Light field reconstruction brain image brain region boundary extraction method based on convex hull fitting

Abstract

The invention discloses a light field reconstruction brain image brain region boundary extraction method based on convex hull fitting. The method comprises the steps that the p percentile of a brain image is calculated, and image cutting, contrast stretching and Gaussian filtering smooth denoising are carried out; carrying out Canny edge detection on the filtered image to obtain a binary edge image; boundary enhancement and connection are carried out through expansion and closed operation, all connected regions are extracted, and the contour with the maximum area is selected as the contour of the brain region boundary; a datum point is determined based on the contour of the brain region boundary, the polar angle and distance of each point except the datum point are calculated and sorted, and convex hull fitting is carried out through Graham Scan scanning based on the datum point and the sorted point set; converting the convex hull into a closed boundary region, and obtaining a mask image of the brain image after light field reconstruction through a mask function; according to the method, the special boundary form of the light field reconstruction image can be accurately adapted, deep learning and data training are not needed, and the generated mask has high geometric consistency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of light field microscopic imaging, and more particularly to a light field reconstruction brain image brain region boundary extraction method based on convex hull fitting. BACKGROUND

[0002] Light field microscopic imaging collects the information of three-dimensional scene space and angle through a single shooting, and restores the angle of each position of the three-dimensional scene through post-reconstruction, is very suitable for fast and high-resolution imaging of dynamic neural activity in the living brain, and three-dimensional data of different depths in the brain can be obtained through reconstruction of the light field image.

[0003] However, the reconstruction process will be disturbed by the signal interference of the image sensor noise, background fluorescence and the motion artifacts introduced by muscle activity or blood vessels and the like, and then the reconstruction error exists in the reconstructed image, which submerges the real signal of the brain image; due to the optical properties of the brain tissue itself and the limitation of the imaging depth, the edge of the brain tissue often has low contrast and is discontinuous; in large brain science research, a large amount of light field imaging data needs to be automatically processed to extract the neuron activity signal in a specific brain region, and manual delineation of the brain region boundary is time-consuming and easy to introduce subjective error.

[0004] Inaccurate brain region mask will lead to pollution of downstream neural signal extraction analysis, specifically: False positive: the noise or non-neuron structure outside the mask is wrongly included, which is misjudged as an active neuron by the neuron recognition algorithm, and an error signal is introduced; false negative: the incomplete mask may cut off the real neuron located at the edge of the brain region, resulting in loss of its signal; signal pollution: when calculating the overall background noise or the fluorescence change of the neuron, if the mask contains a high-intensity noise area, the calculation result will be seriously distorted, and the signal-to-noise ratio of the signal will be reduced.

[0005] Segmentation and boundary extraction of brain images are important prerequisites for brain structure analysis, functional research and lesion detection, and the current common brain boundary extraction methods mainly have the following problems: The method based on global threshold is simple to implement, but sensitive to noise and brightness change, and the boundary is easy to break or shift; the method based on region growing depends on seed point selection, and the degree of automation is not high; the method based on deep learning has high segmentation accuracy, but needs a large amount of labeled data, has high calculation cost, and is difficult to popularize in resource-limited scenarios.

[0006] Therefore, the problem that the person skilled in the art urgently needs to solve is: how to overcome the reconstruction error, abnormal noise, low contrast and boundary fracture existing in the light field reconstructed brain image, realize a full-automatic and high-robust brain region boundary extraction method, so as to generate continuous, complete and accurate binary mask, provide high-quality region of interest basis for subsequent automatic neural cell recognition and fluorescence signal extraction, and finally improve the accuracy and reliability of neural activity analysis results. SUMMARY

[0007] In view of the above problems, the present application is proposed to provide a convex hull fitting-based light field reconstructed brain image brain region boundary extraction method to overcome the above problems or at least partially solve the above problems.

