A method for evaluating cerebral vascular state based on fundus multi-modal image fusion technology

Abstract

The application provides a method for evaluating brain vascular state based on fundus multi-modal image fusion technology, acquires fundus color photograph image and optical coherence tomography image or optical coherence tomography angiography image of the same subject, pre-processes the multi-modal images, extracts vascular registration features, establishes cross-modal registration transformation relationship, maps different modal images to a unified coordinate system to form fusion data, constructs a three-dimensional region of interest in the fusion data and acquires vascular volume data, calculates composite biomarkers such as vascular volume density and / or blood flow vessel ratio, and outputs brain vascular state evaluation results based on the composite biomarkers. The application realizes fusion analysis of multi-modal fundus vascular information, and improves the objectivity and stability of brain vascular state evaluation.

Description

Technical Field

[0001] This invention belongs to the field of medical image processing technology, specifically relating to a method for assessing cerebrovascular status based on fundus multimodal image fusion technology. Background Technology

[0002] Changes in cerebrovascular status are closely related to systemic hemodynamics and microcirculatory function. How to objectively characterize and quantify cerebrovascular status under non-invasive conditions has always been an important focus in the field of medical imaging analysis. Among existing technologies, fundus imaging is considered to have some indirect reference value in assessing cerebrovascular status because it can directly reflect the morphology and blood flow characteristics of human microvessels. Currently, color fundus photography or optical coherence tomography (OCT) images are commonly used in clinical practice and research to observe and analyze fundus vessels. However, different imaging modalities differ significantly in imaging mechanisms, spatial resolution, and information expression dimensions. A single modality often only reflects part of the vascular structure or local features, making it difficult to comprehensively depict the spatial distribution and blood flow status of vessels.

[0003] Under current technological conditions, color fundus photography is mainly used to acquire the outline and course characteristics of blood vessels, while optical coherence tomography (OCT) or vascular imaging focuses on reflecting the interlayer structure of blood vessels or intraluminal blood flow information. Because these various images are typically acquired and analyzed independently, their coordinate systems are not unified, and there is a lack of effective correspondence between the images. This makes it difficult to accurately match and comprehensively utilize vascular information obtained from different modalities, thus limiting the effectiveness of multi-source information fusion analysis. Furthermore, existing analysis methods mostly rely on two-dimensional parameters or local indices, lacking means to perform three-dimensional regional quantification of blood vessels under a unified spatial benchmark, making it difficult to accurately reflect the overall state characteristics of blood vessels within a certain spatial range.

[0004] On the other hand, current quantitative analyses of fundus vascular status often focus on single structural parameters or simple statistical indicators, lacking a systematic characterization of vascular volume distribution, intraluminal blood flow, and the relationship between vascular diameter, making it difficult to form stable and repeatable comprehensive characterization indicators. Furthermore, image quality varies between individuals and even among the same subject at different time points. Current technologies still have shortcomings in image quality control, data consistency processing, and long-term state change analysis, affecting the reliability and comparability of assessment results.

[0005] Therefore, how to effectively fuse multimodal images based on existing fundus imaging technology, comprehensively quantify vascular structure and blood flow characteristics within a unified spatial framework, and output cerebrovascular status assessment results with stable significance remains a technical problem that needs to be further solved by existing technologies. Summary of the Invention

[0006] This invention addresses the shortcomings of existing technologies by providing a method for assessing cerebrovascular status based on fundus multimodal image fusion technology. It effectively solves the problems in existing technologies, such as inconsistent coordinate systems in fundus multimodal images and difficulty in accurately registering and fusing vascular information, which makes it difficult to achieve objective and quantitative assessment of cerebrovascular status.

[0007] The technical solution adopted by the present invention to solve the above problems is as follows:

[0008] A method for assessing cerebrovascular status based on fundus multimodal image fusion technology includes the following steps:

[0009] A) Acquire fundus color photography (CFP) and fundus computed tomography (OCT) or optical coherence tomography (OCTA) images of the same subject;

[0010] B) Preprocess the CFP and the OCT or OCTTA respectively, the preprocessing including at least noise reduction, brightness or contrast normalization and feature enhancement;

[0011] C) Extract the registration features of fundus vessels from the CFP and the OCT or OCTA respectively. The registration features include at least the course of the main vessel, the bifurcation point of the vessel, or the information of the vessel skeleton.

