CN114092464B - OCT image processing method and device - Google Patents

Abstract

The invention discloses a processing method and a processing device of OCT images, wherein the method comprises the following steps: obtaining a B-Scan image corresponding to a target feature, performing image layering processing on the B-Scan image through a preset image processing algorithm to obtain an initial layering result, inputting the B-Scan image into a pre-trained deep neural network model to obtain an output result, wherein the output result comprises probability corresponding to each pixel point corresponding to a target area of the B-Scan image, the probability corresponding to each pixel point is used for representing the probability that each pixel point belongs to interlayer boundaries of certain two adjacent layering layers included in the initial layering result, the target area is an area comprising the target feature, and interlayer boundary information corresponding to the target feature in the B-Scan image is determined according to the probability corresponding to all the pixel points. Therefore, the method and the device can be used for executing image layering processing on the acquired B-Scan image by combining the deep neural network model, and are beneficial to improving the image layering efficiency and improving the accuracy of the image layering result.

Description

OCT image processing method and device

Technical Field

The present invention relates to the field of image processing technology, and in particular to an OCT image processing method and device.

Background technique

With the rapid development of computer and medical technology, OCT (Optical Coherence Tomography) technology has been widely used in diagnostic equipment for fundus diseases, which is of great significance for the detection, experiments and compilation of teaching materials of fundus diseases. OCT is a high-sensitivity, high-resolution, high-speed, non-invasive tomography imaging method, which uses the coherence of light to scan the fundus. Each scan is called an A-scan, and multiple adjacent continuous scans are combined together to form a B-scan image. The B-scan image is also the commonly seen OCT cross-sectional image (it can also be understood as an OCT image), which is the most important imaging method of OCT in medical diagnosis.

In practical applications, the diagnosis of fundus diseases by diagnostic equipment usually relies on the target feature stratification results after OCT image stratification, such as retinal stratification results. However, it is found in practice that the OCT image stratification method based on traditional stratification techniques such as histogram, boundary stratification and regional stratification, although the algorithm performance is relatively robust, the stratification speed of the target feature stratification results obtained by stratifying OCT images is slow. Therefore, it is particularly important to provide an OCT image stratification algorithm to improve the stratification efficiency of OCT images while ensuring the algorithm performance.

Summary of the invention

The technical problem to be solved by the present invention is to provide an OCT image processing method and device, which can be combined with a deep neural network model to improve the stratification efficiency of the OCT image and the accuracy of the stratification results while ensuring the performance of the OCT image stratification algorithm.

In order to solve the above technical problems, the first aspect of the present invention discloses a method for processing an OCT image, the method comprising:

Obtain the B-Scan image corresponding to the target feature;

Performing image layering processing on the B-Scan image by using a preset image processing algorithm to obtain an initial layering result;

Input the B-Scan image into a pre-trained deep neural network model to obtain an output result, wherein the output result includes a probability corresponding to each pixel point corresponding to a target area of the B-Scan image, and the probability corresponding to each pixel point is used to indicate the possibility that each pixel point belongs to a layer boundary between two adjacent layers included in the initial layering result, and the target area is a region including the target feature;

According to the corresponding probabilities of all the pixel points, the inter-layer boundary information corresponding to the target feature in the B-Scan image is determined.

As an optional implementation, in the first aspect of the present invention, performing image layering processing on the B-Scan image by a preset image processing algorithm to obtain an initial layering result includes:

Performing filtering processing on the B-Scan image by using a preset filtering function to obtain a filtered image;

Calculating the positive gradient of the filtered image in the vertical direction of the image, and constructing a first cost function according to the positive gradient;

Determine a first minimum cost path from the left edge to the right edge of the filtered image according to a predetermined path algorithm and the first cost function, and obtain a first layering line;

Determine a second minimum cost path from the left edge to the right edge of the filtered image according to the path algorithm and the first cost function, and obtain a second layering line;

Calculating the negative gradient of the filtered image in the vertical direction of the image, and constructing a second cost function according to the negative gradient;

Determine a search area, where the search area is a lower area corresponding to a lower layer line between the first layer line and the second layer line;

Determine a third minimum cost path from the left edge of the search area to the right edge of the search area according to the path algorithm and the second cost function, and perform a smoothing filter operation on the third minimum cost path to obtain a third layering line;

determining the first stratification line, the second stratification line and the third stratification line as initial stratification results;

And, before determining the second minimum cost path from the left edge to the right edge of the filtered image according to the path algorithm and the first cost function to obtain the second layering line, the method further includes:

The first path is marked as an unreachable path in the filtered image.

As an optional implementation, in the first aspect of the present invention, before inputting the B-Scan image into a pre-trained deep neural network model to obtain an output result, the method further includes:

Determining a target area including the target feature in the B-Scan image;

Wherein, determining the target area including the target feature in the B-Scan image comprises:

The first stratification line and the second stratification line that are located relatively above each other are shifted upward by a first preset distance in the vertical direction of the image to obtain a first boundary line;

The third layering line is shifted downward by a second preset distance in the vertical direction of the image to obtain a second boundary line;

An area below the first boundary line and above the second boundary line is determined as a target area including the target feature in the B-Scan image.

As an optional implementation, in the first aspect of the present invention, after inputting the B-Scan image into a pre-trained deep neural network model to obtain an output result, before determining the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all the pixel points, the method further includes:

Determine whether there is a target pixel point falling outside the target area among all the pixel points;

When it is determined that none of the target pixel points falls outside the target area among all the pixel points, triggering the operation of determining the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all the pixel points;

When it is determined that there are target pixel points that fall outside the target area among all the pixel points, a probability update operation is performed on all the target pixel points that fall outside the target area to update the probabilities corresponding to all the target pixel points, and trigger the execution of the operation of determining the inter-layer boundary information corresponding to the target feature in the B-Scan image based on the probabilities corresponding to all the pixel points.

As an optional implementation manner, in the first aspect of the present invention, the determining whether there is a target pixel point falling outside the target area among all the pixel points includes:

For each column of all the pixel points, determining whether there is a target pixel point in the column that falls outside the target area;

And, performing a probability update operation on all the target pixel points falling outside the target area to update the probabilities corresponding to all the target pixel points falling outside the target area includes:

For each column of all the pixel points, if there is a target pixel point in the column that falls outside the target area, each target pixel point in the column that falls outside the target area is multiplied by a preset value corresponding to the target pixel point to obtain a product result, and the probability corresponding to the target pixel point is updated according to the product result.

