A method for three-dimensional reconstruction and volume quantification of mid-serous edema area

Abstract

The present invention discloses a method for three-dimensional reconstruction of a mid-series edema area and volume quantification thereof, specifically relating to the technical field of medical image processing. The method pre-processes the original image to lay the foundation for the smooth three-dimensional reconstruction and volume quantification of the mid-series edema area; extracts points from the section contour of the OCT B-scan image and maps them to the correct position in the three-dimensional space, and then uses visualization means to present the upper and lower surfaces of the edema area, thus achieving three-dimensional reconstruction of the edema area and obtaining a parametric space of the mid-series edema area, laying the foundation for subsequent volume quantification; by constructing an inter-frame adjacency relationship, using symmetrical frame units to reasonably fill blank areas, further dividing the symmetrical frame units into microelements in the parametric space of the mid-series edema area, calculating the volumes of all microelements and combining them with the image scale to achieve volume quantification of the edema area.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method for three-dimensional reconstruction of a mid-serous edema area and volume quantification thereof. Background Art

[0002] Central Serous Chorioretinopathy (CSC), also known as CSC, is a disease characterized by subepithelial fluid accumulation and characteristic edema caused by impaired barrier function of the pigment epithelium in the macula or posterior pole of the retina. This condition is highly prevalent in my country, and patients often experience symptoms such as decreased vision, blurred vision, and metamorphopsia, which severely impact their quality of life.

[0003] Current clinical assessment of the mid-serous edema area has certain limitations:

[0004] 1) The subretinal fluid accumulation area is unevenly distributed in three-dimensional space. Clinically, only a certain section of the edema area in a single-frame OCT B-scan image is examined, and the three-dimensional morphology of the edema area cannot be observed.

[0005] 2) Single-frame OCT B-scan images can only characterize one-dimensional parameters such as the height and diameter of a certain section of the mid-serous edema area, which is not accurate enough to describe the size of the edema area from a three-dimensional perspective. Summary of the Invention

[0006] Currently, there is a lack of effective methods for 3D reconstruction and volume quantification of mid-serosal edema areas in clinical practice. 3D reconstruction and volume quantification of mid-serosal edema areas can address these deficiencies and provide a more comprehensive assessment basis for clinical decision-making. 3D reconstruction and volume quantification of mid-serosal edema areas based on different scanning methods have different effects: Line scanning offers high volume reconstruction accuracy, but it has a higher sampling density and takes longer to scan. Patient eye movements can easily produce artifacts, which can interfere with edema segmentation algorithms or manual annotation. Furthermore, the parallel sampling pattern of line scanning results in significant variability in edema scale between different sections. Sections near the edges of the mid-serosal edema area are smaller, making them difficult to manually annotate or segment using algorithms. Furthermore, line scanning acquires a large amount of data, requiring significant data processing prior to 3D quantification. In contrast, circular scanning offers the advantages of shorter acquisition time, fewer artifacts, easier annotation or segmentation, and a smaller data volume. Furthermore, the scale of mid-serosal edema sections acquired with circular scanning is similar, which helps ensure higher reconstruction accuracy.

[0007] To this end, the present invention provides a method for three-dimensional reconstruction of the mid-serous edema area and volume quantification thereof to solve the problems raised in the background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for three-dimensional reconstruction of the mid-splenic edema area and volume quantification thereof, comprising the following steps:

[0009] Step 1: Preprocess the original image, including image cropping, section annotation, and edge detection, to lay the foundation for smooth 3D reconstruction and volume quantification of the mid-serosal edema area. The original image consists of IR images and OCT B-scan images.

[0010] Step 2: Points from the cross-sectional contours extracted from the OCT B-scan image are mapped to their correct positions in three-dimensional space. Visualization is then used to present the upper and lower surfaces of the edema area, achieving three-dimensional reconstruction of the edema area. This process includes spatial alignment, contour point serialization, and contour point spatial mapping. Through three-dimensional reconstruction, a parametric space for the mid-splenic edema area is obtained, laying the foundation for subsequent volume quantification.

[0011] Step 3: By building an inter-frame adjacency relationship, symmetrical frame units are used to reasonably fill the blank areas. The symmetrical frame units are further divided into microelements in the parametric space of the mid-serous edema area. The volumes of all microelements are calculated and then combined with the image scale to quantify the edema area volume. This solves the problem of blank areas with no data between frames due to the angular gap between the circular scan frames caused by the discrete sampling characteristics of OCT B-scan image sampling.

[0012] Preferably, the specific operations of step 1 are as follows:

[0013] Step 1.1: Extract the OCT B-scan image from the original image to meet the data requirements for 3D reconstruction and volume quantification of the edema area. Based on the fixed position relationship of the OCT B-scan image in the original image, use the coordinates of the upper left corner and lower right corner of the OCT B-scan image as the cropping boundary to obtain the OCT B-scan image. The cropping method is shown in formula (1);

[0014] (1);

[0015] In formula (1), represents the OCT B-scan image area, is the coordinate of the upper left corner of the OCT B-scan image area, is the coordinate of the lower right corner of the OCT B-scan image area; and The coordinates of are calculated based on the resolution of the infrared image and the overall size of the original image, as shown in formula (2);

[0016] (2);

[0017] In formula (2), and are the width and height of the infrared image, is the width of the original image;

[0018] Step 1.2: Use medical image annotation software to perform pixel-level annotation on the edema area section of the OCT B-scan image to obtain an annotation map of the edema area section in the OCT B-scan image;

[0019] Step 1.3: Use the Canny algorithm to detect the edge of the edema area in the binary image, so as to accurately extract its contour.

