JP2022548453A - Image segmentation method and apparatus, electronic device and storage medium - Google Patents

Abstract

The present disclosure relates to image segmentation methods and apparatuses, electronic devices, and storage media. The image segmentation method comprises the steps of: predicting pixels belonging to a target object in an image to be processed; obtaining a preliminary segmented image corresponding to the image to be processed; adjusting pixel values of predicted pixels not belonging to the target object in closed regions included in edges of the target object in the preliminary segmented image, and adjusting pixel values of pixels not belonging to the target object; and obtaining a result.

Description

(Cross reference to related application)
This disclosure is proposed based on and claims priority from a Chinese patent application numbered 202010827077.1 and filed on Aug. 17, 2020, and the entire contents of the Chinese patent are incorporated herein by reference into the present disclosure.

TECHNICAL FIELD The present disclosure relates to the technical field of image processing, and more particularly to an image segmentation method and apparatus, an electronic device, and a storage medium.

Image segmentation refers to techniques and processes for dividing an image into regions with specific, unique properties to present targets of interest. Image segmentation is an important step from image processing to image analysis. Image segmentation methods in the related art are mainly classified into threshold-based segmentation methods, region-based segmentation methods, edge-based segmentation methods, segmentation methods based on certain theories, and so on.

The present disclosure provides an image segmentation method and apparatus, an electronic device and a storage medium.

An image segmentation method according to one aspect of the present disclosure includes:
predicting pixels belonging to a target object in a pending image and obtaining a rudimentary segmented image corresponding to the pending image;
calculating predicted pixel values of pixels not belonging to the target object in a closed region included in the edge of the target object in the preliminary segmented image based on the edge information of the target object in the image to be processed; adjusting to obtain a first segmentation result corresponding to the pending image.

Predicting pixels belonging to a target object in an image waiting to be processed, obtaining a rudimentary divided image corresponding to the image awaiting processing, and based on edge information of the target object in the image awaiting processing, in the rudimentary divided image, adjusting pixel values of predicted pixels not belonging to the target object in a closed region included in an edge of the target object to obtain a first segmentation result corresponding to the pending image; Accurate and robust segmentation results can be obtained.

In one possible embodiment, the pixel value of the predicted pixel belonging to the target object in the rudimentary segmented image is a first preset value and the predicted pixel belonging to the target object. the pixel values of the pixels that are not at a second preset value;
calculating predicted pixel values of pixels not belonging to the target object in a closed region included in the edge of the target object in the preliminary segmented image based on the edge information of the target object in the image to be processed; The step of adjusting and obtaining a first segmentation result corresponding to the pending image includes:
adjusting pixel values of closed regions whose pixel values in said preliminary segmented image are said second preset value to said first preset value to obtain a filled preliminary segmented image; When,
adjusting pixel values of the filled rudimentary segmented image based on the edge information of the target object in the pending image to obtain the first segmentation result corresponding to the pending image; include.

In the embodiment, the pixel values of the closed regions whose pixel values in the preliminary segmented image are the second preset value are adjusted to the first preset value, and the filled preliminary segmentation is performed. Acquiring the image causes the first segmentation result corresponding to the pending image to cover the interior of the organ of the target object, e.g., the lung parenchyma, such as in the lung, the interior of the gastrointestinal tract (e.g., the gastrointestinal tract), and the like. can be made That is, by adopting the above-described embodiment, it is possible to fill the hollow that leaked inside the target object (for example, inside the human body) after image division. adjusting the pixel values of the filled rudimentary segmented image according to the edge information of the target object in the to-be-processed image to obtain a first segmentation result corresponding to the to-be-processed image; The probability that background parts in the image (ie parts that do not belong to the target object) will be split into those that belong to the target object can be reduced.

In one possible embodiment, pixel values of closed areas whose pixel values in said rudimentary segmented image are said second preset value are adjusted to said first preset value and filled. The step of obtaining a rudimentary segmented image includes:
splicing a preset width edge around the rudimentary segmented image to obtain a spliced rudimentary segmented image, the pixels of the pixels on the spliced preset width edge; the value is the second preset value;
selecting pixels at image edges of the spliced primitive segmented image as seed points and performing a flooding filling operation on the spliced primitive segmented image to obtain the filled primitive segmented image; and including.

Seeds on which a flood-filling operation is performed by splicing a preset width edge around the perimeter of the rudimentary segmented image and selecting pixels at image edges of the spliced rudimentary segmented image as seed points. It can be ensured that the points belong to the background portion (i.e. the portion not belonging to the target object), so that the first segmentation result corresponding to the pending image covers the interior of the target object's organs. and get more accurate segmentation results.

In one possible embodiment, adjusting the pixel values of the filled rudimentary split image based on the edge information of the target object in the to-be-processed image to generate the first split corresponding to the to-be-processed image. The step to get the result is
determining the maximum connected domain included in the edges of the target object in the filled rudimentary segmented image based on the edge information of the target object in the to-be-processed image;
adjusting pixel values of pixels other than the largest connected domain in the filled rudimentary segmented image to the second preset value to obtain the first segmentation result corresponding to the pending image; and including.

Since this embodiment can remove false positive regions that are not connected to the target object, the probability of background parts being erroneously divided into those belonging to the target object can be greatly reduced, thereby improving image segmentation. Accuracy can be improved. For example, if the target object is a human body, this example can remove false-positive regions that are not connected to the human body, thereby erroneously dividing background parts (e.g. bedboards) into those belonging to the human body. can greatly reduce the chances of it happening.

In one possible embodiment, after obtaining a first segmentation result corresponding to the pending image, the image segmentation method comprises:
obtaining an image adjacent to the to-be-processed image and a second segmentation result corresponding to the adjacent image;
adjusting the first segmentation result based on the pixel value of the pixel at the same position in the process-waiting image and the adjacent image, and the second segmentation result, and generating a third segment corresponding to the process-waiting image; and obtaining a split result.

This embodiment can ensure the continuity of the pending image and the second segmentation result, which helps to obtain a smoother and more accurate three-dimensional segmentation result. For example, if the target object is a human body, it can ensure continuity between the pending image and the human body in adjacent images, which helps to obtain a smoother and more accurate 3D human body segmentation result. For example, the embodiment can be adopted to obtain a segmentation result corresponding to each CT image in a CT image sequence, thereby obtaining a smoother and more accurate 3D human body segmentation result.

In one possible embodiment, the first segmentation result is adjusted based on pixel values of co-located pixels in the to-be-processed image and the adjacent image, and the second segmentation result, and the processing The step of obtaining a third division result corresponding to the waiting image includes:
Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. and adjusting the first segmentation result to obtain the third segmentation result corresponding to the pending image.

the first segmentation result based on pixels belonging to the target object in the second segmentation result and located at the same position as the to-be-processed image and having a pixel value difference equal to or less than a third preset value; and obtaining a third segmentation result corresponding to the to-be-processed image, based on segmentation results corresponding to pixels in the adjacent image that are relatively related to the to-be-processed image, the to-be-processed image can be adjusted, which helps improve the accuracy of the final segmentation result corresponding to the to-be-processed image.

In one possible embodiment, the difference between the pixel values of said adjacent images belonging to said target object in said second segmentation result and co-located with said to-be-processed image is a third preset value. adjusting the first segmentation result based on pixels that are less than or equal to a value to obtain the third segmentation result corresponding to the pending image,
obtaining a first set of pixels based on a third preset pixel difference between co-located pixel values in the pending image and the adjacent image;
obtaining a second set of pixels based on pixels belonging to the target object in the second segmentation result from the first set of pixels;
adjusting pixels of the second set of pixels in the first segmentation result to belong to the target object to obtain the third segmentation result corresponding to the pending image.

obtaining a first set of pixels based on pixels for which a difference in pixel value at the same position in the pending image and the adjacent image is less than or equal to a third preset value, and obtaining the first set of pixels; obtain a second set of pixels based on the pixels belonging to the target object in the second segmentation result, and determine the pixels of the second pixel set in the first segmentation result belonging to the target object and obtaining a third segmentation result corresponding to the to-be-processed image, based on pixels belonging to the target object in the second segmentation result and relatively related to the to-be-processed image. can adjust the first segmented result corresponding to the pending image, thereby helping to improve the accuracy of the final segmented result corresponding to the pending image.

In one possible embodiment,
Prior to the step of predicting pixels belonging to a target object in a pending image, the image segmentation method comprises training a neural network based on training images and labeled data of the training images, wherein: wherein the labeled data includes true values of pixels belonging to the target object in the training images;
The step of predicting pixels belonging to a target object in a pending image and obtaining a rudimentary segmented image corresponding to the pending image includes: inputting the pending image to the neural network; and obtaining a rudimentary segmented image corresponding to the to-be-processed image based on the information of pixels belonging to the target object in the to-be-processed image. include.