As an optional implementation manner, in the first aspect of the present invention, determining the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all the pixel points includes:

For each column of pixel points among all the pixel points, normalizing the probability distribution of the column of pixel points to obtain a normalized probability distribution of the column of pixel points;

For each column of pixels among all the pixels, a dot product operation is performed on the normalized probability distribution of the column of pixels and the row number distribution corresponding to the column of pixels to obtain an inter-layer distribution result corresponding to the column of pixels;

According to the inter-layer distribution result corresponding to each column of pixel points in all the pixel points, the inter-layer boundary information corresponding to the target feature in the B-Scan image is determined.

As an optional implementation, in the first aspect of the present invention, the deep neural network model is trained in the following manner:

Acquire a B-Scan image set including annotation information, wherein the annotation information corresponding to each B-Scan image in the B-Scan image set includes label information corresponding to the target feature and boundary information corresponding to the target feature;

Dividing the B-Scan image set to obtain a training set and a test set, wherein the training set is used to train a deep neural network model, and the test set is used to verify the reliability of the trained deep neural network model;

Performing a target processing operation on all B-Scan images included in the training set to obtain a processing result, wherein the target processing operation includes at least one of an up-and-down movement process, a left-and-right flip process, an up-and-down inversion process, and a contrast adjustment process;

Inputting the processing result as input data into a predetermined deep neural network model to obtain an output result;

Analyze and calculate the joint loss according to the output result, the B-Scan image included in the training set, and the boundary information to obtain a joint loss value;

Back-propagating the joint loss value in the deep neural network model, and performing iterative training of a preset cycle length to obtain a trained deep neural network model;

The test set is used to verify the reliability of the trained deep neural network model.

As an optional embodiment, in the first aspect of the present invention, the target feature is a retinal feature.

A second aspect of the present invention discloses an OCT image processing device, the device comprising:

An acquisition module is used to acquire a B-Scan image corresponding to the target feature;

A first processing module, configured to perform image layering processing on the B-Scan image using a preset image processing algorithm to obtain an initial layering result;

A second processing module is used to input the B-Scan image into a pre-trained deep neural network model to obtain an output result, wherein the output result includes a probability corresponding to each pixel point corresponding to a target area of the B-Scan image, and the probability corresponding to each pixel point is used to indicate the possibility that each pixel point belongs to a layer boundary between two adjacent layers included in the initial layering result, and the target area is a region including the target feature;

The first determination module is used to determine the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the corresponding probabilities of all the pixel points.

As an optional implementation, in the second aspect of the present invention, the first processing module includes:

A filtering submodule, used for performing filtering processing on the B-Scan image through a preset filtering function to obtain a filtered image;

A function construction submodule, used for calculating the positive gradient of the filtered image in the vertical direction of the image, and constructing a first cost function according to the positive gradient;

A first determination submodule, determining a first minimum cost path from a left edge to a right edge of the filtered image according to a predetermined path algorithm and the first cost function, to obtain a first layering line;

The first determination submodule is further used to determine a second minimum cost path from the left edge to the right edge of the filtered image according to the path algorithm and the first cost function to obtain a second layering line;

The function construction submodule is further used to calculate the negative gradient of the filtered image in the vertical direction of the image, and construct a second cost function according to the negative gradient;

A second determination submodule is used to determine a search area, where the search area is a lower area corresponding to a lower layer line between the first layer line and the second layer line;

The first determination submodule is further used to determine a third minimum cost path from the left edge of the search area to the right edge of the area according to the path algorithm and the second cost function, and perform a smoothing filter operation on the third minimum cost path to obtain a third layer line;

The second determining submodule is further configured to determine the first stratification line, the second stratification line and the third stratification line as initial stratification results;

And, the first processing module also includes:

A marking submodule is used to mark the first path as an unreachable path in the filtered image before the first determination submodule determines the second minimum cost path from the left edge to the right edge of the filtered image according to the path algorithm and the first cost function to obtain the second layer line.

As an optional implementation, in the second aspect of the present invention,.

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

A second determination module, configured to determine a target area including the target feature in the B-Scan image before the second processing module inputs the B-Scan image into a pre-trained deep neural network model to obtain an output result;

The manner in which the second determination module determines the target area including the target feature in the B-Scan image specifically includes:

The first stratification line and the second stratification line that are located relatively above each other are shifted upward by a first preset distance in the vertical direction of the image to obtain a first boundary line;

The third layering line is shifted downward by a second preset distance in the vertical direction of the image to obtain a second boundary line;

An area below the first boundary line and above the second boundary line is determined as a target area including the target feature in the B-Scan image.

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

A judgment module, used for judging whether there are target pixels falling outside the target area among all the pixels after the second processing module inputs the B-Scan image into a pre-trained deep neural network model to obtain an output result and before the first determination module determines the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all the pixels, and when it is judged that there are no target pixels falling outside the target area among all the pixels, triggering the first determination module to perform the operation of determining the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all the pixels;

The third processing module is used to perform a probability update operation on all the target pixel points that fall outside the target area when the judgment module determines that there are target pixel points among all the pixel points that fall outside the target area, so as to update the probabilities corresponding to all the target pixel points, and trigger the first determination module to perform the operation of determining the inter-layer boundary information corresponding to the target feature in the B-Scan image based on the probabilities corresponding to all the pixel points.

As an optional implementation, in the second aspect of the present invention, the manner in which the judgment module judges whether there is a target pixel point falling outside the target area among all the pixel points specifically includes:

For each column of all the pixel points, determining whether there is a target pixel point in the column that falls outside the target area;

Furthermore, the third processing module performs a probability update operation on all the target pixel points falling outside the target area, so as to update the probabilities corresponding to all the target pixel points falling outside the target area, specifically including:

For each column of all the pixel points, if there is a target pixel point in the column that falls outside the target area, each target pixel point in the column that falls outside the target area is multiplied by a preset value corresponding to the target pixel point to obtain a product result, and the probability corresponding to the target pixel point is updated according to the product result.

As an optional implementation, in the second aspect of the present invention, the first determination module determines the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all the pixel points in a manner specifically including:

For each column of pixel points among all the pixel points, normalizing the probability distribution of the column of pixel points to obtain a normalized probability distribution of the column of pixel points;

For each column of pixels among all the pixels, a dot product operation is performed on the normalized probability distribution of the column of pixels and the row number distribution corresponding to the column of pixels to obtain an inter-layer distribution result corresponding to the column of pixels;

According to the inter-layer distribution result corresponding to each column of pixel points in all the pixel points, the inter-layer boundary information corresponding to the target feature in the B-Scan image is determined.