[0020] Preferably, the process of detecting the edge of the mid-serous edema area using the Canny algorithm in step 1.3 is divided into the following four steps:

[0021] (1) Image denoising: Noise is a place where the grayscale changes greatly and is easily identified as a false edge; therefore, a Gaussian filter is used to smooth the image and reduce the noise in the image, using the formula The 5*5 Gaussian filter shown:

[0022] (3);

[0023] (2) Calculate the image gradient, detect possible edges, and calculate the gradient direction and amplitude of the grayscale value of each pixel in the image: Use the Sobel operator to detect the horizontal direction of the image. and vertical direction The edge of the gradient is calculated using formula (4) and formula (5) and amplitude , where Indicates that the pixel is Grayscale of direction, Indicates that the pixel is The grayscale of the direction; the Sobel operator extracts the image edge through discrete differential method:

[0024] (4);

[0025] (5);

[0026] (3) Non-maximum suppression: The non-maximum suppression algorithm is used to eliminate the stray responses caused by edge detection. The grayscale change area is usually concentrated. The pixels with the largest grayscale gradient in the gradient direction within the local range are retained, and the others are not retained. A large part of the points are eliminated, and the edge with multiple pixel widths is turned into a single pixel width edge.

[0027] (4) Double threshold screening: After non-maximum suppression, there are still many possible edge points. The double threshold method is applied to divide strong edges and weak edges. Pixels with grayscale gradients greater than the high threshold are retained as strong edges, and pixels with grayscale gradients less than the low threshold are removed. Pixels with grayscale gradients between the high and low thresholds are screened as weak edges. Search within the neighborhood of the weak edge. If a strong edge exists, the weak edge is retained. If not, the weak edge is deleted to complete edge detection.

[0028] Preferably, the specific operations of step 2 are as follows:

[0029] Step 2.1: Establish a three-dimensional coordinate system based on the right-handed Cartesian coordinate system Map the points in the slice contour extracted from the OCT B-scan image to the correct position in the three-dimensional coordinate system; where the origin It is set as the lowest point of the vertical centerline of the OCT B-scan image; The axis coincides with the horizontal scan line and points in the horizontal direction indicated by the scan arrow; The axis coincides with the vertical scan line and points in the vertical direction indicated by the scan arrow; Axis perpendicular to Flat surface, perpendicular to the paper and outward;

[0030] The center of the ring scan in the contour IR image corresponds to the vertical center line of each frame of the OCT B-scan image, which can be regarded as the rotation center of the entire scanning sequence in the three-dimensional coordinate system; the lowest point of the vertical center line of each frame of the OCT B-scan image is aligned with the origin of the three-dimensional coordinate system, and the vertical center line of the OCT B-scan image is aligned with the origin of the three-dimensional coordinate system. Axis, unify the rotation base of all OCTB-scan images to Axis, which clarifies the relative position relationship of each frame image in the three-dimensional coordinate system;

[0031] In three-dimensional coordinate system The positive direction of the axis is used as the angle reference, and the clockwise rotation angle is positive. The first OCT B-scan image frame corresponds to an angle of 0°, and the remaining frames are rotated clockwise by corresponding angles based on their corresponding scanning positions in the IR image. After the angle reference is aligned, the absolute position of each OCT B-scan image frame in three-dimensional space can be determined.

[0032] Step 2.2: Order the obtained unordered point set section contour to form an ordered coordinate sequence, and then construct a structured contour description;

[0033] Step 2.3: Map the points in the ordered coordinate sequence from the two-dimensional coordinate system to the three-dimensional coordinate system to obtain the coordinates of all section contour points in the three-dimensional space; spatial mapping maps the section contour points acquired by the circular scan from the two-dimensional space in the OCT B-scan image to the three-dimensional space to obtain the parametric space of the edema area, providing doctors with an intuitive reference for the morphology of the edema area and laying the foundation for subsequent volume quantification of the edema area.

[0034] Preferably, the specific operations of step 2.2 are as follows:

[0035] The two-dimensional coordinate system that defines the section profile is denoted as The origin of the coordinate system coincides with the lowest point of the vertical center line of the OCT B-scan image, and the horizontal right direction is Axis positive direction, vertical upward direction Axis positive direction, The axis is the center of the ring sweep;

[0036] Use a vertical line from the leftmost side of the outline Starting from Move rightward in sequence for the step length, radial step length The length is equal to one pixel, recording the coordinates of the intersection of the vertical line and the upper and lower contours until the rightmost side of the contour Get The coordinates of the upper and lower contour points are completed. Frame slice contour point serialization;

[0037] Serialize all the section contour points to get the point set , as shown in formula (6);

[0038] (6);

[0039] In formula (6) represents an ordered set of section contour points, and Indicates the The first round of sweep pairs of upper and lower contour points, Indicates the The angle of the secondary sweep, Indicates the interval angle rad between two adjacent ring scans, is the total number of ring sweeps, (pixel) means In the second scan The distance between the upper and lower contour points and the center of the ring scan is specified in Left of axis is a negative value, Right of axis is a positive value, and They are The first round of sweep Upper and lower contour points Coordinates; if and only if or The two points in the time point pair coincide.

[0040] Preferably, the specific operations of step 2.3 are as follows:

[0041] The two-dimensional coordinate system is transformed into The points below are mapped to the three-dimensional coordinate system In the coordinate system:

[0042] ;

[0043] and Indicates the The first scan The positions of the upper and lower contour point pairs in the two-dimensional coordinate system; and Indicates the The first The positions of the upper and lower contour point pairs in the three-dimensional coordinate system after mapping.

[0044] Preferably, the specific operations of step 3 are as follows:

[0045] Step 3.1: In the parameterized space, further divide the symmetric frame unit into infinitesimal elements to solve its volume;

[0046] Step 3.2: After dividing the edema area into several microelements, the volume of the edema area can be approximated as the sum of the volumes of all microelements.

[0047] Preferably, step 3.1 specifically includes:

[0048] No. In the symmetrical frame units, the 、 、 The vertical pixel columns between the pixel pairs rotate around the center of the ring scan to form three microelements; each microelement is approximately a cuboid, and its volume is approximately the volume of the cuboid ; For the The first symmetric frame unit The length of the infinitesimal element is Indicates that it is equal to the width of a pixel in the OCT B-scan image. (pixel), height (pixel) means;

[0049] The volume of a single infinitesimal approximation is calculated according to formula (8);

[0050] (8);

[0051] The unit of volume is voxel (volumetric pixel), which is the pixel unit in a three-dimensional image and represents the volume of a cube with a side length of 1 pixel. The volume of a single microelement contains approximation errors, which are high-order infinitesimals and can be ignored. After all the symmetrical frame units that form the edema area are divided into microelements, the entire edema area is divided into several microelements.