In this embodiment, the neural network predicts the portion belonging to the target object in the to-be-processed image. For example, if the awaiting image to be processed is a CT image and the target object is a human body, the embodiment removes various bedboards in the CT image, i.e., does not focus on parts other than the human body, Since the segmentation that emphasizes the human body part in 2 is not considered, the accuracy and robustness of the segmentation result in a large amount of irregular bedboard data can be ensured. In other words, even when an irregularly shaped bedboard is included in an image waiting to be processed, an accurate and robust segmentation result can be obtained by adopting this embodiment.

In one possible embodiment,
the training images are computed tomography (CT) images;
The step of training a neural network based on training images and labeled data of the training images includes normalizing pixel values of the training images according to a preset CT value range to obtain normalized training images. and training the neural network based on the normalized training images and the labeled data of the training images.

In this embodiment, the pixel values of the training images are normalized according to a preset CT value range, normalized training images are obtained, and the normalized training images and the training images labeled Training the neural network based on data helps reduce the computational complexity of the neural network and improve the convergence speed of the neural network.

An image segmentation method according to one aspect of the present disclosure includes:
predicting pixels belonging to a target object in a pending image and obtaining a rudimentary segmented image corresponding to the pending image;
obtaining an image adjacent to the to-be-processed image and a second segmentation result corresponding to the adjacent image;
adjusting the preliminary divided image based on pixel values of pixels at the same position in the to-be-processed image and the adjacent image, and the result of the second division, and forming a fourth division corresponding to the to-be-processed image; and obtaining a result.

In one possible embodiment, the preliminary segmented image is adjusted based on pixel values of co-located pixels in the to-be-processed image and the adjacent image, and the second segmentation result, and The step of obtaining a fourth segmentation result corresponding to the image includes:
Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. and adjusting the preliminary segmented image to obtain the fourth segmented result corresponding to the pending image.

In one possible embodiment, the difference between the pixel values of said adjacent images belonging to said target object in said second segmentation result and co-located with said to-be-processed image is a third preset value. adjusting the rudimentary segmented image based on pixels that are less than or equal to a value to obtain a fourth segmented result corresponding to the pending image,
obtaining a first set of pixels based on pixels whose co-located pixel value difference in the pending image and the adjacent image is less than or equal to a third preset value;
obtaining a second set of pixels based on pixels belonging to the target object in the second segmentation result from the first set of pixels;
adjusting the pixels of the second set of pixels in the preliminary segmented image to those belonging to the target object to obtain a fourth segmented result corresponding to the pending image.

An image segmentation device according to one aspect of the present disclosure includes:
a first segmentation unit configured to predict pixels belonging to a target object in a pending image and obtain a rudimentary segmented image corresponding to the pending image;
calculating predicted pixel values of pixels not belonging to the target object in a closed region included in the edge of the target object in the preliminary segmented image based on the edge information of the target object in the image to be processed; a first adjuster configured to adjust and obtain a first segmentation result corresponding to the to-be-processed image.

In one possible embodiment, the pixel value of the predicted pixel belonging to the target object in the rudimentary segmented image is a first preset value and the predicted pixel belonging to the target object. the pixel values of the pixels that are not at a second preset value;
The first adjustment module includes:
adjusting the pixel values of the closed regions whose pixel values in the preliminary segmented image are the second preset value to the first preset value to obtain a filled preliminary segmented image;
adapted to adjust pixel values of the filled preliminary segmented image based on edge information of the target object in the to-be-processed image to obtain a first segmentation result corresponding to the to-be-processed image. .

In one possible embodiment, said first adjustment module comprises:
splicing a preset width edge around the rudimentary segmented image to obtain a spliced rudimentary segmented image, wherein pixel values of pixels on the spliced rudimentary segmented image are obtained from a preset value of 2;
selecting a pixel at an image edge of the spliced rudimentary segmented image as a seed point, and performing a flooding filling operation on the spliced rudimentary segmented image to obtain a filled rudimentary segmented image; Configured.

In one possible embodiment, said first adjustment module comprises:
determining the maximum connected domain included in the edge of the target object in the filled preliminary segmented image based on the edge information of the target object in the pending image;
adjusting pixel values of pixels other than the largest connected domain in the filled preliminary segmented image to the second preset value to obtain a first segmentation result corresponding to the pending image; Configured.

In one possible embodiment, the image segmentation device comprises:
a second acquisition module configured to acquire an image adjacent to the pending image and a second segmentation result corresponding to the adjacent image;
adjusting the first segmentation result based on the pixel value of the pixel at the same position in the process-waiting image and the adjacent image, and the second segmentation result, and generating a third segment corresponding to the process-waiting image; and a third adjustment module configured to obtain the split result.

In one possible embodiment, said third adjustment module comprises:
Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. , adjusting the first segmentation result to obtain a third segmentation result corresponding to the to-be-processed image.

In one possible embodiment, said third adjustment module comprises:
obtaining a first set of pixels based on pixels for which a difference between co-located pixel values in the pending image and the adjacent image is less than or equal to a third preset value;
Obtaining a second pixel set based on pixels belonging to the target object in the second segmentation result from the first pixel set;
It is configured to adjust the pixels of the second set of pixels in the first segmented result to those belonging to the target object to obtain a third segmented result corresponding to the pending image.

In one possible embodiment,
The image segmentation device is a training module configured to train a neural network based on training images and labeled data of the training images, wherein the labeled data of the training images includes the further comprising a training module containing true values of pixels belonging to the target object;
The first segmentation module inputs an image to be processed into the neural network, predicts information of pixels belonging to the target object in the image to be processed by the neural network, and predicts information of pixels belonging to the target object in the image to be processed by the neural network. It is configured to obtain a rudimentary segmented image corresponding to the to-be-processed image based on the pixel information.

In one possible embodiment,
the training images are computed tomography (CT) images;
The training module normalizes the pixel values of the training images according to a preset CT value range, obtains normalized training images, and labels the normalized training images and the training images. configured to train the neural network based on data;

An image segmentation device according to one aspect of the present disclosure includes:
a second segmentation unit configured to predict pixels belonging to a target object in a pending image and obtain a rudimentary segmented image corresponding to the pending image;
a first acquisition unit configured to acquire an image adjacent to the to-be-processed image and a second division result corresponding to the adjacent image;
adjusting the preliminary divided image based on pixel values of pixels at the same position in the to-be-processed image and the adjacent image, and the result of the second division, and forming a fourth division corresponding to the to-be-processed image; a second coordinator configured to obtain a result.

In one possible embodiment, said second coordination module comprises:
Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. , adjusting the preliminary segmented image to obtain a fourth segmented result corresponding to the to-be-processed image.

In one possible embodiment, said second coordination module comprises:
obtaining a first set of pixels based on pixels for which a difference between co-located pixel values in the pending image and the adjacent image is less than or equal to a third preset value;
Obtaining a second pixel set based on pixels belonging to the target object in the second segmentation result from the first pixel set;
adjusting pixels of the second set of pixels in the preliminary segmented image to those belonging to the target object to obtain a fourth segmented result corresponding to the pending image.

An electronic device according to one aspect of the present disclosure comprises one or more processors and a memory storing executable instructions, the one or more processors storing executable instructions stored in the memory. to execute the above image segmentation method.

A computer-readable storage medium according to one aspect of the present disclosure stores computer program instructions for causing a processor to perform the above method.

In an embodiment of the present disclosure, predicting pixels belonging to a target object in an image to be processed, obtaining a rudimentary divided image corresponding to the image to be processed, based on edge information of the target object in the image to be processed, Adjusting pixel values of predicted pixels not belonging to the target object in closed regions included in edges of the target object in the preliminary segmented image, and obtaining a first segmentation result corresponding to the pending image. By obtaining , a more accurate and robust segmentation result can be obtained.

A computer program according to one aspect of the present disclosure includes computer readable code for causing a processor in the electronic device to perform the above image segmentation method when executed on an electronic device.

It should be understood that the foregoing general description and the following detailed description are exemplary and interpretative only and are not restrictive of the present disclosure.

Other features and aspects of the present disclosure will become apparent from the detailed description of exemplary embodiments based on the drawings that follow.

