reflected light of the measurement light applied to the tissue of the eye to be inspected. In the OCT device, the control unit of the OCT device acquires, as a base image, an ophthalmic image of the tissue taken or an added average image obtained by adding and averaging a plurality of ophthalmic images of the same part of the tissue. By inputting the basic image into the mathematical model trained by the machine learning algorithm, a target image having a higher image quality than the basic image is acquired, and the mathematical model is used for a plurality of trainings in which the same part of the tissue is photographed. Of the ophthalmic images, L (L ≧ 1) data based on the training ophthalmic image is used as input training data, and H (H> L) additional average images obtained by adding and averaging the training ophthalmic images. The data is trained as output training data.

The ophthalmic image processing program provided by the typical embodiment in the present disclosure is an ophthalmic image processing program executed by an ophthalmic image processing apparatus that processes an ophthalmic image which is an image of a tissue of an eye to be inspected, and is the ophthalmic image processing. When the program is executed by the control unit of the ophthalmic image processing device, the ophthalmic images taken by the ophthalmic image taking device or a plurality of ophthalmic images obtained by taking the same part of the tissue by the ophthalmic image taking device are added and averaged. The purpose of acquiring a target image having a higher image quality than the base image by inputting the base image into a mathematical model trained by a machine learning algorithm and a base image acquisition step of acquiring the added average image as a base image. The image acquisition step is executed by the ophthalmic image processing apparatus, and the mathematical model is used for L (L ≧ 1) of the training ophthalmic images among a plurality of training ophthalmic images obtained by photographing the same part of the tissue. The data based on the training data is used as input training data, and the data of the added average image obtained by adding and averaging the H images (H> L) of the training ophthalmic images is used as the output training data.

The mathematical model construction method provided by the typical embodiment in the present disclosure is constructed by being trained by a machine learning algorithm, and a mathematical model construction that constructs a mathematical model that outputs a target image by inputting a base image. It is a mathematical model construction method executed by the device, and is an image set acquisition step of acquiring a plurality of training ophthalmic images obtained by photographing the same part of the tissue of the eye to be inspected, and L images out of the plurality of training ophthalmic images. The data based on the training ophthalmic image of (L ≧ 1) is used as the input training data, and the data of the added average image obtained by adding and averaging the H sheets (H> L) of the training ophthalmic images is used as the output training data. It includes a training step of constructing the mathematical model by training the mathematical model.

According to the ophthalmic image processing apparatus, the OCT apparatus, the ophthalmic image processing program, and the mathematical model construction method according to the present disclosure, high-quality ophthalmic images are appropriately acquired.

The control unit of the ophthalmic image processing device exemplified in the present disclosure is an ophthalmic image taken by the ophthalmic image taking device or an added average image obtained by adding and averaging a plurality of ophthalmic images obtained by taking the same part of the tissue by the ophthalmic image taking device. Is acquired as a base image. The control unit acquires a target image of higher quality than the base image by inputting the base image into the mathematical model trained by the machine learning algorithm. The mathematical model is trained by a training dataset, which is a set of training data for input and training data for output. Of a plurality of training ophthalmic images obtained by photographing the same part of the tissue, data based on L (L ≧ 1) training ophthalmic images is used as input training data. Further, among a plurality of training ophthalmic images, the data of the added average image obtained by adding and averaging H (H> L) training ophthalmic images is regarded as output training data (sometimes referred to as correct answer data). To. In other words, H is a number greater than the number of training ophthalmic images used for input training data.

That is, the mathematical model building apparatus that builds the mathematical model executes the image set acquisition step and the training step. In the image set acquisition step, a plurality of training ophthalmic images obtained by photographing the same part of the tissue of the eye to be inspected are acquired. In the training step, among a plurality of training ophthalmic images, data based on L (L ≧ 1) training ophthalmic images is used as input training data, and H (H> L) training ophthalmic images are used. The mathematical model is constructed by training the mathematical model using the data of the averaging image of the above as the training data for output.

In this case, the constructed mathematical model is as good as an image obtained by adding and averaging a large number of ophthalmic images even when a small number of ophthalmic images or an added average image of a small number of ophthalmic images is input as a base image. It is possible to output a target image in which the influence of noise is suppressed. Therefore, it is easy to shorten the shooting time when shooting the base image. Further, even in the conventional image processing, it is possible to suppress the influence of noise without performing the addition averaging processing. However, in the conventional image processing, it is not possible to distinguish between a weak signal and noise, and there is a case where a weak signal that is originally necessary is erased. On the other hand, the mathematical model illustrated in the present disclosure can output a target image based on a base image in which a portion where the signal of the image is weakened due to an influence other than noise is not erased. Therefore, high quality ophthalmic images are properly acquired.

The data used as the input training data of the training data set (that is, "data based on L training ophthalmic images") is the data of the captured ophthalmic images themselves (that is, the original images that are not added and averaged). It may be the data of the added average image obtained by adding and averaging a plurality of ophthalmic images. Both the original image data and the averaging image data may be used as input training data. When the training data for input is the data of the original image, it is desirable that the base image input to the mathematical model is also the original image not added and averaged. Further, when the input training data is the data of the averaging image, it is desirable that the base image is also the averaging image. Further, the number of images (at least one of the original image and the averaging image) used for the input training data may be one, or may be a plurality of images (for example, about 2 to 5). You may. Further, the number H of training ophthalmic images used when generating the averaging image used as the output training data is at least larger than the number of training ophthalmic images used for the input training data. For example, when L original images are used as input training data, H> L. Further, when an image obtained by adding and averaging L'images is used as the input training data, H> L'. The number of ophthalmic images used for the base image is less than the number of H images.

The ophthalmologic image may be a two-dimensional tomographic image of the tissue of the eye to be inspected taken by an OCT device. The control unit of the ophthalmic image processing apparatus may acquire a plurality of basic images obtained by photographing each of a plurality of positions different from each other in the tissue of the eye to be inspected. The control unit may acquire each target image at a plurality of positions by inputting each of the plurality of basic images into the mathematical model. The control unit generates at least one of a three-dimensional tomographic image of the tissue and a two-dimensional front image when the tissue is viewed from the front (that is, the direction of the line of sight of the eye to be inspected) based on the plurality of target images. May be good. That is, the ophthalmic imaging apparatus has a three-dimensional tomographic image generation step of generating a three-dimensional tomographic image of the tissue based on a plurality of target images, and a front surface that generates a two-dimensional frontal image of the tissue based on the plurality of target images. At least one of the image generation steps may be performed. In this case, a three-dimensional tomographic image or a two-dimensional frontal image of the tissue in which the influence of noise is suppressed is appropriately generated without taking a large number of two-dimensional tomographic images.