[0008] In order to achieve the above purpose, the present application adopts the following technical scheme: A convex hull fitting-based light field reconstructed brain image brain region boundary extraction method, comprising: S1. Calculate the p-th percentile of the brain image after light field reconstruction and clip the image gray value; S2. Contrast stretching is performed on the clipped image, and Gaussian filtering is performed for smoothing and denoising; S3. Canny edge detection is performed on the image after filtering to obtain a binary edge map; S4. Boundary enhancement and connection are performed on the binary edge map through dilation and closing operation; S5. All connected regions are extracted from the binary edge map, and the largest area contour is selected as the contour representing the brain region boundary; S6. A reference point is determined based on the contour of the brain region boundary, the polar angle and distance of each point except the reference point are calculated and sorted, and convex hull fitting is performed through Graham Scan scanning based on the reference point and the sorted point set; S7. The convex hull is converted into a closed boundary region, and a mask image of the light field reconstructed brain image is obtained through a mask function.

[0009] Preferably, the specific content of step S1 comprises: S11. Calculate the p-th percentile of the brain image after light field reconstruction .

[0010] Wherein, represents the two-dimensional gray matrix of the brain image , , is an image pixel set, x and y are pixel coordinates, and the image pixel value represents a gray value; S12. Clip the image gray value to the interval :

[0011] wherein, is the cropped image, is defined as:

[0012] wherein, is the minimum clipping value, is the maximum clipping value.

[0013] Preferably, the specific content of the contrast stretching in step S2 is: the cropped image is linearly normalized to map the dynamic range to the standard 8-bit grayscale range [0, 255] to maximize the image contrast;

[0014] wherein, and are the minimum and maximum grayscale values in the image , respectively; The method of Gaussian filtering is:

[0015] wherein, is the image after contrast stretching, is the image after filtering, is the size of the two-dimensional Gaussian kernel, denotes the Gaussian distribution function, and are the Gaussian kernel coordinates in x and y directions, respectively.

[0016] Preferably, the specific content of step S3 includes: S31. Using the Sobel operator to calculate the first-order gradient components of the image in x and y directions, respectively, on the image after Gaussian filtering; S32. According to the gradient components, the gradient amplitude and gradient direction of each pixel point are calculated; S33. Based on the gradient amplitudes of each pixel point and its two adjacent pixels along the gradient direction, non-maximum suppression is performed on the gradient amplitude image to obtain candidate edge points; S34. A double-threshold processing is applied to the gradient amplitude image after non-maximum suppression to label strong edge points and weak edge points for the candidate edge points; S35. The strong edge points and the weak edge points connected to any strong edge point within the 8-neighborhood are edge-connected to obtain a binary edge map.

[0017] Preferably, the method of step S4 for boundary enhancement and connection of the binary edge map is:

[0018]

[0019] wherein, is the binary edge map, represents dilation, represents erosion, is a 5x5 structuring element matrix.

[0020] Preferably, in step S5, the contour of the brain region boundary is:

[0021] wherein, is all connected regions extracted from the binary edge map after boundary enhancement and connection, is the area of each region, and the largest area is selected as the contour of the brain region.

[0022] Preferably, in step S6, the method of determining the reference point based on the contour of the brain region boundary is: selecting the point set of the contour of the brain region boundary selecting the point with the smallest longitudinal coordinate as the reference point if there are multiple points with the same longitudinal coordinate, selecting the point with the smallest horizontal coordinate as the reference point.

[0023] Preferably, in step S6, the specific content of calculating the polar angle and distance of each point except the reference point and sorting is: for each point in the point set of the contour of the brain region boundary except the reference point , calculating the polar angle and the squared Euclidean distance relative to the reference point :

[0024]

[0025] wherein, is the four-quadrant arctangent function, and the return value ranges from ; sorting all points in ascending order of polar angle; when the polar angle is the same, sorting in ascending order of distance from the reference point , to obtain the sorted point set.

[0026] ​The specific method for convex hull fitting based on the reference point and the ordered point set by the Graham Scan scanning is preferably as follows: Initialize a stack and push the reference point and the first two points of the ordered point set into the stack in turn:

[0027] From the third point of the ordered point set, each point is traversed in turn, and in each step, the turning relationship of three points including the top of the stack top, the next top next and the current point is checked. When , it indicates that the three points form a left turn, and the current point is pushed into the stack. When , it indicates that the three points form a collinear or right turn, and the top element top of the stack is popped out, and the new top three points are rechecked until the left turn condition is met or the number of elements in the stack is less than two. After traversing all points, the point sequence stored in the stack is the vertex of the convex hull. After traversing all points, the point sequence stored in the stack is the vertex of the convex hull. After traversing all points, the point sequence stored in the stack is the vertex of the convex hull. After traversing all points, the point sequence stored in the stack is the vertex of the convex hull. After traversing all points, the point sequence stored in the stack is the vertex of the convex hull.