FIG. 1 is a schematic diagram 1 showing an application scene of an image segmentation method according to an embodiment of the present disclosure; FIG. 2 is a schematic diagram 2 showing an application scene of an image segmentation method according to an embodiment of the present disclosure; 4 is a flow chart illustrating an image segmentation method according to an embodiment of the present disclosure; 1 is a schematic diagram illustrating a U-shaped convolutional neural network according to embodiments of the present disclosure; FIG. FIG. 4 is a schematic diagram illustrating splicing a preset width edge around a rudimentary segmented image to obtain a spliced rudimentary segmented image according to an embodiment of the present disclosure; 4 is a flowchart illustrating an image segmentation method according to another embodiment of the present disclosure; 1 is a block diagram illustrating an image segmentation device according to an embodiment of the present disclosure; FIG. FIG. 4 is another block diagram illustrating an image segmentation device according to an embodiment of the present disclosure; 8 is a block diagram illustrating electronic device 800 according to an embodiment of the disclosure. FIG. 19 is a block diagram illustrating electronic device 1900 according to an embodiment of the disclosure. FIG.

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, illustrate the technical solution of the present disclosure. used to interpret proposals.

Various illustrative embodiments, features, and aspects of the disclosure are described in detail below with reference to the drawings. The same reference numerals in the figures represent elements of the same or similar function. Although the drawings depict various aspects of the illustrative embodiments, the drawings are not necessarily drawn to scale unless otherwise noted.

As used herein, the term "exemplary" means "serving as an example, example, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as superior or better than other embodiments.

The term "and/or" is used herein only to describe the relationship of related objects and indicates that there can be three types of relationships, e.g., A and/or B exists alone, A and B exist simultaneously, and B exists alone. In addition, the term "at least one type" as used herein indicates any one type out of a plurality of types or any combination of at least two types out of a plurality of types, e.g. Including at least one type can indicate including any one or more elements selected from the set consisting of A, B and C.

Also, in order to better explain the present disclosure, numerous details are given in the following specific embodiments. A person of ordinary skill in the art should understand that the present disclosure may equally be practiced without certain details. In some examples, methods, means, elements and circuits that are well known to those of ordinary skill in the art have not been described in detail so as not to obscure the subject matter of the present disclosure.

In the process of performing Computed Tomography (CT) on the human body, the CT machine bedboard becomes an artifact in the scanned CT image sequence. This artifact causes significant interference in the three-dimensional visualization of the human body by computer-aided software (ie, obtaining a three-dimensional human body model from a CT image sequence) and subsequent processing processes. This is because bedboards of various shapes occlude the human body during 3D visualization, and when organs inside the human body are divided, some malformed bedboards outside the human body may be identified as false positives. This is because

In the related art, when performing body segmentation on a CT image, mainly through threshold and morphological operations, the bedboard in the CT image is removed and the body part in the CT image is retained. Generally, the shape of the bedboard, the CT value of the bedboard in the CT image, and the uniformity of the CT value of the bedboard in the CT image are significantly different from the human body, and may be removed by the method of threshold and morphology manipulation. However, in some abnormal situations, related techniques fail to obtain accurate segmentation results. For example, a curved cortical bedboard that is in close contact with the human body is in close contact with the human body in the CT image, has no clear boundary, has relatively close CT values, and is difficult to separate from the human body. In addition, for example, there are baffles on both sides of the bedboard, the arms of the human body are placed outside the baffles, and the baffles are sandwiched between both sides of the body. be. Here, the CT value is a calculation unit for measuring the density of a certain local tissue or organ of the human body, and is also called Hounsfield Unit (HU).

To solve the above technical problems, embodiments of the present disclosure predict pixels belonging to a target object in a pending image, obtain a rudimentary segmented image corresponding to the pending image, and process the Adjusting pixel values of predicted pixels not belonging to the target object in closed regions included in edges of the target object in the preliminary segmented image based on the edge information of the target object in the waiting image. , an image segmentation method and apparatus, an electronic device, and a storage medium capable of obtaining a more accurate and robust segmentation result by obtaining a first segmentation result corresponding to the image waiting to be processed.

The execution body of the image segmentation method may be an image segmentation device. For example, the image segmentation method may be performed by a terminal device or a server or other processing device. Here, the terminal device includes a user device (UE: User Equipment), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (PDA: Personal Digital Assistant), a handheld device, a computing device, an in-vehicle device. Or it may be a wearable device or the like. In some possible embodiments, the image segmentation method may be implemented by a processor calling computer readable instructions stored in memory.

Application scenes of the image segmentation method according to the embodiment of the present disclosure will be described below.

In one possible embodiment, referring to schematic diagram 1 showing the application scene of the image segmentation method shown in FIG. Thus, the image dividing device 11 can collect images waiting to be divided by the image collecting device 12 and divide the images waiting to be divided by the processing device 12 to obtain the first division result. For example, the image segmentation device may be implemented as a CT machine, acquires CT images waiting to be segmented by a CT scanner, and performs image segmentation processing on the acquired CT images awaiting segmentation.

In another possible embodiment, referring to schematic FIG. 2 showing the application scene of the image segmentation method shown in FIGS. The collected images waiting to be divided can be received, and in this way, the image dividing device 10 can perform division processing on the received images waiting to be divided to obtain the first division result. For example, the image segmentation device may be implemented as a smart phone, and the smart phone can collect the split-awaiting CT images sent by the CT machine over the network, thus the smart phone can Image segmentation processing can be performed on CT images waiting to be segmented.

Based on the above application scene, an image segmentation method according to an embodiment of the present disclosure is described, and FIG. 1 shows a flow chart of the image segmentation method according to an embodiment of the present disclosure. As shown in FIG. 2, the image segmentation method includes steps S11 and S12.

In step S11, a pixel belonging to a target object in an image waiting to be processed is predicted, and a rudimentary segmented image corresponding to the image waiting to be processed is obtained.

In embodiments of the present disclosure, a pending image may represent an image for which image segmentation needs to be performed. The image to be processed may be a two-dimensional image or a three-dimensional image. In some possible embodiments, the pending image may be a medical image. For example, the processing-waiting image may be a CT image, an MRI (Magnetic Resonance Imaging) image, or the like. Of course, the awaiting image can be any image requiring image segmentation other than a medical image.

In embodiments of the present disclosure, the target image may represent an object requiring segmentation. For example, the target object may be a human body, an animal body, a human body organ, an animal body organ, and the like.

Embodiments of the present disclosure can predict whether each pixel in the pending image belongs to the target object. For example, the probability that each pixel in the to-be-processed image belongs to the target object can be predicted. For any pixel in the to-be-processed image, if the probability that the pixel belongs to the target object is greater than or equal to a preset threshold, the pixel may be determined to belong to the target object, and the pixel belongs to the target object. If the probability is less than a preset threshold, the pixel may be determined not to belong to the target pixel. For example, the preset threshold may be 0.5.

In embodiments of the present disclosure, a binarized rudimentary segmented image corresponding to the to-be-processed image can be obtained based on the predicted pixels belonging to the target object in the to-be-processed image. Here, the size of the rudimentary divided image may be the same as the size of the to-be-processed image. For example, if the height of the pending image is H and the width is W, the height of the preliminary divided image is also H and the width is W as well. In one possible embodiment, the pixel value of the predicted pixel belonging to the target object in the rudimentary segmented image is a first preset value and the predicted pixel belonging to the target object. The pixel value of the non-selected pixel is a second preset value, and the first preset value is not equal to the second preset value. For example, for any pixel in the to-be-processed image, if the pixel is predicted to belong to the target object, the pixel value of the pixel in the rudimentary segmented image is a first preset value, and the pixel is predicted not to belong to the target object, the pixel value of that pixel in the rudimentary segmented image is a second preset value. For example, the first preset value is 1 and the second preset value is 0, i.e. the pixel value of the predicted pixel belonging to the target object in the rudimentary segmented image is 1. Yes, and predicted pixel values that do not belong to the target object are zero. In embodiments of the present disclosure, the first preset value and the second preset value are not limited, and the first preset value and the second preset value may be different . Also, for example, the first preset value may be 0 and the second preset value may be 255.

In one possible embodiment, prior to the step of predicting pixels belonging to a target object in pending images, said image segmentation method comprises training a neural network based on training images and labeled data of said training images. wherein the labeled data of the training images includes true values of pixels belonging to the target object in the training images; and predicting pixels belonging to the target object in the pending images. and obtaining a rudimentary segmented image corresponding to the awaiting image includes inputting the awaiting image to the neural network and predicting information of pixels belonging to a target object in the awaiting image by the neural network. and obtaining a rudimentary segmented image corresponding to the to-be-processed image based on information of pixels belonging to a target object in the to-be-processed image.

