Corneal epithelium segmentation in optical coherence tomography images

Abstract

Systems and methods described herein provide improved techniques for segmenting a corneal epithelial layer. One method includes receiving an optical coherence tomography (OCT) image of an eye, generating a binarized image of the eye based on the OCT image, generating a binary mask of a cornea of the eye based on the binarized image of the eye and the OCT image, segmenting an anterior cornea of the eye on the OCT image based on the binary mask of the cornea of the eye, generating a binary mask of an epithelial layer of the eye based on the OCT image and the segmented anterior cornea of the eye, segmenting a Bowman's layer in the cornea of the eye on the OCT image based on the binary mask of the epithelial layer of the eye, and making the segmented anterior cornea and the segmented Bowman's layer data available for generating an epithelial map.

Description

Background Technology

[0001] The embodiments disclosed herein generally relate to methods and apparatus for ophthalmic image segmentation, and more specifically to methods and apparatus for segmenting corneal epithelium of the eye in optical coherence tomography (OCT) images generated from scanning data from an ophthalmic scanning device.

[0002] For example, ophthalmic diagnostic systems such as OCT devices are configured to generate OCT images of a patient's eye. Such images can be valuable diagnostic tools for clinicians, such as physicians, and other users of the OCT device. For instance, certain information related to one or more structures of the eye can be extracted from the OCT images. Extracting such information from OCT images typically requires segmenting one or more regions, structures, and / or tissues of the eye within the OCT image.

[0003] However, due to certain inherent characteristics of OCT images, segmenting the regions, structures, and / or tissues of the eye from one or more OCT images presents various challenges. For example, OCT images may contain speckle noise and / or other image interference, which can increase the difficulty for computational systems to identify and segment different structures or tissues of the scanned eye.

[0004] Furthermore, when identifying and segmenting one or more structures and / or tissues of a scanned eye, sufficient relevant image data (e.g., sample and / or reference OCT images of the scanned eye) may not be available to develop and train models (e.g., machine learning models) that meet high-accuracy medical diagnostic standards. This lack of relevant image data prevents existing image segmentation systems and technologies from meeting the accuracy and performance standards and requirements for providing clinicians with real-time and / or near-real-time diagnostic information. Summary of the Invention

[0005] This disclosure generally relates to methods and apparatus for segmenting the corneal epithelium of the eye from OCT images to provide diagnostic information.

[0006] In some embodiments, a method for segmenting optical coherence tomography (OCT) images generally includes receiving an OCT image of an eye. The method further includes generating a binarized image of the eye based on the OCT image. The method also includes generating a binary mask of the cornea of ​​the eye based on the binarized image of the eye and the OCT image. The method further includes segmenting the anterior cornea of ​​the eye on the OCT image based on the binary mask of the cornea. The method further includes generating a binary mask of the epithelial layer of the eye based on the OCT image and the segmented anterior cornea. The method also includes segmenting the layer of Bowman's cornea in the cornea on the OCT image based on the binary mask of the epithelial layer. The method further includes using the segmented anterior cornea and segmented layer of Bowman's cornea data to generate an epithelial map.

[0007] In some embodiments, an optical coherence tomography (OCT) system generally includes a memory containing computer-executable instructions. The OCT system further includes a processor configured to execute the computer-executable instructions and cause the OCT system to generate an OCT image of the eye. The processor is further configured to cause the OCT system to generate a binarized image of the eye based on the OCT image. The processor is further configured to cause the OCT system to generate a binary mask of the cornea of ​​the eye based on the binarized image of the eye and the OCT image. The processor is further configured to cause the OCT system to segment the anterior cornea of ​​the eye on the OCT image based on the binary mask of the cornea. The processor is further configured to cause the OCT system to generate a binary mask of the epithelial layer of the eye based on the OCT image and the segmented anterior cornea. The processor is further configured to cause the OCT system to segment the layer of Bowman's cornea in the cornea of ​​the eye based on the binary mask of the epithelial layer. The processor is further configured to use the segmented anterior cornea and segmented layer of Bowman's cornea data to generate an epithelial map.