The two-dimensional front image may be a so-called "Enface image". The Enface image data includes, for example, integrated image data in which brightness values are integrated in the depth direction (Z direction) at each position in the XY direction, integrated value of spectral data at each position in the XY direction, and a certain depth. It may be brightness data at each position in the XY direction in the direction, brightness data at each position in the XY direction in any layer of the retina (for example, the surface layer of the retinal), and the like.

The ophthalmologic image may be a two-dimensional tomographic image of the fundus of the eye to be inspected taken by an OCT device. The control unit of the ophthalmologic image processing apparatus may identify at least one of a plurality of layers included in the fundus by performing analysis processing on a target image which is a two-dimensional tomographic image. That is, the ophthalmologic image processing apparatus may perform a segmentation step of identifying the layer of the fundus from the acquired target image. In this case, the segmentation result in which the influence of noise is suppressed can be appropriately acquired without taking a large number of two-dimensional tomographic images.

It is also possible to use an image other than the two-dimensional tomographic image as the ophthalmic image. For example, the ophthalmologic image may not be a two-dimensional tomographic image, but may be a two-dimensional frontal image of a tissue (for example, the fundus) taken from the front (that is, the line-of-sight direction of the eye to be inspected). In this case, the two-dimensional front image may be an OCT angio image. The OCT angio image may be, for example, a motion contrast image acquired by processing at least two OCT signals acquired at different times with respect to the same position. In addition, an image obtained by photographing the autofluorescence (FAF) of the fundus of the eye to be inspected by a laser scanning eye examination device (SLO) (hereinafter, referred to as “FAF image”) may be used as an ophthalmic image. Similar to the OCT two-dimensional tomographic image, the OCT angio image and the FAF image are also appropriately suppressed from the influence of noise by performing the addition averaging processing of a plurality of images. Further, the imaging site of the ophthalmologic image may be, for example, the fundus of the eye to be inspected or the anterior segment of the eye. As described above, as the ophthalmic image in the present disclosure, various images capable of suppressing the influence of noise by the addition averaging process of a plurality of images can be used.

Further, the control unit of the ophthalmic image processing apparatus may execute an analysis process different from the segmentation process for identifying the layer on the target image. For example, the control unit acquires a frontal image of the fundus showing blood vessels (for example, the OCT angio image described above) as a target image, and executes a process of analyzing the blood vessel density of the fundus with respect to the acquired target image. You may. In addition, the control unit may perform analysis processing on an image generated based on a plurality of target images (for example, the above-mentioned three-dimensional tomographic image or Enface image).

The control unit of the ophthalmologic image processing apparatus may execute at least one of the simultaneous display processing and the switching display processing. In the simultaneous display process, the control unit causes the display device to display the base image and the target image at the same time. In the switching display process, the control unit switches between the base image and the target image and displays them on the display device in response to the instruction input by the user. In this case, the user can easily confirm both the target image in which the influence of noise is suppressed and the base image (original image or the averaging image of the original image) on which the target image is based. Therefore, the user can perform diagnosis and the like more appropriately. The base image and the target image may be displayed on the same display device or may be displayed on different display devices.

The control unit may acquire a plurality of basic images continuously captured by the ophthalmologic image capturing apparatus. The control unit may acquire a plurality of target images by sequentially inputting the acquired plurality of basic images into the mathematical model. The control unit may generate moving image data of the tissue based on the acquired plurality of target images. That is, the ophthalmic image processing apparatus may execute a moving image generation step of generating moving image data of the tissue based on a plurality of target images. In this case, the user can check the moving image of the tissue displayed in a state where the influence of noise is suppressed. That is, even with the conventional technique, it is possible to suppress the influence of noise by adding and averaging a plurality of ophthalmic images. However, in the conventional technique, since one averaging image is created from a plurality of ophthalmic images, the frame rate tends to decrease remarkably when displaying a moving image of a tissue using the averaging image. In particular, when the shooting position changes with the passage of time, it is difficult to generate a good moving image using the averaging image. On the other hand, the ophthalmic image processing apparatus of the present disclosure can generate a moving image of a tissue in which the influence of noise is suppressed while suppressing a decrease in the frame rate. Therefore, the user can perform diagnosis and the like more appropriately.

The control unit may simultaneously display the moving image of the base image and the moving image of the target image on the display device. Further, the control unit may switch between the moving image of the base image and the moving image of the target image and display them on the display device in response to the instruction input by the user. In this case, the user can easily confirm both the moving image of the target image and the moving image of the base image.

It is a block diagram which shows the schematic structure of the mathematical model construction apparatus 1, the ophthalmologic image processing apparatus 21, and the ophthalmology imaging apparatus 11A, 11B. It is a flowchart of the mathematical model construction process executed by the mathematical model construction apparatus 1. It is explanatory drawing for demonstrating the 1st acquisition pattern of input training data and output training data. It is explanatory drawing for demonstrating the 2nd acquisition pattern of input training data and output training data. It is explanatory drawing for demonstrating the 3rd acquisition pattern of input training data and output training data. It is explanatory drawing for demonstrating the 4th acquisition pattern of input training data and output training data. It is a flowchart of high-quality moving image generation processing executed by ophthalmic image processing apparatus 21. It is a flowchart of the high-quality three-dimensional image generation processing executed by an ophthalmologic image processing apparatus 21. It is a figure which shows an example of a high-quality three-dimensional image 40.

(Device configuration)
Hereinafter, one of the typical embodiments in the present disclosure will be described with reference to the drawings. As shown in FIG. 1, in this embodiment, the mathematical model construction device 1, the ophthalmic image processing device 21, and the ophthalmic imaging devices 11A and 11B are used. The mathematical model construction device 1 constructs a mathematical model by training the mathematical model by a machine learning algorithm. The constructed mathematical model outputs a target image having a higher image quality than the base image (in the present embodiment, the influence of noise is appropriately suppressed) based on the input base image. The ophthalmic image processing apparatus 21 acquires a target image from the base image by using a mathematical model. The ophthalmic imaging devices 11A and 11B capture an ophthalmic image which is an image of the tissue of the eye to be inspected.