[0012] D) Based on the registration features, establish a cross-modal registration transformation relationship between the CFP and the OCT or OCTA, and map the OCT or OCTA to the coordinate system of the CFP to form fused data with a unified coordinate system;

[0013] E) Construct a three-dimensional region of interest (ROI) in the fused data, and segment the blood vessels within the three-dimensional ROI to obtain blood vessel body data;

[0014] F) Calculate at least one composite biomarker based on the vascular body data, wherein the composite biomarker includes at least vascular body density (VVD) or flow-vessel ratio (FVR), wherein:

[0015] The vascular volume density (VVD) is the ratio of the total vascular volume within the three-dimensional region of interest (ROI) to the total volume of the three-dimensional ROI.

[0016] The blood flow-to-vessel ratio (FVR) is a functional relationship between the effective diameter of intraluminal blood flow in the same vessel as measured in the OCT or OCTTA and the outer diameter of the vessel as measured in the CFP.

[0017] G) Input the composite biomarker into the cerebrovascular status mapping model and output the cerebrovascular status assessment results. The cerebrovascular status assessment results include at least one or more of the following: cerebrovascular strength index, stenosis risk level, inadequate perfusion risk indication, and collateral compensation degree.

[0018] Preferably, the cross-modal registration transformation relationship in step D) includes at least one of rigid registration and non-rigid registration, and the registration process includes two stages: coarse registration and fine registration. The coarse registration is based on the positioning of the optic disc center or the direction of the main blood vessel, and the fine registration is based on the matching of blood vessel bifurcation points or the minimization of the error of the blood vessel skeleton point set. When the registration error is greater than a preset threshold, iterative correction or re-registration is performed.

[0019] Preferably, in step E), the three-dimensional region of interest (ROI) covers a preset lateral range with the macula center as a reference, and covers a preset interlayer range of the retina in the axial direction.

[0020] Preferably, the lateral range of the three-dimensional region of interest (ROI) is a 6mm × 6mm area centered on the macula, and the axial interlayer range is the range from the nerve fiber layer to the inner plexiform layer.

[0021] Preferably, step E) segmenting the blood vessels within the three-dimensional ROI includes: extracting the intraluminal blood flow region in the OCT or OCTTA, extracting the outer edge region of the blood vessels in the CFP, and establishing a correspondence between the outer diameter of the blood vessels and the effective diameter of the intraluminal blood flow in a unified coordinate system.

[0022] Preferably, the vascular density (VVD) is calculated as follows:

[0023] VVD= ×100%

[0024] Wherein, Vvessel is the total volume of blood vessels within the three-dimensional ROI, and VROI is the total volume of the three-dimensional ROI.

[0025] Preferably, the blood flow vessel ratio (FVR) is determined as follows: for the same vessel, the effective diameter r of the intraluminal blood flow measured in the OCT or OCTTA and the outer diameter R of the vessel measured in the CFP are obtained, and the FVR is calculated based on r and R, wherein the FVR is the diameter ratio or area ratio of r to R.

[0026] Preferably, the FVR is calculated in the form of an area ratio, satisfying the following:

[0027]

[0028] Where r is the effective diameter of blood flow within the lumen, and R is the outer diameter of the blood vessel.

[0029] Preferably, before step F), an arteriovenous classification step is included. The arteriovenous classification step marks the blood vessels in the fused data as arteries and veins based on the color features, reflection features, or blood vessel course features of the CFP, and calculates the arterial feature set or the vein feature set respectively. The arterial feature set or the vein feature set includes at least one or more of the following: arterial VVD, vein VVD, arteriovenous VVD ratio, arterial FVR statistic, or vein FVR statistic.