As an optional implementation, in the second aspect of the present invention, the deep neural network model is trained in the following manner:

Acquire a B-Scan image set including annotation information, wherein the annotation information corresponding to each B-Scan image in the B-Scan image set includes label information corresponding to the target feature and boundary information corresponding to the target feature;

Dividing the B-Scan image set to obtain a training set and a test set, wherein the training set is used to train a deep neural network model, and the test set is used to verify the reliability of the trained deep neural network model;

Performing a target processing operation on all B-Scan images included in the training set to obtain a processing result, wherein the target processing operation includes at least one of an up-and-down movement process, a left-and-right flip process, an up-and-down inversion process, and a contrast adjustment process;

Inputting the processing result as input data into a predetermined deep neural network model to obtain an output result;

Analyze and calculate the joint loss according to the output result, the B-Scan image included in the training set, and the boundary information to obtain a joint loss value;

Back-propagating the joint loss value in the deep neural network model, and performing iterative training of a preset cycle length to obtain a trained deep neural network model;

The test set is used to verify the reliability of the trained deep neural network model.

As an optional embodiment, in the second aspect of the present invention, the target feature is a retinal feature.

The third aspect of the present invention discloses another OCT image processing device, the device comprising:

A memory storing executable program code;

a processor coupled to the memory;

an input interface and an output interface coupled to the processor;

The processor calls the executable program code stored in the memory to execute the OCT image processing method disclosed in the first aspect of the present invention.

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

In an embodiment of the present invention, a method and device for processing an OCT image are provided, the method comprising: obtaining a B-Scan image corresponding to a target feature, performing image stratification processing on the B-Scan image through a preset image processing algorithm to obtain an initial stratification result, and then inputting the B-Scan image into a pre-trained deep neural network model to obtain an output result, the output result including the probability corresponding to each pixel corresponding to the target area of the B-Scan image, the probability corresponding to each pixel is used to indicate the possibility that each pixel belongs to the interlayer boundary of two adjacent layers included in the initial stratification result, the target area is an area including the target feature, and according to the probabilities corresponding to all pixels, the interlayer boundary information corresponding to the target feature in the B-Scan image is determined. It can be seen that the implementation of the present invention can intelligently obtain a B-Scan image including the target feature, which is conducive to improving the classification efficiency of the B-Scan image; it can also combine the deep neural network model and the initial stratification result to intelligently determine the interlayer boundary information corresponding to the target feature in the B-Scan image, which is conducive to improving the image stratification efficiency and the accuracy of the image stratification result.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

FIG1 is a schematic flow chart of an OCT image processing method disclosed in an embodiment of the present invention;

FIG2 is a schematic flow chart of another OCT image processing method disclosed in an embodiment of the present invention;

FIG3 is a schematic diagram of the structure of an OCT image processing device disclosed in an embodiment of the present invention;

FIG4 is a schematic diagram of the structure of another OCT image processing device disclosed in an embodiment of the present invention;

FIG. 5 is a schematic diagram of the structure of yet another OCT image processing device disclosed in an embodiment of the present invention.

Detailed ways

In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

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 "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or end including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or ends.

Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

The present invention discloses an OCT image processing method and device, which can intelligently obtain a B-Scan image including target features, which is beneficial to improving the classification efficiency of the B-Scan image; it can also combine a deep neural network model and the initial stratification result to intelligently determine the inter-layer boundary information corresponding to the target feature in the B-Scan image, which is beneficial to improving the image stratification efficiency and the accuracy of the image stratification result. The following are detailed descriptions.

Embodiment 1

Please refer to FIG. 1 , which is a flowchart of an OCT image processing method disclosed in an embodiment of the present invention. The OCT method described in FIG. 1 can be applied to the layered processing of retinal B-Scan images, and the layered results obtained by the method can be applied to the compilation of medical textbooks or as auxiliary materials for retinal research, which is not limited in the embodiment of the present invention. As shown in FIG. 1 , the OCT image processing method may include the following operations:

101. Obtain a B-Scan image corresponding to the target feature.

In an embodiment of the present invention, the target feature may include a human eye retinal feature (hereinafter referred to as a retinal feature), and correspondingly, the B-Scan image may include a B-Scan image including retinal features obtained through OCT technology processing. The B-Scan image may be obtained by directly acquiring a B-Scan image including retinal features acquired by a device carrying OCT scanning technology, or by acquiring a B-Scan image including retinal features stored in a system database, which is not limited in the embodiment of the present invention.

102. Perform image layering processing on the B-Scan image using a preset image processing algorithm to obtain an initial layering result.

In an embodiment of the present invention, when the B-Scan image includes the above-mentioned retinal features, the corresponding initial stratification result is the stratification result corresponding to the retinal tissue layer. The preset image processing algorithm includes an algorithm improved from the traditional B-Scan image stratification processing (such as an improved minimum cost path algorithm based on a gradient cost map).

It should be noted that in the OCT image processing method provided by the present invention, since the retina includes many tissue layers, when performing image stratification processing on the retinal B-Scan image through a preset image processing algorithm, not all tissue layers included in the retina are stratified, but only the ILM (inner limiting membrane) layer, ISOS (inner and outer segments of photoreceptor cells) layer and BM (Bruch's membrane) layer, which have more obvious boundaries than other retinal tissue layers, are divided. This improves the stratification efficiency of the retinal tissue layers of the B-Scan image and the accuracy of the stratification results while ensuring the robust performance of the processing algorithm of the OCT image.

It can be seen that in the embodiment of the present invention, by reducing the number of stratification layers of the retinal tissue layers in the B-Scan image, the amount of data that needs to be calculated when processing the B-Scan image is reduced, thereby achieving the purpose of improving the stratification efficiency of the B-Scan image and improving the accuracy of the stratification results.

103. Input the B-Scan image into the pre-trained deep neural network model to obtain the output result.

In an embodiment of the present invention, the output result includes the probability corresponding to each pixel point corresponding to the target area of the B-Scan image, and the probability corresponding to each pixel point is used to indicate the possibility that each pixel point belongs to the inter-layer boundary of two adjacent layers included in the above-mentioned initial stratification result; the target area is the area including the target feature.

