Method and system for measuring blood oxygen saturation based on eye images

Abstract

The embodiment of the present application provides a kind of blood oxygen saturation measurement method and system based on eye image, belong to image processing technical field.The method comprises: the image information of the same eye of user under two different shooting wavelengths is collected, the image of relatively long shooting wavelength and the image of relatively short shooting wavelength are obtained respectively, and the optic disc position of two images is positioned respectively;After the global registration of two images based on the optic disc position of two images, each image is carried out blood vessel segmentation, and the corresponding blood vessel segmentation image is obtained;Based on two blood vessel segmentation images, the segmentation information of arterial blood vessel and venous blood vessel is obtained;The blood oxygen saturation of current user's eye is calculated based on the segmentation information of arterial blood vessel and venous blood vessel.The present application solves the problem that the blood oxygen saturation measurement cannot be applied due to the low accuracy of eye blood vessel division.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a blood oxygen saturation measurement method based on eye images and a blood oxygen saturation measurement system based on eye images. BACKGROUND

[0002] Blood oxygen saturation is the percentage of oxygenated hemoglobin in the blood, which is the concentration of blood oxygen in the blood, and it is an important physiological parameter of respiratory circulation. Accurate measurement of blood oxygen saturation is a problem that has been concerned and researched in the medical field. The method widely used at present is to measure blood oxygen saturation based on optical density measurement method. The absorption spectrum of hemoglobin (a protein that carries oxygen in the blood) is very different in two states, namely its oxygen-containing state and non-oxygen-containing state. In simple terms, the color projected out will change when the blood oxygen saturation is different. Based on this feature, blood oxygen saturation can be measured through blood vessel images.

[0003] The human eye includes a large number of pulse blood vessels. Because of the structure of the eye, these pulse blood vessels are ideal targets for blood oxygen saturation measurement (with obvious color characteristics), so if blood oxygen saturation measurement can be based on eye blood vessels, the accuracy of blood oxygen saturation will be very high. However, because there are many pulse blood vessels in the eye, the venous blood vessels and the arterial blood vessels are distributed in a cross manner, which makes it difficult to distinguish the blood vessels. Because the blood oxygen saturation standards of arterial blood vessels and venous blood vessels are not the same, if the venous blood vessels and the arterial blood vessels cannot be accurately distinguished, the blood oxygen saturation measurement result obtained is also unreliable. At present, the eye blood vessel distinction mainly relies on manual annotation, but this method is not ideal in terms of labor cost and recognition accuracy, and it is the biggest obstacle to the blood oxygen saturation measurement scheme based on eye images. In view of the problem that the existing method cannot apply blood oxygen saturation measurement because of the low accuracy of eye blood vessel distinction, a new blood oxygen saturation measurement method based on eye images needs to be created. SUMMARY

[0004] The purpose of the embodiments of the present application is to provide a blood oxygen saturation measurement method and system based on eye images to at least solve the problem that the existing method cannot apply blood oxygen saturation measurement because of the low accuracy of eye blood vessel distinction.

[0005] In order to achieve the above object, the first aspect of the present application provides a blood oxygen saturation measurement method based on eye images, the method comprising: collecting image information of the same eye of a user under two different shooting wavelengths, obtaining an image with a longer shooting wavelength and an image with a shorter shooting wavelength respectively, and positioning the optic disc positions of the two images respectively; after global registration of the two images based on the optic disc positions of the two images, performing blood vessel segmentation on each image to obtain the corresponding blood vessel segmentation image; obtaining the segmentation information of the arterial blood vessels and the venous blood vessels based on the two blood vessel segmentation images; and calculating the blood oxygen saturation of the eye of the current user based on the segmentation information of the arterial blood vessels and the venous blood vessels.

[0006] Optionally, the two different shooting wavelengths are 570±5nm and 600±5nm respectively.

[0007] Optionally, the positioning of the optic disc positions of the two images comprises: positioning the optic disc region in the image based on the yolov5 algorithm, and fitting the elliptical contour of the optic disc through the circumscribed rectangle; wherein the circumscribed rectangle is the detection box of the yolov5 algorithm.

