Anterior segment OCT image three-dimensional reconstruction method

Abstract

This invention relates to a method for three-dimensional reconstruction of anterior segment OCT images, comprising the following steps: S1, acquiring an OCT image of a rotated section of the anterior segment; S2, preprocessing the anterior segment OCT image acquired in step S1 to obtain anterior segment contour sampling points transformed into a spatial coordinate system; S3, performing local spherical interpolation on the anterior segment contour sampling points transformed into a spatial coordinate system obtained in step S2; S4, performing triangular meshing on the surface point clouds of the superior cornea, inferior cornea, iris, and lens obtained by interpolation in step S3 to obtain a three-dimensional reconstructed anterior segment image; S5, performing mesh smoothing on the three-dimensional reconstructed anterior segment image obtained in step S4. This application can improve image processing accuracy and perform three-dimensional reconstruction of the same anterior segment image in real time.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method for three-dimensional reconstruction of anterior segment OCT images. Background Technology

[0002] Optical coherence tomography (OCT) is one of the most important imaging techniques in ophthalmology in recent years. Due to its non-contact, radiation-free, high-sensitivity, and non-destructive characteristics, OCT has become a standard technique for measuring human eye structures during ophthalmic surgery. The core of an OCT system is a Michelson interferometer. Light emitted from a broadband, low-coherence source passes through the interferometer and enters the sample arm and reference arm respectively. Backscattered light from the biological tissue returning from the sample arm and light reflected back from the reference arm form an interference signal via a fiber optic coupler, which is then acquired by a photodetector. Finally, data processing yields a tomographic image of the sample. Because the coherence length of the broadband source is very short, when the reference arm is fixed at a certain position, it can be assumed that only scattered light from a specific depth within the sample can interfere with the reference light. Thus, by moving the reference arm, scanning at different axial depths within the sample can be achieved, thereby realizing optical coherence tomography of the sample.

[0003] OCT (Optical Coherence Tomography) is classified into A-Scan, B-Scan, and C-Scan (also known as Volume-Scan) based on different scanning methods. A-Scan refers to scanning a single point, acquiring depth information at that point's coordinates, and is the most basic scanning method in OCT systems. B-Scan is a two-dimensional scanning method composed of multiple consecutive A-Scans, capable of reflecting the internal tomographic sections of the object being measured, and is currently the main scanning method used in ophthalmological clinical applications. C-Scan (Volume-Scan) is a three-dimensional scanning method composed of multiple consecutive B-Scans, capable of reflecting all depth information within a volume range of the object being measured. Anterior segment OCT images obtained through B-Scan contain ocular tissue structures such as the cornea, iris, and lens, and are currently widely used for preoperative diagnosis, surgical planning, and surgical navigation in ophthalmology. However, because anterior segment OCT images are two-dimensional data, carrying only limited local information about the anterior segment, this undoubtedly increases the risk of misdiagnosis or surgical accidents. Three-dimensional data, due to its more comprehensive and richer information, represents the future direction of data development in ophthalmic medical applications.

[0004] Currently, there are two main methods for obtaining three-dimensional data of the anterior segment: I. Directly acquiring three-dimensional data of the entire anterior segment. The C-Scan (Volume-Scan) mentioned above is one such method. However, this method has high hardware requirements for the scanning system, and due to the extremely large data volume, the speed of data acquisition and processing is very slow, making it difficult to directly apply in clinical practice. Another method is confocal structured illumination (CSI), which is used in LensAR (one of the current mainstream femtosecond laser cataract surgery systems). This method can acquire high-precision three-dimensional data of the anterior segment, but its hardware cost is very high, hindering its widespread use.