Furthermore, when the target feature is the above-mentioned retinal feature, the target area is the area including the retinal tissue layer; assuming that the above-mentioned initial stratification result includes three tissue layers, and in the vertical direction of the image, from top to bottom they are the ILM layer, the ISOS layer and the BM layer, the two adjacent layers include the ILM layer and the ISOS layer, and the ISOS layer and the BM layer, wherein the two adjacent layers refer to two tissue layers that are adjacent in position among the tissue layers obtained by algorithm or artificial division, and do not refer to two adjacent tissue layers among all the tissue layers included in the actual retina.

104. According to the corresponding probabilities of all pixel points, determine the inter-layer boundary information corresponding to the target features in the B-Scan image.

In an embodiment of the present invention, when the target feature is the above-mentioned retinal feature, the probability corresponding to each pixel refers to the probability that the pixel belongs to the interlayer boundary of two adjacent retinal layers; and determining the interlayer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all pixels may specifically include the following steps:

For each column of pixels among all the pixels, a normalization process is performed on the probability distribution of the pixel columns to obtain a normalized probability distribution of the pixel columns;

For each column of all pixels, a dot product operation is performed between the normalized probability distribution of the column of pixels and the row number distribution corresponding to the column of pixels to obtain the inter-layer distribution result corresponding to the column of pixels;

According to the inter-layer distribution results corresponding to each column of pixels among all pixels, the inter-layer boundary information corresponding to the target features in the B-Scan image is determined.

It should be noted that the function used to perform normalization processing on the probability distribution of the column of pixels may include a Softmax function.

It can be seen that the implementation of the OCT image processing method described in Figure 1 can perform targeted image stratification processing on the acquired B-Scan image including the target features, and by reducing the number of stratification layers of the retinal tissue layers in the B-Scan image, the amount of data required for calculation when processing the B-Scan image is reduced, thereby improving the stratification efficiency of the B-Scan image; it can also combine with a pre-trained deep neural network model to determine the inter-layer boundary information corresponding to the target features in the B-Scan image, thereby improving the stratification efficiency and further improving the accuracy of the stratification results.

In an optional embodiment, performing image layering processing on the B-Scan image by using a preset image processing algorithm to obtain an initial layering result may specifically include the following steps:

Performing filtering processing on the B-Scan image through a preset filtering function to obtain a filtered image;

Calculate the positive gradient of the filtered image in the vertical direction of the image, and construct a first cost function based on the positive gradient;

According to a predetermined path algorithm and a first cost function, a first minimum cost path from a left edge to a right edge of the filtered image is determined to obtain a first layering line;

According to the path algorithm and the first cost function, a second minimum cost path from the left edge to the right edge of the filtered image is determined to obtain a second layering line;

Calculate the negative gradient of the filtered image in the vertical direction of the image, and construct a second cost function based on the negative gradient;

Determine a search area, where the search area is a lower area corresponding to a lower layer line between the first layer line and the second layer line;

According to the path algorithm and the second cost function, a third minimum cost path from the left edge of the search area to the right edge of the area is determined, and a smoothing filter operation is performed on the third minimum cost path to obtain a third layer line;

Determine the first stratification line, the second stratification line and the third stratification line as initial stratification results;

Furthermore, before determining the second minimum cost path from the left edge to the right edge of the filtered image according to the path algorithm and the first cost function to obtain the second layer line, the method further includes the following steps:

The first path is marked as an unreachable path in the filtered image.

In this optional embodiment, the predetermined path algorithm may be a minimum cost path algorithm improved on the basis of the Dijkstra or Bellman-Ford algorithm, which is not limited in the embodiment of the present invention; and, the function expression corresponding to the first cost function may be Cost1=a*exp(-G) or Cost1=a*(-G), and the function expression corresponding to the second cost function may be Cost2=a*exp(-G) or Cost2=a*(-G), where G is the gradient value.

In this optional embodiment, the first path is marked as an unreachable path in the filtered image, so that after obtaining the first layer line, a second layer line different from the first layer line is obtained based on the marked unreachable path and after re-obtaining the minimum cost path through a predetermined path algorithm, thereby ensuring that two different layer lines can be obtained.

In this optional embodiment, filtering processing is performed on the B-Scan image through a preset filtering function, wherein the preset filtering function may be a median filtering function and a mean filtering function; when a smoothing filtering operation is performed on the third minimum cost path, the actual processing may be to perform median filtering and mean filtering smoothing processing on the coordinate points included in the third minimum cost path.

It can be seen that this optional embodiment provides a minimum cost path algorithm, which can divide the required first layer line, second layer line and third layer line in the B-Scan image, and reduce the amount of data required to be calculated when processing the B-Scan image by means of the layering number of retinal tissue layers in the B-Scan image, thereby improving the layering efficiency of the image layering algorithm and improving the accuracy of the layering results.

In another optional embodiment, after inputting the B-Scan image into a pre-trained deep neural network model to obtain an output result, and before determining the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all pixel points, the method may further include the following steps:

Determine whether there is a target pixel point falling outside the target area among all the pixels;

When it is determined that there is no target pixel point falling outside the target area among all the pixels, an operation of determining the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all the pixels is triggered;

When it is determined that there are target pixels falling outside the target area among all the pixels, a probability update operation is performed on all the target pixels falling outside the target area to update the probabilities corresponding to all the target pixels, and trigger the execution of an operation to determine the inter-layer boundary information corresponding to the target feature in the B-Scan image based on the probabilities corresponding to all the pixels.

In this optional embodiment, determining whether there is a target pixel point falling outside the target area among all the pixel points includes:

Determine whether the coordinate value corresponding to each pixel point among all the pixels exceeds the coordinate interval included in the target area.

It can be seen that this optional embodiment can perform a probability update operation for all target pixel points falling outside the target area, which is beneficial to subsequently determine the inter-layer boundary information corresponding to the target features in the B-Scan image based on the probabilities corresponding to all pixel points, thereby improving the accuracy of the determined inter-layer boundary information.

In this optional embodiment, the above-mentioned determining whether there is a target pixel point falling outside the target area among all the pixel points may also include:

Determine whether the probability corresponding to each pixel among all pixels is within a preset probability threshold, where the preset probability threshold is a predetermined probability threshold corresponding to the pixel falling within the target area (such as (0.5, 1), excluding the two endpoint values).

It can be seen that in this optional embodiment, another method for determining whether a pixel point falls within the target area is provided. Instead of analyzing and processing the coordinate values corresponding to each pixel point, including the coordinate data on the x-axis, y-axis and even z-axis, only the probability values corresponding to the pixel points are analyzed and processed, which reduces the amount of data analyzed and processed, expands the method for determining whether a pixel point falls within the target area, and improves the efficiency of determining and obtaining results.

