Retina image segmentation method and device, equipment, storage medium and product
By using a retinal image segmentation method and optimizing the segmentation model with a deep learning model, the fundus images are processed automatically, solving the problem of inaccurate image segmentation in retinopathy of prematurity and improving diagnostic efficiency and accuracy.
Patent Information
- Application Number
- CN202410937382.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-12
- Publication Date
- 2024-11-15
AI Technical Summary
Current technologies for image segmentation of retinopathy of prematurity are inaccurate, have a high misdiagnosis rate, and rely heavily on doctors' clinical experience and visual observation, which is time-consuming.
A retinal image segmentation method is adopted. By inputting fundus images into the target segmentation model, the segmentation results of blood vessels and ridges are obtained. Based on the blood vessel segmentation results, false positive results in the ridge segmentation results are removed, and the target segmentation results are output. The segmentation model is optimized by image preprocessing and deep learning models.
It has enabled automated and intelligent segmentation of the retinal crests of premature infants, shortening the diagnosis time, improving the accuracy of image segmentation results, reducing the misdiagnosis rate, and providing strong support for early detection and treatment.
Smart Images

Figure CN118967709A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of medical imaging technology, and in particular to retinal image segmentation methods, devices, equipment, storage media and products. Background Art
[0002] Retinopathy of prematurity is a serious eye disease that mainly affects the visual development of premature and low-birth-weight infants. If not treated in time, it may lead to irreversible visual impairment or even blindness. Therefore, the development of an efficient and accurate diagnostic system for retinopathy of prematurity has important practical significance and application value. Traditional diagnostic methods mainly rely on the clinical experience and naked eye observation of doctors, which have problems such as strong subjectivity, high misdiagnosis rate, and long time consumption. Now, by applying advanced computer vision and deep learning technologies to the medical field, the misdiagnosis rate can be reduced.
[0003] The image segmentation results used for diagnosis in existing methods are inaccurate. Therefore, how to improve the accuracy of image segmentation results is a problem that needs to be solved urgently. Summary of the invention
[0004] The main purpose of this application is to provide a retinal image segmentation method, device, equipment, storage medium and product, aiming to solve the technical problem of inaccurate image segmentation results.
[0005] To achieve the above objectives, the present application proposes a retinal image segmentation method, the method comprising:
[0006] Input the fundus image to the target segmentation model to obtain the blood vessel segmentation results and the ridge segmentation results;
[0007] Based on the blood vessel segmentation result, the false positive results in the ridge segmentation result are removed, and the target segmentation result is output.
[0008] In one embodiment, the step of removing false positive results in the ridge segmentation result based on the blood vessel segmentation result and outputting the target segmentation result includes:
[0009] Based on the blood vessel segmentation result, obtaining the blood vessel range according to the pixel point set of the blood vessel structure;
[0010] Clustering the ridge segmentation result according to the pixel point set of the ridge structure to obtain at least one ridge category;
[0011] sorting the ridge categories according to the number of pixel points of the ridges to obtain a ridge sorting result;
[0012] Based on the blood vessel range and the ridge sorting result, the false positive results in the ridge segmentation result are identified and removed, and the target segmentation result is output.
[0013] In one embodiment, the step of obtaining the blood vessel range according to the pixel point set of the blood vessel structure based on the blood vessel segmentation result includes:
[0014] Based on the blood vessel segmentation result, identifying a set of pixel points representing the blood vessel structure;
[0015] Clustering the blood vessel segmentation result according to the pixel point set of the blood vessel structure to obtain at least one blood vessel category;
[0016] Based on the blood vessel categories and the pixel point sets of the blood vessel structures, blood vessel categories whose blood vessel structure pixel points are lower than a blood vessel pixel point number threshold are removed to obtain a blood vessel trunk structure;
[0017] Based on the vascular trunk structure, the vascular range is obtained.
[0018] In one embodiment, the step of identifying and removing false positive results in the ridge segmentation result based on the blood vessel range and the ridge sorting result, and outputting the target segmentation result includes:
[0019] Based on the ridge ranking result, the ridge category ranked first and not within the blood vessel range is identified as a positive result;
[0020] The ridge categories with pixel numbers higher than the ridge pixel number threshold and located within the blood vessel range are identified as false positive results;
[0021] False positive results in the ridge segmentation result are removed, and the positive result is output as the target segmentation result.