[0030] Preferably, the preprocessing in step B) further includes quality control, which includes scoring the imaging quality of the CFP, OCT, or OCTA, and performing any one or more of the following when the score is lower than a preset threshold: rejection, prompting re-acquisition, or super-resolution enhancement processing; and step A) can acquire multi-time point image data of the same subject, and step G) further outputs the trend change results of cerebrovascular status for use in assessing the progression of cerebrovascular status or the effectiveness of intervention.

[0031] Compared with the prior art, the present invention has the following advantages:

[0032] This invention addresses the problems of inconsistent coordinate systems and inaccurate correspondence of vascular information in existing technologies by introducing a multimodal fusion processing method that integrates fundus color imaging with optical coherence tomography (OCT) or OCT angiography. It utilizes stable features such as vascular trunk shape, bifurcation points, and vascular skeleton to establish cross-modal registration and transformation relationships, mapping vascular information acquired from different modalities to a unified coordinate system for fusion analysis. This avoids the information fragmentation caused by independent analysis of each modality, improves the spatial consistency and comparability of multi-source vascular information, and provides a reliable foundation for the subsequent comprehensive quantification of vascular features.

[0033] Building upon this foundation, the present invention further constructs a three-dimensional region of interest within the fused data, segments the blood vessels within this region, and acquires vascular body data. By introducing composite biomarkers such as vascular body density and blood flow-to-vascular ratio, the original analysis method based on two-dimensional morphological parameters or single statistical indicators is expanded into a regionalized quantitative assessment method based on three-dimensional space. This method not only reflects the overall distribution of blood vessels within a certain spatial range but also comprehensively characterizes the relationship between the external structure of blood vessels and the intraluminal blood flow state, making the obtained assessment results more comprehensive, stable, and repeatable.

[0034] Furthermore, this invention introduces an imaging quality control mechanism in the image preprocessing stage and supports joint analysis of image data from multiple time points of the same subject. This ensures that data obtained under different acquisition conditions have a consistent foundation before entering fusion and quantitative analysis, reducing the impact of imaging quality differences on the assessment results. At the same time, through unified modeling and output of composite biomarkers, it achieves indexation, grading, and trend characterization of cerebrovascular status, enhancing the applicability of assessment results in long-term follow-up and status change analysis, and improving the reliability and practical value of the overall method in practical applications. Attached Figure Description

[0035] Figure 1 This is a flowchart of a method for assessing cerebrovascular status based on fundus multimodal image fusion technology according to the present invention. Detailed Implementation

[0036] The following are specific embodiments of the present invention, and the technical solutions of the present invention will be further described in conjunction with the accompanying drawings. However, the present invention is not limited to these embodiments.

[0037] like Figure 1 As shown in this embodiment, a method for assessing cerebrovascular status based on fundus multimodal image fusion technology first involves acquiring fundus color photography (CFP) and fundus computed tomography (OCT) or optical coherence tomography (OCTA) images of the same subject. The CFP image can be acquired using existing commercial fundus photography equipment, preferably a color fundus image with a field of view between 30° and 60° and a resolution of at least 2048×2048 pixels, used to clearly present the overall course, bifurcation structure, and outer contour of retinal vessels. The OCT or OCTA image can be acquired using existing commercial optical coherence tomography equipment, preferably volumetric data covering a scanning area of ​​at least 6mm×6mm, with an axial resolution better than 10μm and an inter-slice resolution better than 5μm, used to acquire different inter-slice structures of the retina and intravascular blood flow signals. The aforementioned multimodal images can be acquired in the same examination or separately within a short time interval, as long as the images correspond to the same subject.

[0038] After acquiring the aforementioned multimodal images, preprocessing operations are performed on the CFP and OCT or OCTA images respectively. This preprocessing is used to eliminate the impact of noise, brightness unevenness, and contrast differences introduced during imaging on subsequent analysis. Specifically, it may include denoising the CFP images using Gaussian filtering or median filtering, brightness and contrast normalization using histogram equalization or adaptive contrast enhancement, and edge enhancement or multi-scale feature enhancement for vascular regions. Simultaneously, inter-slice smoothing, speckle noise suppression, and blood flow signal enhancement are performed on the OCT or OCTA volume data. Through these preprocessing operations, image data from different devices and acquisition conditions have a relatively consistent image quality foundation before entering the fusion analysis, thereby reducing the impact of image quality differences on the final evaluation results.