[0005] II. 3D Reconstruction of the Anterior Segment from OCT Image Sequences. Currently, 3D reconstruction methods based on medical images are mainly divided into volume rendering and surface rendering. The basic idea of ​​volume rendering is to simulate the propagation effect of light in volume data and calculate the contribution of each voxel in the 3D data field to the screen pixels in optical projection. Classic methods include ray casting, shear deformation, snowballing, and texture mapping-based methods, but these methods are computationally intensive and sensitive to noise. Surface rendering, on the other hand, in the application of 3D reconstruction of the anterior segment, reconstructs only the surface of the 3D geometry of the anterior segment tissue based on the input OCT image sequence. Common methods include the Marching Cube (MC) method and contour reconstruction. These methods are interactive and have a fast reconstruction speed, but their accuracy is very low. To improve accuracy, B-spline curve interpolation is usually used to interpolate the surface. However, this method depends on the order of the B-spline curve. Low orders cannot guarantee accuracy, while high orders severely affect the reconstruction speed. Moreover, B-spline curves are sensitive to noise, so the robustness of surface reconstruction cannot be guaranteed. Summary of the Invention

[0006] In view of this, it is necessary to provide a method for three-dimensional reconstruction of anterior segment OCT images, which can improve image processing accuracy and perform three-dimensional reconstruction of the same anterior segment image in real time.

[0007] This invention provides a method for three-dimensional reconstruction of anterior segment OCT images, comprising the following steps: S1, acquiring an OCT image of a rotated section of the anterior segment; S2, preprocessing the anterior segment OCT image acquired in step S1 to obtain anterior segment contour sampling points transformed into a spatial coordinate system; S3, performing local spherical interpolation on the anterior segment contour sampling points transformed into a spatial coordinate system obtained in step S2; S4, performing triangular meshing on the surface point clouds of the superior cornea, inferior cornea, iris, and lens obtained by interpolation in step S3 to obtain a three-dimensional reconstructed anterior segment image; S5, performing mesh smoothing on the three-dimensional reconstructed anterior segment image obtained in step S4.

[0008] Specifically, step S1 includes: Radial scanning uses the center of the pupil as the axis of rotation and performs several single-line cross-sectional scans of the human eye with the same angular step size to obtain multiple rotating cross-sectional images.

[0009] Specifically, step S2 includes: Step S21: Extract the anterior segment contour from the anterior segment OCT image obtained in step S1. Step S22: Perform polynomial fitting on the anterior segment contour obtained in step S21; Step S23: Perform equally spaced point sampling on the anterior segment contour fitted in step S22 to obtain anterior segment contour sampling points; Step S24: Transform the anterior segment contour sampling points obtained in step S23 to a spatial coordinate system.

[0010] Specifically, the anterior segment contour includes: the upper corneal contour, the lower corneal contour, the iris contour, the anterior lens capsule contour, the posterior lens capsule contour, and the lens lateral wall contour.

[0011] Specifically, step S22 includes: The contour distribution characteristics are described by a multinomial, and the fitted contour is represented as: Where: n is the order of the polynomial, Let be the horizontal coordinates of the starting point of the contour point set. represents the horizontal coordinates of the endpoint of the contour point set.

[0012] Specifically, step S23 includes: For the fitted curves of the upper corneal contour, lower corneal contour, anterior lens capsule contour, and posterior lens capsule contour, point sampling is performed at equal intervals in the horizontal direction. The horizontal coordinate of the contour starting point is... The coordinates of the termination point are The sampling step size is This yields a set of horizontal coordinates for a series of sampling points. Substituting the corresponding fitting polynomial, we obtain the point set: ; For the fitted curve of the lens lateral wall contour, point sampling is performed at equal intervals in the vertical direction. The vertical coordinate of the contour starting point is... The coordinates of the termination point are The sampling step size is This yields the vertical coordinate set of a series of sampling points. Substituting the corresponding fitting polynomial yields the point set. ; For the fitted curve of the iris contour, point sampling is performed at equal intervals in the horizontal direction: the horizontal coordinate of the starting point of the contour is... The coordinates of the termination point are The sampling step size is For each set of horizontal coordinates Find the point in the contour point set whose horizontal coordinate is closest to the horizontal coordinate of the given coordinate and use it as a sampling point.