[0008] In another embodiment, an imaging system includes: a memory including computer-executable instructions; and a processor configured to execute the computer-executable instructions and cause the imaging system to: generate an optical coherence tomography (OCT) image of an eye; generate a binarized image of the eye based on the OCT image; generate a binary mask of the cornea of ​​the eye based on the binarized image of the eye and the OCT image; segment the anterior cornea of ​​the eye on the OCT image based on the binary mask of the cornea; generate a binary mask of the epithelial layer of the eye based on the OCT image and the segmented anterior cornea; segment the layer of Bowman in the cornea of ​​the eye on the OCT image based on the binary mask of the epithelial layer; and use the segmented anterior cornea and segmented layer of Bowman data to generate an epithelial map. In this embodiment, the segmented layer of Bowman is the upper boundary of the binary mask of the epithelial layer.

[0009] Various aspects of this disclosure provide apparatus, devices, processors, and computer-readable media for performing the methods described herein. Attached Figure Description

[0010] To gain a more detailed understanding of the features described above, reference can be made to embodiments, some of which are illustrated in the accompanying drawings. However, it should be noted that the drawings illustrate exemplary embodiments only and should not be construed as limiting the scope of the disclosure, and may allow for other equally effective embodiments.

[0011] Figure 1 A block diagram of selected components of an example imaging system according to certain embodiments of this disclosure is shown.

[0012] Figure 2 A block diagram of selected components of an OCT controller according to certain embodiments of this disclosure is shown.

[0013] Figure 3 An example flowchart illustrating the segmentation of the corneal epithelium in an OCT image according to certain embodiments of this disclosure is shown.

[0014] Figure 4A Example OCT images of the cornea according to certain embodiments of this disclosure are shown.

[0015] Figure 4B Example binarized images of OCT images of the cornea according to certain embodiments of this disclosure are shown.

[0016] Figure 4C Example enhanced OCT images of the cornea according to certain embodiments of this disclosure are shown.

[0017] Figure 4D An example initial mask for the corneal epithelium of a scanned eye is shown according to certain embodiments of this disclosure.

[0018] Figure 4E Example enhanced OCT images of the cornea according to an illustrative embodiment of this disclosure are shown.

[0019] Figure 4F Example enhanced OCT images of the cornea according to certain embodiments of this disclosure are shown.

[0020] Figure 5 A flowchart illustrating an example method for segmenting the corneal epithelium according to certain embodiments of this disclosure is shown.

[0021] For ease of understanding, the same reference numerals have been used where possible to refer to the same elements common to the figures. It is contemplated that elements and features of one embodiment may be advantageously combined with those of other embodiments without further description. Detailed Implementation

[0022] This disclosure generally relates to methods and apparatus for segmenting the corneal epithelium of the eye from OCT images to provide diagnostic information.

[0023] OCT images of the scanned eye can be valuable diagnostic tools for clinicians because they can provide valuable diagnostic information about the scanned eye. Recent developments in OCT imaging have allowed imaging of certain structures or tissues of the eye that were previously invisible or undetectable in OCT images. An example of such a recent development in OCT imaging is the advancement of corneal OCT, which makes the corneal epithelium visible in OCT images. The visibility of this structure of the eye in OCT images can provide clinicians with useful diagnostic information. For example, the visibility of the corneal epithelium can be used to generate an epithelial map, which clinicians and / or other diagnostic systems can use to determine whether a patient undergoing eye surgery (e.g., refractive eye surgery) is at risk of certain postoperative problems (e.g., postoperative ectasia).

[0024] To extract relevant diagnostic information about the eye, it may be necessary to segment one or more parts or structures of the eye that are diagnostically significant to clinicians from OCT images. However, OCT images may contain interference, such as speckle noise. The presence of such interference can reduce the accuracy and performance of image segmentation systems and techniques. For example, such interference may cause discontinuities or blurring at or near the boundaries of different structures or tissues of the eye. Due to discontinuities or blurring at or near the boundaries of eye structures or tissues in OCT images, existing image segmentation systems and techniques may struggle to accurately segment eye structures or tissues in OCT images. Furthermore, due to the size of some structures and tissues of the eye, when captured in OCT images, these structures and tissues may have low contrast and may be too thin to be accurately segmented using existing image segmentation systems and techniques. For example, the corneal epithelium may be thin, and when captured in OCT images, it may have low contrast and a thin boundary layer, which may lead to errors when segmenting the corneal epithelium in OCT images.

