Optic disc morphology quantification method, apparatus, medium, and device

Abstract

The application discloses a disciform shape quantification method, device, medium and equipment, relates to the technical field of image processing, and the method comprises the steps of identifying and cutting out a to-be-detected image in a target fundus image, inputting the to-be-detected image into a preset preliminary segmentation model, obtaining a disc preliminary segmentation image, a myopic arc preliminary segmentation image and a disc and myopic arc confusion area image, inputting the confusion area image into a preset confusion area segmentation model, obtaining a disc correction image and a myopic arc correction image, fusing the disc preliminary segmentation image and the disc correction image, obtaining a target disc image, fusing the myopic arc preliminary segmentation image and the myopic arc correction image, obtaining a target myopic arc image, calculating a disc shape quantification value based on the high-precision target disc image and the target myopic arc image, improving the accuracy of disc shape quantification, only performing data processing based on the to-be-detected image, avoiding global calculation based on the entire image, and also improving the accuracy of disc shape quantification.

Description

Technical Field

[0001] This invention relates to image processing technology and the field of digital medicine, and in particular to a method, apparatus, medium and device for quantifying optic disc morphology. Background Technology

[0002] Moderate to high myopia can affect people's work and daily life. Therefore, patients with moderate to high myopia need to have their eye health checked regularly to prevent or reduce the progression of myopia. Because the optic disc morphology in the fundus of patients with moderate to high myopia undergoes a series of changes, including changes in disc size, ellipticity, disc tilt, and the appearance of myopic arcs around the disc, these changes can be monitored through regular fundus photography. By detecting and processing medical images, these changes in optic disc morphology can be identified, thereby providing insight into the overall health of the eyes.

[0003] Currently, there are two main methods for quantifying optic disc morphology in the fundus: manual measurement and intelligent measurement. Manual measurement involves professionals manually interpreting images or using software to measure the optic disc morphology, which is inefficient and makes it difficult to accurately measure changes in the optic disc morphology. Intelligent measurement uses computer vision to perform image semantic segmentation on fundus color images, extracting the optic disc and its surrounding myopic arc structure to obtain quantified values ​​of optic disc morphology. However, existing intelligent measurement methods perform global calculations based on the entire image during image semantic segmentation. This method involves a large amount of redundant computation. This redundant computation not only leads to low detection efficiency but also causes parameter confusion due to the introduction of numerous parameters, resulting in low accuracy in optic disc morphology quantification. Summary of the Invention

[0004] In view of this, the present invention provides a method, apparatus, medium and device for quantifying optic disc morphology, the main purpose of which is to solve the problem of low accuracy in optic disc morphology quantification using fundus color ultrasound images.

[0005] According to one aspect of this application, a method for quantifying the shape of a viewing disc is provided, the method comprising:

[0006] The image to be detected is identified and cropped from the target fundus image, and the image to be detected is input into a preset preliminary segmentation model to obtain a preliminary segmentation image of the optic disc and a preliminary segmentation image of the myopic arc.

[0007] Based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc, an image of the confusion region between the optic disc and the myopic arc is obtained. The image of the confusion region is then input into a preset confusion region segmentation model to obtain a corrected image of the optic disc and a corrected image of the myopic arc.

[0008] The preliminary segmented image of the optic disc and the corrected image of the optic disc are fused to obtain a target optic disc image. The preliminary segmented image of the myopic arc and the corrected image of the myopic arc are fused to obtain a target myopic arc image. Based on the target optic disc image and the target myopic arc image, the optic disc morphology quantization value is calculated.

[0009] Optionally, identifying and cropping the image to be detected from the target fundus image includes:

[0010] Heatmaps were generated from the target fundus image to obtain a fundus heatmap.

[0011] Based on the thermal data values ​​in the fundus thermal map, fundus regions that meet the preset fundus thermal value conditions are determined;

[0012] The image to be detected is cropped from the target fundus image to match the fundus region.

[0013] Optionally, the step of performing heatmap rendering on the target fundus image to obtain a fundus heatmap includes:

[0014] Perform grayscale transformation on the target fundus image to obtain a fundus grayscale image;

[0015] Based on the grayscale value of each pixel in the fundus grayscale image, read the corresponding thermal data value of each pixel from a preset color mapping table;

[0016] A fundus thermal map is generated based on the thermal data value corresponding to each pixel.

[0017] Optionally, obtaining an image of the confused region between the optic disc and the myopic arc based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc includes:

[0018] Find the intersection region between the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc;

[0019] Obtain the minimum bounding rectangle of the intersection region, and enlarge the minimum bounding rectangle by a preset magnification ratio with the center point of the minimum bounding rectangle as the center to obtain the magnified rectangular region;

[0020] The image corresponding to the magnified rectangular region in the image to be detected is obtained as the image of the confusion area between the optic disc and the myopic arc.

