Image preservation and stitching for minimal flash eye disease diagnosis

Abstract

Provided herein are systems and methods for minimizing exposure of a retina to a flash during image capture for a diagnosis. In one embodiment, a system captures a plurality of retinal images of different retinal regions. The system determines that a first portion of a first image does not satisfy a criterion, that a second portion of the first image satisfies the criterion, identifies a portion of the retina depicted in the first portion that does not satisfy the criterion, and determines whether the portion of the retina is depicted in a third portion of a second image and whether the third portion satisfies the criterion. In response to determining that the third portion satisfies the criterion, the system performs the diagnosis. In response to determining that the portion of the retina is not depicted in the second image, the system captures an additional image of the retinal region.

Description

Technical Field

[0001] This invention relates generally to the autonomous diagnosis of retinal abnormalities, and more specifically to image preservation that reduces the need for retinal flashes in patients. Background Technology

[0002] Autonomous systems used to diagnose retinal abnormalities capture images of the patient's retina (the term fundus may be used interchangeably here) and analyze these images for abnormalities. Typically, images of different parts of the retina are captured (e.g., images centered on the fovea and images centered on the optic disc). When the captured images are insufficient for diagnosis, additional images are captured until enough images of each part of the retina are obtained. Repeated exposure to flashes by the imaging device can lead to further pupillary constriction, making successful imaging less likely after each flash. Therefore, reducing the need to capture more images before enough are captured reduces the number of flashes and the need to take numerous photographs, or force the patient to wait for diagnosis until the pupil returns to normal dilation. Furthermore, reducing the need for more images enables faster diagnosis because the number of flashes that cause a sharp increase in pupillary constriction is unlikely to be reached all at once, thus preventing delays in diagnosis until a viable image of the patient's eye can be obtained again. Summary of the Invention

[0003] The system and method presented herein are designed to reduce the need for re-flashing a patient's eye to capture images of retinal areas that are insufficiently captured in previously captured images. As an example, two images can be captured to perform a diagnosis on a patient's right eye—an image centered on the fovea of ​​the retina and an image centered on the optic disc. There is some overlap between the portions of the retina depicted in these two images. If a region of the retina depicted in one image is unavailable for any reason (e.g., overexposure or underexposure), instead of recapturing the image, the system can determine whether the same region, without the same defect, is depicted in the other image. If the region is available from the other image, the system avoids the need for re-flashing the patient's right eye to capture another image of that region; instead, portions of the two images can be stitched together to perform the diagnosis.

[0004] In one embodiment, to minimize retinal exposure to flash during image acquisition for diagnostic purposes, a retinal image preprocessing tool captures multiple retinal images (e.g., by instructing an imaging device to capture images and sending them to the retinal image preprocessing tool). Each retinal image may correspond to a different retinal region among multiple retinal regions (e.g., centered on the fovea and centered on the optic disc). The retinal image preprocessing tool may determine that a first portion of the first image (e.g., centered on the fovea) does not meet criteria (e.g., the image is overexposed or underexposed, has shadows or other artifacts, etc.), while a second portion of the first image (e.g., centered on the optic disc) meets criteria (e.g., the second image is properly exposed).

[0005] The retinal image preprocessing tool can identify portions of the retina depicted in the first image that do not meet the criteria, and can determine that the same portion of the retina is depicted in the third image of the second image. The retinal image preprocessing tool can determine whether the third portion meets the criteria. In response to determining that the third portion meets the criteria, a diagnosis can be performed using multiple retinal images. In response to determining that the portion of the retina is not depicted in the second image, the retinal image preprocessing tool can capture an additional image of the retinal region depicted in the first image. Attached Figure Description

[0006] Figure 1 This is an exemplary block diagram of system components in an environment utilizing a retinal image preprocessing tool, according to one embodiment.

[0007] Figure 2 This is an exemplary block diagram of modules and components of an imaging device according to one embodiment.

[0008] Figure 3 This is an exemplary block diagram of modules and components of a retinal image preprocessing tool according to one embodiment.