[0025] Furthermore, when segmenting eye structures or tissues from an image, it is essential to extract the overall features associated with that structure and segment the structure based on these extracted features. Image segmentation models trained on OCT image data including eye structures and / or tissues can be used to identify and extract features associated with these structures. However, for certain eye structures and / or tissues, identifying such deterministic features can be difficult when sufficient image data including such structures and / or tissues is unavailable. For example, there may not be a sufficient amount of annotated OCT image data including the corneal epithelium to train an image segmentation model that can accurately identify and segment the corneal epithelium in OCT images.

[0026] Therefore, to accurately segment the corneal epithelium, a large dataset of OCT images including the corneal epithelium is needed to develop an image segmentation model capable of accurately segmenting the corneal epithelium of a scanned eye. Furthermore, the amount of such annotated OCT image data must be large enough to ensure that these image segmentation models are trained to meet the accuracy and performance requirements for medical diagnosis. However, generating a sufficiently large dataset of annotated OCT images including the corneal epithelium to develop and / or train image segmentation models can consume significant computational and time resources from the image segmentation system, making it impractical in real-world applications to successfully train an image segmentation model to accurately segment OCT images to obtain the desired structures and / or tissues of a scanned eye.

[0027] Furthermore, some embodiments disclosed herein provide various devices, methods, and systems for segmenting the corneal epithelium of the eye in OCT images (such as corneal OCT images). As described herein, a corneal OCT image is a specific type of OCT image presenting the cornea and its sublayers of a scanned eye. In particular, as described herein, these devices, methods, and systems automatically segment the corneal epithelium from corneal OCT images. Moreover, these devices, methods, and systems described herein accurately segment the corneal epithelium in corneal OCT images without extensively training an image segmentation model on a large set of annotated OCT image data or OCT image data including data related to the corneal epithelium. Therefore, the techniques described herein improve the accuracy of image segmentation systems while significantly reducing the amount of training data, computational resources, and time resources required to accurately segment the corneal epithelium in corneal OCT images.

[0028] Figure 1 A block diagram of selected components of an example imaging system 100 is shown. The imaging system 100 includes an optical coherence tomography (OCT) scanner 102, an OCT controller 104, and a display 106.

[0029] OCT scanner 102 may include several OCT components and / or instruments (not shown separately). The OCT components and / or instruments can be of various types, and OCT scanner 102 may be configured differently based on the type of OCT components and / or instruments. In some embodiments, OCT scanner 102 may be configured as time-domain OCT (TD-OCT). In some embodiments, OCT scanner 102 may be configured as frequency-domain OCT (FD-OCT). In some embodiments, OCT scanner 102 may be configured as swept-source OCT (SS-OCT).

[0030] OCT scanner 102 performs an OCT scan on a patient's eye 110. OCT scanner 102 performs the OCT scan by controlling one or more sample beams (not shown) to be output to the eye 110 and receiving one or more measurement beams (not shown) reflected back from the eye 110. The one or more measurement beams can be reflected back from the eye 110 in response to the interaction of photons from the sample beams with tissue in the eye 110. OCT scanner 102 can be configured to move the sample beams to a specific location in the eye in response to receiving commands and / or location information from OCT controller 104.

[0031] OCT scanner 102 can be configured to scan eye 110 at various depths. For example, OCT scanner 102 can be configured to scan the entire depth of eye 110 to perform a whole-eye scan. Similarly, OCT scanner 102 can be configured to scan any part of eye 110, such as the retina of eye 110. In some embodiments, OCT scanner 102 can scan different depths of eye 110 at different resolutions. For example, OCT scanner 102 can scan the entire depth of eye 110 at a lower resolution and can scan a portion of eye 110, such as the retina, cornea, and its sublayers, at a higher resolution.

[0032] OCT scanner 102 can be configured to generate scan data based on one or more measurement beams reflected from the eye. The scan data can represent a depth profile of the scanned tissue. In some embodiments, the scan data generated by OCT scanner 102 may include two-dimensional (2D) scan data of line scans (B-scans). In some embodiments, the scan data generated by OCT scanner 102 may include three-dimensional (3D) scan data of face scans (C-scans). OCT scanner 102 can be configured to send the generated scan data to OCT controller 104. In some embodiments, OCT scanner 102 can be configured to send the generated scan data in real-time or near real-time. In some embodiments, OCT scanner 102 can be configured to send the generated scan data after OCT scanner 102 has completed the entire scanning operation.