Embodiment 2

Please refer to FIG. 2, which is a flowchart of another OCT image processing method disclosed in an embodiment of the present invention. The OCT method described in FIG. 2 can be applied to the layered processing of retinal B-Scan images. The layered results obtained by the method can be applied to the compilation of medical textbooks or as auxiliary materials for retinal research, which is not limited in the embodiment of the present invention. As shown in FIG. 2, the OCT image processing method can include the following operations:

201. Obtain a B-Scan image corresponding to the target feature.

202. Perform image layering processing on the B-Scan image using a preset image processing algorithm to obtain an initial layering result.

203. Input the B-Scan image into a pre-trained deep neural network model to obtain an output result.

204. Determine the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the corresponding probabilities of all the pixel points.

In the embodiment of the present invention, for other descriptions of step 201 to step 204, please refer to other specific descriptions of step 101 to step 104 in embodiment 1, and the embodiment of the present invention will not be repeated here.

205. Determine a target region including target features in the B-Scan image.

In an embodiment of the present invention, before inputting the B-Scan image into a pre-trained deep neural network model to obtain an output result, determining a target area including target features in the B-Scan image may specifically include the following steps:

The first stratification line and the stratification line located relatively above the second stratification line are shifted upward by a first preset distance in the vertical direction of the image to obtain a first boundary line;

The third layering line is shifted downward by a second preset distance in the vertical direction of the image to obtain a second boundary line;

The area below the first boundary line and above the second boundary line is determined as a target area including target features in the B-Scan image.

It can be seen that the OCT image processing method described in Figure 2 can perform image stratification processing on the acquired B-Scan image including the target feature, and by reducing the number of stratification layers of the retinal tissue layer in the B-Scan image, the amount of data required for calculation when processing the B-Scan image is reduced, thereby improving the stratification efficiency of the B-Scan image; it can also determine the inter-layer boundary information corresponding to the target feature in the B-Scan image in combination with a pre-trained deep neural network model, thereby improving the stratification efficiency and further improving the accuracy of the stratification results; in addition, it can also intelligently determine the target area including the target feature, which is beneficial to reduce the amount of data that the image processing algorithm needs to process when performing subsequent image processing operations on the target area, thereby improving the image stratification efficiency to a certain extent; at the same time, after clarifying the target area including the target feature, it is beneficial to reduce the interference of redundant areas that do not include the target feature on the image stratification processing, thereby improving the accuracy of the image stratification results.

In an optional embodiment, determining whether there is a target pixel point falling outside the target area among all the pixel points includes:

For each column of all pixel points, determine whether there is a target pixel point that falls outside the target area in the column of pixel points;

And, performing a probability update operation on all target pixel points falling outside the target area to update the probabilities corresponding to all target pixel points falling outside the target area, including:

For each column of all pixels, if there is a target pixel that falls outside the target area in the column of pixels, each target pixel that falls outside the target area in the column of pixels is multiplied by a preset value corresponding to the target pixel to obtain a product result, and the probability corresponding to the target pixel is updated according to the product result.

In this optional embodiment, the preset value may include a fixed and extremely small coefficient ε (such as 0.0001), and the coefficient ε included in the preset value may also be an attenuation value that changes with position. For example, in the target area composed of the first boundary line and the second boundary line, the area that is located in the center of the target area and has an equal distance from the first boundary line and the second boundary line is used as a reference system, and the area where the reference system is located is used as the starting point, and the starting point is defined as the farthest distance from the first boundary line or the second boundary line. The closer the position of the pixel point corresponding to the coefficient ε is to the first boundary line or the second boundary line, the larger the value corresponding to the coefficient ε is, and the farther the position of the pixel point corresponding to the coefficient ε is from the first boundary line or the second boundary line, the smaller the value corresponding to the coefficient ε is.

It can be seen that in this optional embodiment, a step of processing pixels in columns is provided, which can not only process pixels one by one, but also process pixels in columns, thereby improving the processing efficiency for pixels and improving the layering efficiency of OCT images to a certain extent.

In another optional embodiment, the deep neural network model is trained in the following manner:

Acquire a B-Scan image set including annotation information, wherein the annotation information corresponding to each B-Scan image in the B-Scan image set includes label information corresponding to the target feature and boundary information corresponding to the target feature;

Divide the B-Scan image set into a training set and a test set. The training set is used to train the deep neural network model, and the test set is used to verify the reliability of the trained deep neural network model.

Performing a target processing operation on all B-Scan images included in the training set to obtain a processing result, wherein the target processing operation includes at least one of an up-and-down movement process, a left-and-right flip process, an up-and-down inversion process, and a contrast adjustment process;

Input the processed results as input data into a predetermined deep neural network model to obtain output results;

According to the output results, the B-Scan images included in the training set and the boundary information, the joint loss is analyzed and calculated to obtain the joint loss value;

The joint loss value is back-propagated in the deep neural network model, and iterative training is performed for a preset cycle length to obtain a trained deep neural network model;

Among them, the test set is used to verify the reliability of the trained deep neural network model.

In this optional embodiment, analyzing and calculating the joint loss to obtain the joint loss value includes:

The label loss L _{label_dice} is calculated based on the first data (denoted as M0) included in the output result and the labeled pixel-level label, where the pixel-level label is a label encoded by a preset encoding method (one-hot);

According to the first data (M0) and the pixel-level label, a cross entropy loss L _{label_ce} is calculated;

According to the second data (denoted as B0) and the boundary information included in the output result, the boundary cross entropy loss L _{bd — ce} is calculated;

According to the third data (denoted as B2) and the boundary information included in the above mathematical operation result, the smoothing loss L _{bd — l1} is calculated;

Multiply the label loss L _{label_dice} , the cross entropy loss L _{label_ce} , the boundary cross entropy loss L _{bd_ce} , and the smoothing loss L _{bd_l1} by a coefficient respectively to obtain their respective product results, and sum all the product results to obtain the numerical result corresponding to the joint loss as the joint loss value.

The calculation formula of joint loss in practical application is as follows:

L＝λ _{label_dice} L _{label_dice} +λ _{label_ce} L _{label_ce} +λ _{bd_ce} L _{bd_ce} +λ _{bd_l1} L _{bd_l1}

It can be seen that the deep neural network model obtained through training, combined with the probability distribution corresponding to each column of all pixels in units of columns, performs relevant processing operations (including normalization and dot product operations) to determine the inter-layer boundary information corresponding to the target features in the B-Scan image, thereby achieving the purpose of improving the stratification efficiency of OCT images and improving the accuracy of stratification results.

