Adaptive Repeated Scanning Method and Device

Abstract

The present invention discloses an adaptive repeated scanning method and apparatus. The method includes: scanning a target area corresponding to a target feature at a current scanning position to obtain a first image set of the target feature at the current scanning position; performing a multi-image averaging operation on all the scanned images included in the first image set to obtain an initial image; determining whether there are abnormal features in the initial image; when it is determined that there are abnormal features in the initial image, repeatedly scanning the target area according to a determined additional scanning number to obtain a second image set; performing a multi-image averaging operation on all the scanned images included in the first image set and all the scanned images included in the second image set to obtain a target image. It can be seen that implementing the present invention can obtain more scanned images by increasing the scanning number to obtain a target image with higher imaging quality when and only when there are abnormal features in the initial image.

Description

Technical Field

[0001] The present invention relates to the technical field of scanning imaging, and particularly to an adaptive repeated scanning method and device. Background Art

[0002] Optical Coherence Tomography (OCT) is a tomographic imaging method that utilizes the coherence of light. Each tomographic scan image (cross-sectional image) is called a B-Scan image, and each B-Scan image includes multiple A-Scan lines. In practical applications, B-Scan images can be used for teaching and experimental research to learn and study the diagnosis of fundus diseases. The fundus diseases that can be learned and studied through B-scan images include, but are not limited to, fundus effusion, macular hole, macular edema, retinal hole, retinal detachment, neovascularization, drusen, etc.

[0003] Generally speaking, imaging images obtained by repeatedly sampling the same position have higher imaging quality. Therefore, existing scanning methods perform undifferentiated repeated scanning of the scanning area a fixed number of times to obtain imaging images. However, using such a scanning method will increase the scanning time and affect the patient experience. Thus, it is very important to obtain imaging images with higher imaging quality in the scanning area while ensuring the scanning efficiency. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an adaptive repeated scanning method and device, which can scan a target area corresponding to a target feature at a current scanning position to obtain a plurality of scanning images, and perform a multi-image averaging operation on the plurality of scanning images to obtain an initial image, thereby helping to generally understand the situation of the target feature based on the initial image, and can also determine whether there are abnormal features in the initial image. Moreover, when and only when there are abnormal features in the initial image, more scanning images can be obtained by increasing the scanning times, so as to obtain a target image with higher imaging quality based on more scanning images while ensuring the scanning efficiency, and help to have a more accurate and detailed understanding of the situation of the target feature based on the target image with higher imaging quality.

[0005] To solve the above technical problem, a first aspect of the present invention discloses an adaptive repeated scanning method, and the method includes:

[0006] Scanning a target area corresponding to a target feature at a current scanning position to obtain a first image set of the target feature at the current scanning position; the target area is an area including the target feature at the current scanning position;

[0007] Perform a multi-image averaging operation on all the scanned images included in the first image set to obtain an initial image of the target feature at the current scanning position;

[0008] Determine whether there are abnormal features in the initial image;

[0009] When it is determined that there are abnormal features in the initial image, repeat scanning the target area according to the determined supplementary scanning times to obtain a second image set of the target feature at the current scanning position;

[0010] Perform a multi-image averaging operation on all the scanned images included in the first image set and all the scanned images included in the second image set to obtain a target image of the target feature at the current scanning position.

[0011] As an optional implementation manner, in the first aspect of the present invention, the determining whether there are abnormal features in the initial image includes:

[0012] Obtain at least one first comparison image;

[0013] Perform a difference comparison operation on the initial image and all the first comparison images to obtain a difference comparison operation result, and determine an abnormal value corresponding to the initial image according to the difference comparison operation result;

[0014] Determine whether the abnormal value is greater than a preset abnormal value threshold;

[0015] When it is determined that the abnormal value is greater than the preset abnormal value threshold, determine that there are abnormal features in the initial image.

[0016] As an optional implementation manner, in the first aspect of the present invention, the supplementary scanning times are preset supplementary scanning times or scanning times calculated according to a preset supplementary scanning formula; wherein, the larger the abnormal value corresponding to the initial image, the more scanning times calculated according to the preset supplementary scanning formula;

[0017] And, the preset supplementary scanning formula is:

[0018] M = round[(Q - G) * A + B];

[0019] Wherein, M is the scanning times calculated according to the preset supplementary scanning formula, Q is the abnormal value corresponding to the initial image, G is the preset abnormal value threshold, A is a preset non-negative and non-zero constant, and B is the minimum scanning times determined in advance.

[0020] As an optional implementation manner, in the first aspect of the present invention, the method further includes:

[0021] When it is determined that there are abnormal features in the initial image, obtain the initially determined initial scanning interval distance;

[0022] Adjust the initial scanning interval distance to obtain a target scanning interval distance;

[0023] Determine at least one position to be scanned after the current scanning position according to the target scanning interval distance and the current scanning position;

[0024] Among them, the scanning interval distance between the position to be scanned closest to the current scanning position and the current scanning position is the target scanning interval distance; when the number of all the positions to be scanned is greater than 1, the scanning interval distance between two adjacent positions to be scanned is the target scanning interval distance.

[0025] As an optional implementation manner, in the first aspect of the present invention, the adjusting the initial scanning interval distance to obtain a target scanning interval distance includes:

[0026] Confirm the result of subtracting the preset scanning interval distance from the initial scanning interval distance as the target scanning interval distance; or,

[0027] Calculate the target scanning interval distance according to the preset scanning interval distance formula and the initial scanning interval distance; where, the larger the abnormal value corresponding to the initial image, the larger the target scanning interval distance calculated according to the preset scanning interval distance formula;

[0028] And, the preset scanning interval distance formula is:

[0029] N2 = N1*(Q - G)*C

[0030] Where, N2 is the target scanning interval distance, N1 is the initial scanning interval distance, Q is the abnormal value corresponding to the initial image, G is the preset abnormal value threshold, and C is a preset non - negative and non - zero constant.

[0031] As an optional implementation manner, in the first aspect of the present invention, the performing a difference comparison operation on the initial image and all the first comparison images to obtain a difference comparison operation result, and determining the abnormal value corresponding to the initial image according to the difference comparison operation result includes:

[0032] Obtain the feature vectors of the initial image and all the first comparison images, calculate the difference value between each first comparison image and the initial image according to the feature vectors of the initial image and all the first comparison images to obtain a plurality of difference values, and determine the abnormal value corresponding to the initial image according to the average value or the maximum value of all the difference values;

[0033] Further, calculating a difference value between each of the first comparison images and the initial image based on the feature vector of the initial image and the feature vectors of all the first comparison images to obtain a plurality of difference values, including:

[0034] Calculating a vector distance or a vector angle between the feature vector of the initial image and the feature vector of each of the first comparison images, and obtaining a difference value between the initial image and each of the first comparison images based on the vector distance or the vector angle between the feature vector of the initial image and the feature vector of each of the first comparison images; wherein, when the vector distance or the vector angle between the feature vector of the initial image and the feature vector of any one of the first comparison images is larger, the difference value between the initial image and the first comparison image is larger;

[0035] Further, determining an outlier corresponding to the initial image according to an average value of all the difference values or a maximum value of all the difference values, including:

[0036] Determining an average value of all the difference values or a maximum value of all the difference values as the outlier corresponding to the initial image; or,

[0037] Performing a normalization operation on an average value of all the difference values or a maximum value of all the difference values to obtain a normalized average value or a normalized maximum value, and determining the normalized average value or the normalized maximum value as the outlier corresponding to the initial image.

[0038] As an optional implementation manner, in the first aspect of the present invention, performing a difference comparison operation on the initial image and all the first comparison images, obtaining a difference comparison operation result, and determining an outlier corresponding to the initial image according to the difference comparison operation result, including:

[0039] Obtaining the feature vector of the initial image and the feature vectors of all the first comparison images;

[0040] Calculating a first average value and a first variance value corresponding to the feature vectors of all the first comparison images according to the feature vectors of all the first comparison images, determining a first distribution vector corresponding to all the first comparison images according to the first average value and the first variance value, calculating a first vector distance between the first distribution vector and the feature vector of the initial image, and determining an outlier corresponding to the initial image according to the first vector distance; wherein, when the first vector distance is larger, the determined outlier corresponding to the initial image is larger; the first vector distance is calculated by the following formula:

[0041] L1 = (f1 - μ1) T σ1-1 (f1 - μ1);

[0042] Wherein, L1 is the first vector distance, f1 is the feature vector of the initial image, μ1 is the first average value, σ1 is the first variance value, and T is the transpose operation symbol of the matrix; or,

[0043] Filter out multiple second comparison images from all the first comparison images according to the feature vector of the initial image and the feature vectors of all the first comparison images, and calculate the second variance value of the feature vectors of all the second comparison images and the feature vector of the initial image; randomly select multiple third comparison images from all the first comparison images, calculate the variance values of the feature vectors of each third comparison image and the feature vectors of all the second comparison images to obtain multiple third difference values; calculate the third average value of all the third difference values, calculate the difference between the second variance value and the third average value, and determine the outlier corresponding to the initial image according to the difference; wherein, the larger the difference, the larger the outlier corresponding to the initial image determined; all the second comparison images are multiple first comparison images with the smallest vector distance from the feature vector of the initial image.

[0044] As an optional implementation manner, in the first aspect of the present invention, before repeating the scan of the target area according to the determined additional scan times, the method further includes:

[0045] When it is determined that there are abnormal features in the initial image, determine the abnormal area in the initial image;

[0046] And, repeating the scan of the target area according to the determined additional scan times includes:

[0047] Repeat the scan of the abnormal area according to the determined additional scan times;

[0048] And, repeating the scan of the abnormal area according to the determined additional scan times includes:

[0049] Adjust the A-Scan scan interval distance corresponding to the abnormal area, and scan the abnormal area according to the determined additional scan times and the adjusted A-Scan scan interval distance.