[0021] Optionally, fusing the preliminary segmentation image of the optic disc and the corrected image of the optic disc to obtain a target optic disc image, and fusing the preliminary segmentation image of the myopia arc and the corrected image of the myopia arc to obtain a target myopia arc image, includes:

[0022] In the preliminary segmentation image of the visual disc, a first region corresponding to the corrected image of the visual disc is obtained, and the first region in the preliminary segmentation image of the visual disc is replaced with the corrected image of the visual disc to obtain the target visual disc image.

[0023] In the preliminary segmentation image of the myopia arc, a second region corresponding to the myopia arc correction image is obtained, and the second region in the preliminary segmentation image of the myopia arc is replaced with the myopia arc correction image to obtain the target myopia arc image.

[0024] Optionally, the optic disc morphology quantification value includes the area ratio of the myopic arc to the optic disc, the optic disc ellipticity, and the optic disc tilt. The calculation of the optic disc morphology quantification value based on the target optic disc image and the target myopic arc image includes:

[0025] Based on the target optic disc image, obtain the optic disc pixel area, the longest diameter of the optic disc, the shortest diameter of the optic disc, and the angle between the longest diameter of the optic disc and the vertical axis of the coordinate system where the target fundus image is located; based on the target myopia arc image, obtain the myopia arc pixel area.

[0026] The ellipticity of the optic disc is calculated based on the shortest diameter and the longest diameter of the optic disc. The tilt of the optic disc is calculated based on the angle between the longest diameter of the optic disc and the vertical axis of the coordinate system containing the target fundus image, and the longest diameter of the optic disc.

[0027] The area ratio of the myopic arc to the optic disc is calculated based on the pixel area of ​​the myopic arc and the pixel area of ​​the optic disc.

[0028] Optionally, before inputting the image to be detected into a preset preliminary segmentation model, the method further includes:

[0029] Outline the optic disc and myopic arc in the fundus image sample;

[0030] Based on the optic disc contour and the myopic arc contour, the portion of the fundus image sample outside the optic disc and the myopic arc is masked, and the optic disc and the myopic arc in the masked fundus image sample are labeled.

[0031] The initial segmentation model was trained using labeled fundus image samples.

[0032] According to another aspect of this application, a disc morphology quantization device is provided, comprising:

[0033] The preliminary segmentation module is used to identify and crop out the image to be detected from the target fundus image, and input the image to be detected into the preset preliminary segmentation model to obtain the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc.

[0034] The confusion region segmentation module is used to obtain a confusion region image between the optic disc and the myopic arc based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc, and input the confusion region image into a preset confusion region segmentation model to obtain a corrected image of the optic disc and a corrected image of the myopic arc.

[0035] The optic disc morphology quantization value calculation module is used to fuse the preliminary segmented image of the optic disc and the corrected image of the optic disc to obtain a target optic disc image, fuse the preliminary segmented image of the myopia arc and the corrected image of the myopia arc to obtain a target myopia arc image, and calculate the optic disc morphology quantization value based on the target optic disc image and the target myopia arc image.

[0036] Optionally, identifying and cropping the image to be detected from the target fundus image includes:

[0037] Heatmaps were generated from the target fundus image to obtain a fundus heatmap.

[0038] Based on the thermal data values ​​in the fundus thermal map, fundus regions that meet the preset fundus thermal value conditions are determined;

[0039] The image to be detected is cropped from the target fundus image to match the fundus region.

[0040] Optionally, the step of performing heatmap rendering on the target fundus image to obtain a fundus heatmap includes:

[0041] Perform grayscale transformation on the target fundus image to obtain a fundus grayscale image;

[0042] Based on the grayscale value of each pixel in the fundus grayscale image, read the corresponding thermal data value of each pixel from a preset color mapping table;

[0043] A fundus thermal map is generated based on the thermal data value corresponding to each pixel.

[0044] Optionally, obtaining an image of the confused region between the optic disc and the myopic arc based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc includes:

[0045] Find the intersection region between the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc;

[0046] Obtain the minimum bounding rectangle of the intersection region, and enlarge the minimum bounding rectangle by a preset magnification ratio with the center point of the minimum bounding rectangle as the center to obtain the magnified rectangular region;

[0047] The image corresponding to the magnified rectangular region in the image to be detected is obtained as the image of the confusion area between the optic disc and the myopic arc.

[0048] Optionally, fusing the preliminary segmentation image of the optic disc and the corrected image of the optic disc to obtain a target optic disc image, and fusing the preliminary segmentation image of the myopia arc and the corrected image of the myopia arc to obtain a target myopia arc image, includes:

[0049] In the preliminary segmentation image of the visual disc, a first region corresponding to the corrected image of the visual disc is obtained, and the first region in the preliminary segmentation image of the visual disc is replaced with the corrected image of the visual disc to obtain the target visual disc image.

[0050] In the preliminary segmentation image of the myopia arc, a second region corresponding to the myopia arc correction image is obtained, and the second region in the preliminary segmentation image of the myopia arc is replaced with the myopia arc correction image to obtain the target myopia arc image.