[0052] Preferably, step 3.2 specifically includes:

[0053] For each symmetric frame unit, all the microelements in it are traversed according to formula (9) and their volumes are calculated and summed up to be the volume of the edema area in the parameterized space The unit is voxel; in formula (9), the volume of the edema area in the parameterized space is obtained by adding up the volumes of all micro-elements;

[0054] (9);

[0055] After calculating the volume of the edema area in the parameterized space, it is necessary to find the true volume of the edema area according to the scale; the volume of the edema area in the parameterized space The volume of the edema area is obtained by the present invention The relationship between is defined by formula (10), where It is a scale;

[0056] (10);

[0057] The scale is based on the formula to obtain;

[0058] ;

[0059] In formula (11), , On OCT B-Scan images, and Resolution in direction, and The lengths are exist and The number of pixels corresponding to the direction.

[0060] The present invention aims to provide a three-dimensional evaluation basis for clinical diagnosis and treatment, and this method has important value in clinical applications. First, different forms of edema areas correspond to differentiated laser irradiation strategies, and the three-dimensional reconstruction of the mid-series edema area provides an important reference basis for personalized treatment plans; secondly, the prognosis of mid-series edema areas of different forms is not the same. Three-dimensional reconstruction of the mid-series edema area helps to establish the correlation between three-dimensional morphological parameters and prognosis; finally, the volume quantification of the mid-series edema area can provide an accurate, comprehensive, objective and accurate basis for clinical decision-making. The present invention not only improves the clinical evaluation system of mid-series edema, but also lays an important foundation for the realization of precision medicine of mid-series edema, realizes the leap from one-dimensional evaluation to three-dimensional quantification, provides a comprehensive, objective and accurate basis for the evaluation of mid-series edema area, and promotes precision medicine of mid-series edema. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 The fundus OCT line scan image provided by the present invention; the left bottom of the left IR image and the left bottom of the right OCT B-scan image are the scales of the IR image and the OCT B-scan image respectively;

[0062] Figure 2 The fundus OCT ring scan image provided by the present invention; the left bottom of the left IR image and the left bottom of the right OCT B-scan image are the scales of the IR image and the OCT B-scan image respectively;

[0063] Figure 3 This is a diagram of the diameter and height of a certain section of the edema area measured by the present invention; the scale bar of the OCT B-scan image is at the bottom left of the image;

[0064] Figure 4 A complete technical roadmap for this invention;

[0065] Figure 5 Schematic diagram of image cropping provided by the present invention; the image above the arrow is the image before cropping, and the left bottom of the IR image on the left and the left bottom of the OCT B-scan image on the right are the scales of the IR image and OCT B-scan image, respectively; the image below the arrow is the image after cropping, and the left bottom is the scale of the OCT B-scan image;

[0066] Figure 6 This is the effect diagram of the edema area section in the OCT B-scan image provided by the present invention marked in the Pair software. The red mask is the labeled area; Figure 6 The scale bar of the OCT B-scan image is at the bottom left;

[0067] Figure 7 The present invention provides The binary image obtained after region annotation;

[0068] Figure 8 A contour diagram of a certain section of the mid-serous edema area after edge extraction provided by the present invention;

[0069] Figure 9 A schematic diagram of the definition of the three-dimensional coordinate system provided by the present invention;

[0070] Figure 10 A schematic diagram of the OCT B-scan image provided by the present invention rotating around the center of the circular scan;

[0071] Figure 11 The contour point serialization graph provided by the present invention;

[0072] Figure 12 The upper and lower surface collection points of the edema area and the IR image projection display provided by the present invention

[0073] Figure 13 A schematic diagram of the visualization space of the edema area provided by the present invention;

[0074] Figure 14 A schematic diagram of the adjacency relationship provided by the present invention;

[0075] Figure 15 Schematic diagram of microelement volume calculation provided by the present invention;

[0076] Figure 16 Schematic diagram of the microelement division provided by the present invention;

[0077] Figure 17 Detailed drawings to scale provided for the present invention;

[0078] Figure 18 Schematic diagram of line scan volume solution provided by the present invention;

[0079] Figure 19 A schematic diagram of a specific implementation method of the present invention provided by the present invention;

[0080] Figure 20 The cross-sectional image of the edema area is annotated using Pair software provided by the present invention;

[0081] Figure 21 Schematic diagram of the lesion contour extraction process provided by the present invention;

[0082] Figure 22 An example diagram of contour point serialization provided by the present invention;

[0083] Figure 23 This is a display diagram of seven examples of 3D reconstruction of edema areas and their volume quantification results provided by the present invention. DETAILED DESCRIPTION

[0084] The following describes the implementation of the present invention using specific embodiments. Those skilled in the art will readily understand the other advantages and benefits of the present invention from the disclosure herein. Obviously, the embodiments described are only a portion of the present invention, not all of it. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0085] In clinical diagnosis and treatment, Optical Coherence Tomography (OCT) is the core imaging tool for diagnosing retinal edema. It generates retinal tomographic images through B-scan (brightness scan, B-scan) scanning, which can clearly show the cross-sectional morphology of the retinal edema area. Currently, there are two commonly used OCT B-scan scanning methods in clinical practice: linear scanning (hereinafter referred to as line scanning) and circular scanning (hereinafter referred to as circular scanning). Figure 1 and Figure 2 Shown are fundus OCT images acquired using two scanning methods: the infrared (IR) image on the left and the OCT B-scan image on the right. In line scanning, the scan line moves along a pre-set axis and samples densely, producing parallel and evenly spaced slices. In circular scanning, the scan line rotates clockwise, sampling at pre-set angles around the center of the edema area, producing slices arranged at equal angles around the center of the circular scan.

[0086] Currently, clinical evaluation of mid-scleral edema is mainly based on analysis of OCT B-scan images. Clinicians assess the severity of the disease and formulate treatment plans by measuring one-dimensional parameters such as the height and diameter of the edema area in a single frame of OCT B-scan images; at the same time, by comparing the above parameters before and after surgery, they evaluate the treatment effect. Figure 3 The diagram shows a doctor measuring the diameter and height of a section of the edema area.