Embodiment 3

Please refer to FIG3 , which is a schematic diagram of the structure of an OCT image processing device disclosed in an embodiment of the present invention. The OCT image processing device may be an OCT image processing terminal, an OCT image processing device, an OCT image processing system or an OCT image processing server. The OCT image processing server may be a local server, a remote server, or a cloud server (also known as a cloud server). When the OCT image processing server is a non-cloud server, the non-cloud server can communicate with the cloud server, which is not limited in the embodiment of the present invention. As shown in FIG3 , the OCT image processing device may include an acquisition module 301, a first processing module 302, a second processing module 303 and a first determination module 304, wherein:

The acquisition module 301 is used to acquire a B-Scan image corresponding to the target feature.

The first processing module 302 is used to perform image layering processing on the B-Scan image acquired by the acquisition module 301 through a preset image processing algorithm to obtain an initial layering result.

The second processing module 303 is used to input the B-Scan image acquired by the acquisition module 301 into a pre-trained deep neural network model to obtain an output result, which includes the probability corresponding to each pixel point corresponding to the target area of the B-Scan image. The probability corresponding to each pixel point is used to indicate the possibility that each pixel point belongs to the inter-layer boundary of two adjacent layers included in the initial stratification result. The target area is the area including the target feature.

The first determination module 304 is used to determine the inter-layer boundary information corresponding to the target feature in the B-Scan image acquired by the acquisition module 301 according to the probabilities corresponding to all the pixel points obtained by the second processing module 303 .

It can be seen that the OCT image processing device described in Figure 3 can perform targeted image stratification processing on the acquired B-Scan image including the target features, and by reducing the number of stratification layers of the retinal tissue layers in the B-Scan image, the amount of data required for calculation when processing the B-Scan image is reduced, thereby improving the stratification efficiency of the B-Scan image; it can also combine with a pre-trained deep neural network model to determine the inter-layer boundary information corresponding to the target features in the B-Scan image, thereby improving the stratification efficiency and further improving the accuracy of the stratification results.

In an optional embodiment, as shown in FIG4 , the first processing module 302 may include a filtering submodule 3021, a function building submodule 3022, a first determining submodule 3023, and a second determining submodule 3024, wherein:

The filtering submodule 3021 is used to perform filtering processing on the B-Scan image through a preset filtering function to obtain a filtered image.

The function construction submodule 3022 is used to calculate the positive gradient of the filtered image obtained by the filtering submodule 3021 in the vertical direction of the image, and construct a first cost function based on the positive gradient.

The first determination submodule 3023 determines the first minimum cost path from the left edge to the right edge of the filtered image obtained by the filtering submodule 3021 according to a predetermined path algorithm and the first cost function obtained by the function construction submodule 3022, and obtains a first layering line.

The first determination submodule 3023 is further used to determine the second minimum cost path from the left edge to the right edge of the filtered image obtained by the filtering submodule 3021 according to the predetermined path algorithm and the first cost function obtained by the function construction submodule 3022, and obtain the second layer line.

The function construction submodule 3022 is further used to calculate the negative gradient of the filtered image obtained by the filtering submodule 3021 in the vertical direction of the image, and to construct a second cost function based on the negative gradient.

The second determining submodule 3024 is used to determine a search area, where the search area is a lower area corresponding to the first layer line obtained by the first determining submodule 3023 and the layer line located relatively below the second layer line.

The first determination submodule 3023 is also used to determine the third minimum cost path from the left edge of the search area to the right edge of the area determined by the second determination submodule 3024 according to the predetermined path algorithm and the second cost function obtained by the function construction submodule 3022, and perform a smoothing filtering operation on the third minimum cost path to obtain a third layer line.

The second determining submodule 3024 is further used to determine the first stratification line, the second stratification line and the third stratification line obtained by the first determining submodule 3023 as initial stratification results.

Furthermore, the first processing module 302 may also include a marking submodule 3025, wherein:

The marking submodule 3025 is used to determine the second minimum cost path from the left edge to the right edge of the filtered image obtained by the filtering submodule 3021 according to a predetermined path algorithm and a first cost function in the first determination submodule 3023, and mark the first path as an unreachable path in the filtered image obtained by the filtering submodule 3021 before obtaining the second layer line.

It can be seen that this optional embodiment provides a minimum cost path algorithm, which can divide the required first layer line, second layer line and third layer line in the B-Scan image, and reduces the amount of data required to be calculated when processing the B-Scan image by means of the layering number of retinal tissue layers in the B-Scan image, thereby improving the layering efficiency of the image layering algorithm and improving the accuracy of the layering results.

In another optional embodiment, as shown in FIG4 , the OCT image processing device further includes a second determination module 305, wherein:

The second determination module 305 is used to input the acquired B-Scan image into the pre-trained deep neural network model in the second processing module 303 to determine the target area including the target feature in the B-Scan image before obtaining the output result.

The manner in which the second determining module 305 determines the target area including the target feature in the B-Scan image specifically includes:

The first stratification line and the stratification line located relatively above the second stratification line are shifted upward by a first preset distance in the vertical direction of the image to obtain a first boundary line;

The third layering line is shifted downward by a second preset distance in the vertical direction of the image to obtain a second boundary line;

The area below the first boundary line and above the second boundary line is determined as a target area including target features in the B-Scan image.

It can be seen that this optional embodiment can intelligently determine the target area including the target features, which is beneficial to reducing the amount of data that needs to be processed by the image processing algorithm when subsequent image processing operations are performed on the target area, thereby improving the image stratification efficiency to a certain extent; at the same time, after clarifying the target area including the target features, it is beneficial to reduce the interference of redundant areas that do not include the target features on the image stratification processing, which is beneficial to improving the accuracy of the image stratification results.

In yet another optional embodiment, as shown in FIG4 , the OCT image processing device further includes a judgment module 306 and a third processing module 307, wherein:

The judgment module 306 is used to judge whether there are target pixels falling outside the target area among all the pixels after the second processing module 303 inputs the B-Scan image into a pre-trained deep neural network model to obtain the output result and before the first determination module 304 determines the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all the pixels. When it is determined that there are no target pixels falling outside the target area among all the pixels, the first determination module 304 is triggered to execute an operation of determining the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all the pixels.