[0013] Specifically, step S24 includes: Step S241: Calculate the point corresponding to the minimum y-value in the upper corneal polynomial fitting curve, denoted as [reference needed]. ; Step S242, sample each anterior segment contour point Convert to homogeneous spatial coordinates ; Step S243, construct the translation matrix :

[0014] For each spatial homogeneous coordinate Perform a translation transformation:

[0015] Step S244: Rotate the translated anterior segment contour sampling points by 90° in the opposite direction of the x-axis. This process is equivalent to constructing a rotation matrix. :

[0016] And for each Perform rotational transformation: ; Step S245: Rotate the anterior segment contour sampling points after rotation around the positive z-axis. , The rotation angle of the scan line is the equivalent of constructing a rotation matrix. :

[0017] And for each Perform rotational transformation: .

[0018] Specifically, step S3 includes: For each anterior segment contour sampling point, which corresponds to the nth single-line cross-sectional scan, let this anterior segment contour sampling point be denoted as... The following processing should be performed: Find the point in the anterior segment contour sampling point set corresponding to the (n-1)th single-line cross-sectional scan. The closest point is denoted as ; Find the point in the anterior segment contour sampling point set corresponding to the (n+1)th single-line cross-sectional scan. The closest point is denoted as ; Find the point in the anterior segment contour sampling point set corresponding to the (n+2)th single-line cross-sectional scan. The closest point is denoted as ; calculate , , , The sphere formed by four points is denoted by its center as . The radius is R; exist and Interpolation is performed on the arcs between them.

[0019] This invention improves image processing accuracy and enables real-time 3D reconstruction of the same anterior segment image by performing optical coherence tomography (OCT) on the same anterior segment image, extracting the contour of the OCT image, fitting the image contour, and then performing sampling, registration, interpolation, meshing, and surface smoothing techniques on the fitted contour. This invention can be applied to the fields of ophthalmic diagnosis, navigation for ophthalmic laser surgery, and real-time monitoring of ophthalmic laser surgery procedures. Attached Figure Description

[0020] Figure 1 This is a flowchart of the three-dimensional reconstruction method for anterior segment OCT images according to the present invention; Figure 2 This is a schematic diagram of a radial scan with an angle step of 45° provided for an embodiment of the present invention; wherein 1, 2, 3, and 4 represent four single-line cross sections; Figure 3 This is a schematic diagram of the anterior segment OCT image contour provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of equally spaced points on the anterior segment contour provided in an embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the method of establishing a spatial coordinate system according to an embodiment of the present invention; Figure 6 This is a schematic diagram of local spherical interpolation provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of interpolation in a local coordinate system provided in an embodiment of the present invention. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0022] like Figure 3 As shown, the anterior segment contour includes: the superior corneal contour, the inferior corneal contour, the iris contour, the anterior lens capsule contour, the posterior lens capsule contour, and the lens lateral wall contour. Different contour extraction methods are used for different contour features.

[0023] Please see Figure 1 This is a flowchart of a preferred embodiment of the anterior segment OCT image three-dimensional reconstruction method of the present invention.

[0024] Step S1: Obtain an OCT image of the rotated section of the anterior segment. Specifically, this includes: Radial scanning, with the pupil center as the rotation axis, performs several single-line cross-sectional scans of the human eye at the same angular step size, obtaining multiple rotating cross-sectional images, such as... Figure 2 As shown, four radial scans (single-line cross-sectional scans) were performed with a rotation angle step of 45° to obtain four OCT images of the anterior segment rotating cross-section at different angles.

[0025] Step S2 involves preprocessing the anterior segment OCT image obtained in step S1 to obtain anterior segment contour sampling points transformed into a spatial coordinate system. This includes: Step S21: Extract the anterior segment contour from the anterior segment OCT image obtained in step S1; that is, the contours of the superior cornea, inferior cornea, iris, anterior lens capsule, posterior lens capsule, and lateral lens wall. Specifically, this includes: First, regarding such Figure 3 The extraction of the upper corneal contour, lower corneal contour, and iris contour shown includes: 1. Perform gamma transformation on the image to improve contrast; 2. Perform median filtering on the gamma-transformed image; 3. Perform a closing operation on the filtered image; 4. Binarize the image obtained from the closing operation; 5. Perform contour detection on the binarized image; 6. Calculate the horizontal length of all contour lines, and extract the longest contour line as the upper corneal contour; 7. Extract the second longest contour line, and use the left and right endpoints as dividing points to segment the lower corneal contour and iris contour.