As an example, a personal computer (hereinafter referred to as "PC") is used for the mathematical model building apparatus 1 of the present embodiment. Although the details will be described later, the mathematical model building apparatus 1 constructs a mathematical model by training the mathematical model using the ophthalmic image acquired from the ophthalmic imaging apparatus 11A. However, the device that can function as the mathematical model construction device 1 is not limited to the PC. For example, the ophthalmologic imaging device 11A may function as the mathematical model building device 1. Further, the control units of the plurality of devices (for example, the CPU of the PC and the CPU 13A of the ophthalmologic imaging apparatus 11A) may collaborate to construct a mathematical model.

Further, a PC is used for the ophthalmic image processing device 21 of the present embodiment. However, the device that can function as the ophthalmic image processing device 21 is not limited to the PC. For example, the ophthalmic imaging device 11B or a server may function as the ophthalmic image processing device 21. When the ophthalmologic image capturing device (OCT device in the present embodiment) 11B functions as the ophthalmologic image processing device 21, the ophthalmologic image capturing device 11B captures the ophthalmic image while producing an ophthalmic image having a higher image quality than the captured ophthalmic image. Can be obtained properly. Further, a mobile terminal such as a tablet terminal or a smartphone may function as the ophthalmic image processing device 21. The control units of the plurality of devices (for example, the CPU of the PC and the CPU 13B of the ophthalmologic image capturing device 11B) may cooperate to perform various processes.

Further, in the present embodiment, a case where a CPU is used as an example of a controller that performs various processes will be illustrated. However, it goes without saying that a controller other than the CPU may be used for at least a part of various devices. For example, by adopting a GPU as a controller, the processing speed may be increased.

The mathematical model construction device 1 will be described. The mathematical model construction device 1 is arranged, for example, in an ophthalmic image processing device 21 or a manufacturer that provides an ophthalmic image processing program to a user. The mathematical model building apparatus 1 includes a control unit 2 that performs various control processes and a communication I / F5. The control unit 2 includes a CPU 3 which is a controller that controls control, and a storage device 4 that can store programs, data, and the like. The storage device 4 stores a mathematical model construction program for executing a mathematical model construction process (see FIG. 2) described later. Further, the communication I / F5 connects the mathematical model building apparatus 1 to other devices (for example, an ophthalmic imaging apparatus 11A and an ophthalmic image processing apparatus 21).

The mathematical model construction device 1 is connected to the operation unit 7 and the display device 8. The operation unit 7 is operated by the user in order for the user to input various instructions to the mathematical model construction device 1. For the operation unit 7, for example, at least one of a keyboard, a mouse, a touch panel, and the like can be used. A microphone or the like for inputting various instructions may be used together with the operation unit 7 or instead of the operation unit 7. The display device 8 displays various images. As the display device 8, various devices capable of displaying an image (for example, at least one of a monitor, a display, a projector, and the like) can be used. The "image" in the present disclosure includes both a still image and a moving image.

The mathematical model building apparatus 1 can acquire data of an ophthalmic image (hereinafter, may be simply referred to as an “ophthalmic image”) from the ophthalmic imaging apparatus 11A. The mathematical model building apparatus 1 may acquire ophthalmic image data from the ophthalmic imaging apparatus 11A by, for example, at least one of wired communication, wireless communication, a detachable storage medium (for example, a USB memory), and the like.

The ophthalmic image processing apparatus 21 will be described. The ophthalmologic image processing device 21 is arranged, for example, in a facility (for example, a hospital or a health examination facility) for diagnosing or examining a subject. The ophthalmic image processing device 21 includes a control unit 22 that performs various control processes and a communication I / F 25. The control unit 22 includes a CPU 23, which is a controller that controls control, and a storage device 24 that can store programs, data, and the like. The storage device 24 is provided with an ophthalmic image processing program for executing ophthalmic image processing described later (for example, high-quality moving image generation processing illustrated in FIG. 7, high-quality three-dimensional image generation processing shown in FIG. 8 and the like). It is remembered. The ophthalmic image processing program includes a program that realizes a mathematical model constructed by the mathematical model construction apparatus 1. The communication I / F 25 connects the ophthalmic image processing device 21 to other devices (for example, the ophthalmic imaging device 11B and the mathematical model building device 1).

The ophthalmic image processing device 21 is connected to the operation unit 27 and the display device 28. As the operation unit 27 and the display device 28, various devices can be used in the same manner as the operation unit 7 and the display device 8 described above.

The ophthalmic image processing device 21 can acquire an ophthalmic image from the ophthalmic image capturing device 11B. The ophthalmic image processing device 21 may acquire an ophthalmic image from the ophthalmic image capturing device 11B by, for example, at least one of wired communication, wireless communication, a detachable storage medium (for example, a USB memory), and the like. Further, the ophthalmic image processing device 21 may acquire a program or the like for realizing the mathematical model constructed by the mathematical model building device 1 via communication or the like.

The ophthalmic imaging devices 11A and 11B will be described. As an example, in the present embodiment, a case where an ophthalmic image capturing device 11A for providing an ophthalmic image to the mathematical model building apparatus 1 and an ophthalmologic imaging device 11B for providing an ophthalmic image to the ophthalmic image processing device 21 will be described. .. However, the number of ophthalmic imaging devices used is not limited to two. For example, the mathematical model construction device 1 and the ophthalmic image processing device 21 may acquire ophthalmic images from a plurality of ophthalmic imaging devices. Further, the mathematical model construction device 1 and the ophthalmology image processing device 21 may acquire an ophthalmology image from one common ophthalmology image capturing device. The two ophthalmologic imaging devices 11A and 11B illustrated in this embodiment have the same configuration. Therefore, in the following, the two ophthalmologic imaging devices 11A and 11B will be collectively described.

Further, in the present embodiment, the OCT apparatus is exemplified as the ophthalmologic imaging apparatus 11 (11A, 11B). However, an ophthalmologic imaging apparatus other than the OCT apparatus (for example, a laser scanning optometry apparatus (SLO), a fundus camera, a corneal endothelial cell imaging apparatus (CEM), etc.) may be used.