[0050] As an optional implementation manner, in the first aspect of the present invention, determining the abnormal area in the initial image includes:

[0051] Input the initial image into a pre-determined neural network model to obtain the initial feature map of the initial image output by the neural network model; each pixel position of the initial feature map corresponds to a set of pixel positions of the initial image, each set of pixel positions includes multiple pixel positions of the initial image, all the sets of pixel positions include all the pixel positions of the initial image and the pixel positions included in all the sets of pixel positions are different from each other;

[0052] Obtain at least one fourth comparison image; input all the fourth comparison images into the neural network model to obtain the comparison feature maps of each of the fourth comparison images output by the neural network model; each of the comparison feature maps includes at least one channel, and the number of channels included in each of the comparison feature maps is the same;

[0053] Calculate the average value and variance value of the pixel values corresponding to all the pixel positions of all the comparison feature maps of each channel, determine the fourth average value according to the average values corresponding to all the channels, determine the fourth variance value according to the variance values corresponding to all the channels, and determine the second distribution vector corresponding to all the fourth comparison images according to the fourth average value and the fourth variance value;

[0054] Obtain the feature vectors of all the pixel positions of the initial feature map, calculate the second vector distance between the feature vector of each pixel position of the initial feature map and the second distribution vector to obtain the second vector distances of all the pixel positions of the initial feature map; the second vector distance between the second distribution vector and the feature vector at the target pixel position of the initial feature map is calculated by the following formula:

[0055] L2 = (f2 - μ2) T σ2 -1 (f2 - μ2);

[0056] wherein, L2 is the second vector distance between the feature vector at the target pixel position of the initial feature map and the second distribution vector, f2 is the feature vector at the target pixel position of the initial feature map, μ2 is the fourth average value, σ2 is the fourth variance value, and T is the transpose operation symbol of the matrix;

[0057] Determine the outlier of each pixel position of the initial feature map according to the second vector distance of each pixel position of the initial feature map, and determine the abnormal region in the initial image according to the outlier of each pixel position of the initial feature map; wherein, the larger the second vector distance of a certain pixel position is, the larger the determined outlier of this pixel position is.

[0058] A second aspect of the present invention discloses an adaptive repeated scanning device, the device includes:

[0059] A first scanning module, configured to scan a target area corresponding to a target feature at a current scanning position, so as to obtain a first image set of the target feature at the current scanning position; the target area is an area including the target feature at the current scanning position;

[0060] An image processing module, configured to perform a multi-image averaging operation on all the scanned images included in the first image set, so as to obtain an initial image of the target feature at the current scanning position;

[0061] A judgment module, configured to judge whether there is an abnormal feature in the initial image;

[0062] A second scanning module, configured to, when the judgment module judges that there is an abnormal feature in the initial image, repeatedly scan the target area according to a determined supplementary scanning times, so as to obtain a second image set of the target feature at the current scanning position;

[0063] The image processing module is further configured to perform a multi-image averaging operation on all the scanned images included in the first image set and all the scanned images included in the second image set, so as to obtain a target image of the target feature at the current scanning position.

[0064] As an optional implementation manner, in the second aspect of the present invention, the judgment module includes:

[0065] An acquisition sub-module, configured to acquire at least one first comparison image;

[0066] A comparison sub-module, configured to perform a difference comparison operation on the initial image and all the first comparison images, so as to obtain a difference comparison operation result, and determine an abnormal value corresponding to the initial image according to the difference comparison operation result;

[0067] A judgment sub-module, configured to judge whether the abnormal value is greater than a preset abnormal value threshold, and when it is judged that the abnormal value is greater than the preset abnormal value threshold, determine that there is an abnormal feature in the initial image.

[0068] As an optional implementation manner, in the second aspect of the present invention, the supplementary scanning times is a preset supplementary scanning times or the scanning times calculated according to a preset supplementary scanning formula; wherein, the greater the abnormal value corresponding to the initial image is, the more the scanning times calculated according to the preset supplementary scanning formula is;

[0069] And, the preset supplementary scanning formula is:

[0070] M = round[(Q - G)*A + B];

[0071] Wherein, M is the number of scans calculated according to the preset supplementary scan formula, Q is the outlier corresponding to the initial image, G is the preset outlier threshold, A is a preset non - negative and non - zero constant, and B is the determined minimum number of scans.

[0072] As an alternative implementation, in the second aspect of the present invention, the device further includes:

[0073] An acquisition module, configured to acquire a preset initial scan interval distance when the determination module determines that there are abnormal features in the initial image;

[0074] An adjustment module, configured to adjust the initial scan interval distance to obtain a target scan interval distance;

[0075] A first determination module, configured to determine at least one to - be - scanned position after the current scan position according to the target scan interval distance and the current scan position;

[0076] Wherein, the scan interval distance between the to - be - scanned position closest to the current scan position and the current scan position is the target scan interval distance; when the number of all to - be - scanned positions is greater than 1, the scan interval distance between two adjacent to - be - scanned positions is the target scan interval distance.

[0077] As an alternative implementation, in the second aspect of the present invention, the manner in which the adjustment module adjusts the initial scan interval distance to obtain a target scan interval distance specifically includes:

[0078] Confirming the result of subtracting the preset scan interval distance from the initial scan interval distance as the target scan interval distance; or,

[0079] Calculating the target scan interval distance according to the preset scan interval distance formula and the initial scan interval distance; wherein, the larger the outlier corresponding to the initial image is, the larger the target scan interval distance calculated according to the preset scan interval distance formula is;

[0080] And, the preset scan interval distance formula is:

[0081] N2 = N1*(Q - G)*C

[0082] Wherein, N2 is the target scan interval distance, N1 is the initial scan interval distance, Q is the outlier corresponding to the initial image, G is the preset outlier threshold, and C is a preset non - negative and non - zero constant.

[0083] As an alternative implementation, in the second aspect of the present invention, the comparison sub - module includes:

[0084] An acquisition unit for acquiring the feature vectors of the initial image and the feature vectors of all the first comparison images;

[0085] A calculation unit for calculating the difference values between each of the first comparison images and the initial image according to the feature vectors of the initial image and the feature vectors of all the first comparison images, to obtain a plurality of difference values;

[0086] A determination unit for determining the outlier corresponding to the initial image according to the average value of all the difference values or the maximum value of all the difference values;

[0087] Moreover, the manner in which the calculation unit calculates the difference values between each of the first comparison images and the initial image according to the feature vectors of the initial image and the feature vectors of all the first comparison images, to obtain a plurality of difference values specifically includes:

[0088] Calculating the vector distance or vector angle between the feature vector of the initial image and the feature vector of each of the first comparison images, and obtaining the difference value between the initial image and each of the first comparison images according to the vector distance or vector angle between the feature vector of the initial image and the feature vector of each of the first comparison images; wherein, when the vector distance or vector angle between the feature vector of the initial image and the feature vector of any one of the first comparison images is larger, the difference value between the initial image and the first comparison image is larger;

[0089] Moreover, the manner in which the determination unit determines the outlier corresponding to the initial image according to the average value of all the difference values or the maximum value of all the difference values specifically includes:

[0090] Determining the average value of all the difference values or the maximum value of all the difference values as the outlier corresponding to the initial image; or,

[0091] Performing a normalization operation on the average value of all the difference values or the maximum value of all the difference values to obtain a normalized average value or a normalized maximum value, and determining the normalized average value or the normalized maximum value as the outlier corresponding to the initial image.

[0092] As an optional implementation manner, in the second aspect of the present invention, the comparison sub-module includes an acquisition unit and a comparison processing unit, wherein:

[0093] The acquisition unit for acquiring the feature vectors of the initial image and the feature vectors of all the first comparison images;

[0094] The comparison processing unit is configured to calculate a first average value and a first variance value corresponding to the feature vectors of all the first comparison images according to the feature vectors of all the first comparison images, determine a first distribution vector corresponding to all the first comparison images according to the first average value and the first variance value, calculate a first vector distance between the first distribution vector and the feature vector of the initial image, and determine an outlier corresponding to the initial image according to the first vector distance; wherein, the larger the first vector distance is, the larger the outlier corresponding to the initial image is determined; the first vector distance is calculated by the following formula:

[0095] L1 = (f1 - μ1) T σ1 -1 (f1 - μ1);

[0096] wherein, L1 is the first vector distance, f1 is the feature vector of the initial image, μ1 is the first average value, σ1 is the first variance value, and T is the transpose operation symbol of the matrix; or,

[0097] Filter out a plurality of second comparison images from all the first comparison images according to the feature vector of the initial image and the feature vectors of all the first comparison images, and calculate a second variance value between the feature vectors of all the second comparison images and the feature vector of the initial image; randomly select a plurality of third comparison images from all the first comparison images, and calculate the variance values between the feature vectors of each third comparison image and the feature vectors of all the second comparison images to obtain a plurality of third variance differences; calculate a third average value of all the third variance differences, calculate the difference between the second variance value and the third average value, and determine the outlier corresponding to the initial image according to the difference; wherein, the larger the difference is, the larger the outlier corresponding to the initial image is determined; all the second comparison images are a plurality of first comparison images with the smallest vector distance from the feature vector of the initial image.

[0098] As an optional implementation manner, in the second aspect of the present invention, the device further includes:

[0099] A second determination module, configured to determine an abnormal area in the initial image when the judgment module determines that there are abnormal features in the initial image before the second scanning module repeatedly scans the target area according to the determined supplementary scanning times;

[0100] And, the manner in which the second scanning module repeatedly scans the target area according to the determined supplementary scanning times to obtain a second image set of the target feature at the current scanning position specifically includes:

[0101] Repeat scanning the abnormal area according to the determined number of supplementary scans to obtain a second image set of the target feature at the current scanning position;

[0102] Moreover, the specific manner in which the second scanning module repeats scanning the abnormal area according to the determined number of supplementary scans to obtain a second image set of the target feature at the current scanning position includes:

[0103] Adjust the A-Scan scanning interval distance corresponding to the abnormal area, and scan the abnormal area according to the determined number of supplementary scans and the adjusted A-Scan scanning interval distance to obtain a second image set of the target feature at the current scanning position.