[0039] After preprocessing, vascular registration features for cross-modal registration are extracted from CFP images and OCT or OCTA images, respectively. These registration features are structural information with stable spatial significance across different modalities, preferably including the course of the main vascular trunk, vascular bifurcation points, and vascular skeleton information. Specifically, in CFP images, the vascular skeleton can be extracted using vascular segmentation and thinning algorithms, and the direction of the main vascular trunk and bifurcation nodes can be identified. In OCT or OCTA images, corresponding vascular structural features can be extracted by processing vascular projection maps or blood flow signal layers. Since these features have good correspondences in different modalities of imagery, they can serve as the basis for establishing cross-modal registration relationships.

[0040] Based on the extracted vascular registration features, a cross-modal registration transformation relationship is established between CFP and OCT or OCTA. This registration process preferably includes two stages: coarse registration and fine registration. Coarse registration is used to achieve initial alignment between images of different modalities, and can be performed based on the optic disc center position, the overall direction of the main vascular trunk, or image scale information to complete translation, rotation, and scale unification. After coarse registration, fine registration further improves the registration accuracy. Fine registration can be achieved based on vascular bifurcation point matching or error minimization methods of vascular skeleton point sets to compensate for local nonlinear deformation. During the registration process, the registration error can be evaluated in real time. When the registration error exceeds a preset threshold, iterative correction or re-registration is automatically performed to ensure that the final cross-modal registration result meets the accuracy requirements of fusion analysis. In this way, OCT or OCTA images are mapped to the CFP coordinate system, forming fusion data with a unified coordinate system, thereby solving the problems of inconsistent coordinate systems and difficulty in accurately corresponding vascular information in existing technologies.

[0041] After obtaining the fused data, a three-dimensional region of interest (ROI) is constructed in a unified coordinate system. The ROI preferably uses the macula as a lateral reference, covering a 6mm × 6mm area centered on the macula, and axially covering a predetermined interlayer range from the retinal nerve fiber layer to the inner plexus layer, thus forming a three-dimensional analysis area with clearly defined spatial boundaries. By defining the ROI, irrelevant areas can be avoided from being introduced into the analysis process, enhancing the comparability of evaluation results between different subjects and between different time points of the same subject. Within the ROI, blood vessels are segmented to obtain vascular body data. Specifically, the intravascular blood flow region is extracted from OCT or OCTA images, and the outer edge region of the blood vessels is extracted from CFP images. A correspondence between the outer diameter of the blood vessels and the effective diameter of the intravascular blood flow is established in a unified coordinate system.

[0042] Based on the aforementioned vascular body data, at least one composite biomarker is further calculated. The vascular body density (VVD) is obtained by calculating the ratio of the total vascular volume within a three-dimensional region of interest (ROI) to the total volume of the ROI, reflecting the overall distribution of blood vessels within a certain spatial range. The flow-vessel ratio (FVR) is calculated based on the effective intraluminal blood flow diameter (r) measured in OCT or OCTA for the same blood vessel, and the outer diameter (R) measured in CFP. It can be expressed as a diameter ratio or an area ratio to comprehensively characterize the relationship between the vascular external structure and the intraluminal blood flow state. By introducing the aforementioned composite biomarker, the joint quantification of vascular structural characteristics and blood flow characteristics is achieved, avoiding the problem of insufficient information caused by relying on only a single parameter.

[0043] Finally, the composite biomarkers are input into the cerebrovascular state mapping model, outputting cerebrovascular state assessment results. These results can be represented in an indexed, graded, or trend-based manner, such as outputting information like cerebrovascular strength index, stenosis risk level, inadequate perfusion risk indication, or collateral compensation degree, thereby objectively characterizing cerebrovascular state under non-invasive conditions. Furthermore, by supporting the joint processing of imaging data from multiple time points for the same subject, the model can also output trend results of cerebrovascular state changes over time, which can be used for long-term state change analysis or intervention effect evaluation.