[0022] In one embodiment, before the step of inputting the fundus image to the target segmentation model, the step includes:
[0023] Based on the retinal image data, a segmentation network is used to train the target segmentation model to obtain a first optimized segmentation model;
[0024] An image block whose segmentation index is lower than a preset index is input into the first optimized segmentation model, the first optimized segmentation model is retrained to obtain a second optimized segmentation model, and the second optimized segmentation model is identified as the target segmentation model.
[0025] In one embodiment, the step of inputting the fundus image to the target segmentation model to obtain a blood vessel segmentation result and a ridge segmentation result comprises:
[0026] Preprocessing the retinal image data using an image preprocessing algorithm to obtain preprocessed image data, wherein the preprocessed image data includes image data of blood vessels and image data of ridges;
[0027] Establishing a blood vessel segmentation model according to the image data of the blood vessel, inputting the fundus image into the blood vessel segmentation model, and obtaining a blood vessel segmentation result;
[0028] A ridge segmentation model is established according to the image data of the ridge, and the fundus image is input into the ridge segmentation model to obtain a ridge segmentation result.
[0029] In addition, to achieve the above-mentioned purpose, the present application also proposes a retinal image segmentation device, the retinal image segmentation device comprising:
[0030] An image segmentation module, used for inputting the fundus image into the target segmentation model to obtain a blood vessel segmentation result and a ridge segmentation result;
[0031] The image recognition module is used to remove the false positive results in the ridge segmentation results based on the blood vessel segmentation results, and output the target segmentation results.
[0032] In addition, to achieve the above-mentioned purpose, the present application also proposes a retinal image segmentation device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the retinal image segmentation method described above.
[0033] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the retinal image segmentation method described above are implemented.
[0034] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the retinal image segmentation method described above are implemented.
[0035] One or more technical solutions proposed in this application have at least the following technical effects:
[0036] Input the fundus image to the target segmentation model to obtain the blood vessel segmentation results and ridge segmentation results, and realize the automatic segmentation of the retinal ridges of premature infants. The automated and intelligent segmentation diagnosis system can quickly process a large amount of image data, greatly shortening the diagnosis time. By removing the false positive results in the ridge segmentation results based on the blood vessel segmentation results and outputting the target segmentation results, the accuracy of the image segmentation results is improved, and the misdiagnosis rate is further reduced, providing strong support for early detection and treatment. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0039] Figure 1 This is a flowchart of the first embodiment of the retinal image segmentation method of the present application;
[0040] Figure 2 This is a flow chart of a second embodiment of the retinal image segmentation method of the present application;
[0041] Figure 3 This is a schematic diagram of the flow chart of the third embodiment of the retinal image segmentation method of the present application;
[0042] Figure 4 This is a schematic diagram of the module structure of the retinal image segmentation device according to an embodiment of the present application;
[0043] Figure 5 Schematic diagram of the device structure of the hardware operating environment involved in the retinal image segmentation method in the embodiment of the present application.
[0044] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0045] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.
[0046] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.
[0047] Since traditional diagnostic methods mainly rely on doctors' clinical experience and naked eye observation, they have problems such as strong subjectivity, high misdiagnosis rate, and long time consumption. In addition, existing methods rarely consider ridge segmentation in the diagnosis of this disease. At the same time, there are problems such as insufficient extraction of small target features, class imbalance, and false positives in the segmentation results.
[0048] Therefore, this application provides a solution to input fundus images into the target segmentation model, obtain blood vessel segmentation results and ridge segmentation results, and realize automatic segmentation of premature infant retinal ridges. The automated and intelligent segmentation diagnosis system can quickly process a large amount of image data, greatly shortening the diagnosis time. By removing the false positive results in the ridge segmentation results based on the blood vessel segmentation results and outputting the target segmentation results, the accuracy of the image segmentation results is improved, and further, the misdiagnosis rate is reduced, providing strong support for early detection and treatment.
[0049] Based on this, the present application embodiment provides a retinal image segmentation method, referring to Figure 1 , Figure 1 This is a flowchart diagram of the first embodiment of the retinal image segmentation method of the present application.
[0050] In this embodiment, the retinal image segmentation method includes steps S10 to S20:
[0051] Step S10, inputting the fundus image into the target segmentation model to obtain the blood vessel segmentation result and the ridge segmentation result.