[0104] As an optional implementation manner, in the second aspect of the present invention, the specific manner in which the second determination module determines the abnormal area in the initial image includes:

[0105] Input the initial image into a pre-determined neural network model to obtain an initial feature map of the initial image output by the neural network model; each pixel position of the initial feature map corresponds to a set of pixel positions of the initial image, each set of pixel positions includes multiple pixel positions of the initial image, all the sets of pixel positions include all the pixel positions of the initial image and all the pixel positions included in all the sets of pixel positions are different;

[0106] Obtain at least one fourth comparison image; input all the fourth comparison images into the neural network model to obtain a comparison feature map of each of the fourth comparison images output by the neural network model; each comparison feature map includes at least one channel, and the number of channels included in each comparison feature map is the same;

[0107] Calculate the average value and variance value of the pixel values corresponding to all pixel positions of all the comparison feature maps of each channel, determine a fourth average value according to the average values corresponding to all the channels, determine a fourth variance value according to the variance values corresponding to all the channels, and determine a second distribution vector corresponding to all the fourth comparison images according to the fourth average value and the fourth variance value;

[0108] Obtain the feature vectors of all pixel positions of the initial feature map, calculate the second vector distance between the feature vector of each pixel position of the initial feature map and the second distribution vector to obtain the second vector distances of all pixel positions of the initial feature map; the second vector distance between the second distribution vector and the feature vector at the target pixel position of the initial feature map is calculated by the following formula:

[0109] L2 = (f2 - μ2) T σ2-1 (f2 - μ2);

[0110] Wherein, L2 is the second vector distance between the feature vector of the initial feature map at the target pixel position and the second distribution vector, f2 is the feature vector of the initial feature map at the target pixel position, μ2 is the fourth average value, σ2 is the fourth variance value, and T is the transpose operation symbol of the matrix;

[0111] Determine the outliers at each pixel position of the initial feature map according to the second vector distance at each pixel position of the initial feature map, and determine the abnormal region in the initial image according to the outliers at each pixel position of the initial feature map; wherein, the larger the second vector distance at a certain pixel position, the larger the determined outlier at that pixel position.

[0112] The third aspect of the present invention discloses another adaptive repeated scanning device, and the device includes:

[0113] A memory storing executable program code;

[0114] A processor coupled to the memory;

[0115] The processor calls the executable program code stored in the memory and executes the adaptive repeated scanning method disclosed in the first aspect of the present invention.

[0116] The fourth aspect of the present invention discloses a computer - storable medium, and the computer - storage medium stores computer instructions, which are used to execute the adaptive repeated scanning method disclosed in the first aspect of the present invention when being called.

[0117] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0118] In an embodiment of the present invention, at a current scanning position, a target area corresponding to a target feature is scanned to obtain a first image set of the target feature at the current scanning position; the target area is an area including the target feature at the current scanning position; a multi-image averaging operation is performed on all the scanned images included in the first image set to obtain an initial image of the target feature at the current scanning position; it is determined whether there are abnormal features in the initial image; when it is determined that there are abnormal features in the initial image, the target area is scanned repeatedly according to the determined number of supplementary scans to obtain a second image set of the target feature at the current scanning position; a multi-image averaging operation is performed on all the scanned images included in the first image set and all the scanned images included in the second image set to obtain a target image of the target feature at the current scanning position. It can be seen that implementing the present invention can scan the target area corresponding to the target feature to obtain multiple scanned images, and perform a multi-image averaging operation on the multiple scanned images to obtain an initial image, which helps to generally understand the situation of the target feature based on the initial image. It can also determine whether there are abnormal features in the initial image, and when and only when there are abnormal features in the initial image, more scanned images can be obtained by increasing the number of scans, so as to obtain a target image with higher imaging quality based on more scanned images while ensuring the scanning efficiency, which helps to have a more accurate and detailed understanding of the situation of the target feature based on the target image with higher imaging quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0119] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings can be obtained based on these drawings without creative efforts.

[0120] Figure 1 is a flowchart of an adaptive repeated scanning method disclosed in an embodiment of the present invention;

[0121] Figure 2 is a flowchart of another adaptive repeated scanning method disclosed in an embodiment of the present invention;

[0122] Figure 3 is a structural diagram of an adaptive repeated scanning device disclosed in an embodiment of the present invention;

[0123] Figure 4 is a structural diagram of another adaptive repeated scanning device disclosed in an embodiment of the present invention;

[0124] Figure 5 is a structural diagram of a comparison sub-module disclosed in an embodiment of the present invention;

[0125] Figure 6 It is a schematic structural diagram of another comparison sub-module disclosed in an embodiment of the present invention;

[0126] Figure 7 It is a schematic structural diagram of another adaptive repetitive scanning device disclosed in an embodiment of the present invention. Detailed implementation manners

[0127] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0128] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or terminal including a series of steps or units is not limited to the listed steps or units, but optionally further includes unlisted steps or units, or optionally further includes other steps or units inherent to these processes, methods, products or terminals.

[0129] Referring to "embodiment" herein means that a specific feature, structure or characteristic described in connection with the embodiment may be included in at least one embodiment of the present invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.

[0130] The present invention discloses an adaptive repetitive scanning method and device, which can scan a target area corresponding to a target feature to obtain a plurality of scanned images, and perform a multi-image averaging operation on the plurality of scanned images to obtain an initial image, thereby helping to generally understand the situation of the target feature based on the initial image, and can also determine whether there are abnormal features in the initial image. Moreover, when and only when there are abnormal features in the initial image, more scanned images can be obtained by increasing the number of scans, so as to obtain a target image with higher imaging quality based on more scanned images while ensuring the scanning efficiency, which helps to have a more accurate and detailed understanding of the situation of the target feature based on the target image with higher imaging quality. The following will be described in detail respectively.

[0131] Embodiment 1

[0132] Please refer to Figure 1 , Figure 1 which is a schematic flowchart of an adaptive repetitive scanning method disclosed in an embodiment of the present invention. Among them, Figure 1 the described adaptive repetitive scanning method can be applied to a scanning control system or a scanning control terminal, and the embodiments of the present invention do not make limitations. As Figure 1 shown, the adaptive repetitive scanning method may include the following operations:

[0133] 101. Scan the target area corresponding to the target feature at the current scanning position to obtain a first image set of the target feature at the current scanning position.

[0134] In the embodiments of the present invention, the target area is an area including the target feature at the current scanning position. Further optionally, the area including the target feature may be a rectangular area including the target feature, such as a 3mm * 3mm rectangular area. Optionally, the scanning method may be Optical Coherence Tomography (OCT). Optionally, the target area is repeatedly sampled N times to obtain N B-Scans, and the N B-Scans form the first image set. Specifically, N may be any integer greater than 1. Optionally, the target feature may be a retinal feature.

[0135] 102. Perform a multi-image averaging operation on all the scanned images included in the first image set to obtain an initial image of the target feature at the current scanning position.

[0136] In the embodiments of the present invention, the scanned image is a B-Scan image, and the initial image is obtained by performing a multi-image averaging operation on all the B-Scan images obtained by scanning the target area corresponding to the target feature at the current scanning position.

[0137] In the embodiments of the present invention, optionally, the multi-image averaging operation is to obtain the pixel values of all the scanned images at the same pixel position, sum and average the pixel values corresponding to the same pixel position to obtain new pixel values for all the pixel positions, and thus generate an initial image according to all the pixel positions and the new pixel values corresponding to all the pixel positions. It can be seen that implementing this optional embodiment can obtain an initial image by performing a multi-image averaging operation on all the scanned images included in the first image set, which helps to generally understand the relevant situation of the target feature through the initial image.

[0138] 103. Determine whether there are abnormal features in the initial image.

[0139] In the embodiments of the present invention, the abnormal features may include lesion features.

[0140] 104. When it is determined that there are abnormal features in the initial image, the target area is scanned repeatedly according to the determined number of supplementary scans, and a second image set of the target features at the current scanning position is obtained.

[0141] Optionally, the number of supplementary scans is a preset number of supplementary scans or the number of scans calculated according to a preset supplementary scan formula; among them, the larger the abnormal value corresponding to the initial image, the more the number of scans calculated according to the preset supplementary scan formula. It can be seen that implementing this optional embodiment can determine the number of supplementary scans through the preset number of supplementary scans or the preset supplementary scan formula, which helps to repeatedly scan the target area according to the determined number of supplementary scans subsequently.

[0142] In this optional embodiment, specifically, the preset supplementary scan formula can be:

[0143] M = round[(Q - G) * A + B];

[0144] Wherein, M is the number of scans calculated according to the preset supplementary scan formula, Q is the abnormal value corresponding to the initial image, G is the preset abnormal value threshold, A is a preset non - negative and non - zero constant, and B is the minimum number of scans determined in advance.

[0145] In this optional embodiment, the larger the abnormal value corresponding to the initial image, the larger the preset non - negative and non - zero constant A. Optionally, the preset supplementary scan formula can also be a quadratic function formula. It can be seen that implementing this optional embodiment can determine the number of supplementary scans through the preset supplementary scan formula, which helps to improve the accuracy and adaptability of the number of supplementary scans, and also helps to repeatedly scan the target area according to the calculated number of supplementary scans subsequently.

[0146] 105. Perform a multi - image averaging operation on all the scanned images included in the first image set and all the scanned images included in the second image set to obtain a target image of the target feature at the current scanning position.

