Retina OCTA image enhancement method and system
By combining spectral transformation and grayscale mapping with feature filtering, high-order feature processing is performed on retinal OCTA images, which solves the problem of poor enhancement effect in existing technologies and improves image quality and fidelity.
Patent Information
- Application Number
- CN202510949651.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-10
- Publication Date
- 2025-12-05
AI Technical Summary
Existing OCTA image enhancement methods are not effective in medical image processing, especially for retinal OCTA images. They lack specificity, and traditional methods cannot effectively utilize the data characteristics of OCTA equipment, resulting in limited enhancement effects.
A method combining spectral transformation and grayscale mapping with feature filtering is used to perform high-order feature processing in the frequency and spatial domains of retinal OCTA images. Enhanced images are saved by fusing similarity judgment, including operations such as Fourier transform, Laplace transform, grayscale transformation and feature filtering.
It achieves effective enhancement of retinal OCTA images, improving image quality and specificity, and enhancing image fidelity and noise reduction capabilities.
Smart Images
Figure CN121073784A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] Embodiments of the present disclosure belong to the technical field of OCTA image enhancement, and particularly relate to a retinal OCTA image enhancement method and system. BACKGROUND
[0002] OCTA (Optical Coherence Tomography Angiography, optical coherence tomography angiography) technology realizes visual imaging by detecting OCT signals obtained by repeated scanning at a single location and performing pixel-by-pixel calculation of optical correlation signals. The current research method of OCTA image deep learning is mainly based on the field of digital images, such as supervised learning of image recognition, detection, segmentation, registration, etc. In order to achieve accuracy, many scenes in current medical tasks are very dependent on samples.
[0003] Unlike natural scene images, the unique features of the texture features, shape features, and color histogram features of the fundus retinal images of OCTA have very huge inherent data domain closure. Therefore, if there is not enough OCTA image data, the image processing means of deep learning is difficult to achieve the purpose of effective enhancement. Therefore, how to root in the characteristics of the OCTA data domain and effectively enhance the image is very important for the current deep learning research of OCTA.
[0004] The current image enhancement means is generally based on data enhancement of natural scene digital images, such as gray scale transformation, contrast enhancement, three-dimensional transformation such as rotation, interpolation, noise addition or filtering, etc. These general methods are generally digital images in color natural scenes, which are very limited in the application scene of OCTA and the like. The enhanced OCTA medical images do not have specific potential spatial characteristics, especially the data characteristics generated by the physical acquisition of the OCTA device. Traditional image enhancement is based on a single device or a single OCTA, so that the potential enhanced characteristics do not have the image distribution characteristics of other devices.
[0005] The current image enhancement means is generally based on data enhancement of natural scene digital images, such as gray scale transformation, contrast enhancement, three-dimensional transformation such as rotation, interpolation, noise addition or filtering, etc. These general methods are generally digital images in color natural scenes, which are very limited in the application scene of OCTA and the like. The enhanced OCTA medical images do not have specific potential spatial characteristics, especially the data characteristics generated by the physical acquisition of the OCTA device. Traditional image enhancement is based on a single device or a single OCTA, so that the potential enhanced characteristics do not have the image distribution characteristics of other devices. SUMMARY
[0006] Embodiments of the present disclosure aim to at least solve one of the technical problems existing in the prior art, and provide a retinal OCTA image enhancement method and system.
[0007] One aspect of the present disclosure provides a retinal OCTA image enhancement method, which comprises:
[0008] The method comprises:
[0009] obtaining a retinal OCTA original image;
[0010] performing a spectrum conversion on the original image to obtain a frequency domain image, performing a frequency domain processing operation on the frequency domain image, and performing an inverse conversion on the frequency domain image to obtain a first image;
[0011] performing a gray mapping conversion on the original image to obtain a gray image, performing a feature filtering operation on the gray image, and obtaining a second image;
[0012] fusing the first image and the second image to obtain a fused image;
[0013] calculating a similarity between the fused image and the original image, and saving the fused image if the similarity meets a predetermined requirement.
[0014] Optionally, the spectrum conversion includes at least one of a Fourier transform, a Laplace transform, and a wavelet transform.