[0008] Optionally, the global registration of the two images based on the optic disc positions of the two images comprises: moving the image with the shorter shooting wavelength to register the image with the longer shooting wavelength, comprising: 1) obtaining ΔX and ΔY by differencing the center coordinates of the optic disc positions of the two images; 2) moving the image with the shorter shooting wavelength based on ΔX and ΔY, comprising: moving the image with the shorter shooting wavelength by ±(|ΔX|+2) and ±(|ΔY|+2) along the x-axis and y-axis respectively, and the specific moving distance is determined by the evaluation parameter between the image with the shorter shooting wavelength and the image with the longer shooting wavelength during the moving process; the evaluation parameter is: wherein τ is the evaluation parameter; δ is the pixel difference value of the two images; and SSIM is the structural similarity of the two images. The image coordinates corresponding to the minimum evaluation parameter in the moving process are selected as the final coordinates of the image with the shorter shooting wavelength.

[0009] Optionally, the blood vessel segmentation of each image to obtain the corresponding blood vessel segmentation image comprises: respectively pre-processing the two images by using Gaussian filtering; segmenting the venous blood vessels in the pre-processed image with the longer shooting wavelength based on the U2Net algorithm to obtain the blood vessel segmentation image of the image with the longer shooting wavelength; enhancing the pre-processed image with the shorter shooting wavelength based on the dehazing algorithm to obtain the enhanced image with the shorter shooting wavelength; and segmenting all arterial blood vessels in the enhanced image with the shorter shooting wavelength based on the U2Net algorithm to obtain the blood vessel segmentation image of the image with the shorter shooting wavelength.

[0010] Optionally, the obtaining of the segmentation information of the arterial blood vessels and the venous blood vessels based on the two images comprises: taking the blood vessel segmentation image of the image with the longer wavelength as the segmentation information of the venous blood vessels; registering the blood vessel segmentation image of the image with the shorter wavelength to the blood vessel segmentation image of the image with the longer wavelength; and subtracting the blood vessel segmentation image mask of the image with the shorter wavelength from the blood vessel segmentation image mask of the image with the longer wavelength to obtain the arterial blood vessel segmentation image as the segmentation information of the arterial blood vessels.

[0011] Optionally, the calculating of the blood oxygen saturation of the current user's eye based on the segmentation information of the arterial blood vessels and the venous blood vessels comprises: extracting the skeletons of the arterial blood vessels and the venous blood vessels and obtaining the center lines of the arterial blood vessels and the venous blood vessels; traversing the center lines of the arterial blood vessels and the venous blood vessels based on a preset step length to obtain the blood vessel boundary pixel values of the arterial blood vessels and the venous blood vessels respectively; identifying the macular position in the image with the longer wavelength, taking the line connecting the center point of the optic disc position and the center point of the macular position in the image with the longer wavelength as the horizontal coordinate, and taking the center point of the optic disc as the origin to construct a rectangular coordinate system; and using the optical density measurement method to calculate the venous blood oxygen saturation and the arterial blood oxygen saturation in the four quadrant regions of the rectangular coordinate system based on the blood vessel boundary pixel values of the arterial blood vessels and the venous blood vessels respectively; and calculating the average values of the venous blood oxygen saturation and the arterial blood oxygen saturation in each quadrant to obtain the venous blood oxygen saturation and the arterial blood oxygen saturation of the current user's eye.

[0012] Optionally, the range of the constructed rectangular coordinate system is a region with a 2r-5r distance expanded outward from the center.

[0013] The second aspect of the present application provides an eye image-based blood oxygen saturation measurement system, which comprises: an acquisition unit configured to acquire image information of the same eye of a user under two different shooting wavelengths, to obtain an image with a longer wavelength and an image with a shorter wavelength respectively, and to locate the respective optic disc positions of the two images; a processing unit configured to: after globally registering the two images based on the optic disc positions of the two images, perform blood vessel segmentation on each image to obtain a corresponding blood vessel segmentation image; obtain segmentation information of arterial blood vessels and venous blood vessels based on the two blood vessel segmentation images; and a detection unit configured to calculate the blood oxygen saturation of the current user's eye based on the segmentation information of the arterial blood vessels and the venous blood vessels.