[0009] Figure 4 This is a block diagram illustrating the components of an example machine capable of reading instructions from a machine-readable medium and executing those instructions in a processor (or controller).

[0010] Figure 5 An exemplary image of a patient's eye according to one embodiment is depicted.

[0011] Figure 6 An exemplary flowchart is depicted according to one embodiment for minimizing retinal exposure to flashes during image acquisition for diagnostic purposes.

[0012] The accompanying drawings illustrate various embodiments of the invention for illustrative purposes only. Those skilled in the art will readily recognize from the following discussion that alternative embodiments of the structures and methods shown herein can be employed without departing from the principles of the invention as described herein. Detailed Implementation

[0013] (a) Environmental Overview

[0014] Figure 1 This is an exemplary block diagram of system components in an environment for utilizing a retinal image preprocessing tool, according to one embodiment. Environment 100 includes an imaging device 110, a network 120, a retinal image preprocessing tool 130, and a retinal disease diagnostic tool 140. Imaging device 110 is a device configured to capture one or more images of the retina of a patient's eye. These images can be captured by the imaging device 110 through autonomous operation, such as computer program instructions or external signals (e.g., received from the retinal pigmentation determination tool 130), or a combination thereof. Examples of what these images might look like and how they are derived are described in co-owned U.S. Patent Application No. 15 / 466,636, filed March 22, 2017, the disclosure of which is incorporated herein by reference in its entirety. Reference to this disclosure... Figure 5 Other examples are described.

[0015] After capturing an image, imaging device 110 transmits the image to retinal disease diagnostic tool 140. Retinal disease diagnostic tool 140 can take one or more images as input and can autonomously output a diagnosis based on that input using a machine learning model. Retinal disease diagnostic tool 140 autonomously analyzes the retinal image and uses machine learning analysis of biomarkers within it to determine a diagnosis. The diagnosis may specifically determine that the user has a particular disease, such as diabetic retinopathy, or it may determine that the user may have a disease and therefore should see a doctor for confirmation and treatment. The manner in which retinal disease diagnostic tool 140 performs the analysis and determines the diagnosis is further discussed in co-owned U.S. Patent No. 10,115,194, granted October 30, 2019, the disclosure of which is incorporated herein by reference in its entirety.

[0016] Before performing a diagnosis, the retinal disease diagnostic tool 140 can enable the retinal image preprocessing tool 130 to determine whether the images(s) captured by the imaging device 110 are sufficient to perform the diagnosis. The method by which the retinal image preprocessing tool 130 performs this analysis is described below. Figure 3This is described in further detail. Although described as a component of the retinal disease diagnostic tool 140, the retinal image preprocessing tool 130 can be a separate entity that receives images, performs sufficient review on the images, and, if the images are sufficient for diagnosis, transmits them to the retinal disease diagnostic tool 140. The retinal image preprocessing tool 130 and / or the retinal disease diagnostic tool 140 can be instantiated on one or more servers. Furthermore, the retinal image preprocessing tool 130 and / or the retinal disease diagnostic tool 140 can be instantiated wholly or partially at the imaging device 110, thereby eliminating some or all of the need for communication services on the network 120.

[0017] (b) Exemplary imaging device components

[0018] Figure 2 This is an exemplary block diagram of modules and components of an imaging apparatus according to one embodiment. Imaging apparatus 110 includes an image capture component 211, a flash component 212, a retinal disease diagnostic tool application protocol interface (API) 214, and a user interface 215. Although not depicted, imaging apparatus 110 may include on-board instances of any one or both of retinal image preprocessing tool 130 and retinal disease diagnostic tool 140, and any components thereof. Imaging apparatus 110 may include any database or memory for performing any of the functions described herein. Imaging apparatus 110 may also exclude certain depicted components.

[0019] Image capture component 211 can be any sensor configured to capture an image of a patient's retina. For example, a dedicated lens can be used to capture an image of a patient's retina. Flash component 212 can be any component capable of illuminating the patient's retina during image capture by image capture component 211, and can be configured to emit light in accordance with the image capture operation of image capture component 211.