[0052] It should be noted that fundus images can be understood as retinal images obtained through professional equipment, which are used to observe and analyze retinal structures such as blood vessels and ridges. The target segmentation model is used to separate specific targets in the input image (such as blood vessels, ridges) from the background or other irrelevant structures. The vessel segmentation result refers to the image data of the blood vessel part identified and extracted from the fundus image by the vessel segmentation model, and the ridge segmentation result refers to the image data of the ridge part (such as the optic disc boundary or the retinal nerve fiber layer ridge) identified and extracted from the fundus image by the ridge segmentation model.
[0053] It should be noted that the step of inputting the fundus image to the target segmentation model to obtain the blood vessel segmentation result and the ridge segmentation result includes: using an image preprocessing algorithm to preprocess the retinal image data to obtain preprocessed image data, and the preprocessed image data includes blood vessel image data and ridge image data. Among them, the image preprocessing algorithm is used to preprocess the input fundus image to improve the performance of the subsequent segmentation model. Preprocessing may include operations such as denoising, contrast enhancement, and color correction. Exemplarily, the preprocessed image data can be understood as retinal image data obtained after denoising and contrast enhancement. A blood vessel segmentation model is established based on the blood vessel image data, and the fundus image is input to the blood vessel segmentation model to obtain the blood vessel segmentation result; a ridge segmentation model is established based on the ridge image data, and the fundus image is input to the ridge segmentation model to obtain the ridge segmentation result.
[0054] Step S20, based on the blood vessel segmentation result, remove the false positive results in the ridge segmentation result, and output the target segmentation result.
[0055] It should be noted that the false positive result in image processing refers to the situation where the model mistakenly identifies the non-target area as the target area. For example, the blood vessel segmentation result is mistaken for the ridge segmentation result. The target segmentation result can be understood as the model output image without the false positive result, which can accurately reflect the real distribution of the ridges in the fundus image.
[0056] In this embodiment, the fundus image is input into the target segmentation model to obtain the blood vessel segmentation results and the ridge segmentation results, so as to realize the automatic segmentation of the retinal ridges of premature infants. The automated and intelligent segmentation diagnosis system can quickly process a large amount of image data, greatly shortening the diagnosis time. By removing the false positive results in the ridge segmentation results based on the blood vessel segmentation results and outputting the target segmentation results, the accuracy of the image segmentation results is improved, and further, the misdiagnosis rate is reduced, providing strong support for early detection and treatment.
[0057] Reference Figure 2 , Figure 2 This is a flow chart of the second embodiment of the retinal image segmentation method of the present application. Figure 1 The first embodiment shown provides a second embodiment of the retinal image segmentation method of the present application.
[0058] It should be noted that the ridge is the boundary between the vascular area and the non-vascular area, and the ridge will not appear on the same side as the blood vessel. In addition, the number of pixels occupied by the ridge is much larger than the number of false positive pixels. Based on this, a second embodiment is proposed. In the second embodiment, the step S20 includes:
[0059] Step S201 : based on the blood vessel segmentation result, the blood vessel range is obtained according to the pixel point set of the blood vessel structure.
[0060] It should be noted that step S201 includes: based on the blood vessel segmentation result, identifying a set of pixel points representing the blood vessel structure; clustering the blood vessel segmentation result according to the pixel point set of the blood vessel structure to obtain at least one blood vessel category; based on the blood vessel category and the pixel point set of the blood vessel structure, removing the blood vessel category whose number of pixel points of the blood vessel structure is lower than the blood vessel pixel point number threshold to obtain the blood vessel trunk structure; based on the blood vessel trunk structure, obtaining the blood vessel range. It can be understood that the pixel point set of the blood vessel structure can be understood as a set of two-dimensional coordinate points, each coordinate point represents a pixel in the region where the blood vessel is located in the fundus image, and the blood vessel range can be the distribution range of the blood vessel obtained by the pixel point distribution of the blood vessel structure. Exemplarily, the DBSCAN algorithm is used to cluster the blood vessels, and the class with a very small number of pixel points is removed to retain the blood vessel trunk structure, and the coordinates of the four extreme points of the blood vessel are found to form a blood vessel range.
[0061] Step S202: clustering the ridge segmentation result according to the pixel point set of the ridge structure to obtain at least one ridge category.