[0014] In another aspect, the present application provides a computer-readable storage medium having instructions stored thereon, which, when executed on a computer, cause the computer to perform the above-mentioned eye image-based blood oxygen saturation measurement method.

[0015] By the technical scheme, the present application scheme is based on the display difference of veins and arteries in different shooting wavelength images, carries out vein and artery segmentation, and proposes a registration method for ensuring segmentation accuracy. Based on the veins and arteries obtained by accurate segmentation, blood oxygen saturation measurement is carried out, the measurement accuracy of blood oxygen saturation is ensured by improving the accuracy of blood vessel segmentation, and the problem that the blood oxygen saturation measurement cannot be applied due to the low accuracy of eye blood vessel division is solved.

[0016] Other features and advantages of the present application will be described in detail in the following detailed description of the application. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and serve to explain the principles of the application. In the drawings:

[0018] Figure 1 is a step flow chart of the blood oxygen saturation measurement method based on eye images provided by an embodiment of the present application;

[0019] Figure 2 is an eye blood vessel state schematic diagram under different wavelengths provided by an embodiment of the present application;

[0020] Figure 3 is a blood vessel segmentation image effect schematic diagram of the image with longer shooting wavelength provided by an embodiment of the present application;

[0021] Figure 4 is a blood vessel segmentation image effect schematic diagram of the image with shorter shooting wavelength provided by an embodiment of the present application;

[0022] Figure 5 is a system structure diagram of the blood oxygen saturation measurement system based on eye images provided by an embodiment of the present application. DETAILED DESCRIPTION

[0023] The specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application.

[0024] The blood oxygen saturation is the percentage of the oxygenated hemoglobin combined with oxygen in the blood, that is, the concentration of blood oxygen in the blood, which is an important physiological parameter of the respiratory cycle. The accurate measurement of the blood oxygen saturation is a problem that has been concerned and studied in the medical related field. The widely used method is to measure the blood oxygen saturation based on the light density measurement method. The absorption spectrum of hemoglobin (the protein that carries oxygen in the blood) is very different in two states, namely, its oxygen-containing state and non-oxygen-containing state. In short, the color projected out will change with the different blood oxygen saturation, and based on this feature, the blood oxygen saturation can be measured through the blood vessel image.

[0025] The human eye includes a large number of pulse blood vessels. Because of the structure of the eye, the pulse blood vessels are ideal blood oxygen saturation measurement targets (color features are obvious), so if the blood oxygen saturation can be measured based on the eye blood vessels, the blood oxygen saturation accuracy is very high. However, because there are many pulse blood vessels in the eye, the venous blood vessels and the arterial blood vessels are cross-distributed, which causes the difficulty in distinguishing the blood vessels. Because the blood oxygen saturation standards of the arterial blood vessels and the venous blood vessels are different, if the venous blood vessels and the arterial blood vessels cannot be accurately distinguished, the obtained blood oxygen saturation measurement result is also unreliable. At present, the eye blood vessel distinction mainly relies on manual annotation, but the manual cost and the recognition accuracy are not ideal, which is the biggest obstacle to the blood oxygen saturation measurement scheme based on the eye image.

[0026] In view of the problem that the existing method cannot be applied to the blood oxygen saturation measurement because of the low accuracy of the eye blood vessel distinction, the present scheme provides a new blood oxygen saturation measurement method based on the eye image. The present scheme distinguishes the venous blood vessels and the arterial blood vessels based on the display difference of the venous blood vessels and the arterial blood vessels in different shooting wavelength images, and proposes a registration method for ensuring the segmentation accuracy. The venous blood vessels and the arterial blood vessels obtained based on the accurate segmentation are used for the blood oxygen saturation measurement, the measurement accuracy of the blood oxygen saturation is ensured by improving the blood vessel segmentation accuracy, and the problem that the existing method cannot be applied to the blood oxygen saturation measurement because of the low accuracy of the eye blood vessel distinction is solved.