[0015] Optionally, the frequency domain processing operation includes at least one of a high-pass interception, a low-pass interception, a frequency domain cropping, a frequency splicing, and a random selection.
[0016] Optionally, the gray mapping conversion includes at least one of a gray transform, a contrast enhancement transform, and a gamma transform.
[0017] Optionally, the feature filtering operation includes at least one of a Gaussian filtering, a median filtering, and a Laplace filtering.
[0018] Optionally, the calculating the similarity between the fused image and the original image includes:
[0019] calculating the similarity between the fused image and the original image by:
[0020]
[0021] wherein I is the original image, is the fused image, represents the similarity between the fused image and the original image, λ i is a weight factor, L i is a similarity measurement function.
[0022] Further, the similarity measurement function is used to calculate a mean square error, a cosine similarity, or a structural similarity coefficient between the fused image and the original image.
[0023] Further, after the obtaining the first image, the method further includes:
[0024] determine whether the first image has a fidelity denoising feature, and if not, re-process the frequency domain image through a frequency domain processing operation, inverse transform the frequency domain image to the time domain, and obtain a first image.
[0025] Further, after the second image is obtained, the method further includes:
[0026] determine whether the second image has a fidelity denoising feature, and if not, re-process the grayscale image through a feature filtering operation, and obtain a second image.
[0027] Another aspect of the present disclosure provides a retinal OCTA image enhancement system, which
[0028] The system includes:
[0029] An acquisition module is configured to acquire a retinal OCTA original image.
[0030] A frequency domain conversion module is configured to perform frequency spectrum conversion on the original image to obtain a frequency domain image, process the frequency domain image through a frequency domain processing operation, and inverse transform the frequency domain image to the time domain to obtain a first image.
[0031] A grayscale conversion module is configured to perform grayscale mapping conversion on the original image to obtain a grayscale image, and process the grayscale image through a feature filtering operation to obtain a second image.
[0032] A fusion module is configured to fuse the first image and the second image to obtain a fused image.
[0033] A similarity module is configured to calculate a similarity between the fused image and the original image, and if the similarity meets a predetermined requirement, save the fused image.
[0034] The retinal OCTA image enhancement method and system of an embodiment of the present disclosure are based on high-order feature domains, including but not limited to frequency domains and spatial domains, for high-order feature processing; including but not limited to simple image processing, which can be adjusted according to some change thresholds; and can be fused from a plurality of other potential features. BRIEF DESCRIPTION OF DRAWINGS
[0035] Figure 1 FIG. 1 is a flowchart of a retinal OCTA image enhancement method according to an embodiment of the present disclosure;
[0036] Figure 2 FIG. 2 is a structural diagram of a retinal OCTA image enhancement system according to another embodiment of the present disclosure. DETAILED DESCRIPTION
[0037] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. Based on the embodiments of this disclosure, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this disclosure.
[0038] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this disclosure.
[0039] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.
[0040] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various components, these components should not be limited by these terms. These terms are used to distinguish one component from another. Therefore, the first component discussed below may be referred to as the second component without departing from the teachings of this disclosure. As used in this disclosure, the term "and / or" includes all combinations of any and more of the associated listed items.
[0041] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of exemplary embodiments, and the modules or processes in the drawings are not necessarily necessary for implementing this disclosure, and therefore cannot be used to limit the scope of protection of this disclosure.
[0042] like Figure 1 As shown, one embodiment of this disclosure provides a method for enhancing retinal OCTA images, the method comprising:
[0043] Step S1: Obtain the raw retinal OCTA image.
[0044] Specifically, OCTA technology is used to image the retina to form one or more original retinal OCTA images, or one or more images are randomly selected from an existing group of retinal OCTA images as original images. The original images are then processed using two branch steps, S2.1 and S2.2.
[0045] Step S2.1, performing spectrum conversion on the original image to obtain a frequency domain image, processing the frequency domain image through a frequency domain processing operation, and inversely transforming the frequency domain image to the time domain to obtain a first image.