The ophthalmic imaging apparatus 11 (11A, 11B) includes a control unit 12 (12A, 12B) that performs various control processes, and an ophthalmic imaging unit 16 (16A, 16B). The control unit 12 includes a CPU 13 (13A, 13B) which is a controller that controls control, and a storage device 14 (14A, 14B) capable of storing programs, data, and the like.

The ophthalmic imaging unit 16 includes various configurations necessary for capturing an ophthalmic image of the eye to be inspected. The ophthalmic imaging unit 16 of the present embodiment is provided with an OCT light source, a branched optical element that branches OCT light emitted from the OCT light source into measurement light and reference light, a scanning unit for scanning the measurement light, and measurement light. It includes an optical system for irradiating the eye examination, a light receiving element that receives the combined light of the light reflected by the tissue of the eye to be inspected and the reference light, and the like.

The ophthalmologic imaging apparatus 11 can capture a two-dimensional tomographic image and a three-dimensional tomographic image of the fundus of the eye to be inspected. Specifically, the CPU 13 captures a two-dimensional tomographic image of a cross section intersecting the scan line by scanning the OCT light (measurement light) on the scan line. Further, the CPU 13 can take a three-dimensional tomographic image of the tissue by scanning the OCT light two-dimensionally. For example, the CPU 13 acquires a plurality of two-dimensional tomographic images by scanning measurement light on each of a plurality of scan lines having different positions in a two-dimensional region when the tissue is viewed from the front. Next, the CPU 13 acquires a three-dimensional tomographic image by combining a plurality of captured two-dimensional tomographic images.

Further, the CPU 13 captures a plurality of ophthalmic images of the same site by scanning the measurement light a plurality of times on the same site on the tissue (on the same scan line in the present embodiment). The CPU 13 can acquire an averaging image in which the influence of speckle noise is suppressed by performing an averaging process on a plurality of ophthalmic images of the same portion. The addition averaging process may be performed, for example, by averaging the pixel values of the pixels at the same position in a plurality of ophthalmic images. The larger the number of images to be subjected to the averaging process, the easier it is to suppress the influence of speckle noise, but the longer the shooting time. The ophthalmologic image capturing device 11 executes a tracking process for tracking the scanning position of the OCT light with the movement of the eye to be inspected while taking a plurality of ophthalmic images of the same site.

(Mathematical model construction process)
The mathematical model construction process executed by the mathematical model construction apparatus 1 will be described with reference to FIGS. 2 to 6. The mathematical model construction process is executed by the CPU 3 according to the mathematical model construction program stored in the storage device 4.

In the mathematical model construction process, the mathematical model is trained by the training data set, so that a mathematical model for outputting a high-quality target image from the base image is constructed. Although details will be described later, the training data set includes training data for input and training data for output. In this embodiment, as input training data, data based on one to a plurality of ophthalmic images (in this embodiment, a two-dimensional tomographic image of the fundus) is used. Further, as the output training data, the data of the addition average image based on the number of ophthalmic images (H) larger than the number of ophthalmic images (L) used for the input training data is used. The input training data and the output training data target the same part of the tissue.

As shown in FIG. 2, the CPU 3 acquires a set 30 (see FIGS. 3 to 6) of a plurality of training ophthalmic images 300A to 300X (X is an arbitrary integer of 2 or more) in which the same part of the tissue is photographed. (S1). Specifically, in the present embodiment, a set of a plurality of two-dimensional tomographic images taken by scanning OCT light a plurality of times on the same scan line on the fundus is used as a set of training ophthalmic images for ophthalmology. Obtained from the image capturing device 11A. However, the CPU 3 acquires a signal (for example, an OCT signal) that is a basis for generating a two-dimensional tomographic image from the ophthalmic imaging apparatus 11A, and generates a two-dimensional tomographic image based on the acquired signal to generate a two-dimensional tomographic image. Images may be acquired.

Next, the CPU 3 acquires input training data from a part of the plurality of training ophthalmic images 300A to 300X in the set 30 (S2). A specific acquisition method of the training data for input can be appropriately selected, and the details thereof will be described later with reference to FIGS. 3 to 6.

Next, the CPU 3 acquires the averaging images 31 (see FIGS. 3 to 6) of the plurality of training ophthalmic images 300A to 300X in the set 30 as output training data (S3). As described above, the number H of the training ophthalmic images used for the additional average image 31 which is the output training data is larger than the number L of the training ophthalmic images used for the input training data. As an example, in the present embodiment, as shown in FIGS. 3 to 6, all of the plurality of training ophthalmic images 300A to 300X (for example, 120 images) in the set 30 are used for the averaging image 31. The addition average image 31 may be generated by the mathematical model building apparatus 1. Further, the mathematical model building apparatus 1 may acquire the averaging image 31 generated by the ophthalmologic imaging apparatus 11A.

The first acquisition pattern of the training data for input will be described with reference to FIG. In the first acquisition pattern, the CPU 3 is one of the plurality of training ophthalmic images 300A to 300X included in the set 30 of the training ophthalmic image 300A itself (that is, the original image not subjected to addition averaging processing). The data is acquired as training data for input. In this case, when the ophthalmic image processing apparatus 21 inputs a base image to the constructed mathematical model to acquire a target image, it is preferable that one original image is input to the mathematical model as the base image. Since the ophthalmic image processing device 21 can acquire the target image based on one base image, the shooting time of the base image can be easily shortened. In addition, when generating a moving image of a target image, it is easy to increase the frame rate of the moving image.

The second acquisition pattern of the training data for input will be described with reference to FIG. In the second acquisition pattern, the CPU 3 acquires the data of the original images of the plurality of training ophthalmic images among the plurality of training ophthalmic images 300A to 300X included in the set 30 as the input training data. In the example shown in FIG. 4, the data of the two training ophthalmic images 300A and 300B are acquired as the training data for input. However, the number of training ophthalmic images acquired as input training data can be changed within a range smaller than the number of training ophthalmic images used for the addition average image 31. When the second acquisition pattern is adopted, the ophthalmic image processing apparatus 21 has an number of images as close as possible to the number of training ophthalmic images used as input training data when inputting a base image into the constructed mathematical model. It is preferable to input the original image as a base image into the mathematical model. In this case, the quality of the acquired target image is likely to be improved as compared with the case where one original image is input to the mathematical model as the base image. However, even when the second acquisition pattern is adopted, one original image may be input to the mathematical model as the base image. As described above, when the original image is used as the input training data, the number L of the original images used is 1 or more and smaller than the number used for the averaging image 31.