[0027] Figure 1 is a method flowchart of the blood oxygen saturation measurement method based on the eye image provided by an embodiment of the present application. As shown in Figure 1 the present embodiment provides a blood oxygen saturation measurement method based on the eye image, which comprises the following steps:

[0028] Step S10: Collecting the image information of the same eye of a user under two different shooting wavelengths, respectively obtaining the image with a longer shooting wavelength and the image with a shorter shooting wavelength, and positioning the optic disc positions of the two images respectively.

[0029] Specifically, the present application is also based on image analysis to perform optical density measurement method for blood oxygen saturation measurement, so its object is eye image information. There is a large color distinction between venous blood vessels and arterial blood vessels, and by using this feature and the corresponding relationship between color and wavelength, the venous and arterial blood vessels can be distinguished and processed by image acquisition at different wavelengths. Based on the distinguished image information, the arterial and venous blood vessels are segmented to obtain accurate venous and arterial targets. As shown in Figure 2 As shown in the two images, all blood vessels in the image at 570 nm wavelength are shown, while only venous blood vessels are highlighted in the image at 600 nm wavelength. In theory, as long as all images and images with only venous blood vessels are found, the arterial blood vessels can be obtained based on the difference between the two.

[0030] Preferably, the two different wavelengths are 570±5nm and 600±5nm.

[0031] Specifically, the former is to record all the arterial blood vessels, and the latter is to record the venous blood vessels.

[0032] Preferably, the positioning of the respective optic disc positions comprises: positioning the optic disc area in the image based on the yolov5 algorithm, and fitting the elliptical contour of the optic disc through the circumscribed rectangle; wherein the circumscribed rectangle is the detection box of the yolov5 algorithm.

[0033] Specifically, although the arterial and venous blood vessels can be segmented by two images at different wavelengths, it cannot be guaranteed that the two images match in the global range, that is, if the two images cannot be accurately registered, the arterial blood vessels obtained based on the difference between the two images cannot be realized. The width of the eye blood vessels is small, and accurate registration of the two images is required to ensure that the coordinates of the same points in the two images are the same in the same coordinate system. Based on this, the corresponding registration scheme is proposed in the present application. Because the optic disc information of the same eye is unchanged, the optic disc can be used as a reference for image registration. Therefore, the optic disc positioning is preferred.

[0034] At present, there are generally two schemes for locating the optic disc: finding the brightest area in the picture, and finding the center of the brightest area as the center of the optic disc. This scheme is prone to misjudgment when the shooting is not standardized and there is reflection, and the determined radius of the optic disc is not accurate. The other scheme is to segment the optic disc through deep learning, and find the center and radius according to the segmentation result. This method completely depends on the accuracy of the optic disc segmentation, and the segmentation effect cannot be guaranteed when the boundary of the optic disc is not clear. In addition, there will be deviation when determining the center. The scheme of the present application detects the position of the optic disc through the yolov5 algorithm, and the detection frame is the circumscribed rectangle of the optic disc. The optic disc is fitted as an elliptical shape, the center is the center of the ellipse, and the radius is the distance from the center of the ellipse to the long side of the ellipse.

[0035] Step S20: After global registration of the two images based on the positions of the optic discs of the two images, the blood vessels of each image are segmented to obtain the corresponding blood vessel segmentation images.