The third processing module 307 is used to perform a probability update operation on all target pixel points that fall outside the target area when the judgment module 306 determines that there are target pixel points that fall outside the target area among all the pixel points, so as to update the probabilities corresponding to all the target pixel points, and trigger the first determination module 304 to perform an operation to determine the inter-layer boundary information corresponding to the target feature in the B-Scan image based on the probabilities corresponding to all the pixel points.

It can be seen that this optional embodiment can perform a probability update operation for all target pixel points falling outside the target area, which is beneficial to subsequently determine the inter-layer boundary information corresponding to the target features in the B-Scan image based on the probabilities corresponding to all pixel points, thereby improving the accuracy of the determined inter-layer boundary information.

In this optional embodiment, the manner in which the determination module 306 determines whether there is a target pixel point falling outside the target area among all the pixel points specifically includes:

For each column of all the pixels, it is determined whether there is a target pixel that falls outside the target area in the column of pixels.

And, the third processing module performs a probability update operation on all target pixels falling outside the target area, so as to update the probabilities corresponding to all target pixels falling outside the target area, specifically including:

For each column of all pixels, if there is a target pixel that falls outside the target area in the column of pixels, each target pixel that falls outside the target area in the column of pixels is multiplied by a preset value corresponding to the target pixel to obtain a product result, and the probability corresponding to the target pixel is updated according to the product result.

It can be seen that in this optional embodiment, a step of processing pixel points in columns is provided, which can process pixel points one by one or in columns, thereby improving the processing efficiency of pixel points and improving the image stratification efficiency to a certain extent; in addition, the probability update operation can also reduce the subsequent execution of the operation of determining the inter-layer boundary information corresponding to the target feature in the B-Scan image based on the probabilities corresponding to all the pixel points, and the situation where the probabilities corresponding to all the pixel points include abnormal probabilities.

In another optional embodiment, the first determination module 304 determines the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the corresponding probabilities of all pixels, specifically including:

For each column of pixels among all the pixels, a normalization process is performed on the probability distribution of the pixel columns to obtain a normalized probability distribution of the pixel columns;

For each column of all pixels, a dot product operation is performed between the normalized probability distribution of the column of pixels and the row number distribution corresponding to the column of pixels to obtain the inter-layer distribution result corresponding to the column of pixels;

According to the inter-layer distribution results corresponding to each column of pixels among all pixels, the inter-layer boundary information corresponding to the target features in the B-Scan image is determined.

And, the deep neural network model is trained in the following way:

Acquire a B-Scan image set including annotation information, wherein the annotation information corresponding to each B-Scan image in the B-Scan image set includes label information corresponding to the target feature and boundary information corresponding to the target feature;

Divide the B-Scan image set into a training set and a test set. The training set is used to train the deep neural network model, and the test set is used to verify the reliability of the trained deep neural network model.

Performing a target processing operation on all B-Scan images included in the training set to obtain a processing result, wherein the target processing operation includes at least one of an up-and-down movement process, a left-and-right flip process, an up-and-down inversion process, and a contrast adjustment process;

Input the processed results as input data into a predetermined deep neural network model to obtain output results;

According to the output results, the B-Scan images included in the training set and the boundary information, the joint loss is analyzed and calculated to obtain the joint loss value;

The joint loss value is back-propagated in the deep neural network model, and iterative training is performed for a preset cycle length to obtain a trained deep neural network model;

Among them, the test set is used to verify the reliability of the trained deep neural network model.

It can be seen that the deep neural network model obtained through training, combined with the relevant processing operations (including normalization and dot product operations) performed on the probability distribution corresponding to each column of all pixels in units of columns, determines the inter-layer boundary information corresponding to the target features in the B-Scan image, thereby achieving the purpose of improving the stratification efficiency of OCT images and improving the accuracy of stratification results.

Embodiment 4

Please refer to FIG5 , which is a schematic diagram of the structure of another OCT image processing device disclosed in an embodiment of the present invention. As shown in FIG5 , the OCT image processing device includes:

A memory 401 storing executable program codes;

a processor 402 coupled to the memory 401;

Furthermore, it may also include an input interface 403 and an output interface 404 coupled to the processor 402;

The processor 402 calls the executable program code stored in the memory 401 to execute the steps in the OCT image processing method described in the first embodiment of the present invention or the second embodiment of the present invention.

Embodiment 5

An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer storage medium storing a computer program, and the computer program is operable to enable a computer to execute the steps in the OCT image processing method described in Embodiment 1 or Embodiment 2.

The device embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without paying creative labor.

Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method 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 above technical solution can be essentially or partly contributed to the prior art in the form of a software product, and the computer software product can be stored in a computer storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, a magnetic disk storage, a magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

Finally, it should be noted that the OCT image processing method and device disclosed in the embodiments of the present invention only disclose the preferred embodiments of the present invention, which are only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (9)