[0051] Optionally, the optic disc morphology quantification value includes the area ratio of the myopic arc to the optic disc, the optic disc ellipticity, and the optic disc tilt. The calculation of the optic disc morphology quantification value based on the target optic disc image and the target myopic arc image includes:

[0052] Based on the target optic disc image, obtain the optic disc pixel area, the longest diameter of the optic disc, the shortest diameter of the optic disc, and the angle between the longest diameter of the optic disc and the vertical axis of the coordinate system where the target fundus image is located; based on the target myopia arc image, obtain the myopia arc pixel area.

[0053] The ellipticity of the optic disc is calculated based on the shortest diameter and the longest diameter of the optic disc. The tilt of the optic disc is calculated based on the angle between the longest diameter of the optic disc and the vertical axis of the coordinate system containing the target fundus image, and the longest diameter of the optic disc.

[0054] The area ratio of the myopic arc to the optic disc is calculated based on the pixel area of ​​the myopic arc and the pixel area of ​​the optic disc.

[0055] Optionally, the disc morphology quantization device further includes:

[0056] The outlining module is used to outline the optic disc and myopic arc in fundus image samples.

[0057] The annotation module is used to mask the portion of the fundus image sample other than the optic disc and the myopic arc based on the optic disc contour and the myopic arc contour, and to annotate the optic disc and the myopic arc in the masked fundus image sample.

[0058] The training module is used to train a preliminary segmentation model using labeled fundus image samples.

[0059] According to another aspect of this application, a storage medium is provided that stores at least one executable instruction, which causes a processor to perform the operation corresponding to the above-described disc morphology quantization method.

[0060] According to another aspect of this application, a computer device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus;

[0061] The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described disc morphology quantization method.

[0062] By employing the above-described technical solutions, the technical solutions provided by the embodiments of the present invention have at least the following advantages:

[0063] This application provides a method, apparatus, device, and medium for optic disc morphology quantification. It acquires a target image from a target fundus image, obtains preliminary optic disc segmentation images and preliminary myopic arc segmentation images based on the target image, identifies the confusion area between the optic disc and the myopic arc based on these images, and obtains corrected optic disc and myopic arc images based on the confusion area. This achieves accurate segmentation of the confusion area, resulting in high-precision corrected optic disc and myopic arc images. The preliminary optic disc segmentation images and corrected images are then fused to obtain a target optic disc image, and the preliminary myopic arc segmentation images and corrected images are further fused to obtain a target myopic arc image. This yields high-precision target optic disc and target myopic arc images. Based on these high-precision images, optic disc morphology quantification values ​​are calculated, improving the accuracy of optic disc morphology quantification. Data processing is performed only on the target image, avoiding global calculations based on the entire image. The calculation process is simple and accurate, saving computation time and improving the efficiency and accuracy of optic disc morphology quantification.

[0064] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0065] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0066] Figure 1 A flowchart of a method for quantifying the shape of a viewing disc provided in an embodiment of this application is shown;

[0067] Figure 2A flowchart of another method for quantifying the morphology of a viewing disc provided in an embodiment of this application is shown;

[0068] Figure 3 A flowchart of another method for quantifying the shape of a viewing disk provided in an embodiment of this application is shown;

[0069] Figure 4 A flowchart of another method for quantifying the shape of a viewing disk provided in an embodiment of this application is shown;

[0070] Figure 5 A flowchart of another method for quantifying the morphology of a viewing disc provided in an embodiment of this application is shown;

[0071] Figure 6 A flowchart of another method for quantifying the shape of a viewing disk provided in an embodiment of this application is shown;

[0072] Figure 7 This paper shows a block diagram of a visual disc morphology quantization device provided in an embodiment of this application;

[0073] Figure 8 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention is shown.

[0074] in,

[0075] Figure 7 In the middle: 702 - Preliminary segmentation module; 704 - Confusion area segmentation module; 706 - Visual disk morphology quantification value calculation module;

[0076] Figure 8 In Chinese: 802 - Processor; 804 - Communication interface; 806 - Memory; 808 - Communication bus; 810 - Program. Detailed Implementation

[0077] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present invention can be combined with each other.

[0078] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the specific embodiments, structures, features, and effects according to the present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "an embodiment" or "an embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0079] To address the low accuracy of optic disc morphology quantification using fundus ultrasound images, this application provides a method for optic disc morphology quantification, such as... Figure 1 As shown, the method includes:

[0080] 102: Identify and crop the image to be detected from the target fundus image, input the image to be detected into the preset preliminary segmentation model, and obtain the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc;

[0081] In this embodiment, the target fundus image is a medical image, such as a fundus ultrasound image. The image to be detected is an image in the target fundus image that only contains the optic disc. The cropped image to be detected is uniformly sized to 256*256 using bilinear interpolation. The adjusted image to be detected is then input into a preset preliminary segmentation model to obtain preliminary results for two types of segmented targets (optic disc and myopic arc).