[0033] OCT scanner 102 can be configured to initiate a scan of eye 110 in response to receiving a command and / or instruction from OCT controller 104. OCT controller 104 can be configured to send a scan initiation command to OCT scanner 102 in response to receiving an instruction from a user (e.g., a surgeon, clinician, medical professional, etc.) to initiate a scan of the eye. In some embodiments, the instruction from the user may provide information for scanning related to the depth and / or position of the eye, and OCT controller 104 can be configured to provide the received information related to the depth and / or position of the eye to OCT scanner 102. For example, the instruction received by OCT controller 104 may indicate a whole-eye OCT scan, and OCT controller 104 may send an instruction to OCT scanner 102 instructing a whole-eye OCT scan. Similarly, the instruction received by OCT controller 104 may indicate an OCT scan of the retina of the eye, and OCT controller 104 may send an instruction to OCT scanner 102 instructing an OCT scan of the retina of the eye. Similarly, the instructions received by the OCT controller 104 can instruct an OCT scan of the cornea and / or its sublayers of the eye, and the OCT controller 104 can send instructions to the OCT scanner 102 to instruct an OCT scan of the cornea and / or its sublayers of the eye 110.

[0034] The OCT controller 104 can be configured to receive instructions to initiate an eye scan via a user interface (e.g., a graphical user interface (GUI)) and / or an input device (not shown). The input device can be communicatively coupled to and / or integrated with the imaging system 100. Examples of input devices include, but are not limited to, a keypad, a keyboard, a touchscreen device configured to receive touch input, etc.

[0035] OCT controller 104 may be communicatively coupled to OCT scanner 102 via one or more electrical and / or communication interfaces. In some embodiments, the one or more electrical and / or communication interfaces may be configured to transmit data (e.g., scan data generated by OCT scanner 102) from OCT scanner 102 at a high transmission rate, such that OCT controller 104 may receive data from OCT scanner 102 in real time or near real time.

[0036] OCT controller 104 can be configured to generate one or more OCT images based on generated scan data received from OCT scanner 102. For example, OCT controller 104 can be configured to generate 2D images or B-scan images based on 2D scan data generated from line scans. Similarly, OCT controller 104 can be configured to generate 3D images or C-scans based on 3D scan data generated from area scans. The OCT images generated by OCT controller 104 include structures, tissues, and / or portions of the eye scanned by OCT scanner 102. For example, if OCT scanner 102 scans the cornea and / or its sublayers, the scan data generated by OCT scanner 102 may include data related to the cornea and / or sublayers, and the OCT images generated by OCT controller 104 include the cornea and / or its sublayers. In some embodiments, the horizontal display resolution of the OCT images generated by OCT controller 104 may be approximately 4,096 pixels, less than 4,096 pixels, or greater than 4,096 pixels (collectively referred to herein as 4KOCT images). The OCT controller 104 can be configured to perform image generation and / or image processing in real time and / or near real time.

[0037] OCT controller 104 may be configured with one or more detection algorithms configured to detect the cornea and / or sublayers of the cornea of ​​eye 110, and / or any portion of the cornea and / or its sublayers. Examples of one or more sublayers of the cornea of ​​eye 110 include, but are not limited to, the anterior cornea, Bowman's layer, corneal epithelium, etc. OCT controller 104 may be configured with one or more automatic segmentation algorithms to automatically segment the cornea and / or one or more sublayers of the cornea, such as the anterior cornea, Bowman's layer, corneal epithelium, etc., in OCT images. This document is about Figure 3 , Figure 4A , Figure 4B and Figure 5 Additional details of the automatic segmentation algorithm of the OCT controller 104 are described.