[0044] Based on the above embodiments, this embodiment further explains the arteriovenous classification steps and imaging quality control steps to improve the stability and reliability of the cerebrovascular status assessment method in practical applications.

[0045] In some embodiments, before calculating the composite biomarker, a step of classifying blood vessels in the fused data into arteries and veins is included. Since different types of blood vessels differ in hemodynamic characteristics and spatial distribution features, to avoid affecting the stability of the evaluation results by mixing different blood vessel types in the analysis, this embodiment utilizes the advantage of fundus color photography (CFP) images, which can intuitively reflect the color, reflectivity, and course characteristics of blood vessels, to mark arteries and veins. Specifically, blood vessels can be initially classified in CFP images based on differences in color intensity, reflectivity distribution, and morphological characteristics. Simultaneously, the classification results are corrected for consistency by combining the spatial connectivity of blood vessels in a unified coordinate system, thereby forming relatively stable arterial and venous vessel sets. After completing the arteriovenous classification in the above manner, the corresponding vessel volume density (VVD), flow-vessel ratio (FVR), and their statistics can be calculated for the arterial and venous vessel sets respectively, further obtaining characteristic information such as arterial VVD, venous VVD, arteriovenous VVD ratio, arterial FVR statistics, or venous FVR statistics. This implementation method can make full use of the existing structural and color information in multimodal images without increasing the additional acquisition burden, so that the quantitative results of vascular status are more detailed at the vascular type level, which helps to improve the discrimination and consistency of the assessment results.

[0046] In another embodiment, to reduce the impact of imaging quality differences on the multimodal image fusion and quantization results, an imaging quality control step is introduced in the image preprocessing stage. Specifically, the acquired CFP images and OCT or OCTA images can be scored separately, and the imaging quality score can comprehensively consider factors such as image sharpness, noise level, contrast, and the recognizability of key structures. When the imaging quality score is lower than a preset threshold, different processing strategies can be implemented according to actual application needs, such as directly discarding the corresponding image data, prompting for re-acquisition of images, or using super-resolution enhancement processing on the image data to improve image quality. By introducing the above quality control step before entering the cross-modal registration and fusion analysis, the interference of low-quality images on registration accuracy and subsequent vascular body data calculation can be effectively reduced, thereby improving the stability of the overall evaluation process.

[0047] Furthermore, in some embodiments, the method also supports joint processing of fundus multimodal image data acquired at multiple time points from the same subject. In this embodiment, the aforementioned preprocessing, registration and fusion, 3D ROI construction, vascular segmentation, and composite biomarker calculation steps are performed on image data acquired at different time points. Based on this, the assessment results of cerebrovascular status at different time points are compared and analyzed to output the trend results of cerebrovascular status changes over time. By introducing joint analysis with a time dimension, the applicability of the method in long-term status change assessment and intervention effect assessment can be further improved while ensuring the consistency of single assessment results, making the obtained assessment results more continuous and valuable for reference.

[0048] In a further detailed implementation, the following describes the practical application process of the method of the present invention in single-subject assessment and multi-time-point assessment, with the aid of a complete application example, so as to more intuitively understand the implementation process and effects of the present invention.

[0049] In a single-subject, single-assessment application example, one subject is first selected as the assessment object, and their fundus color photography (CFP) and optical coherence tomography (OCT) or optical coherence tomography angiography (OCTA) images are acquired. The CFP images are acquired using standard fundus photography equipment, ensuring the field of view covers the macular region and surrounding major vascular distribution areas. The OCT or OCTA images are acquired using the corresponding scanning mode, with the scanning range covering at least a 6mm × 6mm area based on the macula center, including interlaminar structural information from the retinal nerve fiber layer to the inner plexus layer. After image acquisition, the acquired multimodal images are preprocessed, including denoising, brightness or contrast normalization, and vascular feature enhancement for CFP images; and noise suppression, interlaminar smoothing, and blood flow signal enhancement for OCT or OCTA images. A preliminary assessment of image quality is also performed to ensure that images entering subsequent processing meet preset quality requirements.