[0082] The preset initial segmentation model training uses a weighted sum of the cross-union model loss (Dice loss) and cross-entropy loss (BCE loss) as the loss function, adopts the Adam optimizer, and has an initial learning rate of 0.001. After 200 rounds of training, the model with the lowest loss is selected as the optimal model.

[0083] In another embodiment of the invention, for further definition and explanation, such as Figure 2 As shown, the process of identifying and cropping the image to be detected from the target fundus image includes:

[0084] 202: Perform heat map drawing on the target fundus image to obtain the fundus heat map;

[0085] 204: Based on the thermal data values ​​in the fundus thermal map, determine the fundus regions that meet the preset fundus thermal value conditions;

[0086] 206: Cropping the image to be detected from the target fundus image that matches the fundus region.

[0087] In this embodiment, the target fundus image is preprocessed before being converted into a fundus heatmap. Preprocessing first converts the target fundus image into a uniform image format, for example, adjusting the image resolution to 800*800. Then, a sliding window is used for mean filtering, for example, a 33*33 sliding window. The preprocessed fundus image is then converted into a fundus heatmap. Based on the heatmap data values, the region corresponding to the lowest heatmap data value is obtained. In the heatmap, the blue region corresponds to the lowest heatmap data value. In the preprocessed fundus image, the region corresponding to the lowest heatmap data value is determined as the region to be detected. The region to be detected, i.e., the image containing the optic disc, is then cropped from the preprocessed fundus image. The approximate area containing the optic disc is pre-cropped, and a preset optic disc structure segmentation model (i.e., a preliminary segmentation model) is established based on the cropped area to avoid generating a large number of invalid model parameters. Training a semantic segmentation model based on the optic disc region is equivalent to specifically enlarging the optic disc region. Therefore, the resolution of the optic disc region used for deep learning model analysis is improved, and the optic disc and myopic arc structure can be segmented more accurately.

[0088] In another embodiment of the invention, for further definition and explanation, such as Figure 3 As shown, a heatmap is created from the target fundus image to obtain the fundus heatmap, including:

[0089] 302: Perform grayscale transformation on the target fundus image to obtain a grayscale fundus image;

[0090] 304: Based on the grayscale value of each pixel in the fundus grayscale image, read the corresponding thermal data value of each pixel from the preset color mapping table;

[0091] 306: Generate a fundus thermal map based on the thermal data value corresponding to each pixel.

[0092] In this embodiment, the target fundus image is first converted into a grayscale image, and then the thermal data value corresponding to each pixel in the grayscale image is obtained according to the JET color mapping table. Based on the thermal data value of each pixel, a fundus thermal map is generated.

[0093] 104: Based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc, obtain the confusion region image of the optic disc and the myopic arc, input the confusion region image into the preset confusion region segmentation model, and obtain the optic disc correction image and the myopic arc correction image;

[0094] In this embodiment, the intersection of the segmented myopic arc and the optic disc (i.e., the region that is prone to inter-class confusion) is calculated, which is the confusion region image of the optic disc and the myopic arc. The confusion region image of the optic disc and the myopic arc is cropped and interpolated to a fixed size. The adjusted confusion region image of the optic disc and the myopic arc is input into a lightweight preset confusion region segmentation model to obtain the optic disc correction image and the myopic arc correction image.

[0095] The preset confusing region segmentation model uses a weighted sum of the cross-union model loss (Dice loss) and cross-entropy loss (BCE loss) as the loss function during training, employs the Adam optimizer, has an initial learning rate of 0.001, and selects the model with the lowest loss as the optimal model after 200 training rounds.

[0096] This application provides a method for quantifying optic disc morphology. Compared with the prior art, it employs two image segmentation models: a preset preliminary segmentation model and a preset confusion region segmentation model. The preset preliminary segmentation model first segments out the main structures of the optic disc and myopic arc, while the preset confusion region segmentation model performs secondary segmentation on the areas where the optic disc and myopic arc are confused, thereby correcting the details of the optic disc morphology, achieving accurate segmentation, and improving the accuracy of optic disc and myopic arc segmentation.

[0097] In another embodiment of the present invention, for further definition and explanation, such as... Figure 4 As shown, based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc, an image of the region of confusion between the optic disc and the myopic arc is obtained, including:

[0098] 402: Find the intersection region between the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc;

[0099] 404: Obtain the minimum bounding rectangle of the intersection region, and enlarge the minimum bounding rectangle by a preset magnification ratio with the center point of the minimum bounding rectangle as the center to obtain the magnified rectangular region;

[0100] 406: Obtain the image corresponding to the magnified rectangular region in the image to be detected, as the image of the confusion area between the optic disc and the myopic arc.

[0101] In this embodiment, the intersection of the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc is first performed to obtain the intersection region. Then, the minimum bounding rectangle of the intersection region is calculated. The minimum bounding rectangle is enlarged by a preset magnification ratio, for example, by expanding the length and width of the minimum bounding rectangle by five times. The image corresponding to the magnified rectangular region is obtained in the image to be detected, which is the image of the confusion region between the optic disc and the myopic arc.