[0020] The retinal disease diagnostic tool API 214 connects to the retinal disease diagnostic tool 130 to translate commands from the retinal image preprocessing tool 130 to the imaging device 110. Exemplary commands may include commands to capture or recapture images, commands to adjust the intensity of light emitted by the flash assembly 212, etc. See below for reference. Figure 3 Let's discuss these commands and how they are generated in more detail.

[0021] User interface 215 is an interface through which the operator of imaging device 110 can command imaging device 110 to perform any function it is capable of performing, such as capturing images, adjusting flash intensity, etc. User interface 215 can be any hardware or software interface and can include physical components (e.g., buttons) and / or graphical components (e.g., on a display such as a touchscreen display). User interface 215 can be located on imaging device 110, can be a device peripheral to imaging device 110, or can be located on a device separated from imaging device 110 via network 120, thereby enabling remote operation of imaging device 110. Reference Figure 5 An exemplary user interface is shown and discussed in more detail.

[0022] (c) Exemplary retinal image preprocessing tool components

[0023] Figure 3 This is an exemplary block diagram of modules and components of a retinal image preprocessing tool according to one embodiment. The retinal image preprocessing tool 130 includes an initial image capture module 331, a preprocessing module 232, a backward image evaluation module 233, an image recapture module 234, and an image stitching module 235. Although not described, the retinal image preprocessing tool 110 may include other components, such as additional modules, and any database or memory for performing any of the functions described herein. The retinal image preprocessing tool 130 may also exclude some of the depicted components.

[0024] The initial image capture module 331 captures retinal images of one or both eyes of the patient. As used herein, the term capture can refer to taking an image, meaning that "capture" can involve commanding the imaging device 110 to take an image from a remote device or server and transmitting the image to the initial image capture module 331 via network 120. Alternatively, if the initial image capture module 331 resides on the imaging device 110, the initial image capture module 331 can capture an image by commanding the imaging device to acquire an image and routing that image to the preprocessing module 332.

[0025] The initial image captured by the initial image capture module 331 may include images of different parts of the patient's retina. When the images of different parts are used together as input to one or more machine learning models of the retinal disease diagnostic tool 140, a retinal disease diagnosis output is produced. The retinal regions can be predefined. For example, the administrator of the retinal disease diagnostic tool 140 can instruct that images of certain retinal regions will be captured for diagnosis. These regions can be defined as images centered on the fovea of ​​the retina and images centered on the optic disc. Therefore, the initial image capture module 331 can capture images of one or both eyes of the patient centered on a specified retinal region.

[0026] In one embodiment, the captured image can be a multi-frame video spanning a time span. For example, similar to a "live photo," frames can be captured from the moment the flash is first fired until the flash has completely disappeared, where the frames form a video if displayed sequentially. Capturing multiple frames increases the chance that one frame includes a portion of the retina depicted by the image that meets preprocessing criteria (described below), even if that portion in another frame does not meet the criteria. For example, according to the flash example, if the flash is adjusted to another intensity (e.g., after 0.05 seconds, the flash is dim when it is turned off), and underexposure, overexposure, or shading caused by one intensity of the flash is corrected, the portion of the retina depicted in the frame at the other intensity can be used to meet the criteria.

[0027] Preprocessing module 332 is used to determine whether the captured images are sufficient to be input into a machine learning model. Satisfaction can also be defined by the administrator and can be based on parameters such as underexposure (e.g., capturing an image using too low a flash intensity), overexposure (e.g., capturing an image using too high a flash intensity), blur (e.g., when the patient moves while the image is being captured, resulting in an unclear image), shadows, and / or any other specified parameters. Criteria for determining whether an image is sufficient can be established based on these parameters—for example, the image must be within a certain exposure range, and / or landmarks (e.g., the boundary of the optic disc) must be narrow enough (where a wide boundary indicates blur). Preprocessing module 332 compares the parameters of the images to these criteria to determine whether each image is sufficient to be input into the machine learning model.