[0028] The mask image of the brain image after light field reconstruction in step S7 is preferably as follows:

[0029]

[0030] , wherein f is a mask function, and the convex hull converted closed boundary region is

[0031] Compared with the prior art, the technical scheme has the following beneficial effects: By polar angle ordering and convex hull construction on the edge point set, global fitting of the overall boundary is realized, and the defect that the traditional threshold segmentation is sensitive to local gray level fluctuation is overcome.

[0032] The present application does not need resource-consuming network training process and can be run in real time on CPU, and is suitable for batch image rapid processing. ​​​​The mask generated by the application has high consistency with the target region boundary, strong continuity and no obvious edge jitter, and can be directly used in subsequent processing steps, including but not limited to brain region structure segmentation, multi-modal registration, signal quantitative analysis, and three-dimensional reconstruction and visualization; in particular, in the light field reconstruction image, the gray level change between the tissue boundary and the background is often relatively slow, and the application can provide more stable boundary positioning than the simple threshold method, and ensure the integrity of the brain region ROI. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only a part of the embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0034] Figure 1 A flow chart of a brain region boundary extraction method based on convex hull fitting for a light field reconstruction brain image is provided in the embodiments of the present application. Figure 2 A neighborhood partitioning schematic diagram for non-maximum suppression is provided in the embodiments of the present application. Figure 3 An intermediate effect diagram in the experiment is provided in the embodiments of the present application. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0036] Embodiment one The embodiments of the present application disclose a brain region boundary extraction method based on convex hull fitting for a light field reconstruction brain image, as shown in Figure 1 , which comprises: S1. Calculate the p percentile of the brain image after light field reconstruction and clip the image gray value; S2. Perform contrast stretching on the clipped image, and perform Gaussian filtering and smoothing denoising; S3. Perform Canny edge detection on the image after filtering to obtain a binary edge map; S4. Perform boundary enhancement and connection on the binary edge map through dilation and closing operation; S5. Extract all connected regions from the binary edge map and select the largest contour as the contour representing the brain region boundary; S6. Determine the fiducial point based on the contour of the brain region boundary, calculate the polar angle and distance of each point except the fiducial point and sort them, and perform convex hull fitting based on the fiducial point and the sorted point set by Graham Scan method; S7. Convert the convex hull into a closed boundary region, and obtain the mask image of the brain image after light field reconstruction by a mask function.

[0037] In order to further implement the above technical solution, the specific content of step S1 includes: S11. Calculate the p-th percentile of the brain image after light field reconstruction , which aims to suppress the high-brightness outliers caused by the possible saturated pixels, fluorescent markers or reconstruction artifacts, and the specific content is:

[0038] wherein, represents the two-dimensional gray matrix of the brain image , and the input of the embodiment is the middle clear layer of the three-dimensional brain image, , is the image pixel set, x and y are pixel coordinates, and the image pixel value represents the gray value, and p is usually set to 99; S12. Clip the image gray value to the interval :

[0039] wherein, is the clipped image, the function is defined as:

[0040] wherein, is the minimum cutoff value, is the maximum cutoff value, and in the embodiment, , .

[0041] In order to further implement the above technical solution, the specific content of step S2 is: linear normalization is performed on the clipped image to map the dynamic range to the standard 8-bit gray range [0, 255] to maximize the image contrast;

[0042] wherein, and are the minimum and maximum gray values in the image , respectively; In the embodiment, and Specifically:

[0043] ; In order to remove noise, Gaussian smoothing is performed on the image. The Gaussian filter can effectively reduce the background noise while preserving the edge information of the image. The method of Gaussian filtering is as follows:

[0044] wherein, is the image after contrast stretching, is the image after filtering, is the size of the two-dimensional Gaussian kernel, is the two-dimensional Gaussian kernel, representing the Gaussian distribution function, and are the Gaussian kernel coordinates in the x and y directions, respectively; In this embodiment, the Gaussian distribution function is specifically:

[0045] wherein, is the action bandwidth of the Gaussian kernel, used to control the smoothing degree of the image after Gaussian filtering, The larger the value is, the smoother the filtered image is, and vice versa.