[0026] Then, for example Figure 3 The extraction of the anterior lens capsule contour, posterior lens capsule contour, and lens lateral wall contour shown includes: 1. Perform gamma transformation on the image to improve contrast; 2. Perform median filtering on the gamma-transformed image; 3. Perform closing, opening, and closing operations on the filtered image step by step; 4. Perform contour line detection on the resulting image; 5. Calculate the pixel length of all contour lines, and extract the longest contour line as the lens contour; 6. Detect inflection points on the lens contour and divide the lens contour into the anterior lens capsule contour, the posterior lens capsule contour, and the lens lateral wall contour based on the detected inflection points.

[0027] Step S22 involves performing polynomial fitting on the anterior segment contour obtained in step S21. Specifically, this includes: Since the above contour is essentially a set of points, and its distribution characteristics are described by a multinomial, the fitted contour is represented as: Where: n is the order of the polynomial, Let be the horizontal coordinates of the starting point of the contour point set. represents the horizontal coordinates of the endpoint of the contour point set.

[0028] Least squares is the most commonly used polynomial fitting method, but it includes all contour points, including noise points, in the calculation, making the fitting result highly susceptible to noise. Therefore, this invention uses the RANSAC (Random Sample Consensus) algorithm for polynomial fitting, which can eliminate erroneous contour points and obtain the theoretically optimal fitting curve. Specifically, the polynomial fitting order for the upper and lower corneal contours is 4, for the anterior and posterior lens capsule contours it is 2, and for the lateral lens wall contour it is 1.

[0029] Step S23 involves sampling the anterior segment contour fitted in step S22 at equally spaced points to obtain anterior segment contour sampling points. Specifically, this includes: like Figure 4 As shown, for the fitted curves of the upper corneal contour, lower corneal contour, anterior lens capsule contour, and posterior lens capsule contour, point sampling is performed at equal intervals in the horizontal direction. The horizontal coordinate of the contour starting point is... The coordinates of the termination point are The sampling step size is This yields a set of horizontal coordinates for a series of sampling points. Substituting the corresponding fitting polynomial, we obtain the point set: .

[0030] For the fitted curve of the lens lateral wall profile, since it is a first-order polynomial, the expression will be... Transformed into Point sampling is performed at equal intervals in the vertical direction. The vertical coordinate of the starting point of the contour is... The coordinates of the termination point are The sampling step size is This yields the vertical coordinate set of a series of sampling points. Substituting the corresponding fitting polynomial yields the point set. .

[0031] For the fitted curve of the iris contour, since it does not undergo polynomial fitting, the method of sampling points at equal intervals in the horizontal direction is as follows: for the horizontal coordinate of the starting point of the contour, the method is... The coordinates of the termination point are The sampling step size is For each set of horizontal coordinates Find the point in the contour point set whose horizontal coordinate is closest to the horizontal coordinate of the given coordinate and use it as a sampling point.

[0032] Step S24 involves transforming the anterior segment contour sampling points obtained in step S23 to a spatial coordinate system. Specifically, this includes: like Figure 5 As shown, the spatial coordinate system refers to a spatial coordinate system established with the corneal vertex as the origin, the rotation axis of the radial scan as the z-axis, and the direction of the first scan line as the x-axis. The method for transforming the anterior segment contour sampling points to the spatial coordinate system is as follows: Step S241: Calculate the point corresponding to the minimum y-value in the upper corneal polynomial fitting curve, denoted as [reference needed]. ; Step S242, sample each anterior segment contour point Convert to homogeneous spatial coordinates ; Step S243, construct the translation matrix :

[0033] For each spatial homogeneous coordinate Perform a translation transformation:

[0034] Step S244: Rotate the translated anterior segment contour sampling points by 90° in the opposite direction of the x-axis. This process is equivalent to constructing a rotation matrix. :

[0035] And for each Perform rotational transformation:

[0036] Step S245: Rotate the anterior segment contour sampling points after rotation around the positive z-axis. , The rotation angle of the scan line is the equivalent of constructing a rotation matrix. :

[0037] And for each Perform rotational transformation: .