[0147] It can be seen that implementing the embodiments of the present invention can scan the target area corresponding to the target feature to obtain multiple scanned images, perform a multi - image averaging operation on the multiple scanned images to obtain an initial image, which helps to generally understand the situation of the target feature based on the initial image, and can also determine whether there are abnormal features in the initial image. When and only when there are abnormal features in the initial image, more scanned images can be obtained by increasing the number of scans, so as to obtain a target image with higher imaging quality based on more scanned images while ensuring the scanning efficiency, which helps to have a more accurate and detailed understanding of the situation of the target feature based on the target image with higher imaging quality.

[0148] Embodiment 2

[0149] Please refer to Figure 2 ,Figure 2 It is a schematic flowchart of an adaptive repeated scanning method disclosed in an embodiment of the present invention. Among them, Figure 2 The described adaptive repeated scanning method can be applied to a scanning control system or a scanning control terminal, and the embodiments of the present invention do not make limitations. As Figure 2 shown, the adaptive repeated scanning method may include the following operations:

[0150] 201. Scan the target area corresponding to the target feature at the current scanning position to obtain a first image set of the target feature at the current scanning position.

[0151] 202. Perform a multi-image averaging operation on all the scanned images included in the first image set to obtain an initial image of the target feature at the current scanning position.

[0152] In the embodiments of the present invention, for other descriptions of steps 201 - 202, please refer to the detailed descriptions of steps 101 - 102 in Embodiment 1, and the embodiments of the present invention will not be elaborated herein.

[0153] 203. Obtain at least one first comparison image.

[0154] In the embodiments of the present invention, the first comparison image may be an image of the target feature being a normal feature; for example, when the target feature is a retinal feature, the first comparison image may be a normal retinal picture without lesions.

[0155] 204. Perform a difference comparison operation on the initial image and all the first comparison images to obtain a difference comparison operation result, and determine an outlier corresponding to the initial image according to the difference comparison operation result.

[0156] As an optional implementation manner, performing a difference comparison operation on the initial image and all the first comparison images to obtain a difference comparison operation result, and determining an outlier corresponding to the initial image according to the difference comparison operation result includes:

[0157] Obtain the feature vector of the initial image and the feature vectors of all the first comparison images, calculate the difference value between each first comparison image and the initial image according to the feature vector of the initial image and the feature vectors of all the first comparison images to obtain a plurality of difference values, and determine the outlier corresponding to the initial image according to the average value or the maximum value of all the difference values.

[0158] In this optional embodiment, the initial image is input into the first neural network model, and a feature vector with a certain length output by the first neural network model is the feature vector of the initial image. Optionally, the first neural network model is composed of multiple convolutional layers and uses a global pooling layer and a fully connected layer.

[0159] It can be seen that implementing this optional embodiment can determine the outliers corresponding to the initial image based on the feature vectors of the initial image and the feature vectors of all the first comparison images, which helps to subsequently determine whether there are abnormal features in the initial image according to the determined outliers corresponding to the initial image.

[0160] As yet another optional implementation manner, calculate the difference value between each first comparison image and the initial image according to the feature vector of the initial image and the feature vectors of all the first comparison images, obtaining a plurality of difference values, including:

[0161] Calculate the vector distance or vector angle between the feature vector of the initial image and the feature vector of each first comparison image, and obtain the difference value between the initial image and each first comparison image according to the vector distance or vector angle between the feature vector of the initial image and the feature vector of each first comparison image; wherein, when the vector distance or vector angle between the feature vector of the initial image and the feature vector of any first comparison image is larger, the difference value between the initial image and the first comparison image is larger.

[0162] In this optional embodiment, the vector distance can be at least one of L1 distance, L2 distance (Euclidean distance), cosine distance, and Hamming distance. Optionally, the calculated outlier can be the original outlier or the normalized outlier, which is not limited in the embodiments of the present invention.

[0163] It can be seen that implementing this optional embodiment can obtain the difference value between the initial image and each first comparison image through the vector distance or vector angle between the feature vector of the initial image and the feature vector of each first comparison image, which helps to subsequently determine the outliers corresponding to the initial image according to the difference value between the initial image and each first comparison image.

[0164] As yet another optional implementation manner, determine the outliers corresponding to the initial image according to the average value of all the difference values or the maximum value of all the difference values, including:

[0165] Determine the average value of all the difference values or the maximum value of all the difference values as the outliers corresponding to the initial image. It can be seen that implementing this optional embodiment can determine the average value of all the difference values or the maximum value of all the difference values as the outliers corresponding to the initial image, which helps to subsequently determine whether there are abnormal features in the initial image according to the determined outliers corresponding to the initial image.

[0166] As yet another optional implementation manner, determine the outliers corresponding to the initial image according to the average value of all the difference values or the maximum value of all the difference values, including:

[0167] Perform a normalization operation on the average value of all difference values or the maximum value of all difference values to obtain a normalized average value or a normalized maximum value, and determine the outlier corresponding to the initial image as the normalized average value or the normalized maximum value.

[0168] In this optional embodiment, the normalization operation may be: input the target value of the normalization operation into the sigmoid function (also called the Logistic function) to obtain the output value of the sigmoid function; this output value is the normalized value obtained after performing the normalization operation on the target value, and the range of the normalized value must be between 0 and 1 excluding 0 and 1. For example, when the outlier before normalization is 1, the outlier after normalization is 0.7311; when the outlier before normalization is 2, the outlier after normalization is 0.8808. Since normalization performs the same data processing on different values to obtain results within the same range, compared with directly comparing outliers, comparing the normalized outliers can more clearly show the relative relationship between outliers.

[0169] It can be seen that implementing this optional embodiment can obtain the outlier corresponding to the initial image by performing a normalization operation on the average value of all difference values or the maximum value of all difference values, which helps to more clearly show the relative relationship between different outliers through normalization, thus facilitating the comparison between the outlier and the preset outlier threshold, and also helps to subsequently determine whether there are abnormal features in the initial image based on the determined outlier corresponding to the initial image.

[0170] As another optional implementation, perform a difference comparison operation on the initial image and all first comparison images to obtain the result of the difference comparison operation, and determine the outlier corresponding to the initial image according to the result of the difference comparison operation, including:

[0171] Obtain the feature vector of the initial image and the feature vectors of all first comparison images;

[0172] Calculate the first average value and the first variance value corresponding to the feature vectors of all first comparison images according to the feature vectors of all first comparison images, determine the first distribution vector corresponding to all first comparison images according to the first average value and the first variance value, calculate the first vector distance between the first distribution vector and the feature vector of the initial image, and determine the outlier corresponding to the initial image according to the first vector distance; where, the larger the first vector distance, the larger the outlier corresponding to the initial image determined; the first vector distance is calculated by the following formula:

[0173] L1=(f1 - μ1) T σ1 -1 (f1 - μ1);

[0174] Among them, L1 is the first vector distance, f1 is the feature vector of the initial image, μ1 is the first average value, σ1 is the first variance value, and T is the transpose operation symbol of the matrix.

[0175] Optionally, the calculated outlier can be the original outlier or the normalized outlier, and the embodiments of the present invention do not make any limitations.

[0176] It can be seen that this optional embodiment provides a method for determining the outlier corresponding to the initial image, which helps to subsequently determine whether there are abnormal features in the initial image according to the determined outlier corresponding to the initial image.

[0177] As another optional implementation manner, performing a difference comparison operation on the initial image and all the first comparison images to obtain the result of the difference comparison operation, and determining the outlier corresponding to the initial image according to the result of the difference comparison operation, including:

[0178] Obtaining the feature vector of the initial image and the feature vectors of all the first comparison images;

[0179] Screening out a plurality of second comparison images from all the first comparison images according to the feature vector of the initial image and the feature vectors of all the first comparison images, and calculating the second variance value of the feature vectors of all the second comparison images and the feature vector of the initial image; randomly selecting a plurality of third comparison images from all the first comparison images, and calculating the variance values of the feature vectors of each third comparison image and the feature vectors of all the second comparison images to obtain a plurality of third difference values; calculating the third average value of all the third difference values, calculating the difference between the second variance value and the third average value, and determining the outlier corresponding to the initial image according to the difference; wherein, the larger the difference is, the larger the outlier corresponding to the initial image is determined; all the second comparison images are a plurality of first comparison images with the smallest vector distance from the feature vector of the initial image.

[0180] It can be seen that this optional embodiment provides another method for determining the outlier corresponding to the initial image, which helps to improve the accuracy and adaptability of the determined outlier corresponding to the initial image, and helps to subsequently determine whether there are abnormal features in the initial image according to the determined outlier corresponding to the initial image.

[0181] 205. Judging whether the outlier is greater than a preset outlier threshold.

[0182] 206. When it is judged that the outlier is greater than the preset outlier threshold, determining that there are abnormal features in the initial image.

[0183] 207. When it is judged that there are abnormal features in the initial image, repeatedly scanning the target area according to the determined additional scanning times to obtain a second image set of the target feature at the current scanning position.

[0184] 208. Perform a multi-image averaging operation on all the scanned images included in the first image set and all the scanned images included in the second image set to obtain a target image of the target feature at the current scanning position.

[0185] In the embodiments of the present invention, for other descriptions of steps 207 - 208, please refer to the detailed descriptions of steps 104 - 105 in Embodiment 1, and the embodiments of the present invention will not elaborate further.

[0186] It can be seen that implementing the embodiments of the present invention can scan the target area corresponding to the target feature to obtain multiple scanned images, and perform a multi-image averaging operation on the multiple scanned images to obtain an initial image, thereby helping to generally understand the situation of the target feature based on the initial image; it can also perform a difference comparison operation on the initial image and all the first comparison images to obtain the result of the difference comparison operation, thereby determining the difference value corresponding to the initial image according to the result of the difference comparison operation, and judging whether there are abnormal features in the initial image according to the difference value corresponding to the initial image; it can also, when and only when there are abnormal features in the initial image, obtain more scanned images by increasing the number of scans, so as to obtain a target image with higher imaging quality based on more scanned images under the condition of ensuring the scanning efficiency, which helps to have a more accurate and detailed understanding of the situation of the target feature based on the target image with higher imaging quality.