[0028] In one embodiment, when the preprocessing module 332 determines that the image is insufficient to be input into the machine learning model, the retinal image preprocessing tool 130 can command the recapture of the insufficient image. However, this embodiment has the disadvantage of potentially affecting the patient's health due to the need for additional flashes to the patient's eye. There are also technical disadvantages because additional bandwidth and processing power are required to capture the image and retransmit it to the retinal image preprocessing tool 130. In one embodiment, a backtracking image evaluation module 333 is used to determine whether a diagnosis can be performed regardless of the inadequacy of a given image without recaptured it.

[0029] The backward image evaluation model 333 identifies one or more depicted portions of the retina in an image that do not meet one or more criteria. For example, an image may be partially overexposed or underexposed because retinal pigmentation in a patient's retina is inconsistent, resulting in uniform flash levels that lead to proper exposure in some portions of the patient's retina while causing inappropriate exposure in others. As another example, an image may have shadows that can obscure one or more biomarker portions of the depicted retina while leaving the remaining portion of the image with sufficient quality. Thus, portions(s) that meet one or more criteria and / or portions(s) that do not meet one or more criteria can be isolated by the backward image evaluation model 333. The backward image evaluation model 333 can identify these portions by examining the portions of the image that meet and do not meet the criteria on any basis (e.g., pixel-by-pixel, quadrant-by-quadrant, region-by-region, etc.).

[0030] If a portion of the retina depicted by an image is determined to be insufficient, the backward image evaluation model 333 examines one or more other images captured by the initial image capture module 331 to determine whether the portion is recoverable without needing to recapture the image to cure the deficiency. For example, if a portion of the retina depicted in an image at the fovea center of the patient's retina is insufficient, the backward image evaluation model 333 can determine whether the same portion of the retina is depicted in an image of the patient's retina centered on the optic disc, which was also captured during the initial image capture. Another example could include evaluating whether frames in a multi-frame video in which the insufficient image is a part can be used. Because the backward image evaluation model 333 evaluates whether each captured image is sufficient for diagnosis, it can easily determine whether the same portion of the retina is depicted in another image and whether that same portion has sufficient quality. If the same portion has sufficient quality, the backward image evaluation module 333 determines that recapture is not necessary because the backward image is available. If the backward image evaluation module 333 determines that the same image of the retina is not depicted in another image with sufficient quality, the backward image evaluation module 333 determines that the image should be recaptured. In one embodiment, two or more images may be used together as a backslide for a single image, wherein different deficient parts of the image are repaired by portions from the two or more backslide images.

[0031] Image recapture module 334 recaptures images where there is insufficient image, and in some embodiments, where there is not enough back-up image to compensate for the insufficiency. Aside from capturing images of the already captured area of ​​the retina, image recapture module 334 operates in the same manner as image capture module 331.

[0032] Image stitching module 335 stitches together sufficient portions of an image with replacements for insufficient portions of the image, the replacements being from a backed-up image. As used herein, the term "stitching" refers to the actual or logical aggregation of portions from two images. Actual stitching refers to generating a synthetic image that includes portions from at least two images. Logical stitching refers to taking different portions from different images and using these separate portions for diagnosis. For example, these portions can be fed separately into a machine learning model without generating a synthetic image, and a diagnosis can be output from them. As another example, sufficient portions of an image can be used to generate a first analysis, regardless of the insufficient portions of that image. The backed-up image may have portions corresponding to the insufficient portions(s) of the first image(s) used to generate a second analysis. Both analyses can be used to perform a diagnosis (e.g., by feeding analyses with or without the image itself into a machine learning model). In one embodiment, the image stitching module 335 may weight regions of insufficient image quality based on the prevalence of biomarkers in those regions relative to their distance from anatomical markers (such as the fovea and optic disc), allowing for higher confidence when processing those examinations, even if some regions of insufficient image quality exist in the stitched image of the fovea. This weighting may be based on training on a representative dataset with random forests, ANNs, etc., to learn and weight the importance of various regions of sufficient image quality in the fovea. While the image stitching module 335 is described as a module of the retinal image preprocessing tool 130, it may be wholly or partially replaced by a module of the retinal disease diagnostic tool 140.