[0050] After image preprocessing, vascular registration features are extracted from CFP and OCT or OCTA images, including the course of the main vessels, bifurcation points, and vascular skeleton information. A cross-modal registration transformation relationship is then established based on these features. Specifically, coarse registration is used to initially align images between different modalities, followed by fine registration to further reduce spatial errors between bifurcation points and skeleton point sets. When the registration error exceeds a preset threshold, iterative correction is automatically performed to improve registration accuracy. After registration, the OCT or OCTA images are mapped to the CFP image coordinate system, forming fused data in a unified coordinate system. In this fused data, a three-dimensional region of interest (ROI) is constructed with the macula center as a reference. The vessels within the ROI are segmented to obtain the outer edge region and the intraluminal blood flow region, thus forming complete vascular body data.

[0051] Based on the vascular body data, composite biomarkers are further calculated. In this application example, the vascular volume density (VVD) is obtained by calculating the ratio of the total vascular volume within the three-dimensional region of interest (ROI) to the total volume of the three-dimensional ROI. For selected vessels, the effective intraluminal blood flow diameter (r) measured in OCT or OCTA images and the vessel outer diameter (R) measured in CFP images are obtained. The flow-vessel ratio (FVR) is calculated based on r and R. Subsequently, the obtained VVD and FVR are input into a preset cerebrovascular state mapping model, which outputs the corresponding cerebrovascular state assessment results, such as cerebrovascular strength index or risk level, thereby completing the cerebrovascular state assessment of the subject at that time point.

[0052] In multi-timepoint assessment application examples, the above image acquisition and assessment process is repeated for the same subject at different time points. For example, after the initial assessment, CFP and OCT or OCTA images of the subject are acquired again at preset time intervals, and the newly acquired image data are sequentially processed through preprocessing, cross-modal registration, fusion, 3D ROI construction, vessel segmentation, and composite biomarker calculation. To ensure the comparability of assessment results between different time points, the 3D ROI construction rules and parameters used at each time point are consistent, and the cerebrovascular state mapping model used remains unchanged. By comparing and analyzing the VVD, FVR, and their corresponding cerebrovascular state assessment results obtained at different time points, trend information on the changes in cerebrovascular state over time can be obtained.

[0053] As can be seen from the above application examples of single-subject assessment and multi-time-point assessment, the method provided by this invention can perform consistent assessment of the cerebrovascular status of the same subject at different time points under non-invasive conditions. By using multimodal image fusion and composite biomarkers, the method reduces the randomness caused by single image modality or single assessment, making the assessment results more stable and intuitive, and facilitating long-term status change analysis or intervention effect assessment in practical applications.

[0054] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to replace them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for assessing cerebrovascular status based on fundus multimodal image fusion technology, characterized in that, Includes the following steps: A) Acquire fundus color photography (CFP) and fundus computed tomography (OCT) or optical coherence tomography (OCTA) images of the same subject; B) Preprocess the CFP and the OCT or OCTTA respectively, the preprocessing including at least noise reduction, brightness or contrast normalization and feature enhancement; C) Extract the registration features of fundus vessels from the CFP and the OCT or OCTA respectively. The registration features include at least the course of the main vessel, the bifurcation point of the vessel, or the information of the vessel skeleton. D) Based on the registration features, establish a cross-modal registration transformation relationship between the CFP and the OCT or OCTA, and map the OCT or OCTA to the coordinate system of the CFP to form fused data with a unified coordinate system; E) Construct a three-dimensional region of interest (ROI) in the fused data, and segment the blood vessels within the three-dimensional ROI to obtain blood vessel body data; F) Calculate at least one composite biomarker based on the vascular body data, wherein the composite biomarker includes at least vascular body density (VVD) or flow-vessel ratio (FVR), wherein: The vascular volume density (VVD) is the ratio of the total vascular volume within the three-dimensional region of interest (ROI) to the total volume of the three-dimensional ROI. The blood flow-to-vessel ratio (FVR) is a functional relationship between the effective diameter of intraluminal blood flow in the same vessel as measured in the OCT or OCTTA and the outer diameter of the vessel as measured in the CFP. G) Input the composite biomarker into the cerebrovascular status mapping model and output the cerebrovascular status assessment results. The cerebrovascular status assessment results include at least one or more of the following: cerebrovascular strength index, stenosis risk level, inadequate perfusion risk indication, and collateral compensation degree.