[0102] In one embodiment, the magnified rectangular area can also be used as an image of the area where the visual disc and the myopic arc are confused.

[0103] 106: Fuse the preliminary segmentation image and the corrected image of the optic disc to obtain the target optic disc image. Fuse the preliminary segmentation image and the corrected image of the myopia arc to obtain the target myopia arc image. Based on the target optic disc image and the target myopia arc image, calculate the morphological quantization value of the optic disc.

[0104] In this embodiment, the preliminary segmented image and the corrected image of the optic disc are fused to obtain the final refined target optic disc image, and the preliminary segmented image and the corrected image of the myopic arc are fused to obtain the final refined target myopic arc image. Based on the refined target optic disc image and the target myopic arc image, the optic disc morphology quantization value is calculated, and the calculation accuracy is high.

[0105] In another embodiment of the present invention, for further definition and explanation, the preliminary optic disc segmentation image and the corrected optic disc image are fused to obtain a target optic disc image, and the preliminary myopia arc segmentation image and the corrected myopia arc image are fused to obtain a target myopia arc image, including:

[0106] In the preliminary segmentation image of the visual disk, the first region corresponding to the corrected image of the visual disk is obtained, and the first region in the preliminary segmentation image of the visual disk is replaced with the corrected image of the visual disk to obtain the target visual disk image.

[0107] In the preliminary segmentation image of the myopic arc, the second region corresponding to the myopic arc correction image is obtained, and the second region in the preliminary segmentation image of the myopic arc is replaced with the myopic arc correction image to obtain the target myopic arc image.

[0108] In this embodiment, the optic disc correction image is adjusted to the same aspect ratio as the preliminary optic disc segmentation image. A first region corresponding to the optic disc correction image is determined in the preliminary optic disc segmentation image, and the first region in the preliminary optic disc segmentation image is replaced with the optic disc correction image to obtain the target optic disc image. Similarly, the myopia arc correction image is adjusted to the same aspect ratio as the preliminary myopia arc segmentation image. A second region corresponding to the myopia arc correction image is determined in the preliminary myopia arc segmentation image, and the second region in the preliminary myopia arc segmentation image is replaced with the myopia arc correction image to obtain the target myopia arc image.

[0109] In another embodiment of the invention, for further definition and explanation, such as Figure 5 As shown, the optic disc morphology quantification values ​​include the area ratio of the myopic arc to the optic disc, the optic disc ellipticity, and the optic disc tilt. Based on the target optic disc image and the target myopic arc image, the optic disc morphology quantification values ​​are calculated, including:

[0110] 502: Based on the target optic disc image, obtain the optic disc pixel area, the longest diameter of the optic disc, the shortest diameter of the optic disc, and the angle between the longest diameter of the optic disc and the vertical axis of the coordinate system where the target fundus image is located; based on the target myopic arc image, obtain the myopic arc pixel area.

[0111] 504: Calculate the ovality of the optic disc based on its shortest and longest diameters. Calculate the tilt of the optic disc based on the angle between the longest diameter of the optic disc and the vertical axis of the coordinate system containing the target fundus image, and the longest diameter of the optic disc.

[0112] 506: Calculate the area ratio of the myopic arc to the optic disc based on the pixel area of ​​the myopic arc and the pixel area of ​​the optic disc.

[0113] Specifically, the calculated optic disc morphology quantification values ​​include optic disc ellipticity, optic disc tilt, and the myopic arc / optic disc area ratio. Optic disc ellipticity is the ratio of the shortest to the longest diameter of the segmented optic disc; optic disc tilt is the angle between the longest diameter of the segmented optic disc and the vertical axis of the fundus image coordinate system; and the myopic arc / optic disc area ratio is the ratio of the pixel area of ​​the myopic arc segmented by the model to the pixel area of ​​the optic disc.

[0114] In another embodiment of the invention, for further definition and explanation, such as Figure 6 As shown, before inputting the image to be detected into the preset preliminary segmentation model, the method further includes:

[0115] 602: Outline the optic disc and myopic arc in the fundus image sample;

[0116] 604: Based on the optic disc contour and myopic arc contour, the part outside the optic disc and myopic arc in the fundus image sample is masked, and the optic disc and myopic arc in the masked fundus image sample are labeled.

[0117] 606: Training a preliminary segmentation model using labeled fundus image samples.

[0118] In this embodiment, multiple fundus ultrasound images are collected as fundus image samples. The optic disc and myopic arc data in each fundus image sample are preprocessed. During preprocessing, the target contours, i.e., the contours of the optic disc and myopic arc, are first delineated. Then, the regions of the fundus image sample where the delineated optic disc and myopic arc are removed are masked. The masked fundus image samples are then labeled to obtain labeled fundus image samples. The labeled fundus image samples are divided into a training set, a validation set, and a test set in a 6:2:2 ratio. The preliminary segmentation model is trained based on samples from the training set, validation set, and test set.