[0033] (d) Exemplary Computer Architecture

[0034] Figure 4 This is a block diagram illustrating the components of an example machine capable of reading instructions from a machine-readable medium and executing them in a processor (or controller). Specifically, Figure 4 A graphical representation of a machine in an example form of computer system 400 is shown, wherein program code (e.g., software) can be executed to cause the machine to perform any one or more methods discussed herein. The program code may include instructions 424 executable by one or more processors 402. In alternative embodiments, the machine operates as a standalone device or may be connected (e.g., networked) to other machines. In a networked deployment, the machine may operate within the capabilities of a server machine or client machine in a server-client network environment, or as a peer machine in a peer-to-peer (or distributed) network environment.

[0035] The machine can be a server computer, client computer, personal computer (PC), tablet PC, set-top box (STB), personal digital assistant (PDA), cellular phone, smartphone, web device, network router, switch or bridge, or any machine capable of executing instructions 424 (sequentially or otherwise) to perform the actions specified for the machine. Furthermore, although only a single machine is shown, the term "machine" should also be understood to include any collection of machines that individually or jointly execute instructions 124 to perform any one or more methods discussed herein.

[0036] Example computer system 400 includes a processor 402 (e.g., a central processing unit (CPU), graphics processing unit (GPU), digital signal processor (DSP), one or more application-specific integrated circuits (ASICs), one or more radio frequency integrated circuits (RFICs), or any combination thereof), main memory 404, and static memory 406, configured to communicate with each other via bus 408. Computer system 400 may further include a visual display interface 410. The visual interface may include a software driver capable of displaying a user interface on a screen (or monitor). The visual interface may display the user interface directly (e.g., on a screen) or indirectly (e.g., via a visual projection unit) on a surface, window, etc. For ease of discussion, the visual interface may be described as a screen. Visual interface 410 may include a touch-enabled screen or may interface with a touch-enabled screen. The computer system 400 may also include an alphanumeric input device 412 (e.g., a keyboard or touchscreen keyboard), a cursor control device 414 (e.g., a mouse, trackball, joystick, motion sensor or other pointing tool), a storage unit 416, a signal generation device 418 (e.g., a speaker), and a network interface device 420, which is also configured to communicate via a bus 408.

[0037] Storage unit 416 includes a machine-readable medium 422 on which instructions 424 (e.g., software) embodying any one or more methods or functions described herein are stored. The instructions 424 (e.g., software) may also reside wholly or at least partially within main memory 404 or processor 402 (e.g., within the processor's cache memory) during execution by computer system 400, which also constitute the machine-readable medium. The instructions 424 (e.g., software) may be transmitted or received on network 426 via network interface device 420.

[0038] Although machine-readable medium 422 is shown as a single medium in the example embodiment, the term "machine-readable medium" should be understood to include a single medium or multiple media (e.g., a centralized or distributed database, or associated caches and servers) capable of storing instructions (e.g., instruction 424). The term "machine-readable medium" should also be understood to include any medium capable of storing instructions (e.g., instruction 424) for execution by a machine and causing the machine to perform any one or more methods disclosed herein. The term "machine-readable medium" includes, but is not limited to, data repositories in the form of solid-state memory, optical media, and magnetic media.

[0039] (e) Sample images and user interface

[0040] Figure 5 An exemplary image of a patient's eye according to one embodiment is depicted. User interface 500 may be displayed to the operator of imaging device 110 as part of user interface 215. User interface 500 includes images 510, 520, 530, and 540. As depicted, the patient already has two images captured for each eye: an image centered on the fovea of ​​the retina and an image centered on the optic disc. These images are merely exemplary; any parameters can be used to indicate the position where the image should be centered, as input by the operator of imaging device 110 and / or an administrator defining what images are needed to perform a diagnosis. Furthermore, although two images are used in the depicted example, any number of images can be used. In the case where the images are centered on the optic disc and fovea of ​​the retina, portions of the images will overlap, and therefore, if any overlapping portion of an image does not meet a specified criterion, each image can be used as a fallback to each of the other images.