[0062] It can be understood that the pixel point set of the ridge structure is a set of two-dimensional coordinate points, each coordinate point represents a pixel in the area where the ridge is located in the fundus image, and the ridge segmentation result is classified according to the characteristics of the pixel point set of the ridge structure to obtain at least one ridge category. It can be understood that when there is no false positive result in the blood vessel segmentation result, only one ridge category can be obtained.
[0063] Step S203, sorting the ridge categories according to the number of pixels of the ridges to obtain a ridge sorting result.
[0064] Exemplarily, the segmentation results of the ridges are clustered using the DBSCAN algorithm, and the clustering results are sorted according to the number of pixels in each category.
[0065] Step S204: based on the blood vessel range and the ridge sorting result, identify and remove the false positive results in the ridge segmentation result, and output the target segmentation result.
[0066] It should be noted that step S204 includes: based on the ridge sorting result, identifying the ridge category that is ranked first and is not within the blood vessel range as a positive result; identifying the ridge category whose number of pixels is higher than the ridge pixel number threshold and is within the blood vessel range as a false positive result; removing the false positive result in the ridge segmentation result, and outputting the positive result as the target segmentation result. Exemplarily, after clustering the ridge segmentation results using the DBSCAN algorithm and sorting the clustering results according to the number of pixels each category has, select the category with the largest number and not within the blood vessel range as the ridge, and for other categories, if they fall within the blood vessel range or have a large difference in the number of pixels from the ridge, they are directly removed.
[0067] In this embodiment, based on the blood vessel segmentation result, the blood vessel range is obtained according to the pixel point set of the blood vessel structure, so as to accurately define the range of the blood vessel. The ridge segmentation result is clustered according to the pixel point set of the ridge structure to obtain at least one ridge category, and the ridge categories are sorted according to the number of ridge pixels to obtain the ridge sorting result. The sorting result will show the size and significance of different ridge structures, providing guidance for subsequent analysis and processing. Based on the blood vessel range and the ridge sorting result, the false positive results in the ridge segmentation result are identified and removed, and the target segmentation result is output. By removing these categories from the segmentation result, the number of false positive results can be reduced and the accuracy of the segmentation result can be improved.
[0068] Reference Figure 3 , Figure 3 This is a flow chart of the third embodiment of the retinal image segmentation method of the present application. Figure 2 The second embodiment shown provides the third embodiment of the retinal image segmentation method of the present application.
[0069] In the third embodiment, before step S10, the following steps are included:
[0070] Step S010: Based on the retinal image data, a segmentation network is used to train the target segmentation model to obtain a first optimized segmentation model.
[0071] It should be noted that the segmentation network is a deep learning model used to segment different areas or objects in an image. The first optimized segmentation model can be understood as a model obtained after optimizing the training of the original segmentation model. Exemplarily, in the design of the network, the dense connection and hole convolution in Densenet are integrated into the U-Net network, and the corresponding modules in the original model are replaced. At the same time, a 1×1 convolution layer can be added to the input of each dense block as a side connection. Exemplarily, the first layer of the network can be a 7×7 convolution layer. For an image X to be segmented that is input into the segmentation network, assuming that Y is its true category label at the pixel level, is the category of each pixel predicted by the segmentation model, where Y and is a matrix composed of 0 and 1, representing the category label of each pixel in X, where 1 represents the ridge category and 0 represents the background. i represents the true value of the i-th pixel in Y, express The predicted value of the i-th pixel in , assuming that X has n pixels in total, then the definition of the cross entropy loss is as shown in formula (1):
[0072]
[0073] The definition of Dice loss is shown in formula (2):
[0074]
[0075] In formulas (1) and (2), l BCE is the cross entropy loss, l Dice is the Dice loss, n is the number of pixels, y i represents the true value of the i-th pixel in Y, express The prediction value of the i-th pixel in . The definition of SSIM loss is shown in Formula 3:
[0076]
[0077] Among them, l SSIM is the SSIM loss, μY and Represents Y and The mean of and Represents Y and The variance of, C1 and C2 are two constants, where C1=(t1L) 2 , C2=(t2L) 2 , L represents the dynamic range of pixel values, t1 and t2 are two constants. The final training loss is the sum of the three loss functions, as shown in formula (4):
[0078] ltrainlBCE+lpice+-lssIM(4)
[0079] Among them, l train is the final training loss.