[0087] To address the issues raised in the background art, the present invention proposes a method for 3D reconstruction and volume quantification of the edema area in the midline based on fundus circumferential scanning OCT images. This method constructs the 3D morphology of the edema area and quantifies its volume, achieving a transition from one-dimensional assessment to three-dimensional quantification. This provides a comprehensive, objective, and accurate basis for the assessment of edema areas in the midline, promoting precision medicine in the midline.

[0088] The complete technical route of the present invention is as follows Figure 4As shown, the present invention describes a method for quantifying the volume of mid-scleral edema areas based on fundus OCT ring scan images, which includes three steps: contour extraction, 3D reconstruction, and volume quantification. The above steps have achieved all the technical objectives of the present invention. To verify the accuracy of this method, the present invention further provides a consistency assessment step. This step is only used to verify the effectiveness of the present invention and is not a necessary process for the actual application of the present invention. The specific methods of each step are detailed below.

[0089] 1. Contour extraction:

[0090] This step preprocesses the raw data, including image cropping, section annotation, and edge detection, laying the foundation for the smooth three-dimensional reconstruction and volume quantification of the mid-serous edema area.

[0091] 1.1 Image Cropping:

[0092] The original data of the present invention is collected from a certain OCT fundus examination device. The image consists of an IR image on the left and an OCTB-scan image on the right. Figure 2 To meet the data requirements for 3D reconstruction and volume quantification of the edema area, it is necessary to accurately extract the OCT B-scan image from the original image, as shown in Figure 5 As shown. According to the fixed position relationship of the OCT B-scan image in the original image, the original data is cropped with the upper left corner and lower right corner coordinates of the OCT B-scan image as the cropping boundary to obtain an independent OCT B-scan image part. The cropping method is as shown in the formula As shown. Formula middle, represents the OCT B-scan image area, The upper left corner of the OCT B-scan image area The coordinates of the point, The lower right corner of the OCT B-scan image area The coordinates of the point. and The coordinates are calculated based on the resolution of the infrared image and the overall size of the original image, as shown in the formula As shown. Formula middle, and are the width and height of the infrared image, is the width of the original image.

[0093] (1);

[0094] (2);

[0095] 1.2 Section marking:

[0096] The present invention uses Pair software to perform pixel-level annotation on the edema area section of the OCT B-scan image. Figure 6 and Figure 7 The figure shows the schematic diagram of the section annotation of the edema area in the OCT B-scan image (hereinafter referred to as section annotation). Figure 6 This is the effect diagram marked in Pair software. The red mask is the label area. Figure 7 It is the aforementioned The binary image obtained after region marking, in which the white point with a grayscale value of 255 represents a certain section of the edema area.

[0097] 1.3 Edge Detection:

[0098] An edge is a region in an image where the grayscale value changes suddenly, typically corresponding to the outline of an object or the boundary of a structure. This paper uses the Canny algorithm to detect edges in edema areas in binary images, thereby accurately extracting their contours. The Canny algorithm is one of the most widely used and high-performance edge detection algorithms. The Canny algorithm detects the edges of edema areas in the mid-sera in four steps:

[0099] (1) Image denoising. Noise is where the grayscale changes greatly and is easily identified as a false edge. Therefore, a Gaussian filter is used to smooth the image and reduce the noise. Generally, a 5*5 Gaussian filter is used, as shown in formula (3).

[0100] (3);

[0101] (2) Calculate the image gradient and detect possible edges. This step requires calculating the gradient direction and amplitude of the grayscale value of each pixel in the image. First, use the Sobel operator to detect the image (horizontal) direction and (vertical) edge, and then use formula (4) and formula (5) to calculate the direction of the gradient and amplitude , where Indicates that the pixel is Grayscale of direction, Similarly, the Sobel operator can effectively extract image edges through discrete differential methods.

[0102] (4);

[0103] (5);

[0104] (3) Non-maximum suppression. The non-maximum suppression algorithm can eliminate the stray responses caused by edge detection. The area where the grayscale changes is usually concentrated is usually concentrated. The pixels with the largest grayscale gradient in the gradient direction within the local range are retained, and the others are not retained. This can eliminate a large number of points and turn an edge with a width of multiple pixels into an edge with a width of one pixel.

[0105] (4) Double threshold screening. After non-maximum suppression, there are still many possible edge points. The double threshold method is applied to divide strong edges into weak edges. Pixels with grayscale gradients greater than the high threshold are retained as strong edges, and pixels with grayscale gradients less than the low threshold are removed. Pixels with grayscale gradients between the high and low thresholds are screened as weak edges. Search within the neighborhood of the weak edge. If a strong edge exists, the weak edge is retained. If not, the weak edge is deleted to complete edge detection.

[0106] like Figure 8 Shown is the contour image of a section of the mid-serous edema region after edge extraction (hereafter referred to as the section contour). The contour lines are bolded for ease of visualization; the actual contour line width is 1 pixel. A series of section contours corresponding to each patient are stored by patient name to facilitate subsequent 3D reconstruction and volume quantification.

[0107] 2. 3D reconstruction:

[0108] 3D reconstruction of the edema region is achieved by mapping points extracted from the cross-sectional contours of OCT B-scan images to their correct positions in 3D space. Visualization of the upper and lower surfaces of the edema region is then performed using visualization techniques. This process involves spatial alignment, contour point serialization, and spatial mapping of contour points. 3D reconstruction yields a parametric space for the mid-splenic edema region, laying the foundation for subsequent volume quantification.

[0109] 2.1 Spatial Alignment:

[0110] In order to accurately map a series of section profiles into three-dimensional space, spatial alignment operations are required, including three-dimensional coordinate system definition, rotation center alignment, and angle base alignment.