[0038] In some implementations, the OCT controller 104 may be configured with one or more additional tissue detection and / or automatic segmentation algorithms to detect and / or automatically segment those structures and / or tissue layers of the eye in the generated OCT images. Examples of such additional structures and / or tissue layers of the eye that the OCT controller 104 is configured to detect and / or automatically segment include, but are not limited to, the fovea, retinal pigment epithelium (RPE), anterior corneal surface, retina, cornea, iris, pupil, anterior and posterior surfaces plus the location of the lens, etc. The OCT controller 104 may be configured to apply one or more tissue detection and / or automatic segmentation algorithms to scan data received from the OCT scanner 102 and / or generated OCT images to detect and / or automatically segment one or more tissue layers of the scanned eye.

[0039] OCT controller 104 can be configured to generate enhanced OCT images by generating and / or displaying one or more virtual markers on one or more OCT images (e.g., generated OCT images, received OCT images, etc.), thereby visually identifying one or more detected and / or automatically segmented layers of eye tissue. For example, OCT controller 104 can be configured to detect and / or automatically segment the anterior cornea in an OCT image and generate and / or display virtual markers on the OCT image, thereby visually identifying the location of the anterior cornea and / or at least segmenting a portion of the anterior cornea. Similarly, OCT controller 104 can be configured to detect and / or automatically segment the Bowman's layer of the cornea in an OCT image and generate and / or display virtual markers on the OCT image, thereby visually identifying and / or segmenting the location of the Bowman's layer of the cornea and / or the Bowman's layer of the cornea.

[0040] OCT controller 104 can be configured to generate and / or display virtual markers of various shapes and / or sizes. For example, OCT controller 104 can be configured to generate and / or display curved virtual markers (such as curves). OCT controller 104 can be configured to generate enhanced OCT images by generating and / or displaying virtual markers on OCT images (e.g., OCT images generated by OCT controller 104).

[0041] OCT controller 104 can be configured to display OCT images and / or enhanced OCT images to a user by providing images to display 106. OCT controller 104 can be communicatively coupled and / or electrically connected to display 106. Display 106 can be configured according to one or more display standards and can be any type of display, such as Video Graphics Array (VGA), Extended Graphics Array (XGA), Digital Visual Interface (DVI), High Definition Multimedia Interface (HDMI), etc.

[0042] OCT controller 104 can be configured to use segmented anterior corneal and segmented Bowman's layer data to generate an epithelial map. In some embodiments, OCT controller 104 can be configured to store the location data of the segmented anterior corneal and / or Bowman's layer from the OCT image in a storage unit (not shown separately) communicatively coupled to OCT controller 104. In some embodiments, OCT controller 104 can be configured to transmit the location data and / or image data of the segmented anterior corneal and segmented Bowman's layer to an epithelial map generation module (not shown). In some embodiments, the epithelial map generation module can be located in a system remote from imaging system 100. For example, in some embodiments, the epithelial map generation module can be located in a computing device (e.g., a server computer) communicatively coupled to imaging system 100, and OCT controller 104 can be configured to transmit the location data and / or image data of the segmented anterior corneal and segmented Bowman's layer to a communicatively coupled server.

[0043] Figure 2 A block diagram illustrating selected components of an implementation of the OCT controller, such as the reference above. Figure 1 The OCT controller 104 is described. (e.g.) Figure 2 As shown, the OCT controller 104 includes a processor 201, a bus 202, a display interface 204, a memory 210, and a communication interface 220.

[0044] Processor 201 is communicatively coupled to memory 210, display interface 204, and communication interface 220 via bus 202. OCT controller 104 can be configured to interface with various external components of an imaging system (e.g., imaging system 100), such as OCT scanner 102 and display 106, via processor 201 and communication interface 220. In some embodiments, communication interface 220 is configured to enable OCT controller 104 to connect to a network (not shown). In some embodiments, OCT controller 104 is connected to one or more displays, such as display 106, via display interface 204.

[0045] Memory 210 may include persistent, volatile, fixed, removable, magnetic, and / or semiconductor media. Memory 210 is configured to store one or more machine-readable commands, instructions, data, and / or the like. In some embodiments, such as Figure 2 As shown, memory 210 includes one or more instruction sets and / or instruction sequences, such as operating system 212, scan control application 214, etc. Examples of operating system 212 may include, but are not limited to, UNIX or UNIX-like operating systems. The operating system can be a series of operating systems or another suitable operating system. The scan control application 214 can be configured to perform the OCT controller operations described herein, including but not limited to operations related to initiating eye scans, generating OCT images, processing OCT images, generating and / or displaying virtual markers on OCT images, and generating enhanced OCT images.