The third acquisition pattern of the training data for input will be described with reference to FIG. In the third acquisition pattern, the CPU 3 obtains the data of the added average image 32 obtained by adding and averaging L (L ≧ 2) training ophthalmic images out of the plurality of training ophthalmic images 300A to 300X included in the set 30. , Acquire as training data for input. In the example shown in FIG. 5, the data of the averaging image 32 of the two training ophthalmic images 300A and 300B are acquired as the training data for input. However, the number of training ophthalmic images used for the averaging image 32 can be changed within a range smaller than the number of training ophthalmic images used for the averaging image 31. When the third acquisition pattern is adopted, the ophthalmic image processing apparatus 21 preferably inputs the averaging image as the base image when inputting the base image into the constructed mathematical model. The number of ophthalmic images used for the averaging of the base images is preferably as close as possible to the number of training ophthalmic images used for the averaging image 32. In this case, the quality of the acquired target image is likely to be improved as compared with the case where one original image is input to the mathematical model as the base image. However, even when the third acquisition pattern is adopted, one original image may be input to the mathematical model as the base image.

The fourth acquisition pattern of the training data for input will be described with reference to FIG. In the fourth acquisition pattern, the CPU 3 acquires a plurality of added average images 32 obtained by adding and averaging L (L ≧ 2) training ophthalmic images, and inputs the data of the acquired plurality of added average images 32 for input training. Use as data. In the example shown in FIG. 6, the data of the two averaging images 32A and 32B are acquired as input training data. However, it is also possible to change the number of additional average images 32. When the fourth acquisition pattern is adopted, the ophthalmic image processing apparatus 21 preferably inputs a plurality of averaging images as the base image when inputting the base image into the constructed mathematical model. In this case, the quality of the acquired target image is likely to be improved as compared with the case where one original image is input as the base image. However, even when the fourth acquisition pattern is adopted, one original image may be input to the mathematical model as a base image.

The four patterns described above are examples, and the above patterns can be changed. For example, the CPU 3 may use both the data of the original image of the P training ophthalmic images and the data of the added average image 32 of the P'training ophthalmic images as the input training data. In this case, the number of training ophthalmic images used for the added average image 31 which is the output training data is based on the number of training ophthalmic images used for the input training data (for example, the sum of P and P'). There are also many.

It is also possible to acquire a plurality of sets of training data sets from a set 30 of one set of training ophthalmic images 300A to 300X. For example, in the example shown in FIG. 3, a set of training data sets is acquired in which the data of the training ophthalmic image 300A is used as the input training data and the data of the addition average image 31 is used as the output training data. Here, the CPU 3 can also acquire a training data set from the same set 30 in which the data of the training ophthalmic image 300B is used as the input training data and the data of the addition average image 31 is used as the output training data. ..

Returning to the description of FIG. The CPU 3 uses a machine learning algorithm to train a mathematical model using a training data set (S4). As the machine learning algorithm, for example, a neural network, a random forest, a boosting, a support vector machine (SVM), and the like are generally known.

Neural networks are a technique that mimics the behavior of biological nerve cell networks. Neural networks include, for example, feed-forward (forward propagation) neural networks, RBF networks (radial basis functions), spiking neural networks, convolutional neural networks, recursive neural networks (recurrent neural networks, feedback neural networks, etc.), and probabilities. Neural networks (Boltzmann machines, Basian networks, etc.).

Random forest is a method of learning based on randomly sampled training data to generate a large number of decision trees. When using a random forest, the branches of a plurality of decision trees learned in advance as a discriminator are traced, and the average (or majority vote) of the results obtained from each decision tree is taken.

Boosting is a method of generating a strong classifier by combining a plurality of weak classifiers. A strong classifier is constructed by sequentially learning simple and weak classifiers.

SVM is a method of constructing a two-class pattern classifier using a linear input element. The SVM learns the parameters of the linear input element based on, for example, the criterion (hyperplane separation theorem) of obtaining the margin maximizing hyperplane that maximizes the distance from each data point from the training data.

A mathematical model refers, for example, to a data structure for predicting the relationship between input data and output data. Mathematical models are constructed by training with training datasets. The training data set is a set of training data for input and training data for output. The mathematical model is trained so that when certain input training data is input, the corresponding output training data is output. For example, training updates the correlation data (eg, weights) for each input and output.

In this embodiment, a multi-layer neural network is used as a machine learning algorithm. A neural network includes an input layer for inputting data, an output layer for generating the data to be predicted, and one or more hidden layers between the input layer and the output layer. A plurality of nodes (also called units) are arranged in each layer. Specifically, in this embodiment, a convolutional neural network (CNN), which is a kind of multi-layer neural network, is used. However, other machine learning algorithms may be used. For example, a hostile generative network (GAN) that utilizes two competing neural networks may be adopted as a machine learning algorithm.

The processes of S1 to S4 are repeated until the construction of the mathematical model is completed (S5: NO). When the construction of the mathematical model is completed (S5: YES), the mathematical model construction process is completed. The program and data for realizing the constructed mathematical model are incorporated in the ophthalmologic image processing apparatus 21.

(Ophthalmic image processing)
The ophthalmic image processing executed by the ophthalmic image processing apparatus 21 will be described with reference to FIGS. 7 to 9. The ophthalmic image processing is executed by the CPU 23 according to the ophthalmic image processing program stored in the storage device 24.

A high-quality moving image generation process, which is an example of ophthalmic image processing, will be described with reference to FIG. 7. In the high-quality moving image generation process, a still image of a high-quality target image is acquired using a mathematical model, and a high-quality moving image of the tissue is generated.

First, the CPU 23 acquires a base image which is an ophthalmic image (still image) (S11). The base image of this embodiment is a two-dimensional tomographic image of the fundus of the eye, similar to the above-mentioned training ophthalmic images 300A to 300X. By inputting the base image into the mathematical model described above, the CPU 23 acquires a target image having a higher image quality than the base image (S12).