[0111] The definition of a 3D coordinate system is the foundation for accurately mapping 2D contours into 3D space. To accurately position the slice contours extracted from OCT B-scan images into 3D space, a 3D coordinate system with spatial reference significance must be established. Figure 9 The definition of the three-dimensional coordinate system in the present invention is shown by taking 8 circular scanning lines as an example. The black arrows in the figure indicate the direction of each scanning line. The scanning is performed in a clockwise direction. The numbers next to the arrows indicate the order of scanning. The coordinate system defined in the present invention is a right-handed Cartesian coordinate system, which is denoted as Among them, the origin It is set as the lowest point of the vertical centerline of the OCT B-scan image; The axis coincides with the horizontal scan line and points in the horizontal direction indicated by the scan arrow; The axis coincides with the vertical scan line and points in the vertical direction indicated by the scan arrow; Axis perpendicular to Plane, perpendicular to the paper and pointing outward. The definition of the coordinate system provides a clear geometric basis for the spatial mapping of subsequent contour points.

[0112] Aligning the rotation center is a key step in determining the relative position relationship between the section contour and the ring scan center. The ring scan center in the contour IR image corresponds to the vertical center line of each frame of the OCT B-scan image and can be regarded as the rotation center of the entire scan sequence in space. Figure 10 As shown in the figure, the dotted line represents the ring scan center, and each frame of OCT B-scan image is rotated around the ring scan center according to the corresponding angle. The lowest point of the vertical center line of each frame of OCT B-scan image is aligned with the origin of the three-dimensional coordinate system, and the vertical center line of the OCT B-scan image is aligned with the origin of the three-dimensional coordinate system. Axis. Unify the rotational basis of all OCT B-scan images to Axis, which clarifies the relative position relationship of each frame image in space.

[0113] Angle-based alignment further clarifies the relative position relationship between a series of section profiles, and ultimately completely determines the spatial position of the section profile. In clinical applications, circular scanning usually starts from the vertical scanning line and samples in a clockwise direction. Therefore, the present invention uses the three-dimensional coordinate system in the The positive direction of the axis serves as the angular reference, with clockwise rotation angles considered positive. The first OCT B-scan image corresponds to an angle of 0°, and the remaining frames are rotated clockwise by corresponding angles based on their corresponding scan positions in the IR image. Once the angular reference is aligned, the absolute position of each OCT B-scan image in 3D space can be determined.

[0114] 2.2 Contour point serialization:

[0115] The cross-sectional contours extracted in the previous step are an unordered set of points. To achieve 3D morphological reconstruction and volume quantification of the edema area, this point set needs to be ordered. This involves sorting the points according to a specific rule, organizing the unordered point set into an ordered coordinate sequence. This allows for a structured contour description, laying the foundation for subsequent spatial mapping and volume quantification.

[0116] Figure 11 Shows the The frame scanning angle is The process of serializing the section contour points corresponding to the OCT B-scan image. Define the two-dimensional coordinate system of the section contour, denoted as The origin of the coordinate system coincides with the lowest point of the vertical center line of the OCT B-scan image, and the horizontal right direction is Axis positive direction, vertical upward direction Axis positive direction, The axis is the center of the aforementioned ring sweep. Use a vertical line from the leftmost side of the profile Starting from Move rightward in sequence for the step size, radial step size The length is equal to one pixel, recording the coordinates of the intersection of the vertical line and the upper and lower contours until the rightmost side of the contour Get The coordinates of the upper and lower contour points are completed. Frame section contour point serialization. All section contour points are serialized to obtain a point set , as shown in the formula As shown. In formula (6) represents an ordered set of contour points of the section, and Indicates the The first round of sweep pairs of upper and lower contour points, Indicates the The angle of the secondary sweep, Indicates the interval angle between two adjacent ring scans (rad), is the total number of ring sweeps, (pixel) means In the second scan The distance between the upper and lower contour points and the center of the ring scan is specified in Left of axis is a negative value, Right of axis is a positive value, and They are The first round of sweep Upper and lower contour points Coordinates. If and only if or The two points in the time point pair coincide.

[0117] (6);

[0118] 2.3 Contour point space mapping:

[0119] Point set The points in the formula Mapping from the two-dimensional coordinate system to the three-dimensional coordinate system, we can get the coordinates of all the contour points of the cross-section in the three-dimensional space. Formula (7) The points in the coordinates are mapped to In the coordinate system. and Indicates the The first The positions of the upper and lower contour point pairs in the two-dimensional coordinate system; and Indicates the The first The positions of the upper and lower contour point pairs in the three-dimensional coordinate system after mapping.

[0120] ;

[0121] Spatial mapping maps the contour points of the slices collected by the ring scan from the two-dimensional space in the OCT B-scan image to the three-dimensional space, and obtains the parametric space of the edema area (hereinafter referred to as the parametric space), such as Figure 12 and Figure 13 As shown, the coordinate system in the figure is Figure 9 Defined three-dimensional coordinate system. Figure 12 The blue and red points are the contour points scanned on the upper and lower surfaces of the edema area, respectively. The black curve on the plane is the outline of the edema area. vertical projection on a plane; Figure 13 The parametric space provides doctors with an intuitive reference for the morphology of the edema area, laying the foundation for subsequent volume quantification of the edema area.

[0122] 3. Volume quantification:

[0123] Due to the discrete sampling characteristics of OCT B-scans, there are angular gaps between circular scan frames, resulting in blank areas with no data between frames. This method establishes an adjacency relationship between frames, uses symmetrical frame units to rationally fill in the blank areas, and further divides the symmetrical frame units into microelements within a parameterized space. The volume of all microelements is calculated and then combined with the image scale to quantify the volume of the edema area.

[0124] 3.1 Adjacency Relationship Construction:

[0125] Since OCT B-scan images are acquired discretely, there are sampling gaps between different frames in the parameterized space, and data blank areas exist between frames during the 3D reconstruction process. In order to achieve continuous modeling and accurate quantification of the entire edema area, it is necessary to build an adjacency relationship between frames to establish a reasonable spatial correlation structure in the data blank area. Figure 14As shown in the figure, the green and red hatching lines respectively represent the clockwise and counterclockwise rotation angles of the points contained in the section contours corresponding to the 1st and 2nd frames of OCT B-scan images around the center of the ring scan. The symmetrical frame unit formed, The gap between two adjacent frames is constructed by half of the two adjacent symmetrical frame units, and all symmetrical frame units constitute the entire edema area.