[0080] Step S020, inputting an image block whose segmentation index is lower than a preset index into a first optimized segmentation model, retraining the first optimized segmentation model to obtain a second optimized segmentation model, and identifying the second optimized segmentation model as a target segmentation model.
[0081] It should be noted that the segmentation index can be an index corresponding to the fundus image currently input, and the preset index can be understood as a standard or metric for evaluating the performance of the segmentation model. An image block refers to a small area cut out from a complete image. In medical image processing, since the image may be very large, directly processing the entire image may consume a lot of computing resources. Therefore, the image is usually divided into multiple smaller image blocks, and then these image blocks are processed separately. The second optimized segmentation model can be understood as a model obtained after optimizing the original first optimized segmentation model. Exemplarily, a sliding window sampling algorithm with a dynamic threshold is used, and each time an image block with a poor segmentation effect is selected to be input into the network for retraining. The advantage of this sampling algorithm is that more local images can be obtained in the early stage for model training, which is used to improve the segmentation accuracy of fine blood vessels.
[0082] In this embodiment, based on retinal image data, the target segmentation model is trained by using a segmentation network to obtain a first optimized segmentation model, so that the model can use feature maps of different scales and improve the feature representation of the model. Image blocks with segmentation indicators lower than the preset indicators are input to the first optimized segmentation model, and the first optimized segmentation model is retrained to obtain a second optimized segmentation model. The second optimized segmentation model is identified as the target segmentation model to improve the segmentation accuracy of fine blood vessels. As the training progresses, the number of image blocks with poor segmentation results becomes smaller and smaller, that is, the number of image blocks that need to be retrained also becomes smaller and smaller, saving the training time of the mid- and late-stage models.
[0083] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the retinal image segmentation method of the present application. More simple transformations based on this technical concept are all within the protection scope of the present application.
[0084] This application also provides a retinal image segmentation device, please refer to Figure 4 , the retinal image segmentation device comprises:
[0085] An image segmentation module 10 is used to input the fundus image into the target segmentation model to obtain a blood vessel segmentation result and a ridge segmentation result;
[0086] The image recognition module 20 is used to remove the false positive results in the ridge segmentation results based on the blood vessel segmentation results, and output the target segmentation results.
[0087] The retinal image segmentation device provided by the present application adopts the retinal image segmentation method in the above embodiment, which can solve the technical problem of inaccurate image segmentation results. Compared with the prior art, the beneficial effects of the retinal image segmentation device provided by the present application are the same as the beneficial effects of the retinal image segmentation method provided by the above embodiment, and the other technical features in the retinal image segmentation device are the same as the features disclosed in the above embodiment method, which will not be repeated here.
[0088] The present application provides a retinal image segmentation device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the retinal image segmentation method in the above-mentioned embodiment one.
[0089] Reference below Figure 5 , which shows a schematic diagram of the structure of a retinal image segmentation device suitable for implementing the embodiment of the present application. The retinal image segmentation device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The retinal image segmentation device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.
[0090] like Figure 5As shown, the retinal image segmentation device may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the retinal image segmentation device are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the retinal image segmentation device to communicate with other devices wirelessly or by wire to exchange data. Figure 5 A retinal image segmentation device with various systems is shown in the figure, but it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have instead.
[0091] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.
[0092] The retinal image segmentation device provided by the present application adopts the retinal image segmentation method in the above embodiment, which can solve the technical problem of inaccurate image segmentation results. Compared with the prior art, the beneficial effects of the retinal image segmentation device provided by the present application are the same as the beneficial effects of the retinal image segmentation method provided by the above embodiment, and the other technical features in the retinal image segmentation device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.
[0093] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.
[0094] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
[0095] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, wherein the computer-readable program instructions are used to execute the retinal image segmentation method in the above-mentioned embodiment.
[0096] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.
[0097] The computer-readable storage medium may be included in the retinal image segmentation device; or may exist independently without being assembled into the retinal image segmentation device.
[0098] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the retinal image segmentation device, the retinal image segmentation device: inputs the fundus image into the target segmentation model to obtain the blood vessel segmentation result and the ridge segmentation result; based on the blood vessel segmentation result, removes the false positive results in the ridge segmentation result, and outputs the target segmentation result.
[0099] Computer program code for performing the operations of the present application may be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0100] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0101] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.