[0036] Specifically, as known above, to achieve accurate segmentation of venous and arterial blood vessels through two images, it is necessary to ensure accurate registration of the two images. The scheme of the present application takes the optic disc as the reference and takes the image with longer shooting wavelength as the target, and moves the image with shorter shooting wavelength for registration, including: taking the difference between the center coordinates of the positions of the optic discs of the two images to obtain ΔX and ΔY; moving the image with shorter shooting wavelength based on ΔX and ΔY, including: moving the image with shorter shooting wavelength by ±(|ΔX|+2) and ±(|ΔY|+2) along the x-axis and y-axis, respectively, and the specific moving distance is determined by the evaluation parameter between the image with shorter shooting wavelength and the image with longer shooting wavelength during the moving process; the evaluation parameter is:

[0037]

[0038] wherein τ is the evaluation parameter; δ is the pixel difference value of the two images; and SSIM is the structural similarity of the two images. The image coordinates corresponding to the minimum evaluation parameter during the moving process are selected as the final coordinates of the image with shorter shooting wavelength.

[0039] Among them, the structural similarity (Structural Similarity, SSIM) is an index for measuring the similarity of two images. The index is first proposed by the Laboratory for Image and Video Engineering of the University of Texas at Austin. SSIM uses two images, one of which is an uncompressed undistorted image, and the other is a distorted image. As the implementation of the structural similarity theory, the structural similarity index defines the structural information as independent of brightness, contrast, reflects the properties of the structure in the scene, and models the distortion as a combination of brightness, contrast and structure. The mean value is used as the estimate of the brightness, the standard deviation is used as the estimate of the contrast, and the covariance is used as the measure of the structural similarity. The above ± (|DeltaX| + 2) and ± (|DeltaY| + 2) are traversed and moved, and the evaluation parameters after each moving point are counted. The smaller the evaluation parameter is, the higher the matching degree of the two is. By finding the smallest evaluation parameter, the registration of the two images can be accurately realized.

[0040] Further, the blood vessels in the corresponding images are segmented based on the respective images to obtain respective blood vessel segmentation images, including: respectively performing Gaussian filtering processing on the two images to obtain preprocessed images; as Figure 3 Based on the U2Net algorithm, the venous blood vessels in the preprocessed image with a longer shooting wavelength are segmented as the blood vessel segmentation image of the image with a longer shooting wavelength;

[0041] Further, as Figure 4 Based on the dehazing algorithm, the preprocessed image with a shorter shooting wavelength is enhanced to obtain an enhanced image with a shorter shooting wavelength; based on the U2Net algorithm, all arterial blood vessels in the enhanced image with a shorter shooting wavelength are segmented as the blood vessel segmentation image of the image with a shorter shooting wavelength.

[0042] Specifically, the directly collected image has a large image noise, which will affect the subsequent pixel statistics of the blood vessel segmentation, and the noise needs to be removed, so image preprocessing needs to be performed on the two images to eliminate image noise. The image denoising processing of the present application is based on Gaussian filtering, which is a linear smoothing filter suitable for eliminating Gaussian noise and widely used in the noise reduction process of image processing. It scans each pixel in the image with a template (or convolution, mask), and replaces the value of the center pixel point with the weighted average gray value of the pixels in the neighborhood determined by the template.

[0043] After completing the image preprocessing, by Figure 2It is known that the display characteristics of blood vessels in the image are obvious, and based on the characteristics, the blood vessels can be simply segmented by using the U2Net algorithm. However, because the brightness of the image with a short wavelength is low, if the blood vessels are directly segmented, the segmentation may fail due to the dark scene, and the image with a long wavelength does not have this problem. Therefore, in order to accurately segment the blood vessels in the image with a short wavelength, the image needs to be enhanced. The present application enhances the preprocessed image with a short wavelength based on a dehazing algorithm to obtain an enhanced image with a short wavelength. For example, MSCNN algorithm, DehazeNet algorithm, dark channel prior algorithm, etc. are used, which are not limited herein.

[0044] Step S30: obtaining the segmentation information of the arterial blood vessels and the venous blood vessels based on the two blood vessel segmentation images.