[0119] During model training, the weighted sum of the cross-union model loss (Dice loss) and cross-entropy loss (BCE loss) is used as the loss function. The Adam optimizer is employed with an initial learning rate of 0.001. After 200 training rounds, the model with the lowest loss is selected as the optimal model.

[0120] This application provides a method for optic disc morphology quantification. Compared with existing technologies, it acquires a target image from a fundus image, obtains a preliminary optic disc segmentation image and a preliminary myopic arc segmentation image based on the target image, identifies the confusion area between the optic disc and the myopic arc based on the preliminary optic disc segmentation image and the preliminary myopic arc segmentation image, obtains a corrected optic disc image and a corrected myopic arc image based on the confusion area, achieving accurate segmentation of the confusion area between the optic disc and the myopic arc, and obtaining high-precision corrected optic disc and myopic arc images. The preliminary optic disc segmentation image and the corrected optic disc image are then fused to obtain a target optic disc image, and the preliminary myopic arc segmentation image and the corrected myopic arc image are then fused to obtain a target myopic arc image, resulting in high-precision target optic disc and target myopic arc images. Based on the high-precision target optic disc and target myopic arc images, the optic disc morphology quantification value is calculated, improving the accuracy of optic disc morphology quantification. Data processing is performed only on the target image, avoiding global calculations based on the entire image. The calculation process is simple and accurate, while saving computation time, thus improving the efficiency and accuracy of optic disc morphology quantification.

[0121] Furthermore, as a response to the above Figure 1 The implementation of the method shown in this invention provides a device for quantizing the shape of a viewing disk, as described in this embodiment. Figure 7 As shown, the device includes:

[0122] The preliminary segmentation module 702 is used to identify and crop out the image to be detected in the target fundus image, input the image to be detected into the preset preliminary segmentation model, and obtain the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc.

[0123] The confusion region segmentation module 704 is used to obtain the confusion region image of the optic disc and the myopic arc based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc, and input the confusion region image into the preset confusion region segmentation model to obtain the optic disc correction image and the myopic arc correction image.

[0124] The optic disc morphology quantization value calculation module 706 is used to fuse the preliminary segmentation image of the optic disc and the corrected image of the optic disc to obtain the target optic disc image, fuse the preliminary segmentation image of the myopia arc and the corrected image of the myopia arc to obtain the target myopia arc image, and calculate the optic disc morphology quantization value based on the target optic disc image and the target myopia arc image.

[0125] This application provides an optic disc morphology quantification device. Compared with the prior art, it acquires a target image from a fundus image, obtains a preliminary optic disc segmentation image and a preliminary myopic arc segmentation image based on the target image, obtains the confusion area between the optic disc and the myopic arc based on the preliminary optic disc segmentation image and the preliminary myopic arc segmentation image, obtains a corrected optic disc image and a corrected myopic arc image based on the confusion area, achieves accurate segmentation of the confusion area between the optic disc and the myopic arc, and obtains high-precision corrected optic disc images and corrected myopic arc images. Then, it fuses the preliminary optic disc segmentation image and the corrected optic disc image to obtain a target optic disc image, and fuses the preliminary myopic arc segmentation image and the corrected myopic arc image to obtain a target myopic arc image, obtaining high-precision target optic disc images and target myopic arc images. Based on the high-precision target optic disc images and target myopic arc images, it calculates the optic disc morphology quantification value, which improves the accuracy of optic disc morphology quantification. It only performs data processing based on the target image, avoiding global calculation based on the entire image, making the calculation process simple and accurate, while saving calculation time and improving the efficiency and accuracy of optic disc morphology quantification.

[0126] In one embodiment, identifying and cropping the image to be detected from the target fundus image includes:

[0127] Heatmaps are generated from the target fundus image to obtain a fundus heatmap.

[0128] Based on the thermal data values ​​in the fundus thermal map, fundus regions that meet the preset fundus thermal value conditions are identified;

[0129] The image to be detected is cropped from the target fundus image to match the fundus region.

[0130] In one embodiment, performing heat mapping on a target fundus image to obtain a fundus heat map includes:

[0131] Perform grayscale transformation on the target fundus image to obtain a grayscale fundus image;

[0132] Based on the grayscale value of each pixel in the fundus grayscale image, read the corresponding thermal data value of each pixel from the preset color mapping table;

[0133] A fundus thermal map is generated based on the thermal data value corresponding to each pixel.

[0134] In one embodiment, based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc, an image of the region of confusion between the optic disc and the myopic arc is obtained, including:

[0135] Find the intersection region between the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc;

[0136] Obtain the minimum bounding rectangle of the intersection region, and enlarge the minimum bounding rectangle by a preset magnification ratio using the center point of the minimum bounding rectangle as the center to obtain the magnified rectangular region;

[0137] The image corresponding to the magnified rectangular region in the image to be detected is used as the image of the confusion area between the optic disc and the myopic arc.