[0046] In order to further implement the above technical solution, step S3 is performed through a multi-stage optimal edge detection algorithm Canny edge detection, which aims to suppress interference while accurately positioning the real edge. The Canny edge detection is performed on the image after filtering, and the data is compressed into boundary information for subsequent processing. The specific content of Canny edge detection includes: S31. On the image after Gaussian filtering, the Sobel operator is used to calculate the first-order gradient components of the image in the x and y directions, respectively; In this embodiment, the first-order gradient approximation and are:

[0047]

[0048] wherein, represents convolution operation, and are the Sobel convolution kernels in the x and y directions, respectively:

[0049]

[0050] S32. Calculate the gradient magnitude and the gradient direction of each pixel point based on the gradient components ;

[0051]

[0052] S33. Perform non-maxima suppression on the gradient magnitude image based on the gradient magnitude of each pixel point and its two neighboring pixels along its gradient direction, to obtain the candidate edge points; Specifically, for each pixel point, check its two neighboring pixels along its gradient direction , if the gradient magnitude of the current pixel is not the local maximum along the gradient direction, then suppress it to zero. In actual calculation, the gradient direction is discretized into four main directions (0°, 45°, 90°, 135°), and the gradient magnitudes of the current pixel and its two neighboring pixels along the direction are compared, as shown in the following formula: Figure 2 If is smaller than the gradient magnitude of any neighboring pixel, then , otherwise, keep the original value . After this step, only the points with the local maximum gradient are kept as the candidate edge points, thus achieving the purpose of liking the edge points; S34. Apply double threshold processing to the gradient magnitude image after non-maxima suppression, to label the strong edge points and weak edge points for the candidate edge points; Specifically, set the high threshold and the low threshold , usually , can be set to , if , then the point is labeled as a strong edge point; if , then the point is labeled as a weak edge point; if , then the point is suppressed; S35. Perform edge connection for the strong edge points and the weak edge points connected to any strong edge point within the 8-neighborhood, to obtain the binary edge map; In this embodiment, all weak edge points are checked, and if any weak edge point is connected to any strong edge point within the 8-neighborhood, then the weak edge point is promoted to a strong edge point and included in the final edge, and the final binary edge map is defined as: .

[0053] ​In order to further implement the above technical solution, because the vascular morphology of different brain regions is quite different, in order to prevent the influence of blood vessels on the detection of the brain region contour, step S4 performs dilation and closing operation on the binary edge map to perform boundary enhancement and connection to eliminate the blood vessel cavity in the brain region, and the method is as follows:

[0054]

[0055] wherein, is a binary edge map, represents dilation, represents erosion, is a 5*5 structure element matrix.

[0056] In order to further implement the above technical solution, in step S5, the contour of the extracted brain region boundary is as follows:

[0057] wherein, is all connected regions extracted from the binary edge map after boundary enhancement and connection, is the area of each region, and the largest area is selected as the contour of the brain region.

[0058] The convex hull H is the smallest convex polygon containing all points of the point set S, that is, the smallest convex polygon containing all constituent points in the point set In the present embodiment, the Graham Scan method is used for convex hull fitting.

[0059] In order to further implement the above technical solution, in step S6, the method for determining the reference point based on the contour of the brain region boundary is as follows: select the longitudinal coordinate of the point set of the contour points of the brain region boundary as the reference point, if there are multiple points with the same longitudinal coordinate, select the point with the smallest horizontal coordinate as the reference point:

[0060] wherein, is defined as comparing in ascending order first, and if is the same, comparing in ascending order.

[0061] In order to further implement the above technical solution, in step S6, the specific content of calculating the polar angle and distance of each point except the reference point and sorting is as follows: for the point set​​ each point of the set of points except the reference point , the polar angle of each point with respect to the reference point and the squared Euclidean distance :

[0062]

[0063] where is the four-quadrant arctangent function, the return value range is , which can accurately reflect the direction of the point; sort all points in ascending order of polar angle , and when the polar angle is the same, sort in ascending order of distance from the reference point , to get the sorted point set ; The comparison rule is as follows: if , is in front; if , compare the distance, if , is in front, and the polar angle sorting ensures that the access order of the point set conforms to the geometric loop order of the convex hull.