[0045] Specifically, after step S20, two blood vessel segmentation images are obtained, which are the venous blood vessel segmentation image corresponding to the image with a long wavelength and the full-pulse blood vessel image segmentation image corresponding to the image with a short wavelength. The venous blood vessel segmentation information has been obtained, and the arterial blood vessel segmentation information needs to be obtained based on the two images. Specifically, the blood vessel segmentation image of the image with a long wavelength is taken as the segmentation information of the venous blood vessels; the blood vessel segmentation image of the image with a short wavelength is registered to the blood vessel segmentation image of the image with a long wavelength; the mask image of the blood vessel segmentation image of the image with a short wavelength is subtracted from the mask image of the blood vessel segmentation image of the image with a long wavelength to obtain the arterial blood vessel segmentation image, which is taken as the segmentation information of the arterial blood vessels.

[0046] Step S40: calculating the blood oxygen saturation of the current user's eye based on the segmentation information of the arterial blood vessels and the venous blood vessels.

[0047] Specifically, the skeletons of the venous blood vessels and the arterial blood vessels are extracted, and the center lines of the venous blood vessels and the arterial blood vessels are obtained; based on a preset step length, the center lines of the venous blood vessels and the arterial blood vessels are traversed to obtain the blood vessel boundary pixel values of the venous blood vessels and the arterial blood vessels, respectively; the macular position in the image with a long wavelength is identified, and a straight coordinate system is constructed with the center line of the optic disc position and the center line of the macular position in the image as the horizontal coordinate and the center point of the optic disc as the origin; the optical density measurement method is used to calculate the venous blood oxygen saturation and the arterial blood oxygen saturation in the four quadrant regions of the straight coordinate system based on the blood vessel boundary pixel values of the venous blood vessels and the arterial blood vessels; and the venous blood oxygen saturation and the arterial blood oxygen saturation in each quadrant are averaged to obtain the venous blood oxygen saturation and the arterial blood oxygen saturation of the current user's eye.

[0048] In the embodiment of the present application, according to the divided arteriovenous, skeleton extraction is performed, skeleton intersection is found, and the intersection is divided into different sections for processing according to the intersection, so as to exclude the influence of inaccurate value at the intersection. Then, according to the small picture framed, the transition value of the pixel mean value from left to right, the pixel value on the blood vessel boundary is obtained. Compared with the conventional direct use of the boundary pixel of the segmented blood vessel, this method can ensure that the taken pixel is the value on the blood vessel, avoid the error caused by inaccurate blood vessel segmentation, and also can exclude the reflection point on the blood vessel. The vein vessel pixel is in U-shaped distribution, and the artery is in W-shaped distribution.

[0049] Specifically, the whole blood vessel is traversed, and points are taken along the way, a rectangle (such as 20*10 pixel size) is intercepted along the gradient direction of the blood vessel at each point, the rectangle is taken as a recognition frame, and the blood vessel boundary is found in the recognition frame. In the recognition frame, the transition of the pixel gray value in the rectangle from left to right is collected, and the boundary pixel point is found according to the gradient descent.

[0050] Preferably, the range of the rectangular coordinate system constructed is: a region with a center of origin and an outward expansion distance of 2r-5r; wherein r is the optic disc radius.

[0051] Further, the macular position is located according to yolov5 to determine the center point. A rectangular coordinate system is established according to the positions of the 600 wavelength macula and the optic disc, the optic disc is taken as the center, and a region with a radius of 2 times the optic disc-5 times the optic disc is set; the four quadrants are divided into upper nose, lower nose, upper temple and lower temple, and the blood oxygen saturation of each quadrant is calculated, and then the venous blood oxygen saturation and arterial blood oxygen saturation of each quadrant are averaged to obtain the venous blood oxygen saturation and arterial blood oxygen saturation of the current user's eye.

[0052] The calculation principle of blood oxygen saturation is optical density measurement. Taking the 570 wavelength image and the 600 wavelength image of the present application as an example, the optical density (OD) represents the attenuation of the outgoing light intensity I relative to the incident light intensity I0. The algorithm scans the pixel points in the blood vessel to find the minimum light intensity brightness I in the blood vessel; then the average brightness I0 outside the blood vessel is determined according to the blood vessel diameter, and then the optical density is calculated according to the formula OD = log (I0 / I). Finally, the optical density ratio ODR is calculated according to the formula ODR = OD600 / OD570. Because the size of the optical density ratio is directly related to the blood oxygen saturation (SO2), after obtaining the optical density ratio, the corresponding blood oxygen saturation can be directly obtained based on the table lookup method.