[0126] 3.2 Microelement Division:

[0127] In the parameterized space, the symmetric frame unit is further divided into infinitesimal elements to solve its volume. Figure 15 Shows the In the symmetrical frame units, the 、 、 The vertical pixel columns between the pixel pairs rotate around the center of the ring scan to form three microelements. Each microelement is approximately a cuboid, and its volume is approximately the volume of the cuboid. For the The first of the symmetrical frame units The length of the infinitesimal element is Indicates that it is equal to the width of a pixel in the OCT B-scan image (pixel), and the width is equal to (pixel), height (pixel) represents the volume of a single infinitesimal element. The volume unit for calculation is the voxel (volumetric pixel), which is the pixel unit in a 3D image and represents the volume of a cube with a side length of 1 pixel. The volume of a single microelement includes approximation errors, which are high-order infinitesimals and therefore negligible. After dividing all symmetrical frame units that form the edema area into microelements, the entire edema area is divided into a number of microelements.

[0128] (8);

[0129] 3.3 Discrete integral approximate calculation:

[0130] In the parameterized space, after the edema area is divided into several microelements, the volume of the edema area can be approximated as the sum of the volumes of all microelements. Figure 16 It shows that one quarter of the edema area is divided into several micro-elements. For each symmetric frame unit, according to the formula Traverse all the infinitesimals and calculate their volumes and sum them up as the volume of the edema area in the parameterized space The unit is voxel. In formula (9), the volume of the edema area in the parameterized space is obtained by adding up the volumes of all micro-elements.

[0131] (9);

[0132] After calculating the volume of the edema area in the parametric space, the actual volume of the edema area needs to be calculated based on the scale. The volume of edema area is obtained by the present invention The relationship between is defined by formula (10), where It's a scale.

[0133] (10);

[0134] The scale of the cross section of the edema area is as follows Figure 17 As shown, Figure 17 Details of the scale bar in OCT B-Scan images are given in . The scale bar is calculated based on the formula In formula (11), , On OCT B-Scan images, and Resolution in direction, and The lengths are exist and The number of pixels corresponding to the direction.

[0135] ;

[0136] 4.Consistency assessment:

[0137] After obtaining the volume of the edema area, in order to verify the accuracy of the algorithm, it is necessary to compare it with the volume obtained by dense line scanning to conduct consistency evaluation. A series of line scan images are processed and annotated to obtain a series of binary images of the edema area section. According to the formula Calculate the volume. It represents the gold standard of edema area volume calculated based on OCT line scan images. represents the total number of binary images of line scan sections, Indicates the Linear section The area of ​​pixels with grayscale value equal to 255 in the binary image within the region , the resolution of the line scan OCT B-Scan image is equal to the resolution of the circular scan OCT B-Scan image and , Indicates the line scan spacing ( ), represents the area of ​​the line scan cross-section binary image, represents a point within this region, express Grayscale value of the point.

[0138] (12);

[0139] Figure 18 This is a diagram of the volume solution of line scan, showing the volume of the edema area obtained from three binary images of the edema area section. The distance between two adjacent binary images of the edema area segmentation is equal to the line scan spacing. The line scan has a small scanning interval, and the volume of the edema area solved by the line scan image is accurate enough to be used as a gold standard.

[0140] Table 1 shows the volume obtained by line scanning as the gold standard The present invention obtains volume based on ring scan image In comparison, among 7 patients, the average error of the edema area volume obtained by the present invention is -2.84%, the maximum error is -5.66%, and the minimum error is -0.36%. The small error fully demonstrates the effectiveness of the present invention in quantifying the edema area volume under sparse ring scanning sampling.

[0141] Table 1 Volume comparison (results of the present invention compared with the gold standard)

[0142]

[0143] Table 2 compares the total number of scans and effective number of scans of line scans and circular scans for 7 patients who underwent the two scanning methods. The effective number of scans of line scans refers to the total number of times the line scans scanned the edema area. Combining the scanning results of the seven samples: in terms of effective scan times, the line scans scanned the edema area approximately 55.4% of the time, and the circular scans scanned the edema area 100% of the time; in terms of total scan times, the total number of scans of line scans was approximately 2.27 times that of the circular scans, that is, the time consumed by the patient's line scans was approximately 2.27 times that of the circular scans. The present invention quantitatively evaluates the volume of the mid-serous edema area based on OCT circular scan images, realizes the three-dimensional reconstruction of the mid-serous edema area and its volume quantification with fewer scans, and ensures high accuracy.

[0144] Table 2 Comparison of total scan times and effective scan times (line scan vs. circular scan)

[0145]

[0146] The specific implementation method of the present invention is as follows Figure 19 shown.

[0147] 1. Image Cropping:

[0148] The original data of the present invention is collected from a certain OCT fundus examination device. The image consists of an IR image on the left and an OCTB-scan image on the right. Figure 2To meet the data requirements for 3D reconstruction and volume quantification of the edema area, it is necessary to accurately extract the OCT B-scan image from the original image. Based on the fixed position relationship of the OCT B-scan image in the original image, the present invention uses the upper left and lower right corner coordinates of the OCT B-scan image as the cropping boundaries to crop the original data, thereby obtaining an independent OCT B-scan image.

[0149] 2. Section marking:

[0150] The present invention uses Pair software to perform pixel-level annotation on the edema area section of the OCT B-scan image. Figure 20 As shown, in the present invention, the "contour filling" function of the software is used to mark the section of the mid-serous edema area.

[0151] 3. Edge detection:

[0152] Decompress the file annotated by Pair software to obtain the cross-section binary image, in which the white area represents the cross-section of the edema area. The Canny algorithm is used to detect the edge of the binary image and extract the cross-section contour. The cross-section contour extraction process is as follows: Figure 21 shown.