[0138] In one embodiment, the preliminary optic disc segmentation image and the corrected optic disc image are fused to obtain a target optic disc image, and the preliminary myopia arc segmentation image and the corrected myopia arc image are fused to obtain a target myopia arc image, including:

[0139] In the preliminary segmentation image of the visual disk, the first region corresponding to the corrected image of the visual disk is obtained, and the first region in the preliminary segmentation image of the visual disk is replaced with the corrected image of the visual disk to obtain the target visual disk image.

[0140] In the preliminary segmentation image of the myopic arc, the second region corresponding to the myopic arc correction image is obtained, and the second region in the preliminary segmentation image of the myopic arc is replaced with the myopic arc correction image to obtain the target myopic arc image.

[0141] In one embodiment, the optic disc morphology quantification value includes the area ratio of the myopic arc to the optic disc, the optic disc ellipticity, and the optic disc tilt. Based on the target optic disc image and the target myopic arc image, the optic disc morphology quantification value is calculated, including:

[0142] Based on the target optic disc image, obtain the optic disc pixel area, the longest diameter of the optic disc, the shortest diameter of the optic disc, and the angle between the longest diameter of the optic disc and the vertical axis of the coordinate system where the target fundus image is located. Based on the target myopic arc image, obtain the myopic arc pixel area.

[0143] The ovality of the optic disc is calculated based on its shortest and longest diameters. The tilt of the optic disc is calculated based on the angle between the longest diameter of the optic disc and the vertical axis of the coordinate system containing the target fundus image, and the longest diameter of the optic disc.

[0144] Calculate the area ratio of the myopic arc to the optic disc based on the pixel area of ​​the myopic arc and the pixel area of ​​the optic disc.

[0145] In one embodiment, the disc morphology quantization device further includes:

[0146] The outlining module is used to outline the optic disc and myopic arc in fundus image samples.

[0147] The annotation module is used to mask the portion of the fundus image sample other than the optic disc and myopic arc based on the optic disc contour and myopic arc contour, and to annotate the optic disc and myopic arc in the masked fundus image sample.

[0148] The training module is used to train a preliminary segmentation model using labeled fundus image samples.

[0149] According to one embodiment of the present invention, a storage medium is provided, the storage medium storing at least one executable instruction, the computer executable instruction being able to execute the disc morphology quantization method in any of the above method embodiments.

[0150] Figure 8 The diagram illustrates a structural schematic of a computer device according to an embodiment of the present invention. The specific embodiments of the present invention do not limit the specific implementation of the computer device.

[0151] like Figure 8 As shown, the computer device may include: a processor 802, a communications interface 804, a memory 806, and a communications bus 808.

[0152] The processor 802, communication interface 804, and memory 806 communicate with each other via communication bus 808.

[0153] The communication interface 804 is used to communicate with other network elements such as clients or other servers.

[0154] The processor 802 is used to execute program 810, which can specifically execute the relevant steps in the above-described embodiment of the visual disc morphology quantization method.

[0155] Specifically, program 810 may include program code that includes computer operation instructions.

[0156] Processor 802 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The computer device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0157] Memory 806 is used to store program 810. Memory 806 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0158] Specifically, program 810 can be used to cause processor 802 to perform the following operations:

[0159] The image to be detected is identified and cropped from the target fundus image, and the image to be detected is input into the preset preliminary segmentation model to obtain the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc.

[0160] Based on the preliminary segmentation images of the optic disc and the myopic arc, the confusion region images of the optic disc and the myopic arc are obtained. The confusion region images are then input into the preset confusion region segmentation model to obtain the optic disc correction image and the myopic arc correction image.

[0161] The preliminary segmented image and the corrected image of the optic disc are fused to obtain the target optic disc image. The preliminary segmented image and the corrected image of the myopia arc are fused to obtain the target myopia arc image. Based on the target optic disc image and the target myopia arc image, the morphological quantization value of the optic disc is calculated.

[0162] It will be apparent to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. In one embodiment, they can be implemented using device-executable program code, thereby allowing them to be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular hardware and software combination.

[0163] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A method for quantifying the shape of a viewing disk, characterized in that, include: The image to be detected is identified and cropped from the target fundus image, and the image to be detected is input into a preset preliminary segmentation model to obtain a preliminary segmentation image of the optic disc and a preliminary segmentation image of the myopic arc. Based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc, an image of the confusion region between the optic disc and the myopic arc is obtained. The image of the confusion region is then input into a preset confusion region segmentation model to obtain a corrected image of the optic disc and a corrected image of the myopic arc. The preliminary segmented image of the optic disc and the corrected image of the optic disc are fused to obtain a target optic disc image. The preliminary segmented image of the myopic arc and the corrected image of the myopic arc are fused to obtain a target myopic arc image. Based on the target optic disc image and the target myopic arc image, the optic disc morphology quantification value is calculated, wherein the optic disc morphology quantification value includes the area ratio of the myopic arc to the optic disc, the ellipticity of the optic disc, and the tilt of the optic disc. The step of obtaining an image of the confused region between the optic disc and the myopic arc based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc includes: Find the intersection region between the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc; Obtain the minimum bounding rectangle of the intersection region, and enlarge the minimum bounding rectangle by a preset magnification ratio with the center point of the minimum bounding rectangle as the center to obtain the magnified rectangular region; The image corresponding to the magnified rectangular region in the image to be detected is obtained as the image of the confusion area between the optic disc and the myopic arc.