As an example of such an embodiment, the labeled data of the training images may include a mask corresponding to the training images, the size of the mask corresponding to the training images being the same as the size of the training images. good too. If the true value of any pixel in the training image belongs to the target object, then in the mask corresponding to the training image, the pixel value of that pixel may be a first preset value. , for example, the first preset value may be 1, and in the training image, if the true value of the pixel does not belong to the target object, the mask corresponding to the training image is , the pixel value of the pixel may be a second preset value, for example, the second preset value may be zero. Of course, the labeled data of the training images are not limited to being represented by masks. For example, the labeled data of the training images may be represented in a matrix, table, or other manner.

In such an embodiment, the training images may be input to the neural network, and a predicted segmentation result of the training image may be output via the neural network, wherein the predicted segmentation result of the training image may be obtained from each of the training images. obtaining a loss function value corresponding to the training image, based on the labeled data of the training image and the predicted segmentation result of the training image; training the neural network based on the value of the loss function corresponding to .

As an example of such an embodiment, a Dice loss function value can be obtained based on the predicted segmentation result of the training image obtained by the neural network and the labeled data of the training image. For example, the predicted segmentation result of the training image obtained by the neural network is P, the labeled data of the training image is M, and the value of the dice loss function

である。他の例では、クロスエントロピー損失関数などの損失関数も使用することができる。

is. In other examples, loss functions such as cross-entropy loss functions can also be used.

As an example of such an embodiment, the value of the loss function can be passed to each parameter of the neural network layer by layer by back derivation, and adaptive matrix estimation (Adam) (e.g. with a learning rate of 0.0003 The parameters of the neural network can be updated by optimizers such as Stochastic Gradient Descent (SGD).

As an example of such an embodiment, the information of pixels belonging to a target object in the to-be-processed image predicted by the neural network may include a probability that each pixel in the to-be-processed image belongs to the target object. In this example, the step of obtaining a rudimentary divided image corresponding to the to-be-processed image based on the information of the pixels belonging to the target object in the to-be-processed image includes, for any pixel, the pixel in the to-be-processed image is if the probability of belonging to the target object is greater than or equal to a preset threshold, the pixel value of the pixel in the rudimentary segmented image corresponding to the awaiting image becomes a first preset value; If the probability that the pixel in the image belongs to the target object is less than a preset threshold, the pixel value of the pixel in the preliminary segmented image corresponding to the pending image is a second preset value. and can include

As another example of this embodiment, the information of pixels belonging to the target object in the to-be-processed image predicted by the neural network may include position information of pixels belonging to the target object in the to-be-processed image. In this example, the step of obtaining a rudimentary segmented image corresponding to the to-be-processed image based on the information of the pixels belonging to the target object in the to-be-processed image includes, for any pixel, the target object in the to-be-processed image. when the position information of the pixel belonging to the image includes the position of the pixel, the pixel value of the pixel in the elementary divided image corresponding to the image to be processed becomes a first preset value; if the position information of the pixel belonging to the target object in does not include the position of the pixel, the pixel value of the pixel in the rudimentary segmented image corresponding to the to-be-processed image is a second preset value; can include

In this embodiment, the neural network predicts the portion belonging to the target object in the to-be-processed image. For example, if the awaiting image to be processed is a CT image and the target object is a human body, the embodiment removes various bedboards in the CT image, i.e., does not focus on parts other than the human body, Since the segmentation that emphasizes the human body part in 2 is not considered, the accuracy and robustness of the segmentation result in a large amount of irregular bedboard data can be ensured. In other words, even when an irregularly shaped bedboard is included in an image waiting to be processed, an accurate and robust segmentation result can be obtained by adopting this embodiment.

As an example of this embodiment, the neural network may be a neural network based on deep learning. For example, the neural network may be a U-shaped convolutional neural network. FIG. 3 is a schematic diagram illustrating a U-shaped convolutional neural network according to an embodiment of the present disclosure; In FIG. 3, the direction of data flow is from left to right, and the U-shaped convolutional neural network includes compression and decompression processes. As shown in FIG. 3, after the image to be processed is trimmed or zoomed to a size of 512×512, it is input to the U-shaped convolutional neural network, and the human body part in the image to be processed is fitted by the U-shaped convolutional neural network, Finally, a rudimentary split image can be output. In the example shown in Figure 3, the convolution-normalization-activation-pooling operation can be performed four times on the pending image, each convolution doubling the number of channels in the image, and when pooling The size is halved, the number of channels in the image is increased from 32 to 256, the image size is reduced from 512×512 to 64×64, and the “convolution-normalization-activation-pooling” operation is reduced to 4 times to restore the image to its original size and before upsampling, we need to merge the feature maps of the same size from the previous compression process, and each convolution halves the number of channels, about merging A concatenate operation can be employed, and a single convolution and activation operation restores the number of channels in the image to 1 and normalizes the image. Here, 'convolution-normalization-activation' can be replaced with Residual Block, Deep Convolution Module, Dense Block, and so on. For pooling, max pooling or average pooling can be adopted, and can be replaced by a convolutional layer with a step size of two.

In one example, the training images are 2D CT images and the neural network is a 2D convolutional neural network.

As an example of such an embodiment, the training images may be amplified. For example, randomly zoom the training images by a factor of 0.6-1.4 and then crop the zoomed images from the center by a size of 512x512 to get the same size training images at different zoom scales. be able to. A similar operation is performed on the mask corresponding to the training image accordingly.

As an example of such an embodiment, the training images can be split into a training set and a validation set. For example, the training images can be split into a training set and a validation set in a 4:1 ratio.

As an example of such an embodiment, training images can be used to repeatedly train the neural network until the loss in the neural network's validation set drops below 0.03.

Since the related art divides the image by operations such as morphology, there are a large number of hyperparameters, such as the threshold selected when binarizing, the number of on/off operations, and the structure selected when eroding/expanding. size etc., and for different parts of the human body (head, torso, hands, pedals) the threshold needs to be changed in order to obtain a normal split. In this embodiment, by segmenting the target object in the training image by means of a neural network, it can be widely applied to similar tasks and does not need to set superparameters, so it is more robust.

In one example of such an embodiment, the training images are computed tomography (CT) images, and the step of training a neural network based on training images and labeled data of the training images comprises: normalizing the pixel values of the training images to obtain normalized training images according to; and training the neural network based on the normalized training images and the labeled data of the training images. and a step.

In one example, a preset CT value range can be determined according to the CT value range of the target object. For example, if the target object is a human body, the preset CT value range can be set to [-500, 1200] according to the CT value range of the organs of the human body.

In one example, the step of normalizing the pixel values of the training image according to a preset CT value range to obtain a normalized training image includes, for any pixel in the training image, the preset preprocessing the pixel value of the pixel according to a set CT value range to obtain a preprocessed pixel value of the pixel, wherein the preprocessed pixel value of the pixel is the preprocessed pixel value of the pixel; within a set CT value range; and using a ratio of a first difference and a second difference as a normalized pixel value for said pixel, wherein said first difference is said pixel's equal to the difference between the preprocessed pixel value and the lower limit of the preset CT number range, wherein the second difference is the upper limit of the preset CT number range and the preprocessed pixel of the pixel. and a step equal to the difference between the values. For example, when the preprocessed pixel value of the pixel is h, the lower limit of the preset CT value range is _hmin , and the upper limit of the preset CT value range is _hmax . , the normalized pixel value of the pixel is

に等しくすることができる。前記トレーニング画像における各画素の正規化された画素値に基づいて、正規化されたトレーニング画像を取得することができる。即ち、前記正規化されたトレーニング画像では、任意の画素の画素値は当該画素の正規化された画素値である。

can be equal to A normalized training image can be obtained based on the normalized pixel value of each pixel in the training image. That is, in the normalized training images, the pixel value of any pixel is the normalized pixel value of that pixel.

Here, for any pixel in the training image, the step of preprocessing the pixel value of the pixel according to the preset CT value range and obtaining the preprocessed pixel value of the pixel includes: for any pixel in the image, if the pixel value of said pixel is less than the lower limit of said preset CT value range, said lower limit may be used as the preprocessed pixel value of said pixel; , if the pixel value of said pixel is greater than the upper limit of said preset CT value range, said upper limit can be used as a preprocessed pixel value of said pixel; is within the preset CT value range, the pixel value of the pixel can be used as the preprocessed pixel value of the pixel. For example, the preset CT value range is [−500, 1200], the preset CT value range has a lower limit of −500, and the preset CT value range has an upper limit of 1200. be. If a pixel in the training image has a pixel value of −505, then −500 can be used as the preprocessed pixel value of that pixel, and if a pixel in the training image has a pixel value of 1250, then 12000 can be used as the preprocessed pixel value of the pixel, and if the pixel value of a pixel in the training image is 800, 800 can be used as the preprocessed pixel value of the pixel. .