[0046] Specifically, first, spectrum conversion is performed on the original image obtained in step S1. The processing logic of spectrum conversion can include but is not limited to Fourier transform, Laplace transform, wavelet transform, etc. Specifically, it can be as follows:
[0047]
[0048] Where I(x, y) is the original image, f is the specific mathematical rule of spectrum conversion, and F(u, v) is the frequency domain image obtained after conversion. The mathematical function f of the spectrum conversion corresponding to the processing logic is selected, and the image of each frequency component in the original image is decomposed, and each original image is converted into a frequency domain image.
[0049] Subsequently, the frequency domain image is processed. Appropriate frequency domain processing operations are selected to process the components of the frequency domain image, including image calculation processing such as strengthening, screening, discrimination, and retention. The specific operation can include but is not limited to high-pass interception, low-pass interception, frequency domain cropping, frequency splicing, random selection, etc. After obtaining the processed frequency domain image, a corresponding inverse transformation of the frequency domain is performed to convert it from the frequency domain image to the time domain image. The final time domain image is defined as the first image.
[0050] Step S2.2, performing gray scale mapping conversion on the original image to obtain a gray scale image, and processing the gray scale image through a feature filtering operation to obtain a second image.
[0051] Specifically, first, general gray scale mapping conversion is performed on the original image obtained in step S1. The processing logic of gray scale mapping conversion can include but is not limited to gray scale conversion, contrast enhancement conversion, gamma conversion, etc. Specifically, it can be as follows:
[0052]
[0053] Where I(x, y) is the original image, g is the specific mathematical rule of gray scale mapping conversion, and G(x, y) is the frequency domain image obtained after conversion. The mathematical function g of the gray scale mapping conversion corresponding to the processing logic is selected, and each original image is converted into a gray scale image, and the original image is preprocessed in terms of gray scale.
[0054] Subsequently, image processing operations in the image gray scale domain are performed on the gray scale image. Appropriate feature filtering operations are selected to perform image calculation processing including strengthening, screening, distinguishing, and retaining, etc. on the gray scale features, such as filtering, noise reduction, convolution, etc. The specific operations can include, but are not limited to, Gaussian filtering, median filtering, Laplace filtering, etc. The final gray scale image obtained by the above operations is defined as a second image.
[0055] Finally, it is determined whether the first image and the second image have the fidelity denoising feature, which can be determined according to the following details: 1. whether the background and the local regions of each layer in the image have block-shaped, sheet-shaped, or point-shaped non-uniform artifacts or noise; 2. whether the positions of the boundary lines of each layer in the image are distorted or obviously unclear; 3. whether the average gray scale values of each layer in the image are obviously distorted relative to the contrast of the background. If not, the frequency domain image is processed again by the frequency domain processing operation to obtain the first image, or the gray scale image is processed again by the feature filtering operation to obtain the second image, and then step S3 is performed; if so, the next step S3 is directly performed.
[0056] Step S3, fusing the first image and the second image to obtain a fused image.
[0057] Specifically, the first image and the second image that satisfy the judgment after step S2.1 and step S2.2 are fused to obtain a fused image.
[0058] Step S4, calculating the similarity between the fused image and the original image, and if the similarity meets the predetermined requirement, saving the fused image.
[0059] Specifically, the similarity between the fused image obtained in step S3 and the original image is evaluated. The similarity evaluation is to calculate the coincidence and consistency of the potential features between the input images. The similarity between the fused image and the original image is calculated by the following formula:
[0060]
[0061] wherein I is the original image, is the fused image, represents the similarity between the fused image and the original image, λ i is a weight factor, L i is a similarity measurement function.
[0062] Theoretically, the weight factor λ i satisfies the following conditions:
[0063] Σλ i 2 = 1
[0064] The form of the similarity measure function L can be selected as: Mean Squared Error, Cosine Similarity, Structure Similarity Index Measure, etc.
[0065] The result obtained by the similarity calculation is judged whether it meets the requirements of the principle, if not, the result is discarded; if it meets, the fusion image obtained in step S3, i.e. the enhanced image, is saved to the reasonable path position in the enhanced image group.
[0066] The retinal OCTA image enhancement method of the embodiment of the present disclosure is based on high-order feature fields, including but not limited to frequency domain and spatial domain for high-order feature processing; including but not limited to simple image processing, which can be adjusted by itself for some change threshold; and can be fused from a plurality of other potential features.