In the example shown in FIG. 7, the ophthalmologic image capturing device 11B captures a moving image by continuously capturing a two-dimensional tomographic image of the fundus. In S11, the CPU 23 sequentially acquires each still image constituting the moving image taken by the ophthalmologic image capturing device 11B. The CPU 23 acquires at least one of the original image of the acquired still image and the added average image obtained by adding and averaging a plurality of original images as a base image to be input to the mathematical model. The CPU 23 continuously acquires the target image by sequentially inputting the continuously acquired basic images into the mathematical model.

The type of the base image acquired in S11 (at least one of the original image and the averaging image) and the number of original images used for the base image to be input to the mathematical model are determined in the above-mentioned acquisition pattern of input training data and the like. It may be set as appropriate according to the situation. When the averaging image is used as the base image to be input to the mathematical model, the CPU 23 may acquire the averaging image by averaging a plurality of ophthalmic images obtained by photographing the same imaging region. Further, the CPU 23 may acquire an averaging image generated by the ophthalmologic image capturing apparatus 11B.

In the high-quality moving image generation process illustrated in FIG. 7, a moving image of the tissue is generated. Here, in order to generate a good moving image, it is desirable to increase the frame rate (the number of still images per unit time) as much as possible. That is, in order to increase the frame rate of the moving image of the target image, it is necessary to increase the number of target images (still images) acquired per unit time as much as possible. Therefore, in S11 of FIG. 7, it is preferable that as few original images as possible are acquired as the base image to be input to the mathematical model. For example, each time the CPU 23 acquires one original image from the ophthalmic image capturing device 11B, the CPU 23 inputs the acquired original image into a mathematical model to generate a target image, so that the ophthalmic image capturing device 11B can generate the target image. It is possible to generate a moving image having the same frame rate as the moving image to be shot by using a high-quality target image.

Next, the CPU 23 determines whether or not an instruction for displaying the target image is input to the display device 28 (S14). In the present embodiment, the user operates the operation unit 27 to instruct the display device 28 to display the target image, the base image, or both the target image and the base image, and the ophthalmic image processing device 21. Can be entered in. If the instruction to display the target image is not input (S14: NO), the process proceeds to S17 as it is. When an instruction to display the target image is input (S14: YES), the CPU 23 causes the display device 28 to display the target image (S15). Specifically, in S15 of the present embodiment, a plurality of target images continuously acquired in S12 are sequentially displayed on the display device 28, so that a moving image of the tissue is displayed. That is, based on the moving image of the original image taken at that time by the ophthalmologic image taking device 11B, a high-quality moving image of the tissue being photographed is displayed on the display device 28 in real time. The CPU 23 may generate moving image data of the target image based on the plurality of target images continuously acquired in S12 and store the data in the storage device.

The ophthalmic image processing device 21 can input an instruction to display the target image, the base image, and both the target image and the base image on the display device 28 by various methods. For example, the CPU 23 may display the icon for inputting the display instruction of the target image, the icon for inputting the display instruction of the base image, and the icon for inputting the display instruction of both images on the display device 28. The user may input a display instruction to the ophthalmologic image processing device 21 by selecting a desired icon.

Next, the CPU 23 determines whether or not an instruction for displaying the base image is input to the display device 28 (S17). If it is not input (S17: NO), the process proceeds to S20 as it is. When an instruction to display the base image is input (S17: YES), the CPU 23 causes the display device 28 to display the base image (S18). Specifically, in S18 of the present embodiment, a plurality of base images continuously acquired in S11 are sequentially displayed on the display device 28, so that a moving image of the tissue is displayed. As described above, the CPU 23 can switch between the target image and the base image and display them on the display device 28 in response to the instruction input by the user. Therefore, the user can easily confirm both the target image and the base image.

Next, the CPU 23 determines whether or not an instruction to display the target image and the base image on the display device 28 at the same time is input (S20). If it is not input (S20: NO), the process proceeds to S23 as it is. When the simultaneous display instruction is input (S20: YES), the CPU 23 causes the display device 28 to display the target image and the base image at the same time. Specifically, in S20 of the present embodiment, both the moving image of the target image and the moving image of the base image are displayed on the display device 28.

In S15, S18, and S21, a still image may be displayed instead of a moving image. The CPU 23 may switch between displaying a moving image and displaying a still image in response to an instruction input by the user. Further, the CPU 23 may store at least one data of the moving image and the still image in the storage device 24 without displaying the moving image and the still image on the display device 28.

Further, when the moving image or the still image of the target image is displayed on the display device 28, the CPU 23 can display the target image corresponding to the averaging image of the number of ophthalmic images on the display device 28. .. For example, when the averaging image 32 used as the output training data when constructing the mathematical model is an averaging image based on H ophthalmic images, the target image acquired in S12 is also H. It becomes an image corresponding to the summed average image of one ophthalmic image. Therefore, the CPU 23 causes the display device 28 to display that the target image being displayed corresponds to the averaging image of the H ophthalmic images. Therefore, the user can appropriately grasp the quality of the target image.

The timing of displaying the target image includes, for example, the timing of displaying the target image in real time during shooting by the ophthalmologic image capturing device 11B, the timing of causing the user to confirm whether or not the captured image is good, and the analysis result after shooting. Or there are multiple timings such as the timing to display with the report. The CPU 23 separately gives an instruction to display the target image, the base image, and both the target image and the base image on the display device 28 for each of the plurality of timings at which the target image can be displayed. Can be entered in. That is, the user sets in advance whether to display the target image, the base image, or both the target image and the base image on the display device 28 according to the timing at which the target image can be displayed. Can be left. As a result, the convenience of the user is further improved.

Next, the CPU 23 determines whether or not an instruction to end the display has been input (S23). If it is not input (S23: NO), the process returns to S11, and the processes of S11 to S23 are repeated. When an instruction to end the display is input (S23: YES), the CPU 23 executes a segmentation process on the target image (in the present embodiment, at least one of a plurality of continuously acquired target images) (S24). ). The segmentation process is a process of identifying at least one of a plurality of layers of tissue shown in an image. The layer identification may be performed, for example, by image processing on the target image, or a mathematical model trained by a machine learning algorithm may be utilized. By performing the segmentation processing on the target image, the segmentation result in which the influence of noise is suppressed is appropriately acquired. In some cases, the user operates the operation unit 27 to manually perform the segmentation process. In this case as well, the CPU 23 may display the target image on the display device 28 as an image to be subjected to the segmentation process.