[0046] Figure 3 An example flowchart illustrating the segmentation of the corneal epithelium in an OCT image according to an illustrative embodiment of this disclosure is shown. Operation 300 can be performed, for example, by an OCT controller (e.g., OCT controller 104 of imaging system 100). Operation 300 can be implemented as a software component that executes on one or more processors (e.g., processor 201).

[0047] Operation 300 may begin with operation 302, in which an OCT image (e.g., a corneal OCT image) corresponding to the scanned cornea of ​​the eye is generated and / or received by the OCT controller 104. As described above, the OCT controller 104 may be configured to generate the corneal OCT image based on data received from the OCT scanner 102. In some embodiments, the OCT controller 104 may be configured to receive the corneal OCT image from another component of the imaging system 100. For example, the imaging system 100 may include an image generator communicatively coupled to the OCT scanner 102, and this image generator may be configured to generate the corneal OCT image based on the scan data from the OCT scanner 102. In some embodiments, the image generator and / or the OCT controller 104 may be configured to generate a 4K corneal OCT image including the cornea based on the scan data from the OCT scanner 102.

[0048] Figure 4A An example of a corneal OCT image generated by the image generator or OCT controller 104 of the image system 100 is shown. Figure 4A The OCT image shown includes the cornea of ​​the eye. (As...) Figure 4A As shown, OCT images include speckle noise in certain portions. Return Figure 3 In operation 304, the corneal OCT image is binarized to generate a corneal mask. The corneal OCT image is binarized to determine the location and / or portion of the cornea in the corneal OCT image. Figure 4B An example of a binarized corneal OCT image representing an initial binary mask of the cornea is shown. Figure 4B As shown, the portion of the OCT image showing the cornea is displayed in a brighter color at a prominent position in the binarized OCT image, while the portion of the OCT image that does not show the cornea (e.g., the background of the image) is displayed in a darker color. Figure 4BThe background of the binarized image is shown.

[0049] In operation 306, Figure 4B The initial mask shown is further adjusted and / or refined to generate a final mask of the cornea in the OCT image, which accurately captures the shape and size of the cornea in the OCT image. In some embodiments, the final mask can be generated by iteratively adjusting the initial mask of the cornea using one or more segmentation and / or extraction algorithms (such as the Chan-Vese active contouring algorithm). In some embodiments, the final mask of the cornea can be generated by shrinking (e.g., inward) or expanding (e.g., outward) a significant portion of the initial mask of the cornea. For example, the OCT controller 104 can iteratively apply the Chan-Vese active contouring algorithm to the initial mask generated in operation 304 until a termination condition is met. While the Chan-Vese active contouring algorithm is applied iteratively, the OCT controller 104 can shrink (e.g., inward) or expand (e.g., outward) a significant portion of the mask representing the corneal region. Examples of termination conditions include, but are not limited to, a predetermined number of iterations, a predetermined number of expansions and / or shrinkages of the corneal region, etc.

[0050] OCT controller 104 can be configured to identify the upper boundary of the final mask of the cornea as the upper boundary of the anterior cornea of ​​the scanned eye. In operation 308, OCT controller 104 applies the final mask of the cornea to the corneal OCT image to segment the portion of the corneal OCT image corresponding to the upper boundary of the final mask as the anterior cornea. OCT controller 104 can store the location information of the detected and segmented anterior cornea in a storage unit communicatively coupled to OCT controller 104. In some embodiments, OCT controller 104 can display virtual indicators corresponding to the segmented anterior cornea on the OCT image, such as... Figure 4C As shown. In Figure 4C In the image, the virtual indicator 420 displays the segmented anterior cornea.

[0051] return Figure 3 In operation 310, the OCT controller 104 generates an initial mask of the corneal epithelium of the scanned eye. The OCT controller 104 can be configured to generate the initial mask of the corneal epithelium based on the location of the segmented anterior cornea and the determined location of Bowman's layer. The OCT controller 104 can be configured to determine the location of Bowman's layer based on a predetermined thickness of the corneal epithelium and / or a predetermined distance from the anterior cornea that Bowman's layer is expected to be at. Figure 4D An example of an initial mask for scanning the corneal epithelium of the eye is shown.