1. A method for processing an OCT image, characterized in that the method comprises: Obtain the B-Scan image corresponding to the target feature; Performing image layering processing on the B-Scan image by using a preset image processing algorithm to obtain an initial layering result; Input the B-Scan image into a pre-trained deep neural network model to obtain an output result, wherein the output result includes a probability corresponding to each pixel point corresponding to a target area of the B-Scan image, and the probability corresponding to each pixel point is used to indicate the possibility that each pixel point belongs to a layer boundary between two adjacent layers included in the initial layering result, and the target area is a region including the target feature; Determining the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the corresponding probabilities of all the pixel points; The performing image layering processing on the B-Scan image by using a preset image processing algorithm to obtain an initial layering result includes: Performing filtering processing on the B-Scan image by using a preset filtering function to obtain a filtered image; Calculating the positive gradient of the filtered image in the vertical direction of the image, and constructing a first cost function according to the positive gradient; Determine a first minimum cost path from the left edge to the right edge of the filtered image according to a predetermined path algorithm and the first cost function, and obtain a first layering line; Determine a second minimum cost path from the left edge to the right edge of the filtered image according to the path algorithm and the first cost function, and obtain a second layering line; Calculating the negative gradient of the filtered image in the vertical direction of the image, and constructing a second cost function according to the negative gradient; Determine a search area, where the search area is a lower area corresponding to a lower layer line between the first layer line and the second layer line; Determine a third minimum cost path from the left edge of the search area to the right edge of the search area according to the path algorithm and the second cost function, and perform a smoothing filter operation on the third minimum cost path to obtain a third layering line; determining the first stratification line, the second stratification line and the third stratification line as initial stratification results; And, before determining the second minimum cost path from the left edge to the right edge of the filtered image according to the path algorithm and the first cost function to obtain the second layering line, the method further includes: The first minimum cost path is marked as an unreachable path in the filtered image. 2. The OCT image processing method according to claim 1, characterized in that before inputting the B-Scan image into a pre-trained deep neural network model to obtain an output result, the method further comprises: Determining a target area including the target feature in the B-Scan image; Wherein, determining the target area including the target feature in the B-Scan image comprises: The first stratification line and the second stratification line that are located relatively above each other are shifted upward by a first preset distance in the vertical direction of the image to obtain a first boundary line; The third layering line is shifted downward by a second preset distance in the vertical direction of the image to obtain a second boundary line; An area below the first boundary line and above the second boundary line is determined as a target area including the target feature in the B-Scan image. 3. The OCT image processing method according to claim 2, characterized in that after inputting the B-Scan image into a pre-trained deep neural network model to obtain an output result, before determining the inter-layer boundary information corresponding to the target feature in the B-Scan image based on the probabilities corresponding to all the pixel points, the method further comprises: Determine whether there is a target pixel point falling outside the target area among all the pixel points; When it is determined that none of the target pixel points falls outside the target area among all the pixel points, triggering the operation of determining the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the probabilities corresponding to all the pixel points; When it is determined that there are target pixel points that fall outside the target area among all the pixel points, a probability update operation is performed on all the target pixel points that fall outside the target area to update the probabilities corresponding to all the target pixel points, and trigger the execution of the operation of determining the inter-layer boundary information corresponding to the target feature in the B-Scan image based on the probabilities corresponding to all the pixel points. 4. The OCT image processing method according to claim 3, characterized in that the step of determining whether there is a target pixel point falling outside the target area among all the pixel points comprises: For each column of all the pixel points, determining whether there is a target pixel point in the column that falls outside the target area; And, performing a probability update operation on all the target pixel points falling outside the target area to update the probabilities corresponding to all the target pixel points falling outside the target area includes: For each column of all the pixel points, if there is a target pixel point in the column that falls outside the target area, each target pixel point in the column that falls outside the target area is multiplied by a preset value corresponding to the target pixel point to obtain a product result, and the probability corresponding to the target pixel point is updated according to the product result. 5. The OCT image processing method according to claim 3 or 4, characterized in that the step of determining the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the corresponding probabilities of all the pixel points comprises: For each column of pixel points among all the pixel points, normalizing the probability distribution of the column of pixel points to obtain a normalized probability distribution of the column of pixel points; For each column of pixels among all the pixels, a dot product operation is performed on the normalized probability distribution of the column of pixels and the row number distribution corresponding to the column of pixels to obtain an inter-layer distribution result corresponding to the column of pixels; According to the inter-layer distribution result corresponding to each column of pixel points in all the pixel points, the inter-layer boundary information corresponding to the target feature in the B-Scan image is determined. 6. The method for processing OCT images according to any one of claims 1 to 4, characterized in that the deep neural network model is trained in the following manner: Acquire a B-Scan image set including annotation information, wherein the annotation information corresponding to each B-Scan image in the B-Scan image set includes label information corresponding to the target feature and boundary information corresponding to the target feature; Dividing the B-Scan image set to obtain a training set and a test set, wherein the training set is used to train a deep neural network model, and the test set is used to verify the reliability of the trained deep neural network model; Performing a target processing operation on all B-Scan images included in the training set to obtain a processing result, wherein the target processing operation includes at least one of an up-and-down movement process, a left-and-right flip process, an up-and-down inversion process, and a contrast adjustment process; Inputting the processing result as input data into a predetermined deep neural network model to obtain an output result; Analyze and calculate the joint loss according to the output result, the B-Scan image included in the training set, and the boundary information to obtain a joint loss value; Back-propagating the joint loss value in the deep neural network model, and performing iterative training of a preset cycle length to obtain a trained deep neural network model; The test set is used to verify the reliability of the trained deep neural network model. 7. The method for processing an OCT image according to any one of claims 1 to 4, characterized in that the target feature is a retinal feature. 8. An OCT image processing device, characterized in that the device comprises: An acquisition module is used to acquire a B-Scan image corresponding to the target feature; A first processing module, configured to perform image layering processing on the B-Scan image using a preset image processing algorithm to obtain an initial layering result; A second processing module is used to input the B-Scan image into a pre-trained deep neural network model to obtain an output result, wherein the output result includes a probability corresponding to each pixel point corresponding to a target area of the B-Scan image, and the probability corresponding to each pixel point is used to indicate the possibility that each pixel point belongs to a layer boundary between two adjacent layers included in the initial layering result, and the target area is a region including the target feature; A first determination module is used to determine the inter-layer boundary information corresponding to the target feature in the B-Scan image according to the corresponding probabilities of all the pixel points; The first processing module comprises: A filtering submodule, used for performing filtering processing on the B-Scan image through a preset filtering function to obtain a filtered image; A function construction submodule, used for calculating the positive gradient of the filtered image in the vertical direction of the image, and constructing a first cost function according to the positive gradient; A first determination submodule is used to determine a first minimum cost path from a left edge to a right edge of the filtered image according to a predetermined path algorithm and the first cost function, so as to obtain a first layering line; The first determination submodule is further used to determine a second minimum cost path from the left edge to the right edge of the filtered image according to the path algorithm and the first cost function to obtain a second layering line; The function construction submodule is further used to calculate the negative gradient of the filtered image in the vertical direction of the image, and construct a second cost function according to the negative gradient; A second determination submodule is used to determine a search area, where the search area is a lower area corresponding to a lower layer line between the first layer line and the second layer line; The first determination submodule is further used to determine a third minimum cost path from the left edge of the search area to the right edge of the area according to the path algorithm and the second cost function, and perform a smoothing filter operation on the third minimum cost path to obtain a third layer line; The second determining submodule is further configured to determine the first stratification line, the second stratification line and the third stratification line as initial stratification results; And, the first processing module also includes: A marking submodule is used to mark the first minimum cost path as an unreachable path in the filtered image before the first determination submodule determines the second minimum cost path from the left edge to the right edge of the filtered image according to the path algorithm and the first cost function to obtain the second layer line. 9. An OCT image processing device, characterized in that the device comprises: A memory storing executable program code; a processor coupled to the memory; an input interface and an output interface coupled to the processor; The processor calls the executable program code stored in the memory to execute the OCT image processing method according to any one of claims 1 to 7.