[0187] In yet another alternative embodiment, the method further includes:

[0188] When it is determined that there are abnormal features in the initial image, obtain a pre-determined initial scanning interval distance;

[0189] Adjust the initial scanning interval distance to obtain a target scanning interval distance;

[0190] According to the target scanning interval distance and the current scanning position, determine at least one position to be scanned after the current scanning position;

[0191] Wherein, the scanning interval distance between the position to be scanned closest to the current scanning position and the current scanning position is the target scanning interval distance; when the number of all positions to be scanned is greater than 1, the scanning interval distance between two adjacent positions to be scanned is the target scanning interval distance.

[0192] It can be seen that implementing this alternative embodiment can adjust the scanning interval distance when it is determined that there are abnormal features in the initial image, improve the flexibility and adaptability of scanning, and can also adaptively obtain target feature information to provide suitable images for experimental research or teaching activities.

[0193] In yet another alternative embodiment, adjusting the initial scanning interval distance to obtain a target scanning interval distance includes:

[0194] The result of subtracting the preset scanning interval distance from the initial scanning interval distance is confirmed as the target scanning interval distance.

[0195] In this alternative embodiment, more scanned images of adjacent positions can be obtained by reducing the scanning interval distance. The target features corresponding to the scanned images of adjacent positions are generally relatively close. Therefore, obtaining more scanned images of adjacent positions can improve the accuracy of studying or learning the target features.

[0196] It can be seen that implementing this alternative embodiment can reduce the initial scanning interval distance according to the preset scanning interval distance, so that more images with abnormal features can be obtained in subsequent scans, thereby improving the accuracy of studying or learning the target features corresponding to the images with abnormal features.

[0197] In yet another alternative embodiment, adjusting the initial scanning interval distance to obtain the target scanning interval distance includes:

[0198] Calculating the target scanning interval distance according to the preset scanning interval distance formula and the initial scanning interval distance; wherein, the larger the abnormal value corresponding to the initial image, the larger the target scanning interval distance calculated according to the preset scanning interval distance formula.

[0199] It can be seen that implementing this alternative embodiment can adaptively reduce the initial scanning interval distance according to the abnormal value of the initial image through the preset scanning interval distance formula, improve the adaptability and flexibility of adjusting the scanning interval distance, and at the same time, corresponding numbers of images with abnormal features can be obtained according to different degrees of abnormal values in subsequent scans, thereby improving the adaptability and flexibility of the number of obtained images, contributing to the subsequent study and learning of the target features corresponding to the images with abnormal features, and improving the accuracy of the study and learning.

[0200] In this alternative embodiment, specifically, the preset scanning interval distance formula can be:

[0201] N2 = N1 * (Q - G) * C

[0202] Wherein, N2 is the target scanning interval distance, N1 is the initial scanning interval distance, Q is the abnormal value corresponding to the initial image, G is the preset abnormal value threshold, and C is a preset non - negative and non - zero constant.

[0203] In this alternative embodiment, the larger the abnormal value corresponding to the initial image, the larger the preset non - negative and non - zero constant C.

[0204] It can be seen that implementing this alternative embodiment can determine the adjusted scanning interval distance through the preset scanning interval distance formula, which helps to improve the accuracy and adaptability of the adjusted scanning interval distance.

[0205] In yet another alternative embodiment, before repeating the scanning of the target area according to the determined number of supplementary scans, the method further includes:

[0206] When it is determined that there are abnormal features in the initial image, determine the abnormal area in the initial image.

[0207] It can be seen that implementing this alternative embodiment can determine the abnormal area in the initial image when it is determined that there are abnormal features in the initial image, which helps to adjust the scanning area according to the abnormal area in the initial image subsequently, and also helps to study the relevant information of the target features according to the abnormal area in the initial image subsequently.

[0208] In yet another alternative embodiment, repeating the scanning of the target area according to the determined number of supplementary scans includes:

[0209] Repeat scanning the abnormal area according to the determined number of supplementary scans.

[0210] In this alternative embodiment, each scanned image (i.e., B-Scan) is composed of multiple A-Scans. Suppose each B-Scan consists of 500 A-Scans, and through anomaly detection, the abnormal area in the B-Scan can be identified, for example, the area between the 305th and 351st. Then, in the next scan, we can perform targeted repeated sampling on this abnormal area, while other non-abnormal areas do not need to be scanned again.

[0211] It can be seen that implementing this alternative embodiment can reduce the target area that needs to be repeatedly scanned to the abnormal area, so as to obtain more scanned images to improve the imaging quality of the initial image while ensuring the scanning efficiency.

[0212] In yet another alternative embodiment, repeating the scanning of the abnormal area according to the determined number of supplementary scans includes:

[0213] Adjust the scanning interval distance of the A-Scans corresponding to the abnormal area, and scan the abnormal area according to the determined number of supplementary scans and the adjusted scanning interval distance of the A-Scans.

[0214] It can be seen that implementing this alternative embodiment can adjust the scanning interval distance of the A-Scans corresponding to the abnormal area, improve the flexibility and adaptability of the scanning, and help to obtain more required A-Scans to support the subsequent research and learning of the A-Scans, and improve the accuracy of the research and learning of the A-Scans.

[0215] In this alternative embodiment, the scanning interval distance of a conventional A-Scan is 5 μm. The scanning of the abnormal area can have a denser scanning interval, which can be a constant or a function of the abnormal values corresponding to the initial image. When the scanning interval is a function of the abnormal values, the larger the abnormal value, the smaller the scanning distance.

[0216] It can be seen that implementing this alternative embodiment can adjust the A-Scan scanning interval distance corresponding to the abnormal area, improve the flexibility and adaptability of scanning, and help obtain more A-Scans in the abnormal area in subsequent scans to support subsequent research and study of the A-Scans in the abnormal area, thereby improving the accuracy of researching and studying A-Scans.

[0217] In yet another alternative embodiment, determining the abnormal area in the initial image includes:

[0218] Inputting the initial image into a pre-determined neural network model to obtain the initial feature map of the initial image output by the neural network model. Each pixel position of the initial feature map corresponds to a set of pixel positions of the initial image. Each set of pixel positions includes multiple pixel positions of the initial image. All sets of pixel positions include all pixel positions of the initial image and the pixel positions included in all sets of pixel positions are all different.

[0219] Obtaining at least one fourth comparison image; inputting all fourth comparison images into the neural network model to obtain the comparison feature maps of each fourth comparison image output by the neural network model. Each comparison feature map includes at least one channel, and the number of channels included in each comparison feature map is the same.

[0220] Calculating the average value and variance value of the pixel values corresponding to all pixel positions of all comparison feature maps of each channel, determining the fourth average value according to the average values corresponding to all channels, determining the fourth variance value according to the variance values corresponding to all channels, and determining the second distribution vector corresponding to all fourth comparison images according to the fourth average value and the fourth variance value.

[0221] Obtaining the feature vectors of all pixel positions of the initial feature map, calculating the second vector distance between the feature vector of each pixel position of the initial feature map and the second distribution vector to obtain the second vector distances of all pixel positions of the initial feature map. The second vector distance between the second distribution vector and the feature vector at the target pixel position of the initial feature map is calculated by the following formula:

[0222] L2 = (f2 - μ2) T σ2 -1 (f2 - μ2);

[0223] Wherein, L2 is the second vector distance between the feature vector of the initial feature map at the target pixel position and the second distribution vector, f2 is the feature vector of the initial feature map at the target pixel position, μ2 is the fourth average value, σ2 is the fourth variance value, and T is the transpose operation symbol of the matrix;

[0224] Determine the outliers at each pixel position of the initial feature map according to the second vector distance at each pixel position of the initial feature map, and determine the abnormal region in the initial image according to the outliers at each pixel position of the initial feature map; wherein, the larger the second vector distance at a certain pixel position, the larger the determined outlier at that pixel position.

[0225] In this optional embodiment, the pre-determined neural network model is the second neural network model. Optionally, the second neural network model is composed of multiple convolutional layers. The main difference from the first neural network model is that it does not use a global pooling layer and a fully connected layer. Therefore, the output of the second neural network model is a feature map of i*j*k, where i, j, and k are the length, width, and number of channels of the feature map respectively. Optionally, both i and j are smaller than the length and width of the original image. For example, i can be one-fourth of the length dimension of the original image, and j can be one-fourth of the width dimension of the original image. In this optional embodiment, f2, μ2, and σ2 are all vectors of size 1*k. Optionally, the calculated outlier can be the original outlier or the normalized outlier, which is not limited in the embodiments of the present invention. Optionally, all the fourth comparison images can be all the same as, partially the same as, or all different from all the first comparison images, which is not limited in the embodiments of the present invention.

[0226] It can be seen that this optional embodiment provides a method for determining the abnormal region in the initial image, which helps to understand the distribution of the abnormal region in the initial image and helps to perform repeated scanning operations on the abnormal region determined according to this method subsequently, thereby improving the efficiency and accuracy of scanning.

[0227] Embodiment III

[0228] Please refer to Figure 3 , Figure 3 which is a schematic structural diagram of an adaptive repeated scanning device disclosed in the embodiments of the present invention. Wherein, Figure 3 The described adaptive repeated scanning device can be applied to a scanning control system or a scanning control terminal, which is not limited in the embodiments of the present invention. As Figure 3 shown, the adaptive repeated scanning device may include:

[0229] The first scanning module 301 is configured to scan the target region corresponding to the target feature at the current scanning position to obtain a first image set of the target feature at the current scanning position; the target region is the region including the target feature at the current scanning position;

[0230] An image processing module 302 is configured to perform a multi-image averaging operation on all the scanned images included in the first image set to obtain an initial image of the target feature at the current scanning position;

[0231] A judgment module 303 is configured to judge whether there are abnormal features in the initial image;

[0232] A second scanning module 304 is configured to, when the judgment module 303 determines that there are abnormal features in the initial image, repeatedly scan the target area according to the determined number of supplementary scans to obtain a second image set of the target feature at the current scanning position;

[0233] The image processing module 302 is further configured to perform a multi-image averaging operation on all the scanned images included in the first image set and all the scanned images included in the second image set to obtain a target image of the target feature at the current scanning position.