[0153] 4. 3D reconstruction:

[0154] A series of section contours extracted from the OCT B-scan image are rotated around the center of the ring scan at corresponding angles to achieve three-dimensional reconstruction of the edema area. This process includes spatial alignment, contour point serialization, and contour point spatial mapping. First, the rotation axis and rotation angle of all OCT B-scan images are unified to the defined three-dimensional coordinate system to determine the absolute position of each frame of OCT B-scan image in three-dimensional space. Secondly, in the two-dimensional coordinate system of the contour image, the section contour points extracted in the previous step are organized into an ordered coordinate sequence. The contour point serialization example is as follows: Figure 22 As shown, in In the coordinate system, use a vertical line in the direction of the blue arrow from the leftmost side of the contour to intersect the contour at a point pair and step to the right each time The coordinates of the point pairs are recorded until the rightmost edge of the contour. Finally, the coordinate sequence of the section contour points is sequentially mapped into a three-dimensional coordinate system to achieve three-dimensional reconstruction of the edema area and form a parameterized space, laying the foundation for volume quantification.

[0155] 5. Volume quantization algorithm:

[0156] The volume quantification algorithm is divided into three steps: the first step is to establish an adjacency relationship in the parametric space of the edema area according to the circular scanning order, and fill the sampling gap between two adjacent frames with half of the two adjacent symmetrical frame units; the second step is to further divide the symmetrical frame units into microelements; the third step is to use the sum of the volumes of all microelements to approximate the volume of the edema area in the parametric space, and combine the scale to obtain the true volume of the edema area, thereby realizing the volume quantification of the mid-serous edema area. Figure 23 The three-dimensional reconstruction and volume quantification results of 7 edema areas are shown.

[0157] Although the present invention has been described in detail above using general descriptions and specific embodiments, it will be apparent to those skilled in the art that modifications and improvements may be made thereto. Therefore, such modifications and improvements, without departing from the spirit of the present invention, are intended to be within the scope of protection claimed herein.

Claims

1. A method for three-dimensional reconstruction of mid-splenic edema and volume quantification thereof, characterized by: The following steps are involved: Step 1: Preprocess the original image, including image cropping, section annotation, and edge detection, to lay the foundation for smooth 3D reconstruction and volume quantification of the mid-serosal edema area. The original image consists of IR images and OCT B-scan images. Step 2: Points from the cross-sectional contours extracted from the OCT B-scan image are mapped to their correct positions in three-dimensional space. Visualization is then used to present the upper and lower surfaces of the edema area, achieving three-dimensional reconstruction of the edema area. This process includes spatial alignment, contour point serialization, and contour point spatial mapping. Through three-dimensional reconstruction, a parametric space for the mid-splenic edema area is obtained, laying the foundation for subsequent volume quantification. Step 3: By building an inter-frame adjacency relationship and filling the blank area with symmetrical frame units, the symmetrical frame units are further divided into microelements in the parametric space of the mid-serum edema area. The volume of all microelements is calculated and then combined with the image scale to quantify the volume of the edema area. The specific operations of microelement partitioning are as follows: Step 3.1: In the parameterized space, further divide the symmetric frame unit into infinitesimal elements to solve its volume; Step 3.2: After the edema area is divided into several microelements, the volume of the edema area is approximately the sum of the volumes of all microelements.

2. The method for three-dimensional reconstruction of mid-serous edema area and volume quantification thereof according to claim 1, characterized in that: Step 1: Step 1.1: Extract the OCT B-scan image from the original image to meet the data requirements for 3D reconstruction and volume quantification of the edema area. Based on the fixed position relationship of the OCT B-scan image in the original image, use the coordinates of the upper left corner and lower right corner of the OCT B-scan image as the cropping boundary to obtain the OCT B-scan image. The cropping method is shown in formula (1); R={(x,y)|x1≤x≤x2,y1≤y≤y2} (1); In formula (1), R represents the OCT B-scan image area, (x1, y1) is the coordinate of the upper left corner point P1 of the OCT B-scan image area, and (x2, y2) is the coordinate of the lower right corner point P2 of the OCT B-scan image area. The coordinates of P1 and P2 are calculated based on the resolution of the infrared image and the overall size of the original image, as shown in formula (2). In formula (2), w and h are the width and height of the infrared image, respectively, and W is the width of the original image; Step 1.2: Use medical image annotation software to perform pixel-level annotation on the edema area section of the OCT B-scan image to obtain an annotation map of the edema area section in the OCT B-scan image; Step 1.3: Use the Canny algorithm to detect the edge of the edema area in the binary image, so as to accurately extract its contour.

3. The method for three-dimensional reconstruction of mid-serous edema area and volume quantification thereof according to claim 2, characterized in that: The process of detecting the edge of the mid-serous edema area using the Canny algorithm in step 1.3 is divided into the following four steps: (1) Image denoising: Use Gaussian filtering to smooth the image and reduce the noise in the image. Use a 5*5 Gaussian filter as shown in formula (3): (2) Calculate the image gradient, detect possible edges, and calculate the gradient direction and amplitude of the grayscale value of each pixel in the image: Use the Sobel operator to detect the edges in the horizontal direction x and vertical direction y of the image, and then use formulas (4) and (5) to calculate the gradient direction θ grad and amplitude G xy , where I x Indicates the grayscale of the pixel in the x direction, I y Represents the grayscale of the pixel in the y direction; the Sobel operator extracts the image edge through discrete differential method: (3) Non-maximum suppression: The non-maximum suppression algorithm is used to eliminate the stray responses caused by edge detection. The pixels with the largest grayscale gradient in the gradient direction within the local range are retained, while the others are not retained. A large number of points are eliminated, and the edge with multiple pixel widths is turned into an edge with a single pixel width. (4) Double threshold screening: After non-maximum suppression, the double threshold method is applied to divide strong edges and weak edges. Pixels with grayscale gradients greater than the high threshold are retained as strong edges, and pixels with grayscale gradients less than the low threshold are removed. Pixels with grayscale gradients between the high and low thresholds are screened as weak edges. Search within the neighborhood of the weak edge. If a strong edge exists, the weak edge is retained. If not, the weak edge is deleted to complete edge detection.