A high-quality three-dimensional tomographic image generation process, which is an example of ophthalmic image processing, will be described with reference to FIGS. 8 and 9. In the high-quality three-dimensional tomographic image generation process, high-quality two-dimensional tomographic images and three-dimensional tomographic images are acquired using a mathematical model. When a high-quality three-dimensional tomographic image generation process is executed, the ophthalmologic image capturing apparatus 11B captures a three-dimensional tomographic image in the tissue by scanning the OCT light (measurement light) two-dimensionally. .. Specifically, the ophthalmologic imaging apparatus 11B has a plurality (first) by causing each of a plurality of scan lines having different positions to scan the measurement light in a two-dimensional region when the tissue is viewed from the front. ~ Zth) 2D tomographic image is acquired. A three-dimensional tomographic image is acquired by combining a plurality of two-dimensional tomographic images.

The CPU 23 uses the Nth (initial value of N is "1") two-dimensional tomographic image as the base image among the plurality of (1st to Zth) two-dimensional tomographic images taken by the ophthalmic imaging apparatus 1B. Acquire (S31). The CPU 23 acquires the Nth target image by inputting the acquired base image into the mathematical model (S32). Next, the CPU 23 determines whether or not the processing for all of the plurality of (first to Zth) base images is completed (S33). If it is not completed (S33: NO), "1" is added to the value of N (S34), and the process returns to S31. The processes of S31 to S33 are repeated until the processes for all the base images are completed.

When the processing for all the base images is completed (S33: YES), the acquired three-dimensional tomographic image 40 (see FIG. 9) is generated by combining the acquired plurality of target images (S35). According to the high-quality three-dimensional tomographic image generation process illustrated in FIG. 8, even if the additive average image is not used as each of the two-dimensional tomographic images constituting the three-dimensional tomographic image, the influence of noise is equivalent to that of the additive average image. The suppressed target image produces a three-dimensional tomographic image.

The CPU 23 can also generate an image other than the three-dimensional tomographic image based on the plurality of target images. For example, the CPU 23 may generate a two-dimensional front image (so-called “Enface image”) when the tissue is viewed from the front based on the plurality of target images acquired in S32.

The CPU 23 can also perform analysis processing on the three-dimensional tomographic image 40 generated based on the plurality of target images. For example, the CPU 23 may perform a segmentation process on the three-dimensional tomographic image 40 to extract at least one of a plurality of layers contained in the fundus. The CPU 23 may generate a thickness map that two-dimensionally shows the thickness distribution of at least one of the layers based on the result of the segmentation process. In this case, the thickness map is appropriately generated based on the three-dimensional tomographic image 40 in which the influence of noise is suppressed.

Further, the CPU 23 may generate a two-dimensional tomographic image at an arbitrary position in the tissue included in the imaging range of the three-dimensional tomographic image 40 based on the three-dimensional tomographic image 40 generated based on the plurality of target images. Good. The CPU 23 may display the generated two-dimensional tomographic image on the display device 28. As an example, the CPU 23 of the present embodiment inputs an instruction from the user for designating a position for generating a tomographic image. For example, the user may specify the tomographic image generation position on the two-dimensional front image or the three-dimensional tomographic image 40 displayed on the display device 28 by operating the operation unit 27. The tomographic image generation position may be specified by, for example, lines of various shapes (lines on a straight line, circular lines, etc.). The CPU 23 may generate (extract) a two-dimensional tomographic image at a position specified by the user from the three-dimensional tomographic image 40. In this case, a tomographic image at an arbitrary position desired by the user is appropriately generated based on the three-dimensional tomographic image 40 in which the influence of noise is suppressed.

Further, the CPU 23 may determine whether or not the ophthalmologic image captured by the ophthalmologic image capturing apparatus 11B is good or not every time one or a plurality of ophthalmic images are captured. When the CPU 23 determines that the captured ophthalmic image is good, the CPU 23 may input the good ophthalmic image into the mathematical model as a base image. Further, the CPU 23 may output an instruction to re-photograph the ophthalmologic image at the same position to the ophthalmologic image capturing device 11B when it is determined that the captured ophthalmic image is not good. In this case, the target image is acquired based on a good base image. A specific method for determining whether or not the captured ophthalmic image is good can be appropriately selected. For example, the CPU 23 uses an index (for example, SSI (Signal Strength Index) or QI (Quality Index)) indicating the signal strength or signal goodness of the captured ophthalmic image to obtain a good ophthalmic image. It may be determined whether or not it is. Further, the CPU 23 may determine whether or not the ophthalmic image is good by comparing the captured ophthalmic image with the template.

Further, when the CPU 23 performs an analysis process (for example, a segmentation process) on any of the plurality of target images, the CPU 23 is good from the plurality of target images or the plurality of base images that are the basis of the plurality of target images. Image may be extracted. The CPU 23 may perform analysis processing on the extracted good target image or the target image based on the good base image. In this case, the accuracy of the analysis process is improved. As a method for extracting a good image, various methods such as the above-mentioned method using SSI or the like can be adopted.

Further, when generating a motion contrast image, the CPU 23 may acquire a target image using the mathematical model exemplified in the present disclosure and generate a motion contrast image based on the acquired target image. As described above, when generating a motion contrast image, a plurality of OCT signals are acquired at different times with respect to the same position, and arithmetic processing is performed on the plurality of OCT signals. The CPU 23 may acquire a motion contrast image by acquiring a target image based on each of a plurality of OCT signals (two-dimensional tomographic images) and performing arithmetic processing on the acquired plurality of target images. In this case, a motion contrast image in which the influence of noise is suppressed is appropriately generated.

The techniques disclosed in the above embodiments are merely examples. Therefore, it is possible to modify the techniques exemplified in the above embodiments. First, it is also possible to implement only some of the plurality of techniques exemplified in the above embodiments. For example, the ophthalmic image processing apparatus 21 may execute only one of the high-quality moving image generation process (see FIG. 7) and the high-quality three-dimensional tomographic image generation process (see FIG. 8).