[0053] The present application can avoid calculation errors caused by regional differences, reduce the calculation region, ensure the same calculation conditions in each calculation region, and finally obtain the overall blood oxygen saturation through each region to improve the calculation accuracy of the overall blood oxygen saturation.

[0054] Figure 5 is a system structure diagram of a blood oxygen saturation measurement system based on an eye image provided by an embodiment of the present application. As shown in Figure 5 The present application provides a blood oxygen saturation measurement system based on an eye image, which comprises: an acquisition unit configured to acquire image information of the same eye of a user at two different wavelengths and locate the optic disc position of each image; a processing unit configured to: perform global registration of the two images based on the optic disc positions of the two images, and perform blood vessel segmentation in the corresponding images based on each image to obtain a blood vessel segmentation image of each image; obtain segmentation information of arterial blood vessels and venous blood vessels based on the blood vessel segmentation images of the two images; and a detection unit configured to calculate the blood oxygen saturation of the eye of the user based on the segmentation information of the arterial blood vessels and the venous blood vessels.

[0055] The present application also provides a computer-readable storage medium having instructions stored thereon, which, when executed on a computer, cause the computer to perform the blood oxygen saturation measurement method based on an eye image described above.

[0056] Those skilled in the art can understand that all or part of the steps of the method of the above-mentioned embodiments can be completed by programs instructing related hardware, the programs being stored in a storage medium and including a plurality of instructions for causing a single-chip microcomputer, a chip or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0057] The above describes the optional embodiments of the present application in detail in combination with the drawings, but the embodiments of the present application are not limited to the specific details in the above-mentioned embodiments. Within the technical concept range of the embodiments of the present application, the technical solutions of the embodiments of the present application can be subjected to various simple modifications, and these simple modifications all belong to the protection range of the embodiments of the present application. In addition, it should be noted that each specific technical feature described in the above-mentioned specific embodiments can be combined in any appropriate manner without contradiction. In order to avoid unnecessary repetition, the embodiments of the present application do not further describe various possible combination manners.

[0058] Furthermore, the various embodiments can also be combined, where appropriate, with one another.