[0041] (f) An exemplary data stream for minimizing retinal flash during image acquisition for diagnostic purposes

[0042] Figure 6An exemplary flowchart is depicted according to one embodiment for minimizing retinal exposure to flash during image acquisition for diagnostic purposes. Process 600 begins with one or more processors (e.g., processor 402) of a device running a retinal image preprocessing tool 130 capturing 602 (e.g., using an initial image capture module 331) multiple retinal images, each corresponding to a different retinal region among multiple retinal areas. The multiple retinal images include a first image and a second image. The multiple retinal images may include, for example, an image centered on the fovea of ​​the retina for each eye and an image centered on the optic disc. The first image may be an image 510 centered on the fovea of ​​the patient's left eye, while the second image may be an image 530 centered on the optic disc of the patient's left eye, and different retinal regions may correspond to the retinal region centered in each image.

[0043] The retinal image preprocessing tool 130 determines that a first portion of the first image 604 does not meet the criteria, while a second portion of the first image meets the criteria. For example, due to retinal streaks, overexposure, underexposure, shadows, or other problems, a portion of the left-eye image 510 centered on the fovea of ​​the retina is determined to be insufficient, while the remaining portion of the left-eye image 510 centered on the fovea of ​​the retina is determined to be sufficient. The satisfaction determination can be performed by the preprocessing module 332 using any of the methods described herein.

[0044] The retinal image preprocessing tool 130 identifies portions of the retina that do not meet the criteria depicted in the first portion 606. That is, portions of the retina itself can be depicted at different coordinates in the backward image, and therefore, it is portions of the retina (rather than portions of the image) that are identified. The backward image evaluation module 333 can perform this determination in any manner described herein. The retinal image preprocessing tool 130 then determines whether this portion of the retina is depicted in the third portion of the second image 608. That is, the backward image evaluation module 333 determines whether the other captured image includes a depiction of the same portion of the retina at a certain location in the other captured image. For example, the retinal image preprocessing tool 130 determines whether the left-eye image 530 at the center of the optic disc depicts the same portion of the retina that is insufficient in the left-eye image 510 at the center of the fovea.

[0045] The retinal image preprocessing tool 130 determines whether a third portion of the second image 610 (e.g., a portion of the left-eye image 530 centered on the optic disc, capturing the same portion of the retina as the insufficient portion in the left-eye image 510 centered on the fovea of ​​the retina) meets a criterion. In response to the third portion meeting the criterion, the retinal image preprocessing tool 130 performs a diagnosis 612 using multiple retinal images (e.g., using the image stitching module 335). In response to the third portion not meeting the criterion, the retinal image preprocessing tool 130 captures an additional image 614 of the retinal region depicted in the first image (e.g., using the image recapture module 334).

[0046] (g) Summary of the Invention

[0047] The embodiments of the invention have been described above for illustrative purposes; however, they are not intended to be exhaustive or to limit the invention to the precise forms disclosed. Those skilled in the art will understand that many modifications and variations are possible based on the foregoing disclosure.

[0048] While this disclosure focuses specifically on retinal diseases, it is generally applicable to disease diagnosis, as well as the capture (and potential recapture) and preprocessing of images of other parts of a patient's body. For example, the disclosed preprocessing can be applied to the diagnosis of other organs, such as the kidneys, liver, brain, or heart of a person, or parts thereof, and the images discussed herein can be images of the relevant organs.

[0049] Certain portions of this specification describe embodiments of the invention based on algorithms and symbolic representations of operations on information. Those skilled in the art of data processing commonly use these algorithmic descriptions and representations to effectively communicate the essence of their work to others skilled in the art. While these operations are described functionally, computationally, or logically, they are understood to be implemented by computer programs or equivalent circuits, microcode, etc. Furthermore, without loss of generality, it has sometimes proven convenient to arrange these operations as modules. The described operations and their associated modules can be implemented in software, firmware, hardware, or any combination thereof.

[0050] Any step, operation, or process described herein may be performed or implemented, alone or in combination with other devices, using one or more hardware or software modules. In one embodiment, a software module is implemented using a computer program product, which includes a computer-readable medium containing computer program code that can be executed by a computer processor to perform any or all of the described steps, operations, or processes.