[0052] In step 312, the OCT controller 104 generates a final mask of the corneal epithelium. The OCT controller 104 can be configured to generate the final mask of the corneal epithelium based on a corneal OCT image, an initial mask, and / or one or more segmentation and / or extraction algorithms (such as the Chan-Vese active contouring algorithm). In some embodiments, the OCT controller 104 can generate the final mask of the corneal epithelium by iteratively shrinking and / or expanding the portion of the initial mask representing the corneal epithelium. In some embodiments, similar to the method of generating the final mask of the cornea, the OCT controller 104 can generate the final mask of the corneal epithelium by iteratively shrinking and / or expanding the portion of the mask representing the corneal epithelium by applying the Chan-Vese active contouring algorithm to the initial mask of the corneal epithelium. The OCT controller 104 can be configured to apply the Chan-Vese active contouring algorithm until terminal conditions are met.

[0053] In operation 314, the OCT controller 104 segments Bowman's layer in the OCT image based on the final mask of the corneal epithelium. The OCT controller 104 can be configured to segment the upper boundary of the final mask as the boundary of Bowman's layer or the boundary of Bowman's layer. In some embodiments, the OCT controller 104 can be configured to segment the upper boundary of the final mask of the corneal epithelium as the bottom boundary of Bowman's layer. The OCT controller 104 can be configured to display a virtual indicator (such as virtual indicator 440) on the OCT image that identifies the boundary of Bowman's layer (e.g., the bottom boundary of Bowman's layer). Figure 4E As shown.

[0054] The OCT controller 104 can display virtual indicators (e.g., 420 and 440) on the OCT image to identify the segmented anterior cornea and the segmented Bowman's layer, such as... Figure 4F As shown, the OCT controller 104 can be configured to segment the portion of the cornea between the segmented anterior cornea and the segmented layer of Bowman's cornea in the generated and / or received OCT images as the corneal epithelium.

[0055] Therefore, the technique described in this paper for binarizing corneal OCT images and generating masks to accurately identify and segment the anterior cornea and Bowman's layer allows for accurate segmentation of the corneal epithelium without generating or training a large dataset of image data including the cornea and its sublayers. Thus, the technique described in this paper improves the accuracy of image segmentation systems while significantly reducing the computational, data, and time resources used.

[0056] Figure 5A flowchart illustrating an example method for segmenting the corneal epithelium in an ophthalmic image according to an illustrative embodiment of this disclosure is provided. Operation 500 may be performed, for example, by an OCT controller (e.g., OCT controller 104 of imaging system 100). Operation 500 may be implemented as a software component that executes and runs on one or more processors (e.g., processor 201).

[0057] Operation 500 may begin at 502, in which the OCT controller 104 receives an OCT image of the eye. As described above, in some embodiments, the OCT controller 104 may generate an OCT image of the eye based on scan data from the OCT scanner 102, and in some embodiments, the OCT controller 104 may receive an OCT image from an image generation component of an imaging system (e.g., imaging system 100). At 504, the OCT controller 104 generates a binarized image of the eye based on the OCT image. At 506, the OCT controller 104 generates a binary mask of the cornea based on the binarized image of the eye and the OCT image. At 508, the OCT controller 104 segments the anterior cornea of ​​the eye on the OCT image based on the binary mask of the cornea. At 510, the OCT controller 104 generates a binary mask of the ocular epithelium based on the OCT image and the segmented anterior cornea. At 512, the OCT controller 104 segments the layer of Bowman's cornea in the OCT image based on a binary mask of the eye's epithelium. At 514, the OCT controller 104 uses the segmented anterior cornea and segmented layer of Bowman's data to generate an epithelial map. In some embodiments, as described above, the OCT controller 104 may send the positional data and / or image data of the segmented anterior cornea and segmented layer of Bowman's data in the OCT image to a computing module and / or a device configured to generate an epithelial map, in order to generate an epithelial map or a corneal epithelial map.

[0058] In some embodiments, the OCT controller 104 identifies the location of the cornea based on an OCT image of the eye. In some embodiments, the OCT controller 104 generates a binary mask of the cornea by iteratively adjusting the salient portions of a binary image of the eye. In some embodiments, the OCT controller 104 iteratively adjusts the salient portions of a binary image of the eye by iteratively shrinking or expanding them.