4. The method for three-dimensional reconstruction of mid-serous edema area and volume quantification thereof according to claim 1, characterized in that: Step 2: Step 2.1: Based on a right-handed Cartesian coordinate system, establish a three-dimensional coordinate system (x, y, z) to map points in the slice contour extracted from the OCT B-scan image to their correct positions in the three-dimensional coordinate system. The origin O is set to the lowest point of the vertical centerline of the OCT B-scan image. The x-axis coincides with the horizontal scan line and points in the horizontal direction indicated by the scan arrow. The y-axis coincides with the vertical scan line and points in the vertical direction indicated by the scan arrow. The z-axis is perpendicular to the xOy plane and points outward perpendicular to the paper. The center of the circular scan in the contour IR image corresponds to the vertical centerline of each OCT B-scan image frame and is considered the rotation center of the entire scan sequence in the three-dimensional coordinate system. The lowest point of the vertical centerline of each OCT B-scan image frame is aligned with the origin of the three-dimensional coordinate system, and the vertical centerline of the OCT B-scan image is aligned with the z-axis of the three-dimensional coordinate system. The rotation reference of all OCT B-scan images is unified to the z-axis, clarifying the relative position relationship of each image frame in the three-dimensional coordinate system. The positive direction of the y-axis in the 3D coordinate system is used as the angle reference, with clockwise rotation angles considered positive. The first OCT B-scan image frame corresponds to an angle of 0°, and the remaining frames are rotated clockwise by corresponding angles based on their corresponding scanning positions in the IR image. Once the angle reference is aligned, the absolute position of each OCT B-scan image frame in 3D space can be determined. Step 2.2: Order the obtained unordered point set section contour to form an ordered coordinate sequence, and then construct a structured contour description; Step 2.3: Map the points in the ordered coordinate sequence from the 2D coordinate system to the 3D coordinate system to obtain the coordinates of all section contour points in the 3D space. Spatial mapping maps the section contour points acquired by the circular scan from the 2D space in the OCT B-scan image to the 3D space to obtain the parameterized space of the edema area, laying the foundation for subsequent volume quantification of the edema area.

5. The method for three-dimensional reconstruction of mid-serous edema area and volume quantification thereof according to claim 4, characterized in that: Step 2.2: Define the two-dimensional coordinate system (r, h) of the section contour. The origin of the coordinate system coincides with the lowest point of the vertical centerline of the OCT B-scan image. The horizontal right direction is the positive r-axis, and the vertical upward direction is the positive h-axis. The h-axis is the center of the circular scan. Use a vertical line from the leftmost side of the contour r = r min Starting from the right, move to the right in steps of Δr, where the radial step length Δr is equal to one pixel, and record the coordinates of the intersection of the vertical line and the upper and lower contours until the rightmost point of the contour is r = r max At, get M i The coordinates of the upper and lower contour point pairs are obtained, and the serialization of the contour points of the i-th frame section is completed; After serializing all the section contour points, we get the point set C, as shown in formula (6); C={(θ i ,r ij ,h 1,ij ),(θ i ,r ij ,h 2,ij )|i=1,...N,j=1,...,M i ,i i =(i-1)·Δθ,Δθ=π / N,M i =(r max -r min ) / Δr,r ij =r min +j·Δr} (6); In formula (6), C represents the ordered set of contour points of the section, (θ i ,r ij ,h 1,ij ) and (θ i ,r ij ,h 2,ij ) represents the jth upper and lower contour point pair in the i-th ring scan, θ i represents the angle of the i-th circular sweep, Δθ represents the interval angle between two adjacent circular sweeps, N is the total number of circular sweeps, r ij Indicates the distance between the jth upper and lower contour point pair and the center of the ring scan in the i-th scan, specified to be r to the left of the h-axis ij is a negative value, r to the right of the h axis ij is a positive value, h 1,ij and h 2,ij are the h coordinates of the jth upper and lower contour points in the i-th ring scan respectively; if and only if r ij =r max or r ij =r min The two points in the time point pair coincide.

6. The method for three-dimensional reconstruction and volume quantification of mid-serous edema area according to claim 5, characterized in that: The specific operations of step 2.3 are as follows: The point in the two-dimensional coordinate system (r, h) is mapped to the three-dimensional coordinate system (x, y, z) using formula (7): (r ij ,h 1,ij ) and (r ij ,h 2,ij ) represents the position of the jth upper and lower contour point pair in the i-th scan in the two-dimensional coordinate system; (x ij ,y ij ,z 1,ij ) and (x ij ,y ij ,z 2,ij ) represents the position of the j-th upper and lower contour point pair in the i-th scan in the three-dimensional coordinate system after mapping.

7. The method for three-dimensional reconstruction and volume quantification of mid-serous edema area according to claim 5, characterized in that: Step 3.1 specifically includes: In the i-th symmetrical frame unit, three microelements are formed by the vertical pixel columns between the j-th, j+1-th, and j+2-th pixel pairs rotating around the center of the circular scan; each microelements is approximately a cuboid, and its volume is approximately the volume of the cuboid ΔV ij For the jth microelement in the i-th symmetric frame unit, its length is represented by Δr, which is equal to the width of a pixel in the OCT B-scan image, and the width is equal to |r ij |·Δθ, Δh for height ij express; The volume of a single infinitesimal approximation is calculated according to formula (8); The volume unit is voxel, which is the pixel unit in the three-dimensional image, representing the volume of a cube with a side length of 1 pixel. After all the symmetrical frame units combined into the edema area are divided into microelements, the entire edema area is divided into several microelements.

8. The method for three-dimensional reconstruction and volume quantification of mid-serous edema area according to claim 7, characterized in that: Step 3.2 specifically includes: For each symmetric frame unit, all the microelements in it are traversed according to formula (9) and their volumes are calculated and summed up to be the volume of the edema area in the parameterized space The unit is voxel; in formula (9), the volume of the edema area in the parameterized space is obtained by adding up the volumes of all micro-elements; After calculating the volume of the edema area in the parameterized space, the true volume of the edema area is calculated according to the scale; the volume of the edema area in the parameterized space and the volume of edema area The relationship between is defined by formula (10), where S is the scale; The scale is obtained according to formula (11); In formula (11), S r ,S h are the resolutions in the r and h directions on the OCT B-Scan image, and n rpixel and n hpixel are the number of pixels corresponding to the length d in the r and h directions respectively.

Images