[0234] In an embodiment of the present invention, the target area is an area including the target feature at the current scanning position. Further optionally, the area including the target feature may be a rectangular area including the target feature, such as a rectangular area of 3mm * 3mm. Optionally, the scanning method may be optical coherence tomography (OCT). Optionally, the target area is repeatedly sampled N times to obtain N B-Scans, and the N B-Scans form the first image set. Specifically, N may be any integer greater than 1. Optionally, the target feature may be a retinal feature. In an embodiment of the present invention, the scanned image is a B-Scan image, and the initial image is obtained by performing a multi-image averaging operation on all the B-Scan images obtained by scanning the target area corresponding to the target feature at the current scanning position.

[0235] It can be seen that implementing the embodiment of the present invention can scan the target area corresponding to the target feature to obtain multiple scanned images, perform a multi-image averaging operation on the multiple scanned images to obtain an initial image, thereby helping to generally understand the situation of the target feature based on the initial image, and can also judge whether the initial image meets the imaging quality conditions. When and only when the initial image does not meet the imaging quality conditions, more scanned images can be obtained by increasing the number of scans, so as to obtain a target image with higher imaging quality based on more scanned images while ensuring the scanning efficiency, which helps to have a more accurate and detailed understanding of the situation of the target feature based on the target image with higher imaging quality.

[0236] Optionally, the additional scan times are the preset additional scan times or the scan times calculated according to a preset additional scan formula; wherein, the larger the outlier corresponding to the initial image is, the more scan times are calculated according to the preset additional scan formula. It can be seen that implementing this optional embodiment can determine the additional scan times through the preset additional scan times or the preset additional scan formula, which helps to repeatedly scan the target area according to the determined additional scan times subsequently.

[0237] In this optional embodiment, specifically, the preset additional scan formula can be:

[0238] M = round[(Q - G) * A + B];

[0239] wherein, M is the scan times calculated according to the preset additional scan formula, Q is the outlier corresponding to the initial image, G is the preset outlier threshold, A is a preset non - negative and non - zero constant, and B is the minimum scan times determined in advance.

[0240] In this optional embodiment, the larger the outlier corresponding to the initial image is, the larger the preset non - negative and non - zero constant A is. Optionally, the preset additional scan formula can also be a quadratic function formula. It can be seen that implementing this optional embodiment can determine the additional scan times through the preset additional scan formula, which helps to improve the accuracy and adaptability of the additional scan times, and also helps to repeatedly scan the target area according to the calculated additional scan times subsequently.

[0241] In yet another optional embodiment, as Figure 4 shown, the judgment module 303 includes:

[0242] An acquisition sub - module 3031, configured to acquire at least one first comparison image;

[0243] A comparison sub - module 3032, configured to perform a difference comparison operation on the initial image and all first comparison images, obtain a difference comparison operation result, and determine the outlier corresponding to the initial image according to the difference comparison operation result;

[0244] A judgment sub - module 3033, configured to judge whether the outlier is greater than a preset outlier threshold, and when it is judged that the outlier is greater than the preset outlier threshold, determine that there are abnormal features in the initial image.

[0245] It can be seen that implementing the embodiments of the present invention can scan the target area corresponding to the target feature to obtain multiple scanned images, and perform a multi-image averaging operation on the multiple scanned images to obtain an initial image, thereby helping to generally understand the situation of the target feature based on the initial image; it can also perform a difference comparison operation on the initial image and all the first comparison images to obtain the result of the difference comparison operation, thereby determining the difference value corresponding to the initial image according to the result of the difference comparison operation, and judging whether there are abnormal features in the initial image according to the difference value corresponding to the initial image; it can also, when and only when there are abnormal features in the initial image, obtain more scanned images by increasing the number of scans, thereby obtaining a target image with higher imaging quality based on more scanned images while ensuring the scanning efficiency, which helps to have a more accurate and detailed understanding of the situation of the target feature based on the target image with higher imaging quality.

[0246] In another optional embodiment, as Figure 4 shown, the device further includes:

[0247] An acquisition module 305, configured to acquire a pre-determined initial scanning interval distance when the determination module 303 determines that there are abnormal features in the initial image;

[0248] An adjustment module 306, configured to adjust the initial scanning interval distance to obtain a target scanning interval distance;

[0249] A first determination module 307, configured to determine at least one position to be scanned after the current scanning position according to the target scanning interval distance and the current scanning position;

[0250] Wherein, the scanning interval distance between the position to be scanned closest to the current scanning position and the current scanning position is the target scanning interval distance; when the number of all positions to be scanned is greater than 1, the scanning interval distance between two adjacent positions to be scanned is the target scanning interval distance.

[0251] It can be seen that implementing this optional embodiment can adjust the scanning interval distance when it is determined that there are abnormal features in the initial image, improve the flexibility and adaptability of scanning, and can also adaptively obtain target feature information to provide suitable images for experimental research or teaching activities.

[0252] In another optional embodiment, as Figure 4 shown, the manner in which the adjustment module 306 adjusts the initial scanning interval distance to obtain a target scanning interval distance specifically includes:

[0253] Confirming the result of subtracting the preset scanning interval distance from the initial scanning interval distance as the target scanning interval distance; or,

[0254] Calculate the target scanning interval distance according to the preset scanning interval distance formula and the initial scanning interval distance; among them, when the outlier corresponding to the initial image is larger, the target scanning interval distance calculated according to the preset scanning interval distance formula is larger;

[0255] And, the preset scanning interval distance formula is:

[0256] N2 = N1 * (Q - G) * C

[0257] Wherein, N2 is the target scanning interval distance, N1 is the initial scanning interval distance, Q is the outlier corresponding to the initial image, G is the preset outlier threshold, and C is a preset non - negative and non - zero constant.

[0258] In this optional embodiment, more scanned images at adjacent positions can be obtained by reducing the scanning interval distance. The target features corresponding to the scanned images at adjacent positions are generally relatively close. Therefore, obtaining more scanned images at adjacent positions can improve the accuracy of studying or learning the target features.

[0259] It can be seen that implementing this optional embodiment can reduce the initial scanning interval distance according to the preset scanning interval distance, and can also adaptively reduce the initial scanning interval distance according to the outlier of the initial image through the preset scanning interval distance formula, which helps to improve the adaptability and flexibility of adjusting the scanning interval distance. At the same time, in subsequent scans, corresponding numbers of images with abnormal features can be obtained according to different degrees of outliers, thereby improving the adaptability and flexibility of the number of acquired images, which is helpful for subsequent research and learning of the target features corresponding to the images with abnormal features and improving the accuracy of research and learning.

[0260] In another optional embodiment, as Figure 5 shown, the comparison sub - module 3032 includes:

[0261] An acquisition unit 30321, configured to acquire the feature vector of the initial image and the feature vectors of all first comparison images;

[0262] A calculation unit 30322, configured to calculate the difference value between each first comparison image and the initial image according to the feature vector of the initial image and the feature vectors of all first comparison images, and obtain a plurality of difference values;

[0263] A determination unit 30323, configured to determine the outlier corresponding to the initial image according to the average value of all difference values or the maximum value of all difference values;

[0264] And, the manner in which the calculation unit 30322 calculates the difference value between each first comparison image and the initial image according to the feature vector of the initial image and the feature vectors of all first comparison images to obtain a plurality of difference values specifically includes:

[0265] Calculate the vector distance or vector angle between the feature vector of the initial image and the feature vectors of each first comparison image, and obtain the difference value between the initial image and each first comparison image according to the vector distance or vector angle between the feature vector of the initial image and the feature vectors of each first comparison image; wherein, when the vector distance or the vector angle between the feature vector of the initial image and the feature vector of any first comparison image is larger, the difference value between the initial image and the first comparison image is larger.

[0266] Moreover, the specific manner in which the determination unit 30323 determines the outlier corresponding to the initial image according to the average value of all the difference values or the maximum value of all the difference values specifically includes:

[0267] Determine the average value of all the difference values or the maximum value of all the difference values as the outlier corresponding to the initial image; or,

[0268] Perform a normalization operation on the average value of all the difference values or the maximum value of all the difference values to obtain a normalized average value or a normalized maximum value, and determine the normalized average value or the normalized maximum value as the outlier corresponding to the initial image.

[0269] In this optional embodiment, the initial image is input into the first neural network model, and a feature vector with a certain length output by the first neural network model is the feature vector of the initial image. Optionally, the first neural network model is composed of multiple convolutional layers and uses a global pooling layer and a fully connected layer. In this optional embodiment, the vector distance can be at least one of the L1 distance, the L2 distance (Euclidean distance), the cosine distance, and the Hamming distance. Optionally, the calculated outlier can be the original outlier or the normalized outlier, and the embodiments of the present invention do not make any limitations. In this optional embodiment, the normalization operation can be: input the target value of the normalization operation into the sigmoid function (also called the Logistic function) to obtain the output value of the sigmoid function; this output value is the normalized value obtained after performing the normalization operation on the target value, and the range of the normalized value must be between 0 and 1 excluding 0 and 1. For example, when the outlier before normalization is 1, the outlier after normalization is 0.7311; when the outlier before normalization is 2, the outlier after normalization is 0.8808. Since normalization performs the same data processing on different values to obtain results within the same range, compared with directly comparing the outliers, comparing the normalized outliers can more clearly show the relative relationship between the outliers.