[0038] Step S3 involves performing local spherical interpolation on the anterior segment contour sampling points transformed to the spatial coordinate system obtained from the preprocessing in step S2. Specifically, this includes: like Figure 6 As shown, the local sphere refers to a sphere composed of four adjacent anterior segment contour sampling points. The specific interpolation method is as follows: for each anterior segment contour sampling point, corresponding to the nth single-line cross-sectional scan, this anterior segment contour sampling point is denoted as... Perform the following steps: 1. Find the point in the anterior segment contour sampling point set corresponding to the (n-1)th single-line cross-sectional scan. The closest point is denoted as ; 2. Find the point in the anterior segment contour sampling point set corresponding to the (n+1)th single-line cross-sectional scan. The closest point is denoted as ; 3. Find the point in the anterior segment contour sampling point set corresponding to the (n+2)th single-line cross-sectional scan. The closest point is denoted as ; 4. Calculation , , , The sphere formed by four points is denoted by its center as . The radius is R; 5. In and Interpolation is performed on the arcs between them, specifically as follows: ①With the center of the ball With the origin as the point, and connection The x-axis passes through , , The normal direction of the plane at the three points is the z-axis; establish a local coordinate system. ; ② Construct the translation transformation matrix:

[0039] The homogeneous coordinates are ,right Perform a translation transformation to obtain ; ③ Calculate the origin and The angle between the projection of the line on the xoy plane and the x-axis. Construct the rotation matrix: ; ④ Calculate the origin and The angle between the line and the z-axis Construct the rotation matrix:

[0040] ⑤ The homogeneous coordinates are ,right Transform to obtain and to Transform to obtain ; ⑥ Calculate the origin and The angle between the line connecting the two points and the xoy plane Construct the rotation matrix: ; ⑦ The homogeneous coordinates of the anterior segment contour sampling points in the original coordinate system are: Then the anterior segment contour sampling points will be... Transform to local coordinate system Equivalent to left multiplication ,Right now ,in For the anterior segment contour sampling points in the local coordinate system Homogeneous coordinates below; ⑧ For example Figure 7 As shown, calculate In the local coordinate system The homogeneous coordinates below are , origin and The angle between the line connecting the two points and the x-axis is . Then in and The problem of interpolating along the arc between points is equivalent to: interpolating along the arc in the xoy plane with the origin as the center, R as the radius, and an angle of θ. Interpolation is performed along the arc, with an angle step size of . The calculated interpolation point set is as follows: ; ⑨ Through the formula:

[0041] Will Transform to the original coordinate system to obtain the interpolation points in the original coordinate system. .

[0042] Step S4: Perform triangular meshing processing on the surface point clouds of the upper cornea, lower cornea, iris and lens obtained by interpolation in step S3 to obtain a three-dimensional reconstructed anterior segment image.

[0043] Step S5: Perform mesh smoothing processing on the three-dimensional reconstructed anterior segment image obtained in step S4.

[0044] Although the present invention has been described with reference to the present preferred embodiments, those skilled in the art should understand that the above preferred embodiments are only used to illustrate the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for three-dimensional reconstruction of anterior segment OCT images, characterized in that, The method includes the following steps: S1, Obtain the OCT image of the rotated section of the anterior segment; S2, preprocess the anterior segment OCT image obtained in step S1 to obtain the anterior segment contour sampling points transformed into the spatial coordinate system; S3, perform local spherical interpolation on the anterior segment contour sampling points transformed into the spatial coordinate system obtained from the preprocessing in step S2; S4. The surface point clouds of the upper cornea, lower cornea, iris and lens obtained by interpolation in step S3 are processed into triangular meshes to obtain a three-dimensional reconstructed anterior segment image. S5, perform mesh smoothing processing on the three-dimensional reconstructed anterior segment image obtained in step S4.