2. The method for quantifying the shape of a viewing disc as described in claim 1, characterized in that, The step of identifying and cropping the image to be detected from the target fundus image includes: Heatmaps are generated from the target fundus image to obtain a fundus heatmap. Based on the thermal data values ​​in the fundus thermal map, fundus regions that meet the preset fundus thermal value conditions are determined; The image to be detected that matches the fundus region is cropped from the target fundus image.

3. The method for quantifying the shape of a viewing disc as described in claim 2, characterized in that, The step of creating a heatmap of the target fundus image to obtain a fundus heatmap includes: Perform grayscale transformation on the target fundus image to obtain a fundus grayscale image; Based on the grayscale value of each pixel in the fundus grayscale image, read the corresponding thermal data value of each pixel from a preset color mapping table; A fundus thermal map is generated based on the thermal data value corresponding to each pixel.

4. The method for quantifying the shape of a viewing disc as described in claim 1, characterized in that, The step of fusing the preliminary segmented image of the optic disc and the corrected image of the optic disc to obtain a target optic disc image, and fusing the preliminary segmented image of the myopia arc and the corrected image of the myopia arc to obtain a target myopia arc image, includes: In the preliminary segmentation image of the visual disc, a first region corresponding to the corrected image of the visual disc is obtained, and the first region in the preliminary segmentation image of the visual disc is replaced with the corrected image of the visual disc to obtain the target visual disc image. In the preliminary segmentation image of the myopia arc, a second region corresponding to the myopia arc correction image is obtained, and the second region in the preliminary segmentation image of the myopia arc is replaced with the myopia arc correction image to obtain the target myopia arc image.

5. The method for quantifying the shape of a viewing disc as described in claim 1, characterized in that, The step of calculating the optic disc morphology quantization value based on the target optic disc image and the target myopic arc image includes: Based on the target optic disc image, obtain the optic disc pixel area, the longest diameter of the optic disc, the shortest diameter of the optic disc, and the angle between the longest diameter of the optic disc and the vertical axis of the coordinate system where the target fundus image is located; based on the target myopia arc image, obtain the myopia arc pixel area. The ellipticity of the optic disc is calculated based on the shortest diameter and the longest diameter of the optic disc. The tilt of the optic disc is calculated based on the angle between the longest diameter of the optic disc and the vertical axis of the coordinate system containing the target fundus image, and the longest diameter of the optic disc. The area ratio of the myopic arc to the optic disc is calculated based on the pixel area of ​​the myopic arc and the pixel area of ​​the optic disc.

6. The method for quantifying the shape of a viewing disc as described in any one of claims 1 to 5, characterized in that, Before inputting the image to be detected into a preset preliminary segmentation model, the method further includes: The optic disc and myopic arc in the fundus image sample are outlined. Based on the optic disc outline and myopic arc outline, the part outside the optic disc and myopic arc in the fundus image sample is masked, and the optic disc and myopic arc in the masked fundus image sample are labeled. The initial segmentation model was trained using labeled fundus image samples.

7. A device for quantifying the shape of a viewing disk, characterized in that, include: The preliminary segmentation module is used to identify and crop out the image to be detected from the target fundus image, and input the image to be detected into the preset preliminary segmentation model to obtain the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc. The confusion region segmentation module is used to obtain a confusion region image between the optic disc and the myopic arc based on the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc, and input the confusion region image into a preset confusion region segmentation model to obtain a corrected image of the optic disc and a corrected image of the myopic arc. The optic disc morphology quantification module is used to fuse the preliminary segmented image of the optic disc and the corrected image of the optic disc to obtain a target optic disc image, fuse the preliminary segmented image of the myopic arc and the corrected image of the myopic arc to obtain a target myopic arc image, and calculate the optic disc morphology quantification value based on the target optic disc image and the target myopic arc image. The optic disc morphology quantification value includes the area ratio of the myopic arc to the optic disc, the ellipticity of the optic disc, and the tilt of the optic disc. The obfuscation segmentation module is also used for: Find the intersection region between the preliminary segmentation image of the optic disc and the preliminary segmentation image of the myopic arc; Obtain the minimum bounding rectangle of the intersection region, and enlarge the minimum bounding rectangle by a preset magnification ratio with the center point of the minimum bounding rectangle as the center to obtain the magnified rectangular region; The image corresponding to the magnified rectangular region in the image to be detected is obtained as the image of the confusion area between the optic disc and the myopic arc.

8. A storage medium storing at least one executable instruction that causes a processor to perform an operation corresponding to the morphological quantization method of any one of claims 1-6.

9. A computer device, comprising: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the visual disc morphology quantization method as described in any one of claims 1-6.

Images