Claims

1. A method of blood oxygen saturation measurement based on eye images, characterized by, The method comprises: Collecting image information of the same eye of a user under two different shooting wavelengths, obtaining an image with a longer shooting wavelength and an image with a shorter shooting wavelength respectively, and positioning the optic disc positions of the two images respectively; After global registration of the optic disc positions of the two images, blood vessel segmentation is performed on each image to obtain the corresponding blood vessel segmentation images; wherein, The blood vessel segmentation performed on each image to obtain the corresponding blood vessel segmentation images comprises: performing Gaussian filtering on the two images respectively for pretreatment; based on the U2Net algorithm, the venous blood vessels are segmented in the pretreated image with the longer shooting wavelength to serve as the blood vessel segmentation image of the image with the longer shooting wavelength; based on the defogging algorithm, the pretreated image with the shorter shooting wavelength is enhanced to obtain an enhanced image with the shorter shooting wavelength; based on the U2Net algorithm, all arterial blood vessels are segmented in the enhanced image with the shorter shooting wavelength to serve as the blood vessel segmentation image of the image with the shorter shooting wavelength; The segmentation information of the arterial blood vessels and the venous blood vessels based on the blood vessel segmentation images of the two images comprises: taking the blood vessel segmentation image of the image with the longer shooting wavelength as the segmentation information of the venous blood vessels; registering the blood vessel segmentation image of the image with the shorter shooting wavelength to the blood vessel segmentation image of the image with the longer shooting wavelength; subtracting the mask image of the blood vessel segmentation image of the image with the longer shooting wavelength from the mask image of the blood vessel segmentation image of the image with the shorter shooting wavelength to obtain an arterial blood vessel segmentation image, which serves as the segmentation information of the arterial blood vessels; Segmentation information of arterial blood vessels and venous blood vessels is obtained based on two blood vessel segmentation images; Blood oxygen saturation of the current user's eye is calculated based on the segmentation information of the arterial blood vessels and the venous blood vessels; The calculation of the blood oxygen saturation of the current user's eye based on the segmentation information of the arterial blood vessels and the venous blood vessels comprises: extracting the skeletons of the venous blood vessels and the arterial blood vessels, and obtaining the center lines of the venous blood vessels and the arterial blood vessels; based on a preset step length, the center lines of the venous blood vessels and the arterial blood vessels are traversed to obtain the blood vessel boundary pixel values of the venous blood vessels and the arterial blood vessels respectively; the macular position in the image with the longer shooting wavelength is identified based on the yolov5 algorithm, and a straight line connecting the center points of the optic disc position and the macular position in the image with the longer shooting wavelength is taken as the abscissa, and a rectangular coordinate system is constructed with the center point of the optic disc as the origin; the blood oxygen saturation of each single venous blood vessel and single arterial blood vessel is calculated based on the blood vessel boundary pixel values of the venous blood vessels and the arterial blood vessels by using the optical density measurement method, and the blood oxygen saturation of the venous blood vessels and the arterial blood vessels in the four quadrant regions of the rectangular coordinate system is further calculated respectively; the average values of the venous blood vessel blood oxygen saturation and the arterial blood vessel blood oxygen saturation in each quadrant are calculated to obtain the venous blood vessel blood oxygen saturation and the arterial blood vessel blood oxygen saturation of the current user's eye.

2. The method of claim 1, wherein, The two different shooting wavelengths are 570±5nm and 600±5nm respectively.

3. The method of claim 1, wherein, The positioning of the optic disc positions of the two images comprises: The optic disc region in the image is located based on the yolov5 algorithm, and an elliptical contour of the optic disc is fitted through an external rectangle. The external rectangle is a detection box of the yolov5 algorithm.

4. The method of claim 1, wherein, The two images are globally registered based on the positions of the optic discs in the two images, including: taking the image with the longer wavelength as the target, and moving the image with the shorter wavelength for registration, including: 1) Difference between the center coordinates of the disc positions of the two images, resulting in and ; 2) based on and moving the image of a shorter wavelength, including: moving the image with the shorter wavelength along the x-axis and the y-axis and , the specific moving distance being determined by an evaluation parameter between the image with the shorter wavelength and the image with the longer wavelength during the moving; the evaluation parameter is: ; wherein, is the evaluation parameter; is the pixel difference value of the two images; is the structural similarity of the two images; The image coordinates corresponding to the minimum evaluation parameter in the moving process are screened out as the final coordinates of the image with the shorter wavelength.

5. The method of claim 1, wherein, The range of the constructed rectangular coordinate system is: the area with a 2r-5r distance outwardly expanded from the center of the origin; Wherein, r is the radius of the optic disc.

6. An oximetry system based on eye images, characterized by The system is used to perform the method for measuring blood oxygen saturation based on eye images according to any one of claims 1-5, and the system comprises: An acquisition unit is configured to acquire image information of the same eye of a user under two different wavelengths, to obtain an image with a longer wavelength and an image with a shorter wavelength, and to locate the positions of the optic discs in the two images respectively; A processing unit is configured to: After globally registering the two images based on the positions of the optic discs in the two images, perform blood vessel segmentation on each image to obtain corresponding blood vessel segmentation images; Obtain segmentation information of arterial blood vessels and venous blood vessels based on the two blood vessel segmentation images; A detection unit is configured to calculate blood oxygen saturation of the eye of the current user based on the segmentation information of the arterial blood vessels and the venous blood vessels.

7. A computer-readable storage medium having instructions stored thereon, which, when executed on a computer, cause the computer to perform the method for measuring blood oxygen saturation based on eye images according to any one of claims 1-5.

Images