In this example, according to a preset CT value range, normalize the pixel values of the training image, obtain a normalized training image, and obtain the normalized training image and the labeled data of the training image. By training the neural network based on , it helps to reduce the computational complexity of the neural network and improve the convergence speed of the neural network.

In step S12, predicted pixels not belonging to the target object in a closed region included in the edge of the target object in the preliminary segmented image based on the edge information of the target object in the awaiting image; is adjusted to obtain a first division result corresponding to the awaiting image.

In embodiments of the present disclosure, an edge detection method may be used to determine edge information of the target object in the to-be-processed image. For example, edge detection methods such as Canny algorithm, Sobel algorithm, etc. can be used to determine the edge information of the target object in the pending image. Here, the edge information of the target object in the image waiting to be processed can include position information of pixels belonging to the edge of the target object in the image waiting to be processed.

In one possible embodiment, the first segmentation result can be used as the final segmentation result corresponding to the to-be-processed image.

In an embodiment of the present disclosure, predicting pixels belonging to a target object in an image to be processed, obtaining a rudimentary divided image corresponding to the image to be processed, based on edge information of the target object in the image to be processed, Adjusting pixel values of predicted pixels not belonging to the target object in closed regions included in edges of the target object in the preliminary segmented image, and obtaining a first segmentation result corresponding to the pending image. can divide the pixels inside the organs of the target object into those belonging to the target object when the target object is a human or animal body, thereby obtaining a more accurate and robust segmentation result be able to. For example, the to-be-processed image is a CT image, the target object is a human body, and segmenting the CT image using the image segmentation method provided by the embodiments of the present disclosure accurately identifies the human body part in the CT image. so that extracorporeal interferers (eg, bedboards, ventilator pipelines, head fixation devices, etc.) in CT images can be accurately removed.

In one possible embodiment, based on the edge information of the target object in the to-be-processed image, in the rudimentary segmented image, the predicted target object in a closed region contained in the edge of the target object. The step of adjusting pixel values of non-participating pixels to obtain a first segmentation result corresponding to the pending image includes: closing pixel values in the preliminary segmented image being the second preset value; adjusting pixel values of a region to the first preset value to obtain a filled rudimentary segmented image; adjusting pixel values of a preliminary segmented image to obtain a first segmented result corresponding to the pending image.

Since the interior of some organs (e.g., the interior of the lungs, the interior of the digestive tract) contains low-density air, and the exterior of the target object (e.g., the human body) is also air, these The interior of the organ may be divided into those belonging to the background part. In the embodiment, the pixel values of the closed regions whose pixel values in the preliminary segmented image are the second preset value are adjusted to the first preset value, and the filled preliminary segmentation is performed. Acquiring the image causes the first segmentation result corresponding to the pending image to cover the interior of the organ of the target object, e.g., the lung parenchyma, such as in the lung, the interior of the gastrointestinal tract (e.g., the gastrointestinal tract), and the like. can be made That is, by adopting the above-described embodiment, it is possible to fill the hollow that leaked inside the target object (for example, inside the human body) after image division. adjusting the pixel values of the filled rudimentary segmented image according to the edge information of the target object in the to-be-processed image to obtain a first segmentation result corresponding to the to-be-processed image; The probability that background parts in the image (ie parts not belonging to the target object) will be split into those belonging to the target object can be reduced.

As an example of the embodiment, the pixel values of the closed regions whose pixel values in the preliminary segmented image are the second preset values are adjusted to the first preset values, and the filled primitives are adjusted to the first preset values. The step of obtaining a target segmented image includes splicing a preset width edge around the primitive segmented image to obtain a spliced primitive segmented image, wherein the spliced preset pixel values of pixels on the width side are the second preset values; performing a flooding filling operation on the rudimentary segmented image to obtain a filled rudimentary segmented image.

In this example, the preset width may be one pixel or more. For example, the preset width may be 1 pixel. FIG. 4 is a schematic diagram of splicing a preset width edge around a rudimentary segmented image to obtain a spliced rudimentary segmented image. In the example shown in FIG. 4, the preset width is 1 pixel. As shown in FIG. 4, edges of preset width can be spliced around the perimeter of the rudimentary segmented image. In another example, a side of a preset width can be spliced to one, two or three sides of the rudimentary segmented image.

In this example, the pixels at the image edge of the spliced rudimentary segmented image are the pixels of the edge on the position of the spliced rudimentary segmented image, e.g. , bottom-most pixel, left-most pixel, right-most pixel, and so on. For example, the upper left corner pixel of the spliced rudimentary segment image may be used as a seed point.

In this example, a flood fill operation is performed by splicing a preset width edge around the perimeter of the rudimentary segmented image and selecting pixels at the image edge of the spliced rudimentary segmented image as seed points. It can be ensured that the performed seed points belong to the background portion (i.e. the portion not belonging to the target object), so that the first segmentation result corresponding to the pending image includes the interior of the target object's organs. can be covered and a more accurate segmentation result can be obtained.

As an example of the embodiment, adjusting pixel values of the filled preliminary segmented image based on edge information of the target object in the to-be-processed image to generate a first segmentation result corresponding to the to-be-processed image. The obtaining step includes determining a maximum connected domain included in edges of the target object in the filled preliminary segmented image based on edge information of the target object in the pending image; and adjusting pixel values of pixels outside the largest connected domain in the rudimentary segmented image to the second preset value to obtain a first segmentation result corresponding to the pending image. In this example, false-positive regions that are not connected to the target object can be removed, so the probability of background parts being erroneously divided into those belonging to the target object can be greatly reduced, thereby increasing the accuracy of image segmentation. can be improved. For example, if the target object is a human body, this example can remove false-positive regions that are not connected to the human body, thereby erroneously dividing background parts (e.g. bedboards) into those belonging to the human body. can greatly reduce the chances of it happening.

In one possible embodiment, after the step of obtaining a first segmentation result corresponding to the pending image, the image segmentation method comprises an image adjacent to the pending image and an image corresponding to the adjacent image. obtaining a second segmentation result; adjusting the first segmentation result based on pixel values of pixels at the same position in the pending image and the adjacent image and the second segmentation result; , obtaining a third segmentation result corresponding to the to-be-processed image.

In such embodiments, the image adjacent to the to-be-processed image may be an image belonging to the same image sequence as the to-be-processed image and adjacent to the to-be-processed image. For example, the to-be-processed image may be a CT image, and the adjacent image may be an image belonging to the same CT image sequence as the to-be-processed image and adjacent to the to-be-processed image. The second segmentation result may refer to a final segmentation result corresponding to the adjacent image.

This embodiment can ensure continuity between the pending image and the second segmentation result, which helps to obtain a smoother and more accurate three-dimensional segmentation result. For example, if the target object is a human body, it can ensure continuity between the pending image and the human body in adjacent images, which helps to obtain a smoother and more accurate 3D human body segmentation result. For example, the embodiment can be adopted to obtain a segmentation result corresponding to each CT image in a CT image sequence, thereby obtaining a smoother and more accurate 3D human body segmentation result.

As an example of the embodiment, the first segmentation result is adjusted based on the pixel value of the pixel at the same position in the image awaiting processing and the adjacent image and the second segmentation result, and the awaiting processing is performed. The step of obtaining a third segmentation result corresponding to the image includes determining, among the adjacent images, a pixel value difference belonging to the target object in the second segmentation result and at the same position as the awaiting image. adjusting the first segmentation result based on pixels that are less than or equal to a third preset value to obtain a third segmentation result corresponding to the pending image.

In this example, the difference between co-located pixel values in the adjacent image and the to-be-processed image can refer to the difference between co-located normalized pixel values in the adjacent image and the to-be-processed image. For example, the third preset value may be 0.1. Of course, it is also possible to compare the original pixel values at the same location in the adjacent image and the pending image.

In this example, based on the pixels belonging to the target object in the second division result and located at the same position as the processing-waiting image and having a difference in pixel value equal to or smaller than a third preset value, the second adjusting one segmentation result to obtain a third segmentation result corresponding to the pending image, based on segmentation results corresponding to pixels in the adjacent images that are relatively related to the pending image; A first segmented result corresponding to the pending image can be adjusted, which helps improve the accuracy of the final segmented result corresponding to the pending image. Here, the division result corresponding to an arbitrary pixel of the adjacent image can refer to whether or not the pixel belongs to the target object in the second division result.