[0059] In some embodiments, the segmented anterior cornea is the upper boundary of a binary mask of the eye's cornea. In some embodiments, the OCT controller 104 generates a binary mask of the eye's epithelial layer by identifying the location of the epithelial layer based on the position of the segmented anterior cornea on the OCT image of the eye and a predetermined epithelial thickness. In some embodiments, the location of the binary mask of the epithelial layer is based on the location of the epithelial layer. In some embodiments, the segmented Bowman's layer is the upper boundary of the binary mask of the epithelial layer.

[0060] While the foregoing embodiments described herein are directed to, other and further embodiments of the present disclosure may be devised without departing from its essential scope, the scope of which is defined by the following claims.

Claims

1. A method for segmenting the corneal epithelium in an optical coherence tomography (OCT) image of the eye, the method comprising: Receive the OCT image of the eye; Generate a binarized image of the eye based on the OCT image; A binary mask of the cornea of ​​the eye is generated based on the binarized image of the eye and the OCT image; The anterior cornea of ​​the eye is segmented on the OCT image based on a binary mask of the cornea of ​​the eye; A binary mask of the eye's epithelial layer is generated using the OCT image and the segmented anterior cornea. Using a binary mask of the eye's epithelial layer, the Bowman's layer in the cornea of ​​the eye is segmented on the OCT image; as well as The segmented anterior cornea and segmented Bowman's layer data are used to generate an epithelial map. The binary mask used to generate the cornea of ​​the eye further includes iteratively adjusting the salient portions of the binarized image of the eye.

2. The method of claim 1, further comprising: The location of the cornea of ​​the eye is identified based on the OCT image of the eye.

3. The method of claim 1, wherein iteratively adjusting the salient portion of the binarized image of the eye further comprises: Iteratively shrink or expand the salient portion of the binarized image of the eye.

4. The method of claim 1, wherein the segmented anterior cornea is the upper boundary of a binary mask of the cornea of ​​the eye.

5. The method of claim 1, wherein generating the binary mask for the epithelial layer of the eye further comprises: The location of the epithelial layer is identified based on the position of the segmented anterior cornea on the OCT image of the eye and the predetermined epithelial thickness.

6. The method of claim 5, wherein the position of the binary mask of the epithelial layer is based on the position of the epithelial layer.

7. The method of claim 1, wherein the segmented Bowman layer is the upper boundary of the binary mask of the epithelial layer.

8. An imaging system, the imaging system comprising: The memory includes computer-executable instructions; A processor configured to execute computer-executable instructions and enable the imaging system to: Generate optical coherence tomography (OCT) images of the eye; Generate a binarized image of the eye based on the OCT image; A binary mask of the cornea of ​​the eye is generated based on the binarized image of the eye and the OCT image; The anterior cornea of ​​the eye is segmented on the OCT image based on a binary mask of the cornea of ​​the eye; A binary mask of the eye's epithelial layer is generated using the OCT image and the segmented anterior cornea. Using a binary mask of the eye's epithelial layer, the Bowman's layer in the cornea of ​​the eye is segmented on the OCT image; as well as The segmented anterior cornea and segmented Bowman's layer data are used to generate an epithelial map. The processor is further configured to generate a binary mask of the cornea of ​​the eye by iteratively adjusting salient portions of a binarized image of the eye.

9. The imaging system of claim 8, wherein, The processor is further configured to enable the imaging system to: The location of the cornea of ​​the eye is identified based on the OCT image of the eye.

10. The imaging system of claim 8, wherein, The processor is further configured to iteratively adjust the salient portion of the binarized image of the eye by iteratively shrinking or expanding the salient portion of the binarized image of the eye.

11. The imaging system of claim 8, wherein, The segmented anterior cornea is the upper boundary of the binary mask of the eye's cornea.

12. The imaging system of claim 8, wherein, The processor is further configured to generate a binary mask of the eye's epithelial layer by identifying the position of the epithelial layer based on the position of the segmented anterior cornea on the OCT image of the eye and a predetermined epithelial thickness.

13. The imaging system of claim 12, wherein, The position of the binary mask for the epithelial layer is based on the position of the epithelial layer.

Images