[0051] Embodiments of the invention may also relate to means for performing the operations described herein. Such means may be specifically constructed for the desired purpose, and / or may comprise a general-purpose computing device selectively activated or reconfigured by a computer program stored in a computer. Such a computer program may be stored in a non-transient, tangible, computer-readable storage medium or in any type of medium suitable for storing electronic instructions, and may be coupled to a computer system bus. Furthermore, any computing system mentioned herein may include a single processor, or may be an architecture employing a multiple-processor design to increase computing power.

[0052] Embodiments of the present invention may also relate to products generated by the computational processes described herein. Such products may include information generated by the computational processes, wherein the information is stored on a non-transient tangible computer-readable storage medium, and may include any embodiment of the computer program product or other combination of data described herein.

[0053] Finally, the language used in this specification has been chosen primarily for readability and instruction purposes and may not have been selected to depict or limit the subject matter of the invention. Therefore, the scope of the invention is not limited to this detailed description, but is defined by any of the claims based on this application. Thus, the disclosure of embodiments of the invention is intended to illustrate, rather than limit, the scope of the invention, which is set forth in the appended claims.

Claims

1. A method for minimizing retinal exposure to flashes during image acquisition for diagnostic purposes, the method comprising: A first retinal image of the first retinal region is captured by flashing a light onto the first retinal region in response to determining that the capturing device is focused on the first retinal region. A second retinal image of the second retinal region is captured by flashing a light onto the second retinal region in response to determining that the capturing device is focused on the second retinal region, and the first retinal image and the second retinal image together form a plurality of retinal images; The following method is used to determine whether to flash the first retinal region again: It is determined that a first portion of the first retinal image does not meet the criteria, while a second portion of the first retinal image does meet the criteria; Identify the portions of the retina depicted in the first part that do not meet the criteria; Determine whether the portion of the retina is depicted in the third portion of the second retinal image; Determine whether the third part meets the criteria; In response to determining that the third part meets the criteria, it is determined that no further flashes will be made on the first retinal region and the diagnosis is performed using the plurality of retinal images, wherein performing the diagnosis includes passing the plurality of retinal images to a fully autonomous machine learning model, the fully autonomous machine learning model outputting the probability of a disease condition based on the plurality of retinal images; as well as In response to determining that the portion of the retina is not depicted in the second retinal image, it is determined to flash the first retinal region again to capture an additional image of the first retinal region.

2. The method of claim 1, wherein capturing the first retinal image comprises capturing multiple frames of video while the first retinal region is illuminated by the flash in the first retinal region, each frame of the multiple frames of video capturing an image at a different flash exposure level.

3. The method of claim 1, wherein determining that the first portion of the first retinal image does not meet the criterion includes determining that the first portion of the first retinal image is overexposed or underexposed.

4. The method of claim 1, wherein the second retinal image is an image of the retina of the same eye depicted by the first retinal image.

5. The method of claim 1, wherein performing the diagnosis using the plurality of retinal images comprises: Generate a composite image, the composite image comprising the first portion of the first retinal image and the third portion of the second retinal image stitched into the first retinal image; as well as The diagnosis is performed using the synthesized image.

6. The method of claim 1, wherein performing the diagnosis using the plurality of retinal images comprises: Analyze the first retinal image while ignoring the first portion to generate a first analysis; Analyze the third portion of the second retinal image to generate a second analysis; as well as The diagnosis is performed using the first analysis and the second analysis.

7. The method of claim 1, wherein capturing the additional image comprises: The application protocol interface (API) is used to transmit commands to the imaging device to capture additional images using the viewpoint used to capture the first retinal image; as well as Receive the additional image from the imaging device.

8. The method of claim 1, wherein performing the diagnosis using the plurality of retinal images includes using the third part in the diagnosis.