[0067] As shown in Figure 2 Another embodiment of the present disclosure provides a retinal OCTA image enhancement system, which comprises:
[0068] The acquisition module 210 is configured to acquire a retinal OCTA original image.
[0069] The frequency domain conversion module 221 is configured to perform frequency spectrum conversion on the original image to obtain a frequency domain image, perform processing on the frequency domain image through a frequency domain processing operation, and then inverse transform the frequency domain image to the time domain to obtain a first image.
[0070] The gray scale conversion module 222 is configured to perform gray scale mapping conversion on the original image to obtain a gray scale image, and perform processing on the gray scale image through a feature filtering operation to obtain a second image.
[0071] The fusion module 230 is configured to fuse the first image and the second image to obtain a fusion image.
[0072] The similarity module 240 is configured to calculate the similarity between the fusion image and the original image, and save the fusion image if the similarity meets a predetermined requirement.
[0073] Specifically, the retinal OCTA image enhancement system of the embodiment of the present disclosure is used to implement the retinal OCTA image enhancement method described in the above embodiment, and the specific implementation process has been described in detail in the previous embodiment, which will not be repeated here.
[0074] It is understood that the above embodiments are only exemplary for illustrating the principles of the present disclosure, and the present disclosure is not limited thereto. Various modifications and improvements can be made by those of ordinary skill in the art without departing from the spirit and essence of the present disclosure, and these modifications and improvements are also considered to be within the scope of protection of the present disclosure.
Claims
1. A retinal OCTA image enhancement method, characterized in that, The method comprises: obtaining a retinal OCTA original image; performing spectral conversion on the original image to obtain a frequency domain image, performing frequency domain processing operation on the frequency domain image, and then performing inverse conversion on the frequency domain image to the time domain to obtain a first image; performing gray mapping conversion on the original image to obtain a gray image, and performing feature filtering operation on the gray image to obtain a second image; fusing the first image and the second image to obtain a fused image; calculating the similarity between the fused image and the original image, and if the similarity meets predetermined requirements, saving the fused image.
2. The method of claim 1, wherein, The spectral conversion comprises at least one of Fourier transform, Laplace transform and wavelet transform.
3. The method of claim 1, wherein, The frequency domain processing operation comprises at least one of high-pass interception, low-pass interception, frequency domain clipping, frequency splicing and random selection.
4. The method of claim 1, wherein, The gray mapping conversion comprises at least one of gray transform, contrast enhancement transform and gamma transform.
5. The method of claim 1, wherein, The feature filtering operation comprises at least one of Gaussian filtering, median filtering and Laplace filtering.
6. The method of claim 1, wherein, The calculation of the similarity between the fused image and the original image comprises: calculating the similarity between the fused image and the original image by the following formula: where I is the original image, is the fused image, represents the similarity between the fused image and the original image, λ i is a weight factor, L i is a similarity measure function.
7. The method of claim 6, wherein, The similarity measure function is used to calculate the mean square error or cosine similarity or structural similarity coefficient between the fused image and the original image.
8. The method according to any one of claims 1 to 7, characterized in that, After the first image is obtained, the method further comprises: judging whether the first image has a fidelity denoising feature, and if not, re-performing frequency domain processing operation on the frequency domain image, and then performing inverse conversion on the frequency domain image to the time domain to obtain a first image.
9. The method according to any one of claims 1 to 7, characterized in that, After the second image is obtained, the method further comprises: judging whether the second image has a fidelity denoising feature, and if not, re-performing feature filtering operation on the gray image to obtain a second image.
10. A retinal OCTA image enhancement system, characterized by, The system comprises: an acquisition module configured to obtain a retinal OCTA original image; a frequency domain conversion module configured to perform spectral conversion on the original image to obtain a frequency domain image, perform frequency domain processing operation on the frequency domain image, and then perform inverse conversion on the frequency domain image to the time domain to obtain a first image; a gray conversion module configured to perform gray mapping conversion on the original image to obtain a gray image, and perform feature filtering operation on the gray image to obtain a second image; a fusion module configured to fuse the first image and the second image to obtain a fused image; a similarity module configured to calculate the similarity between the fused image and the original image, and if the similarity meets predetermined requirements, save the fused image.