In one example, a difference in pixel values among the adjacent images belonging to the target object in the second segmentation result and at the same position as the to-be-processed image is equal to or less than a third preset value. adjusting the first segmentation result based on a pixel to obtain a third segmentation result corresponding to the to-be-processed image, wherein pixel values at the same position in the to-be-processed image and the adjacent image are adjusted; obtaining a first set of pixels based on the pixels for which the difference in is less than or equal to a third preset value; obtaining a second set of pixels based on the belonging pixels; adjusting the pixels of the second set of pixels in the first segmentation result to belong to the target object and correspond to the pending image; and obtaining a third split result.

In this example, the difference in pixel value of any pixel in the first pixel set in the pending image and the adjacent image is less than or equal to a third preset value. a pixel value difference between the to-be-processed image and the adjacent image of any pixel in a second pixel set is less than or equal to a third preset value, and the target object in the second segmentation result belongs to For example, if the first segmentation result is A, the first pixel set is B, and the second segmentation result is C, the third segmentation result is S=A∪(B∩C) may be

In the above example, obtaining a first set of pixels based on pixels whose pixel values at the same position in the pending image and the adjacent image have a difference equal to or less than a third preset value, and obtaining a second pixel set based on pixels belonging to the target object in the second segmentation result from among the first pixel set, and obtaining pixels of the second pixel set in the first segmentation result; belonging to the target object and obtaining a third segmentation result corresponding to the to-be-processed image, belonging to the target object in the second segmentation result and relatively to the to-be-processed image. A first segmented result corresponding to the pending image may be adjusted based on the relevant pixels, thereby helping to improve the accuracy of the final segmented result corresponding to the pending image.

In some possible embodiments, the third segmentation result may be used as the final segmentation result corresponding to the pending image.

FIG. 5 is another flow chart illustrating an image segmentation method according to an embodiment of the present disclosure. The execution body of the image segmentation method may be an image segmentation device. For example, the image segmentation method may be performed by a terminal device or a server or other processing device. Here, the terminal device may be a user device, a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant, a handheld device, a computing device, an in-vehicle device, a wearable device, or the like. In some possible embodiments, the image segmentation method may be implemented by a processor calling computer readable instructions stored in memory. As shown in FIG. 5, the image segmentation method includes steps S41 to S43.

In step S41, a pixel belonging to a target object in an image to be processed is predicted, and a preliminary segmented image corresponding to the image to be processed is obtained.

In step S42, an image adjacent to the image to be processed and a second division result corresponding to the adjacent image are obtained.

In embodiments of the present disclosure, the image adjacent to the pending image may be an image belonging to the same image sequence as the pending image and adjacent to the pending image. For example, the to-be-processed image may be a CT image, and the adjacent image may be an image belonging to the same CT image sequence as the to-be-processed image and adjacent to the to-be-processed image. The second segmentation result may refer to a final segmentation result corresponding to the adjacent image.

In step S43, the preliminary divided image is adjusted based on the pixel value of the pixel at the same position in the image to be processed and the adjacent image and the result of the second division, and the second division image corresponding to the image to be processed is adjusted. 4 division results are obtained.

The embodiments of the present disclosure can ensure continuity between the pending image and the second segmentation result, thereby helping to obtain a smoother and more accurate three-dimensional segmentation result. For example, if the target object is a human body, it can ensure continuity between the pending image and the human body in adjacent images, which helps to obtain a smoother and more accurate 3D human body segmentation result. For example, the embodiments of the present disclosure can be employed to obtain a segmentation result corresponding to each CT image in a CT image sequence, resulting in a smoother and more accurate 3D body segmentation result.

In one possible embodiment, the preliminary segmented image is adjusted based on pixel values of co-located pixels in the to-be-processed image and the adjacent image, and the second segmentation result, and The step of obtaining a fourth segmentation result corresponding to the image includes determining, among the adjacent images, the pixel value difference belonging to the target object in the second segmentation result and at the same position as the pending image. adjusting the preliminary segmented image based on pixels that are less than or equal to a third preset value to obtain a fourth segmented result corresponding to the pending image.

In this example, the difference in pixel values at the same location in the adjacent image and the to-be-processed image may refer to the difference in normalized pixel values at the same location in the adjacent image and the to-be-processed image. can. For example, the third preset value may be 0.1. Of course, it is also possible to compare original pixel values at the same location in the adjacent image and the pending image.

In the embodiment, a difference in pixel value among the adjacent images belonging to the target object in the second division result and at the same position as the processing-waiting image is equal to or less than a third preset value. adjusting the rudimentary segmented image based on a pixel to obtain a fourth segmented result corresponding to the to-be-processed image, whereby pixels of the adjacent image relatively related to the to-be-processed image; can adjust the rudimentary segmented image corresponding to the pending image based on the segmented result corresponding to , which helps improve the accuracy of the final segmented result corresponding to the pending image. Here, the division result corresponding to an arbitrary pixel of the adjacent image can refer to whether or not the pixel belongs to the target object in the second division result.

As an example of the embodiment, a difference in pixel value among the adjacent images belonging to the target object in the second segmentation result and at the same position as the processing-waiting image is a third preset value. The step of adjusting the preliminary segmented image to obtain a fourth segmented result corresponding to the to-be-processed image based on pixels that are at the same position in the to-be-processed image and the adjacent image are: obtaining a first set of pixels based on pixels whose value difference is less than or equal to a third preset value; and among the first set of pixels, the target object in the second segmentation result. and adjusting the pixels of said second set of pixels in said initial segmented image to belong to said target object, corresponding to said to-be-processed image. and obtaining a fourth split result.

In this example, the difference in pixel values in the pending image and the adjacent image for any pixel in the first pixel set is less than or equal to a third preset value. a pixel value difference between the to-be-processed image and the adjacent image of any pixel in a second pixel set is less than or equal to a third preset value, and the target object in the second segmentation result belongs to For example, if the first segmentation result is A, the first pixel set is B, and the second segmentation result is C, the third segmentation result is S=A∪(B∩C) may be

In this example, obtaining a first set of pixels based on pixels for which a difference in pixel value at the same position in the to-be-processed image and the adjacent image is equal to or less than a third preset value; obtaining a second pixel set based on the pixels belonging to the target object in the second segmentation result from among one pixel set, and converting the pixels of the second pixel set in the first segmentation result to adjusting to those belonging to the target object and obtaining a third segmentation result corresponding to the to-be-processed image belonging to the target object in the second segmentation result and relatively related to the to-be-processed image. Based on pixels, the first segmented result corresponding to the pending image can be adjusted, thereby helping to improve the accuracy of the final segmented result corresponding to the pending image.

In some possible embodiments, the fourth segmentation result may be used as the final segmentation result corresponding to the pending image.

In one possible embodiment, after obtaining a fourth segmentation result corresponding to the to-be-processed image, the image segmentation method comprises, based on edge information of the target object in the to-be-processed image, the fourth segmentation result. 4, adjusting the pixel values of the predicted pixels not belonging to the target object in the closed area included in the edge of the target object, and generating a fifth segmentation result corresponding to the pending image; Further comprising the step of obtaining.

As an example of the embodiment, based on the edge information of the target object in the pending image, the predicted target object in a closed area included in the edge of the target object in the fourth segmentation result. Adjusting pixel values of pixels that do not belong to obtain a fifth segmentation result corresponding to the pending image, wherein the pixel value in the fourth segmentation result is the second preset value. adjusting pixel values of closed regions to the first preset value to obtain a filled preliminary segmented image corresponding to the fourth segmentation result; adjusting pixel values of the filled preliminary segmented image based on the edge information to obtain a fifth segmented result corresponding to the to-be-processed image.

In one example, the pixel value of the closed area in which the pixel value in the fourth segmentation result is the second preset value is adjusted to the first preset value, and the pixel value in the fourth segmentation result is adjusted to the first preset value. Obtaining a corresponding filled rudimentary segmented image comprises splicing a preset width edge around the fourth segmented result to obtain a spliced fourth segmented result. , wherein the pixel value of the spliced pixels on the preset width side is the second preset value; selecting as seed points and performing a flood filling operation on the spliced fourth segmentation result to obtain a filled rudimentary segmented image corresponding to the fourth segmentation result.