9. A computer program product for minimizing retinal exposure to flashes during image acquisition for diagnostic purposes, the computer program product comprising a non-transient computer-readable storage medium containing computer program code for: A first retinal image of the first retinal region is captured by flashing a light onto the first retinal region in response to determining that the capturing device is focused on the first retinal region. A second retinal image of the second retinal region is captured by flashing a light onto the second retinal region in response to determining that the capturing device is focused on the second retinal region, and the first retinal image and the second retinal image together form a plurality of retinal images; The following method is used to determine whether to flash the first retinal region again: It is determined that a first portion of the first retinal image does not meet the criteria, while a second portion of the first retinal image does meet the criteria; Identify the portions of the retina depicted in the first part that do not meet the criteria; Determine whether the portion of the retina is depicted in the third portion of the second retinal image; Determine whether the third part meets the criteria; In response to determining that the third part meets the criteria, it is determined that no further flashes will be applied to the first retinal region and the diagnosis will be performed using the plurality of retinal images; as well as In response to determining that the portion of the retina is not depicted in the second retinal image, it is determined to flash the first retinal region again to capture an additional image of the first retinal region.

10. The computer program product of claim 9, wherein the computer program code for performing the diagnosis includes computer program code for passing the plurality of retinal images to a fully autonomous machine learning model, the fully autonomous machine learning model outputting the probability of a disease condition based on the plurality of retinal images.

11. The computer program product of claim 9, wherein capturing the first retinal image comprises capturing multiple frames of video while the first retinal region is illuminated by the flash in the first retinal region, each frame of the multiple frames of video capturing an image at a different flash exposure level.

12. The computer program product of claim 9, wherein the computer program code for determining that the first portion of the first retinal image does not meet the standard includes computer program code for determining that the first portion of the first retinal image is overexposed or underexposed.

13. The computer program product of claim 9, wherein the second retinal image is an image of the retina of the same eye depicted by the first retinal image.

14. The computer program product of claim 9, wherein the computer program code for performing the diagnosis using the plurality of retinal images comprises computer program code for: Generate a composite image, the composite image comprising a first portion of the first retinal image and a third portion of the second retinal image stitched together from the first retinal image; and The diagnosis is performed using the synthesized image.

15. The computer program product of claim 9, wherein the computer program code for performing the diagnosis using the plurality of retinal images comprises computer program code for: Analyze the first retinal image while ignoring the first portion to generate a first analysis; Analyze the third portion of the second retinal image to generate a second analysis; as well as The diagnosis is performed using the first analysis and the second analysis.

16. The computer program product of claim 9, wherein the computer program code for capturing the additional image comprises computer program code for: The application programming interface (API) is used to transmit commands to the imaging device to capture additional images using the viewpoint used to capture the first retinal image; and Receive the additional image from the imaging device.

17. The computer program product of claim 9, wherein performing the diagnosis using the plurality of retinal images includes using the third part in the diagnosis.

18. A computer program product for minimizing retinal exposure to flash during image acquisition for diagnostic purposes, the computer program product comprising a non-transitory computer-readable storage medium containing computer program code including: The first module is used for: A first retinal image of the first retinal region is captured by flashing a light onto the first retinal region in response to determining that the capturing device is focused on the first retinal region. as well as A second retinal image of the second retinal region is captured by flashing a light onto the second retinal region in response to determining that the capturing device is focused on the second retinal region, and the first retinal image and the second retinal image together form a plurality of retinal images; The second module is configured to determine whether to flash the first retinal region again by: It is determined that a first portion of the first retinal image does not meet the criteria, while a second portion of the first retinal image does meet the criteria; Identify the portions of the retina depicted in the first part that do not meet the criteria; Determine whether the portion of the retina is depicted in the third portion of the second retinal image; as well as Determine whether the third part meets the criteria; A third module is configured to, in response to determining that the third part meets the criteria, determine not to flash the first retinal region again and perform the diagnosis using the plurality of retinal images; as well as A fourth module, configured to, in response to determining that the portion of the retina is not depicted in the second retinal image, determine to flash the first retinal region again to capture an additional image of the first retinal region.

19. The computer program product of claim 18, wherein performing the diagnosis includes passing the plurality of retinal images to a fully autonomous machine learning model, the fully autonomous machine learning model outputting a probability of a disease condition based on the plurality of retinal images.

Images