2. The method for assessing cerebrovascular status based on fundus multimodal image fusion technology as described in claim 1, characterized in that: The cross-modal registration transformation relationship in step D) includes at least one of rigid registration and non-rigid registration, and the registration process includes two stages: coarse registration and fine registration. Coarse registration is based on the positioning of the optic disc center or the direction of the main blood vessel, while fine registration is based on the matching of blood vessel bifurcation points or the minimization of the error of the blood vessel skeleton point set. When the registration error is greater than a preset threshold, iterative correction or re-registration is performed.

3. The method for assessing cerebrovascular status based on fundus multimodal image fusion technology as described in claim 1, characterized in that: In step E), the three-dimensional region of interest (ROI) covers a preset lateral range with the macula center as the reference, and covers a preset interlayer range of the retina in the axial direction.

4. The method for assessing cerebrovascular status based on fundus multimodal image fusion technology as described in claim 3, characterized in that: The lateral extent of the three-dimensional region of interest (ROI) is a 6mm × 6mm area centered on the macula, and the axial interlayer extent is the range from the nerve fiber layer to the inner plexiform layer.

5. The method for assessing cerebrovascular status based on fundus multimodal image fusion technology as described in claim 1, characterized in that: Step E) segmenting the blood vessels within the three-dimensional ROI includes: extracting the intraluminal blood flow region in the OCT or OCTTA, extracting the outer edge region of the blood vessels in the CFP, and establishing a correspondence between the outer diameter of the blood vessels and the effective diameter of the intraluminal blood flow in a unified coordinate system.

6. The method for assessing cerebrovascular status based on fundus multimodal image fusion technology as described in claim 1, characterized in that: The vascular density (VVD) is calculated as follows: VVD= ×100% Among them, V vessel V represents the total volume of blood vessels within the three-dimensional ROI. ROI The total volume of the three-dimensional ROI is given.

7. The method for assessing cerebrovascular status based on fundus multimodal image fusion technology as described in claim 1, characterized in that: The blood flow vessel ratio (FVR) is determined as follows: For the same vessel, the effective diameter r of the intraluminal blood flow measured in the OCT or OCTTA and the outer diameter R of the vessel measured in the CFP are obtained, and the FVR is calculated based on r and R, where FVR is the diameter ratio or area ratio of r to R.

8. The method for assessing cerebrovascular status based on fundus multimodal image fusion technology as described in claim 7, characterized in that: The FVR is calculated using the area ratio method, satisfying the following: Where r is the effective diameter of blood flow within the lumen, and R is the outer diameter of the blood vessel.

9. The method for assessing cerebrovascular status based on fundus multimodal image fusion technology as described in claim 1, characterized in that: Before step F), there is also an arteriovenous classification step, which marks the blood vessels in the fused data as arteries and veins based on the color features, reflection features or blood vessel course features of the CFP, and calculates the arterial feature set or the vein feature set respectively. The arterial feature set or the vein feature set includes at least one or more of the following: arterial VVD, vein VVD, arteriovenous VVD ratio, arterial FVR statistic or vein FVR statistic.

10. The method for assessing cerebrovascular status based on fundus multimodal image fusion technology as described in claim 1, characterized in that: The preprocessing in step B) also includes quality control, which includes scoring the imaging quality of the CFP, OCT, or OCTA. When the score is lower than a preset threshold, the system will perform rejection, prompt for re-acquisition, or use any one or more of super-resolution enhancement processing. Furthermore, step A) can acquire multi-time point image data of the same subject, and step G) further outputs the trend change results of cerebrovascular status for use in assessing the progression of cerebrovascular status or the effectiveness of intervention.

Images