[0064] In order to further implement the above technical scheme, the specific method for fitting the convex hull by Graham Scan scanning based on the reference point and the sorted point set is as follows: Initialize a stack , and push the reference point and the first two points of the sorted point set into the stack in turn:

[0065] From the third point of the sorted point set, traverse each point , , in each step, check the turning relationship of the three points including the top of the stack top, the next top next, and the current point by calculating the two-dimensional cross product ; define the vector , , and the two-dimensional cross product is:

[0066] When , it means that the three points form a left turn (counterclockwise), and the current point is pushed into the stack ; when ​​When the three points are collinear or rotate to the right (clockwise), the top element of the stack is popped, and the new top three points are rechecked until the left rotation condition is met or the number of elements in the stack is less than two. This ensures that the point sequence in the stack always maintains the geometric property of a strict left rotation, thus corresponding to the convex hull boundary of the currently scanned point set. After traversing all points, the stack is... The sequence of points stored in the convex hull forms the convex hull. The vertices are arranged in a counter-clockwise direction;

[0067] in, .

[0068] To further implement the above technical solution, in step S7, the brain image after light field reconstruction... mask image for:

[0069]

[0070] in, For mask function, As a closed boundary region of convex hull transformation, this mask image accurately covers the brain region and can be directly used for subsequent analysis tasks such as neuronal signal extraction.

[0071] Example 2 This embodiment uses light field reconstruction images of the mouse cerebral cortex as an example. The experimental data comes from adult mouse brain images acquired by a light field microscopy system, with an image resolution of [missing information]. Pixels, after light field reconstruction, yield a depth of approximately The volumetric data is used to select mid-layer slices with clear signals as two-dimensional input images. The brain region boundary extraction method for light field reconstruction based on convex hull fitting proposed in this invention was implemented. The intermediate experimental results are shown in the figure below. Figure 3 As shown, from left to right, the original image, edge map, convex hull mask map, and mask-filtered posterior brain region image are displayed.

[0072] Example 3 A computer-readable storage medium having a computer program stored thereon, characterized in that, when executed by a processor, the computer program implements a method for extracting brain region boundaries from light field reconstructed brain images based on convex hull fitting.

[0073] Example 4 A processing terminal includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, characterized in that the processor executes the computer program to implement a method for extracting brain region boundaries from light field reconstruction brain images based on convex hull fitting.

[0074] The various embodiments described in this specification are presented by way of example, and each embodiment is presented with the understanding that it will not limit the present application to that embodiment alone. Each embodiment is presented in its own right, and the various embodiments are not presented as alternatives to one another. The same part or feature can be present in more than one embodiment. The various embodiments described in this specification can be implemented in any combination.

[0075] The above description of disclosed embodiments is intended to be illustrative, and not restrictive. Many embodiments will be apparent to those of skill in the art upon reviewing the above description. The scope of the application should, therefore, be determined not with reference to the above description, but instead with reference to the appended claims, along with their full scope of equivalents.

Claims

1. A method for extracting brain region boundaries in light field reconstruction brain images based on convex hull fitting, characterized in that, include: S1. Calculate the p percentile of the brain image after light field reconstruction and crop the image grayscale values; S2. Perform contrast stretching on the cropped image and smooth it with Gaussian filtering to remove noise; S3. Perform Canny edge detection on the filtered image to obtain a binary edge map; S4. Enhance the boundary and improve connectivity of a binary edge graph through dilation and closing operations; S5. Extract all connected regions from the binary edge map and select the contour with the largest area as the contour representing the boundary of the brain region. S6. Determine the reference point based on the contour of the brain region boundary, calculate the polar angle and distance of each point except the reference point and sort them, and perform convex hull fitting based on the reference point and the sorted point set by Graham Scan. S7. Convert the convex hull into a closed boundary region, and obtain the mask image of the brain image after light field reconstruction through the mask function.

2. The method for extracting brain region boundaries in light field reconstruction brain images based on convex hull fitting as described in claim 1, characterized in that, The specific content of step S1 includes: S11. Calculate the brain image after light field reconstruction. The p percentile is: in, Representing brain images A two-dimensional grayscale matrix, , This is a set of image pixels, where x and y are pixel coordinates, and the image pixel values ​​represent grayscale values. S12. Crop the image grayscale values ​​to the specified range. : in, The cropped image. The function is defined as: in, The minimum cutoff value, This is the maximum truncation value.