In one example, pixels not belonging to the target object are predicted in a closed region included in the edge of the target object in the fourth segmentation result based on the edge information of the target object in the pending image. and obtaining a fifth segmentation result corresponding to the to-be-processed image, based on edge information of the target object in the to-be-processed image, in the filled rudimentary segmented image. determining a maximum connected domain included in an edge of a target object; adjusting pixel values of pixels other than the maximum connected domain in the filled rudimentary segmented image to the second preset value; and obtaining a fifth segmentation result corresponding to the pending image.

Here, "based on the edge information of the target object in the pending image, in the fourth segmentation result, a predicted closed area included in the edge of the target object that does not belong to the target object A specific embodiment of "Adjusting the pixel values of the pixels to obtain a fifth division result corresponding to the pending image" is the above "based on the edge information of the target object in the pending image, Adjusting pixel values of predicted pixels not belonging to the target object in closed regions included in edges of the target object in the preliminary segmented image, and obtaining a first segmentation result corresponding to the pending image. is similar to the specific embodiment of "obtaining", so the description is omitted here.

In one possible embodiment, the fifth segmentation result can be used as the final segmentation result corresponding to the to-be-processed image.

It is understandable that any of the method embodiments referred to in this disclosure can be combined with each other to form combined embodiments without violating the principle logic, and is limited to the width of the paper. Therefore, the description is omitted in this disclosure. Persons skilled in the art can understand that in the above method of specific embodiments, the specific execution order of each step should be determined by its function and internal logic.

The embodiments of the present disclosure are described below with one specific application scene.

First, get the training images and the masks corresponding to the training images. The training images are CT images of the human body, where the preset CT value range can be set to [-500, 1200] based on the CT values of all tissues and organs of the human body, covers all tissues and organs of the human body.

Based on a preset CT value range, preprocess any pixel in the training image to obtain a preprocessed pixel value. Specifically, for any pixel in the training image, if the pixel value of said pixel is less than the lower limit of said preset CT value range, said lower limit is used as the preprocessed pixel value of said pixel. and if the pixel value of the pixel is greater than the upper value of the preset CT value range, the upper value can be used as the preprocessed pixel value of the pixel; If the pixel value is within the preset CT value range, the pixel value of the pixel can be used as the preprocessed pixel value of the pixel, for example, if the pixel value of a pixel in the training image is − 505, -500 can be used as the preprocessed pixel value of the pixel, and if the pixel value of a pixel in the training image is 1250, then 12000 can be used as the preprocessed pixel value of the pixel. can be used, and if the pixel value of a pixel in the training image is 800, then 800 can be used as the preprocessed pixel value of the pixel.

Next, pixel values of arbitrary pixels of the training image are normalized according to a preset CT value range to obtain a normalized training image. Now, we can use equation (1) to normalize the pixel values of any pixel in the training images.

ここで、ｈは、前記画素の前処理された画素値であり、ｈ_ｍｉｎは、前記予め設定されたＣＴ値範囲の下限値であり、ｈ_ｍａｘは、前記予め設定されたＣＴ値範囲の上限値である。このように、トレーニング画像の各画素に対して上記の処理を行うことにより、正規化されたトレーニング画像を取得することができる。

where h is the preprocessed pixel value of the pixel, h _min is the lower limit of the preset CT value range, and h _max is the upper limit of the preset CT value range. value. By performing the above processing on each pixel of the training image in this manner, a normalized training image can be obtained.

Here, the normalized training images can be amplified. For example, randomly zoom the normalized training images by a factor of 0.6-1.4, and then crop the zoomed images from the center by a size of 512x512 to obtain the same size training at different zoom scales. Images can be acquired. A similar operation is performed on the mask corresponding to the training image accordingly.

Furthermore, the normalized and amplified training images can be separated into a training set and a validation set. For example, the processed training images can be split into training and validation sets in a 4:1 ratio.

Thus, the training set is used to iteratively train the U-shaped convolutional neural network until the loss in the validation set of said U-shaped convolutional neural network drops below 0.03, and the trained U-shaped convolutional neural network can be obtained.

In a practical application, obtaining a pending CT image, inputting the pending CT image into a trained U-shaped convolutional neural network, and by means of the U-shaped convolutional neural network, among the pending CT image, Predicting pixel information and obtaining a preliminary segmented image corresponding to the pending CT image based on pixel information belonging to the target object in the pending CT image.

After the rudimentary segmented image is obtained, splice a 1-pixel wide edge around the rudimentary segmented image to obtain a spliced rudimentary segmented image; A pixel in a corner can be selected as a seed point and a flooding filling operation can be performed on the spliced rudimentary segmented image to obtain a filled rudimentary segmented image. determining a maximum connected domain included in the edge of the target object in the filled elementary segmentation image based on the edge information of the target object in the pending CT image; Pixel values of pixels outside the largest communicating domain in the image may be adjusted to the second preset value to obtain a first segmentation result corresponding to the pending CT image. After the first segmentation result is obtained, an image adjacent to the pending CT image and a second segmentation result corresponding to the adjacent image can be obtained.

Further, obtaining a first set of pixels based on pixels for which a difference in pixel value at the same position between the pending CT image and the adjacent image is equal to or less than a third preset value; obtain a second pixel set based on the pixels belonging to the target object in the second segmentation result, and divide the pixels of the second pixel set in the first segmentation result into the target object. A third segmentation result corresponding to the pending CT image can be obtained by adjusting it to that belonging to the object.

In addition, the present disclosure further provides an image segmentation device, an electronic device, a computer-readable storage medium, and a program that can be used to implement any of the image segmentation methods provided in the present disclosure, and provides corresponding technical solutions. For the measures and technical effects, the corresponding description in the method part can be referred to, so the description is omitted.

FIG. 6 is a block diagram illustrating an image segmentation device according to an embodiment of the present disclosure. As shown in FIG. 6, the image segmentation device includes a first segmentation unit configured to predict pixels belonging to a target object in a pending image and obtain a preliminary segmented image corresponding to the pending image. 51, based on the edge information of the target object in the awaiting image, predicted number of pixels not belonging to the target object in a closed region included in the edge of the target object in the rudimentary segmented image; a first adjuster 52 configured to adjust pixel values to obtain a first segmentation result corresponding to the to-be-processed image.

In one possible embodiment, in said rudimentary segmented image, pixel values of pixels belonging to said predicted target object are said first preset value and do not belong to said predicted target object. A pixel value of a pixel is a second preset value, and the first adjusting unit 52 adjusts a pixel value in a closed region where a pixel value in the elementary divided image is the second preset value. to the first preset value to obtain a filled rudimentary segmented image, and based on the edge information of the target object in the pending image, pixels of the filled rudimentary segmented image adjusting a value to obtain a first segmentation result corresponding to the to-be-processed image.

In one possible embodiment, the first adjuster 52 splices a preset width edge around the rudimentary segmented image to obtain a spliced rudimentary segmented image and spliced rudimentary segmented image. selecting pixels at the image edge of the spliced rudimentary segmented image as seed points, the pixel values of the pixels on the preset width side being the second preset value, and the splicing; It is configured to perform a flooding filling operation on the filled rudimentary segmented image to obtain a filled rudimentary segmented image.

In one possible embodiment, the first adjuster 52 determines, based on the edge information of the target object in the pending image, the maximum determining a connected domain, adjusting pixel values of pixels other than the largest connected domain in the filled rudimentary segmented image to the second preset value, and a first segmented corresponding to the pending image; Configured to retrieve results.

In one possible embodiment, the image segmentation device comprises a second obtaining unit configured to obtain an image adjacent to the to-be-processed image and a second segmentation result corresponding to the adjacent image. , adjusting the first segmentation result based on the pixel value of the pixel at the same position in the process-waiting image and the adjacent image, and the second segmentation result, and adjusting the third segmentation corresponding to the process-waiting image and a third coordinator configured to obtain a division result of the .

In one possible embodiment, the third adjuster determines that the difference in pixel values belonging to the target object in the second segmentation result and located at the same position as the awaiting image among the adjacent images is the It is configured to adjust the first segmentation result based on pixels that are less than or equal to a preset value of three to obtain a corresponding third segmentation result for the to-be-processed image.

In one possible embodiment, the third adjuster is based on pixels for which a difference between values of co-located pixels in the to-be-processed image and the adjacent image is less than or equal to a third preset value. , obtaining a first pixel set, obtaining a second pixel set based on pixels belonging to the target object in the second division result among the first pixel set, and performing the first division Adjusting the pixels of the second set of pixels in the result to those belonging to the target object, and configured to obtain a third segmentation result corresponding to the pending image.