2. The method as described in claim 1, characterized in that, Step S1 includes: Radial scanning uses the center of the pupil as the axis of rotation and performs several single-line cross-sectional scans of the human eye with the same angular step size to obtain multiple rotating cross-sectional images.

3. The method as described in claim 2, characterized in that, Step S2 includes: Step S21: Extract the anterior segment contour from the anterior segment OCT image obtained in step S1. Step S22: Perform polynomial fitting on the anterior segment contour obtained in step S21; Step S23: Perform equally spaced point sampling on the anterior segment contour fitted in step S22 to obtain anterior segment contour sampling points; Step S24: Transform the anterior segment contour sampling points obtained in step S23 to a spatial coordinate system.

4. The method as described in claim 3, characterized in that, The anterior segment contour includes: the upper corneal contour, the lower corneal contour, the iris contour, the anterior lens capsule contour, the posterior lens capsule contour, and the lens lateral wall contour.

5. The method as described in claim 4, characterized in that, Step S22 includes: The contour distribution characteristics are described by a multinomial, and the fitted contour is represented as: Where: n is the order of the polynomial, Let be the horizontal coordinates of the starting point of the contour point set. represents the horizontal coordinates of the endpoint of the contour point set.

6. The method as described in claim 5, characterized in that, Step S23 includes: For the fitted curves of the upper corneal contour, lower corneal contour, anterior lens capsule contour, and posterior lens capsule contour, point sampling is performed at equal intervals in the horizontal direction. The horizontal coordinate of the contour starting point is... The coordinates of the termination point are The sampling step size is This yields a set of horizontal coordinates for a series of sampling points. Substituting the corresponding fitting polynomial, we obtain the point set: ； For the fitted curve of the lens lateral wall contour, point sampling is performed at equal intervals in the vertical direction. The vertical coordinate of the contour starting point is... The coordinates of the termination point are The sampling step size is This yields the vertical coordinate set of a series of sampling points. Substituting the corresponding fitting polynomial yields the point set. ; For the fitted curve of the iris contour, point sampling is performed at equal intervals in the horizontal direction: the horizontal coordinate of the starting point of the contour is... The coordinates of the termination point are The sampling step size is For each set of horizontal coordinates Find the point in the contour point set whose horizontal coordinate is closest to the horizontal coordinate of the given coordinate and use it as a sampling point.

7. The method as described in claim 5, characterized in that, Step S24 includes: Step S241: Calculate the point corresponding to the minimum y-value in the upper corneal polynomial fitting curve, denoted as [reference needed]. ; Step S242, sample each anterior segment contour point Convert to homogeneous spatial coordinates ; Step S243, construct the translation matrix : For each spatial homogeneous coordinate Perform a translation transformation: Step S244: Rotate the translated anterior segment contour sampling points by 90° in the opposite direction of the x-axis. This process is equivalent to constructing a rotation matrix. : And for each Perform rotational transformation: ; Step S245: Rotate the anterior segment contour sampling points after rotation around the positive z-axis. , The rotation angle of the scan line is the equivalent of constructing a rotation matrix. : And for each Perform rotational transformation: .

8. The method as described in claim 5, characterized in that, Step S3 includes: For each anterior segment contour sampling point, which corresponds to the nth single-line cross-sectional scan, let this anterior segment contour sampling point be denoted as... The following processing should be performed: Find the point in the anterior segment contour sampling point set corresponding to the (n-1)th single-line cross-sectional scan. The closest point is denoted as ; Find the point in the anterior segment contour sampling point set corresponding to the (n+1)th single-line cross-sectional scan. The closest point is denoted as ; Find the point in the anterior segment contour sampling point set corresponding to the (n+2)th single-line cross-sectional scan. The closest point is denoted as ; calculate , , , The sphere formed by four points is denoted by its center as . The radius is R; exist and Interpolation is performed on the arcs between them.

Images