[0102] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned retinal image segmentation method, and can solve the technical problem of inaccurate image segmentation results. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the retinal image segmentation method provided in the above-mentioned embodiment, and will not be repeated here.
[0103] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned retinal image segmentation method when executed by a processor.
[0104] The computer program product provided in this application can solve the technical problem of inaccurate image segmentation results. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the retinal image segmentation method provided in the above embodiment, which will not be repeated here.
[0105] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.
Claims
1. A retinal image segmentation method, characterized in that: The method includes: Input the fundus image to the target segmentation model to obtain the blood vessel segmentation results and the ridge segmentation results; Based on the blood vessel segmentation result, the false positive results in the ridge segmentation result are removed, and the target segmentation result is output.
2. The method according to claim 1, characterized in that The step of removing the false positive results in the ridge segmentation results based on the blood vessel segmentation results and outputting the target segmentation results comprises: Based on the blood vessel segmentation result, obtaining the blood vessel range according to the pixel point set of the blood vessel structure; Clustering the ridge segmentation result according to the pixel point set of the ridge structure to obtain at least one ridge category; sorting the ridge categories according to the number of pixel points of the ridges to obtain a ridge sorting result; Based on the blood vessel range and the ridge sorting result, the false positive results in the ridge segmentation result are identified and removed, and the target segmentation result is output.
3. The method according to claim 2, characterized in that The step of obtaining the blood vessel range according to the pixel point set of the blood vessel structure based on the blood vessel segmentation result comprises: Based on the blood vessel segmentation result, identifying a set of pixel points representing the blood vessel structure; Clustering the blood vessel segmentation result according to the pixel point set of the blood vessel structure to obtain at least one blood vessel category; Based on the blood vessel categories and the pixel point sets of the blood vessel structures, blood vessel categories whose blood vessel structure pixel points are lower than a blood vessel pixel point number threshold are removed to obtain a blood vessel trunk structure; Based on the vascular trunk structure, the vascular range is obtained.
4. The method according to claim 3, characterized in that The step of identifying and removing false positive results in the ridge segmentation results based on the blood vessel range and the ridge sorting results, and outputting the target segmentation results includes: Based on the ridge ranking result, the ridge category ranked first and not within the blood vessel range is identified as a positive result; The ridge categories with pixel numbers higher than the ridge pixel number threshold and located within the blood vessel range are identified as false positive results; False positive results in the ridge segmentation result are removed, and the positive result is output as the target segmentation result.
5. The method according to claim 1, characterized in that Before the step of inputting the fundus image to the target segmentation model, the method includes: Based on the retinal image data, a segmentation network is used to train the target segmentation model to obtain a first optimized segmentation model; An image block whose segmentation index is lower than a preset index is input into the first optimized segmentation model, the first optimized segmentation model is retrained to obtain a second optimized segmentation model, and the second optimized segmentation model is identified as the target segmentation model.
6. The method according to any one of claims 1 to 5, characterized in that The step of inputting the fundus image to the target segmentation model to obtain a blood vessel segmentation result and a ridge segmentation result comprises: Preprocessing the retinal image data using an image preprocessing algorithm to obtain preprocessed image data, wherein the preprocessed image data includes image data of blood vessels and image data of ridges; Establishing a blood vessel segmentation model according to the image data of the blood vessel, inputting the fundus image into the blood vessel segmentation model, and obtaining a blood vessel segmentation result; A ridge segmentation model is established according to the image data of the ridge, and the fundus image is input into the ridge segmentation model to obtain a ridge segmentation result.
7. A retinal image segmentation device, characterized in that: The device comprises: An image segmentation module, used for inputting the fundus image into the target segmentation model to obtain a blood vessel segmentation result and a ridge segmentation result; The image recognition module is used to remove the false positive results in the ridge segmentation results based on the blood vessel segmentation results, and output the target segmentation results.
8. A retinal image segmentation device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the retinal image segmentation method according to any one of claims 1 to 6.
9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the retinal image segmentation method according to any one of claims 1 to 6 are implemented.
10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the retinal image segmentation method according to any one of claims 1 to 6 are implemented.
Citation Information
Patent Citations
Newborn ROP identification and labeling method and system, storage medium and electronic equipment
CN116863224A
Systems and methods for automated segmentation of retinal fluid in optical coherence tomography
US20180263490A1