3. The method for extracting brain region boundaries in light field reconstruction brain images based on convex hull fitting as described in claim 1, characterized in that, In step S2, the specific details of contrast stretching are as follows: For the cropped image Linear normalization is performed to map the dynamic range to the standard 8-bit grayscale range [0, 255] to maximize image contrast; in, and Images The minimum and maximum gray values ​​in the range; The Gaussian filtering method is as follows: in, The image after contrast stretching. The image after filtering. For a two-dimensional Gaussian kernel, Represents the Gaussian distribution function. and These are the Gaussian kernel coordinates in the x and y directions, respectively.

4. The method for extracting brain region boundaries in light field reconstruction brain images based on convex hull fitting as described in claim 1, characterized in that, The specific content of step S3 includes: S31. On the image after Gaussian filtering, use the Sobel operator to calculate the first-order gradient components of the image in the x and y directions respectively; S32. Calculate the gradient magnitude and gradient direction of each pixel based on the gradient components; S33. Based on the gradient magnitude of each pixel and its two neighboring pixels along its gradient direction, non-maximum suppression is performed on the gradient magnitude image to obtain candidate edge points; S34. Apply double thresholding to the gradient magnitude map after non-maximum suppression, and mark strong edge points and weak edge points for candidate edge points; S35. Connect the strong edge points and the weak edge points that are connected to any strong edge point in the 8-neighborhood to obtain a binary edge map.

5. The method for extracting brain region boundaries in light field reconstruction brain images based on convex hull fitting as described in claim 1, characterized in that, Step S4 involves performing boundary enhancement and connectivity improvements on the binary edge graph using dilation and closing operations. in, This is a binary edge map. Indicates expansion. Indicates corrosion. It is a 5×5 structure element matrix.

6. The method for extracting brain region boundaries in light field reconstruction brain images based on convex hull fitting as described in claim 1, characterized in that, In step S5, the extracted brain region boundary contours are as follows: in, This refers to all connected regions extracted from the binary edge map after boundary enhancement and connectivity. For each region, the largest area is selected as the outline of the brain region.

7. The method for extracting brain region boundaries in light field reconstruction brain images based on convex hull fitting as described in claim 1, characterized in that, In step S6, the method for determining the reference point based on the contour of the brain region boundary is as follows: Select the contour point set of brain region boundaries The point with the smallest ordinate in the middle is used as the reference point. If there are multiple points with the same vertical coordinate, the point with the smallest horizontal coordinate is selected as the reference point.

8. The method for extracting brain region boundaries in light field reconstruction brain images based on convex hull fitting as described in claim 1, characterized in that, In step S6, the specific steps for calculating and sorting the polar angle and distance of each point other than the reference point are as follows: For the contour point set of brain region boundaries Except for the reference point every point Calculate its relative to the reference point polar angle Square distance from Euclidean distance : in, This is the arctangent function in the four quadrants, and its return value range is... ; Sort all points in ascending order of polar angle; when polar angles are the same, sort them according to their relative position to the reference point. Sort the points in ascending order by distance to obtain the sorted set of points.

9. The method for extracting brain region boundaries in light field reconstruction brain images based on convex hull fitting as described in claim 1, characterized in that, The specific method for convex hull fitting based on the baseline points and the sorted point set using Graham Scan is as follows: Initialize a stack Then, push the pivot point and the first two points of the sorted point set onto the stack in sequence: Starting from the third point in the sorted point set, traverse each point sequentially, calculating the two-dimensional cross product at each step. The check includes the top of the stack (top), the next top of the stack (next), and the current point. The turning relationship between the three points; when When, it indicates that three points form a left turn, and the current point... Push onto stack ;when When the three points are collinear or turn right, pop the top element of the stack `top`, and recheck the new top three points of the stack until the left turn condition is met or the stack contains fewer than two elements. After traversing all points, the stack is... The sequence of points stored in the convex hull forms the convex hull. The vertices are arranged in a counter-clockwise direction.

10. The method for extracting brain region boundaries in light field reconstruction brain images based on convex hull fitting as described in claim 1, characterized in that, Step S7, Brain image after light field reconstruction mask image for: in, For mask function, This is the closed boundary region for convex hull transformation.

Images