[0270] It can be seen that implementing this optional embodiment can determine the outlier corresponding to the initial image based on the feature vector of the initial image and the feature vectors of all the first comparison images, which helps to determine whether there are abnormal features in the initial image based on the determined outlier corresponding to the initial image; it can obtain the difference value between the initial image and each first comparison image through the vector distance or vector angle between the feature vector of the initial image and the feature vector of each first comparison image, which helps to determine the outlier corresponding to the initial image based on the difference value between the initial image and each first comparison image; it can determine the average value of all the difference values or the maximum value of all the difference values as the outlier corresponding to the initial image, which helps to determine whether there are abnormal features in the initial image based on the determined outlier corresponding to the initial image; it can obtain the outlier corresponding to the initial image by performing a normalization operation on the average value of all the difference values or the maximum value of all the difference values, which helps to more clearly display the relative relationship between different outliers through normalization, thus facilitating the comparison between the outlier and the preset outlier threshold, and also helps to determine whether there are abnormal features in the initial image based on the determined outlier corresponding to the initial image.

[0271] In yet another optional embodiment, as Figure 6 shown, the comparison sub-module 3032 includes an acquisition unit 30324 and a comparison processing unit 30325, where:

[0272] The acquisition unit 30324 is configured to acquire the feature vector of the initial image and the feature vectors of all the first comparison images;

[0273] The comparison processing unit 30325 is configured to calculate the first average value and the first variance value corresponding to the feature vectors of all the first comparison images according to the feature vectors of all the first comparison images, determine the first distribution vector corresponding to all the first comparison images according to the first average value and the first variance value, calculate the first vector distance between the first distribution vector and the feature vector of the initial image, and determine the outlier corresponding to the initial image according to the first vector distance; where, when the first vector distance is larger, the determined outlier corresponding to the initial image is larger; the first vector distance is calculated by the following formula:

[0274] L1 = (f1 - μ1) T σ1 -1 (f1 - μ1);

[0275] where, L1 is the first vector distance, f1 is the feature vector of the initial image, μ1 is the first average value, σ1 is the first variance value, and T is the transpose operation symbol of the matrix; or,

[0276] Multiple second comparison images are screened out from all the first comparison images according to the eigenvectors of the initial image and the eigenvectors of all the first comparison images, and the second variance value of the eigenvectors of all the second comparison images and the eigenvectors of the initial image is calculated; multiple third comparison images are randomly selected from all the first comparison images, and the variance values of the eigenvectors of each third comparison image and the eigenvectors of all the second comparison images are calculated to obtain multiple third difference values; the third average value of all the third difference values is calculated, the difference between the second variance value and the third average value is calculated, and the outlier corresponding to the initial image is determined according to the difference; among them, the larger the difference, the larger the outlier corresponding to the initial image determined; all the second comparison images are multiple first comparison images with the smallest vector distance from the eigenvectors of the initial image.

[0277] Optionally, the calculated outlier can be the original outlier or the normalized outlier, which is not limited in the embodiments of the present invention.

[0278] It can be seen that this optional embodiment provides two methods for determining the outlier corresponding to the initial image, which helps to improve the accuracy and adaptability of the outlier corresponding to the determined initial image, and helps to determine whether there are abnormal features in the initial image according to the outlier corresponding to the determined initial image.

[0279] In another optional embodiment, as Figure 4 shown, the device further includes:

[0280] A second determination module 308, configured to determine the abnormal area in the initial image when the judgment module 303 determines that there are abnormal features in the initial image before the second scanning module 304 repeats scanning the target area according to the determined supplementary scanning times;

[0281] And the manner in which the second scanning module 304 repeats scanning the target area according to the determined supplementary scanning times to obtain the second image set of the target feature at the current scanning position specifically includes:

[0282] Repeating scanning the abnormal area according to the determined supplementary scanning times to obtain the second image set of the target feature at the current scanning position;

[0283] And the manner in which the second scanning module 304 repeats scanning the abnormal area according to the determined supplementary scanning times to obtain the second image set of the target feature at the current scanning position specifically includes:

[0284] Adjusting the A-Scan scanning interval distance corresponding to the abnormal area, and scanning the abnormal area according to the determined supplementary scanning times and the adjusted A-Scan scanning interval distance to obtain the second image set of the target feature at the current scanning position.

[0285] In this alternative embodiment, each scanned image (i.e., B-Scan) is composed of multiple A-Scans. Suppose each B-Scan consists of 500 A-Scans, and through anomaly detection, an abnormal area in the B-Scan can be identified, for example, the area between the 305th and 351st A-Scans. Then, in the next scan, we can perform targeted repeated sampling on this abnormal area, while other non-abnormal areas do not need to be scanned again. In this alternative embodiment, the scanning interval distance of a conventional A-Scan is 5 μm. The scanning of the abnormal area can have a denser scanning interval, which can be a constant or a function of the abnormal value corresponding to the initial image; when the scanning interval is a function of the abnormal value, the scanning distance becomes smaller as the abnormal value increases.

[0286] It can be seen that implementing this alternative embodiment can, when an abnormal feature is determined to exist in the initial image, identify the abnormal area in the initial image, which helps to adjust the scanning area according to the abnormal area in the initial image subsequently, and also helps to study the relevant information of the target feature based on the abnormal area in the initial image; it can narrow down the target area that needs to be repeatedly scanned to the abnormal area, thereby obtaining more scanned images to improve the imaging quality of the initial image while ensuring the scanning efficiency; it can adjust the scanning interval distance of the A-Scans corresponding to the abnormal area, improve the flexibility and adaptability of scanning, and help to obtain more required A-Scans to support the subsequent research and learning of A-Scans, and improve the accuracy of researching and learning A-Scans.

[0287] In another alternative embodiment, as Figure 4 shown, the specific manner in which the second determination module 308 determines the abnormal area in the initial image includes:

[0288] Input the initial image into a pre-determined neural network model to obtain the initial feature map of the initial image output by the neural network model; each pixel position in the initial feature map corresponds to a set of pixel positions in the initial image, each set of pixel positions includes multiple pixel positions in the initial image, all sets of pixel positions include all pixel positions in the initial image and the pixel positions included in all sets of pixel positions are all different;

[0289] Obtain at least one fourth comparison image; input all the fourth comparison images into the neural network model to obtain the comparison feature map of each fourth comparison image output by the neural network model; each comparison feature map includes at least one channel, and the number of channels included in each comparison feature map is the same;

[0290] Calculate the average value and variance value of the pixel values corresponding to all pixel positions of all contrast feature maps of each channel. Determine the fourth average value according to the average values corresponding to all channels, determine the fourth variance value according to the variance values corresponding to all channels, and determine the second distribution vector corresponding to all the fourth contrast images according to the fourth average value and the fourth variance value;

[0291] Obtain the feature vectors of all pixel positions of the initial feature map, calculate the second vector distance between the feature vector of each pixel position of the initial feature map and the second distribution vector, and obtain the second vector distance of all pixel positions of the initial feature map; The second vector distance between the second distribution vector and the feature vector of the initial feature map at the target pixel position is calculated by the following formula:

[0292] L2 = (f2 - μ2) T σ2 -1 (f2 - μ2);

[0293] where, L2 is the second vector distance between the feature vector of the initial feature map at the target pixel position and the second distribution vector, f2 is the feature vector of the initial feature map at the target pixel position, μ2 is the fourth average value, σ2 is the fourth variance value, and T is the transpose operation symbol of the matrix;

[0294] Determine the outlier of each pixel position of the initial feature map according to the second vector distance of each pixel position of the initial feature map, and determine the abnormal region in the initial image according to the outlier of each pixel position of the initial feature map; Among them, when the second vector distance of a certain pixel position is larger, the determined outlier of this pixel position is larger.

[0295] It can be seen that this optional embodiment provides a method for determining the abnormal region in the initial image, which helps to understand the distribution of the abnormal region in the initial image, and helps to perform repeated scanning operations on the abnormal region determined according to this method subsequently, thereby improving the efficiency and accuracy of scanning.

[0296] Embodiment 4

[0297] Please refer to Figure 5 , Figure 5 which is a schematic structural diagram of another adaptive repeated scanning device disclosed in the embodiments of the present invention. As Figure 5 shown, the adaptive repeated scanning device may include:

[0298] A memory 401 storing executable program code;

[0299] A processor 402 coupled to the memory 401;

[0300] The processor 402 calls the executable program code stored in the memory 401 and executes the steps in the adaptive repeated scanning method described in Embodiment 1 or Embodiment 2 of the present invention.

[0301] Embodiment 5

[0302] An embodiment of the present invention discloses a computer-readable storage medium. The computer-readable storage medium stores computer instructions, which, when called, are used to execute the steps in the adaptive repeated scanning method described in Embodiment 1 or Embodiment 2 of the present invention.

[0303] Embodiment 6

[0304] An embodiment of the present invention discloses a computer program product. The computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute the steps in the adaptive repeated scanning method described in Embodiment 1 or Embodiment 2.

[0305] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0306] Through the specific descriptions of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, and the storage medium includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc memories, magnetic disk memories, tape memories, or any other computer-readable medium capable of carrying or storing data.