The process of acquiring the base image in S11 of FIG. 7 and S31 of FIG. 8 is an example of the “base image acquisition step”. The process of acquiring the target image in S12 of FIG. 7 and S32 of FIG. 8 is an example of the “target image acquisition step”. The process of generating a three-dimensional tomographic image in S35 of FIG. 8 is an example of the “three-dimensional tomographic image generation step”. The segmentation process shown in S24 of FIG. 7 is an example of the “segmentation step”. The process of switching and displaying the target image and the base image in S14 to S18 of FIG. 7 is an example of the “switching display process” and the “switching display step”. The process of displaying the target image and the base image at the same time in S21 of FIG. 7 is an example of the “simultaneous display process” and the “simultaneous display step”. The process of generating the moving image data of the target image in S15 of FIG. 7 is an example of the “moving image generation step”. The process of acquiring the training ophthalmic image set 30 in S1 of FIG. 2 is an example of the “image set acquisition step”. The process of constructing a mathematical model in S2 to S4 of FIG. 2 is an example of a “training step”.

1 Mathematical model construction device 3 CPU
4 Storage device 11 Ophthalmic imaging device 21 Ophthalmic image processing device 23 CPU
24 Storage device 28 Display device 30 Set 31 Additive average image 32 Additive average image 300 Training ophthalmic image

Claims (8)

An ophthalmic image processing device that processes an ophthalmic image, which is an image of the tissue of the eye to be inspected.
The control unit of the ophthalmic image processing device
An ophthalmic image taken by the ophthalmic imaging apparatus or an added average image obtained by adding and averaging a plurality of ophthalmic images obtained by photographing the same part of the tissue by the ophthalmic imaging apparatus is acquired as a base image.
By inputting the base image into a mathematical model trained by a machine learning algorithm, a target image having a higher image quality than the base image can be obtained.
In the mathematical model, out of a plurality of training ophthalmic images obtained by photographing the same part of the tissue, L (L ≧ 1) data based on the training ophthalmic images are used as input training data, and H (H). > L) An ophthalmic image processing apparatus characterized in that the data of the added average image obtained by adding and averaging the training ophthalmic images is trained as output training data. The ophthalmic image processing apparatus according to claim 1.
The ophthalmic image is a two-dimensional tomographic image of the eye to be inspected taken by an OCT apparatus.
The control unit
A plurality of two-dimensional tomographic images obtained by photographing different positions of tissues are acquired as the basic images, and the images are obtained.
By inputting each of the plurality of basic images into the mathematical model, a plurality of the target images can be obtained.
An ophthalmic image processing apparatus characterized in that at least one of a three-dimensional tomographic image of the tissue and a two-dimensional front image when the tissue is viewed from the front is generated based on the plurality of target images. The ophthalmic image processing apparatus according to claim 1 or 2.
The ophthalmologic image is a two-dimensional tomographic image of the fundus of the eye to be inspected taken by an OCT device.
The control unit
An ophthalmologic image processing apparatus characterized in that at least one of a plurality of layers contained in the fundus is identified by performing analysis processing on the target image acquired by using the mathematical model. The ophthalmic image processing apparatus according to any one of claims 1 to 3.
The control unit performs simultaneous display processing for displaying the basic image and the target image on the display device at the same time, and switching display processing for switching between the basic image and the target image and displaying them on the display device in response to an input instruction. An ophthalmic image processing apparatus characterized in performing at least one of the above. The ophthalmic image processing apparatus according to any one of claims 1 to 4.
The control unit
A plurality of the basic images continuously taken by the ophthalmologic imaging apparatus are acquired, and
By sequentially inputting the acquired plurality of basic images into the mathematical model, a plurality of the target images can be acquired.
An ophthalmic image processing apparatus characterized in that moving image data of the tissue is generated based on the acquired plurality of target images. An OCT device that captures an ophthalmic image of the tissue by processing an OCT signal from the reference light and the reflected light of the measurement light applied to the tissue of the eye to be inspected.
The control unit of the OCT device
An ophthalmic image of the tissue taken or an added average image obtained by adding and averaging a plurality of ophthalmic images of the same part of the tissue was acquired as a base image.
By inputting the base image into a mathematical model trained by a machine learning algorithm, a target image having a higher image quality than the base image can be obtained.
In the mathematical model, out of a plurality of training ophthalmic images obtained by photographing the same part of the tissue, L (L ≧ 1) data based on the training ophthalmic images are used as input training data, and H (H). > L) An OCT apparatus characterized in that the data of the added average image obtained by adding and averaging the training ophthalmic images is trained as output training data. An ophthalmic image processing program executed by an ophthalmic image processing device that processes an ophthalmic image that is an image of the tissue of the eye to be inspected.
When the ophthalmic image processing program is executed by the control unit of the ophthalmic image processing device,
A basic image acquisition step of acquiring as a basic image an ophthalmic image taken by an ophthalmic imaging apparatus or an added average image obtained by adding and averaging a plurality of ophthalmic images obtained by photographing the same part of a tissue by the ophthalmic imaging apparatus.
By inputting the base image into a mathematical model trained by a machine learning algorithm, a target image acquisition step of acquiring a target image having a higher image quality than the base image, and
To the ophthalmic image processing apparatus
In the mathematical model, out of a plurality of training ophthalmic images obtained by photographing the same part of the tissue, L (L ≧ 1) data based on the training ophthalmic images are used as input training data, and H (H). > L) An ophthalmic image processing program characterized in that the data of the added average image obtained by adding and averaging the training ophthalmic images is trained as output training data. It is a mathematical model construction method executed by a mathematical model construction device that is constructed by training by a machine learning algorithm and constructs a mathematical model that outputs a target image by inputting a base image.
An image set acquisition step to acquire a plurality of training ophthalmic images obtained by photographing the same part of the tissue of the eye to be inspected, and
Of the plurality of training ophthalmic images, L (L ≧ 1) data based on the training ophthalmic images are used as input training data, and H (H> L) training ophthalmic images are added and averaged. The training step of constructing the mathematical model by training the mathematical model using the data of the added average image as output training data, and
How to build a mathematical model including.

Images