[0307] Finally, it should be noted that: The adaptive repetitive scanning method and device disclosed in the embodiments of the present invention only disclose the preferred embodiments of the present invention, and are only used to illustrate the technical solutions of the present invention, rather than to limit them; Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive repetitive scanning method, characterized in that, The method includes: Scanning a target area corresponding to a target feature at a current scanning position to obtain a first image set of the target feature at the current scanning position; the target area is an area including the target feature at the current scanning position; Performing a multi-image averaging operation on all the scanned images included in the first image set to obtain an initial image of the target feature at the current scanning position; Determining whether there are abnormal features in the initial image; When it is determined that there are abnormal features in the initial image, scanning the target area repeatedly according to the determined additional scanning times to obtain a second image set of the target feature at the current scanning position; Performing a multi-image averaging operation on all the scanned images included in the first image set and all the scanned images included in the second image set to obtain a target image of the target feature at the current scanning position; The additional scanning times are preset additional scanning times or scanning times calculated according to a preset additional scanning formula; wherein, the larger the abnormal value corresponding to the initial image, the more scanning times calculated according to the preset additional scanning formula; And, the preset additional scanning formula is: M = round[(Q - G)*A + B]; Wherein, M is the scanning times calculated according to the preset additional scanning formula, Q is the abnormal value corresponding to the initial image, G is the preset abnormal value threshold, A is a preset non-negative and non-zero constant, and B is the determined minimum scanning times.

2. The adaptive repetitive scanning method according to claim 1, wherein The determining whether there are abnormal features in the initial image includes: Obtaining at least one first comparison image; Performing a difference comparison operation on the initial image and all the first comparison images to obtain a difference comparison operation result, and determining the abnormal value corresponding to the initial image according to the difference comparison operation result; Determining whether the abnormal value is greater than a preset abnormal value threshold; When it is determined that the abnormal value is greater than the preset abnormal value threshold, determining that there are abnormal features in the initial image.

3. The adaptive repetitive scanning method according to claim 2, wherein The method further includes: When it is determined that there are abnormal features in the initial image, obtaining a determined initial scanning interval distance; Adjusting the initial scanning interval distance to obtain a target scanning interval distance; Determining at least one to-be-scanned position after the current scanning position according to the target scanning interval distance and the current scanning position; Wherein, the scanning interval distance between the to-be-scanned position closest to the current scanning position and the current scanning position is the target scanning interval distance; when the number of all the to-be-scanned positions is greater than 1, the scanning interval distance between two adjacent to-be-scanned positions is the target scanning interval distance.

4. The adaptive repetitive scanning method according to claim 3, wherein The adjusting the initial scanning interval distance to obtain a target scanning interval distance includes: Confirming the result of subtracting the preset scanning interval distance from the initial scanning interval distance as the target scanning interval distance; or, Calculate a target scanning interval distance according to a preset scanning interval distance formula and the initial scanning interval distance; wherein, the larger the outlier value corresponding to the initial image is, the larger the target scanning interval distance calculated according to the preset scanning interval distance formula is; And, the preset scanning interval distance formula is: N2 = N1 * (Q - G) * C Wherein, N2 is the target scanning interval distance, N1 is the initial scanning interval distance, Q is the outlier value corresponding to the initial image, G is the preset outlier threshold, and C is a preset non - negative and non - zero constant.

5. The adaptive repetitive scanning method according to claim 2, wherein The performing a difference comparison operation on the initial image and all the first comparison images to obtain a difference comparison operation result, and determining the outlier value corresponding to the initial image according to the difference comparison operation result includes: Obtain the feature vector of the initial image and the feature vectors of all the first comparison images, calculate the difference value between each first comparison image and the initial image according to the feature vector of the initial image and the feature vectors of all the first comparison images to obtain a plurality of difference values, and determine the outlier value corresponding to the initial image according to the average value of all the difference values or the maximum value of all the difference values; And, the calculating the difference value between each first comparison image and the initial image according to the feature vector of the initial image and the feature vectors of all the first comparison images to obtain a plurality of difference values includes: Calculate the vector distance or vector angle between the feature vector of the initial image and the feature vector of each first comparison image, and obtain the difference value between the initial image and each first comparison image according to the vector distance or vector angle between the feature vector of the initial image and the feature vector of each first comparison image; wherein, the larger the vector distance or the larger the vector angle between the feature vector of the initial image and the feature vector of any first comparison image is, the larger the difference value between the initial image and the first comparison image is; And, the determining the outlier value corresponding to the initial image according to the average value of all the difference values or the maximum value of all the difference values includes: Determine the average value of all the difference values or the maximum value of all the difference values as the outlier value corresponding to the initial image; or, Perform a normalization operation on the average value of all the difference values or the maximum value of all the difference values to obtain a normalized average value or a normalized maximum value, and determine the normalized average value or the normalized maximum value as the outlier value corresponding to the initial image.

6. The adaptive repetitive scanning method according to claim 2, wherein The performing a difference comparison operation on the initial image and all the first comparison images to obtain a difference comparison operation result, and determining the outlier value corresponding to the initial image according to the difference comparison operation result includes: Obtain the feature vector of the initial image and the feature vectors of all the first comparison images; Calculate the first average value and the first variance value corresponding to the feature vectors of all the first comparison images according to the feature vectors of all the first comparison images. Determine the first distribution vector corresponding to all the first comparison images according to the first average value and the first variance value. Calculate the first vector distance between the first distribution vector and the feature vector of the initial image. Determine the outlier corresponding to the initial image according to the first vector distance. Wherein, the larger the first vector distance is, the larger the outlier corresponding to the initial image is determined. The first vector distance is calculated by the following formula: L1 = (f1 - μ1) T σ1 -1 (f1 - μ1); Wherein, L1 is the first vector distance, f1 is the feature vector of the initial image, μ1 is the first average value, σ1 is the first variance value, and T is the transpose operation symbol of the matrix; or, Select multiple second comparison images from all the first comparison images according to the feature vector of the initial image and the feature vectors of all the first comparison images, and calculate the second variance value of the feature vectors of all the second comparison images and the feature vector of the initial image. Randomly select multiple third comparison images from all the first comparison images, and calculate the variance values of the feature vectors of each third comparison image and the feature vectors of all the second comparison images to obtain multiple third variance differences. Calculate the third average value of all the third variance differences, calculate the difference between the second variance value and the third average value, and determine the outlier corresponding to the initial image according to the difference. Wherein, the larger the difference is, the larger the outlier corresponding to the initial image is determined. All the second comparison images are multiple first comparison images with the smallest vector distance from the feature vector of the initial image.

7. The adaptive repetitive scanning method according to claim 2, wherein Before repeating the scanning of the target area according to the determined supplementary scanning times, the method further includes: When it is determined that there are abnormal features in the initial image, determine the abnormal area in the initial image; And, repeating the scanning of the target area according to the determined supplementary scanning times includes: Repeating the scanning of the abnormal area according to the determined supplementary scanning times; And, repeating the scanning of the abnormal area according to the determined supplementary scanning times includes: Adjust the A-Scan scanning interval distance corresponding to the abnormal area, and scan the abnormal area according to the determined supplementary scanning times and the adjusted A-Scan scanning interval distance.

8. The adaptive repeated scanning method according to claim 7, wherein Determining the abnormal area in the initial image includes: Input the initial image into a pre-determined neural network model to obtain the initial feature map of the initial image output by the neural network model. Each pixel position of the initial feature map corresponds to a set of pixel positions of the initial image. Each set of pixel positions includes multiple pixel positions of the initial image. All the sets of pixel positions include all the pixel positions of the initial image and all the pixel positions included in all the sets of pixel positions are different; Obtain at least one fourth comparison image; input all the fourth comparison images into the neural network model to obtain the comparison feature maps of each of the fourth comparison images output by the neural network model; each of the comparison feature maps includes at least one channel, and the number of channels included in each of the comparison feature maps is the same; Calculate the average value and variance value of the pixel values corresponding to all pixel positions of all the comparison feature maps of each channel, determine the fourth average value according to the average values corresponding to all the channels, determine the fourth variance value according to the variance values corresponding to all the channels, and determine the second distribution vector corresponding to all the fourth comparison images according to the fourth average value and the fourth variance value; Obtain the feature vectors of all pixel positions of the initial feature map, calculate the second vector distance between the feature vector of each pixel position of the initial feature map and the second distribution vector, and obtain the second vector distances of all the pixel positions of the initial feature map; the second vector distance between the second distribution vector and the feature vector of the initial feature map at the target pixel position is calculated by the following formula: L2 = (f2 - μ2) T σ2 -1 (f2 - μ2); Where L2 is the second vector distance between the feature vector of the initial feature map at the target pixel position and the second distribution vector, f2 is the feature vector of the initial feature map at the target pixel position, μ2 is the fourth average value, σ2 is the fourth variance value, and T is the transpose operation symbol of the matrix; Determine the outliers of each pixel position of the initial feature map according to the second vector distances of each pixel position of the initial feature map, and determine the abnormal region in the initial image according to the outliers of each pixel position of the initial feature map; wherein, the larger the second vector distance of a certain pixel position is, the larger the outlier determined for this pixel position is.

9. An adaptive repetitive scanning device, characterized in that, The device includes: A first scanning module, configured to scan a target region corresponding to a target feature at a current scanning position to obtain a first image set of the target feature at the current scanning position; the target region is a region including the target feature at the current scanning position; An image processing module, configured to perform a multi-image averaging operation on all the scanned images included in the first image set to obtain an initial image of the target feature at the current scanning position; A judgment module, configured to judge whether there is an abnormal feature in the initial image; A second scanning module, configured to, when the judgment module judges that there is an abnormal feature in the initial image, repeatedly scan the target region according to the determined number of supplementary scans to obtain a second image set of the target feature at the current scanning position; The image processing module is further configured to perform a multi-image averaging operation on all the scanned images included in the first image set and all the scanned images included in the second image set to obtain a target image of the target feature at the current scanning position; The number of supplementary scans is a preset number of supplementary scans or the number of scans calculated according to a preset supplementary scan formula; wherein, the larger the outlier corresponding to the initial image is, the more the number of scans calculated according to the preset supplementary scan formula is; And, the preset supplementary scanning formula is as follows: M = round[(Q - G) * A + B]; Wherein, M is the number of scans calculated according to the preset supplementary scanning formula, Q is the outlier corresponding to the initial image, G is the preset outlier threshold, A is a preset non - negative and non - zero constant, and B is the determined minimum number of scans.

Images