In one possible embodiment, the image segmentation device is configured to train a neural network based on training images and labeled data of the training images, wherein the labeled data of the training images includes: , the first dividing unit 51 inputs the processing-waiting image to the neural network, and the neural network determines the target object in the processing-waiting image by the neural network. It is configured to predict information of pixels belonging to an object and obtain a preliminary segmented image corresponding to the to-be-processed image based on the information of pixels belonging to the target object in the to-be-processed image.

In one possible embodiment, the training images are computed tomography (CT) images, and the training unit normalizes the pixel values of the training images according to a preset CT value range, obtaining normalized training images and training the neural network based on the normalized training images and labeled data of the training images.

In an embodiment of the present disclosure, a pixel belonging to a target object in a pending image is predicted, a rudimentary segmented image corresponding to the pending image is obtained, and a rudimentary segmented image is obtained based on edge information of the target object in the pending image. Adjusting pixel values of predicted pixels not belonging to the target object in a closed region included in an edge of the target object in the divided image to obtain a first division result corresponding to the awaiting image. can obtain more accurate and robust segmentation results.

FIG. 7 is another block diagram illustrating an image segmentation apparatus according to an embodiment of the present disclosure. As shown in FIG. 7, the image segmentation device includes a second segmentation unit configured to predict pixels belonging to a target object in a pending image and obtain a rudimentary segmented image corresponding to the pending image. 61, a first obtaining unit 62 configured to obtain an image adjacent to the to-be-processed image and a second division result corresponding to the adjacent image, the to-be-processed image and the adjacent image. and the second segmentation result to obtain a fourth segmentation result corresponding to the pending image. and a second adjuster 63 .

In one possible embodiment, the second adjustment unit 63 selects, among the adjacent images, the pixel values belonging to the target object in the second segmentation result and at the same position as the to-be-processed image. The rudimentary segmented image is adjusted based on pixels whose difference is less than or equal to a third preset value to obtain a fourth segmented result corresponding to the to-be-processed image.

In one possible embodiment, the second adjustment unit 63 is based on pixels for which the difference between values of co-located pixels in the to-be-processed image and the adjacent image is less than or equal to a third preset value. obtaining a first set of pixels based on pixels belonging to the target object in the second division result among the first set of pixels; obtaining a second set of pixels from the first set of pixels; adjusting pixels of the second set of pixels in the image to belong to the target object to obtain a fourth segmentation result corresponding to the pending image.

In one possible embodiment, the image segmentation device, in the fourth segmentation result, based on edge information of the target object in the to-be-processed image, predicts a fourth adjuster configured to adjust pixel values of the pixels that do not belong to the target object to obtain a fifth segmentation result corresponding to the to-be-processed image.

In one possible embodiment, the fourth adjuster adjusts the pixel values of the closed region, in which the pixel values in the fourth division result are the second preset values, from the first preset values. and obtain a filled preliminary segmented image corresponding to the fourth segmentation result, and based on the edge information of the target object in the pending image, perform the filling of the preliminary segmented image. and configured to adjust pixel values to obtain a fifth segmentation result corresponding to the pending image.

In one possible embodiment, the fourth adjuster splices a preset width edge around the fourth split result to obtain a spliced fourth split result, spliced selecting pixels at the edge of the spliced fourth segmented result image as seed points, the pixel values of the pixels on the preset width side of the second preset value being the second preset value; It is configured to perform a flood filling operation on the spliced fourth segmented result to obtain a filled rudimentary segmented image corresponding to the fourth segmented result.

In one possible embodiment, the fourth adjuster determines, based on the edge information of the target object in the to-be-processed image, the maximum connectivity included in the edges of the target object in the filled rudimentary segmented image. determining a domain, adjusting pixel values of pixels other than the largest connected domain in the filled elementary segmented image to the second preset value, and a fifth segmentation result corresponding to the pending image; is configured to obtain

The embodiments of the present disclosure can ensure continuity between the pending image and the second segmentation result, thereby helping to obtain a smoother and more accurate three-dimensional segmentation result. For example, if the target object is a human body, it can ensure continuity between the pending image and the human body in adjacent images, which helps to obtain a smoother and more accurate 3D human body segmentation result. For example, the embodiments of the present disclosure can be employed to obtain a segmentation result corresponding to each CT image in a CT image sequence, resulting in a smoother and more accurate 3D body segmentation result.

In some embodiments, the functionality provided by the apparatus provided by the embodiments of the present disclosure, or the portions included therein, are configured to perform the methods described in the above method embodiments, and the specific The implementation and technical effects can refer to the descriptions of the above method embodiments, and the descriptions are omitted here for the sake of brevity.

In the embodiments of the present disclosure and other embodiments, a "portion" may be a circuit of a portion, a processor of a portion, a program or software of a portion, etc., and of course may be a unit, or may be a module. It may exist or may not be modularized.

Embodiments of the present disclosure further provide a computer-readable storage medium storing computer program instructions for causing a processor to perform the above method. Here, the computer-readable storage medium may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium.

A computer program according to an embodiment of the present disclosure comprises computer readable code for causing a processor in the electronic device to perform the above image segmentation method when executed on an electronic device.

A computer program product according to embodiments of the present disclosure is used to store computer readable instructions which, when executed, cause a computer to perform operations of the image segmentation method according to any of the embodiments above.

An electronic device according to embodiments of the present disclosure comprises one or more processors and a memory storing executable instructions, wherein the one or more processors execute executable instructions stored in the memory. Call it and follow the steps above.

An electronic device may be provided as a terminal, server or other form of device.

FIG. 8 is a block diagram illustrating an electronic device 800 according to an embodiment of the disclosure. For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, and the like.

Referring to FIG. 8, electronic device 800 includes processing component 802, memory 804, power component 806, multimedia component 808, audio component 810, input/output (I/O) interface 812, sensor component 814, and communication component 816. can include one or more of

The processing component 802 typically controls the overall operation of the electronic device 800, for example, operations associated with display, telephone calls, data communications, camera operations and recording operations. Processing component 802 may include one or more processors 820 that execute instructions to complete all or some steps of the methods described above. To facilitate the interaction of processing component 802 with other components, processing component 802 can include one or more portions. For example, processing component 802 can include multimedia components to facilitate interaction between multimedia component 808 and processing component 802 .

Memory 804 is configured to store various types of data to support operations on electronic device 800 . Examples of these data include instructions for any application programs or methods running on electronic device 800, contact data, phonebook data, messages, images, videos, and the like. Memory 804 can be static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic It may be implemented by any type of volatile or non-volatile storage such as memory, flash memory, magnetic or optical disks, or combinations thereof.

Power component 806 provides power to the various components of electronic device 800 . Power supply components 806 can include a power management system, one or more power supplies, and other components associated with generating, managing, and allocating power for electronic device 800 .

Multimedia component 808 includes a screen that provides one output interface between the electronic device 800 and a user. In some implementations, the screen can include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. A touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor can detect the duration and pressure associated with the touch or slide motion as well as sensing the boundaries of the touch or slide motion. In some embodiments, multimedia component 808 includes one front camera and/or one rear camera. When the electronic device 800 is in an operational mode, such as photography mode or video mode, the front camera and/or the rear camera can receive external multimedia data. Each front and/or rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

Audio component 810 is configured to output and/or input audio signals. For example, audio component 810 includes one microphone (MIC), which is configured to receive external audio signals when electronic device 800 is in operational modes such as call mode, recording mode, and voice identification mode. be done. The received audio signal can also be stored in memory 804 or transmitted via communication component 816 . In some examples, audio component 810 includes a speaker for outputting audio signals.

I/O interface 812 provides an interface between processing component 802 and peripheral interface modules, which may be keyboards, click wheels, buttons, and the like. Buttons can include, but are not limited to, home button, volume button, start button and lock button.

Sensor component 814 includes one or more sensors for providing various aspects of condition assessment to electronic device 800 . For example, the sensor component 814 can detect the on/off state of the electronic device 800, the relative position of the components, for example the component is the display and keypad of the electronic device 800, and the sensor component 814 further Changes in the position of the electronic device 800 or one component of the electronic device 800, presence or absence of contact between the user and the electronic device 800, orientation or acceleration/deceleration of the electronic device 800, and temperature changes of the electronic device 800 can be detected. Sensor component 814 can include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact. The sensor component 814 can further include optical sensors such as complementary metal oxide semiconductor (CMOS) or charge coupled device (CCD) image sensors used in imaging applications. In some examples, the sensor component 814 can further include an acceleration sensor, gyro sensor, magnetic sensor, pressure sensor, or temperature sensor.

Communication component 816 is configured to facilitate wired or wireless communication between electronic device 800 and other devices. The electronic device 800 is a wireless network (Wi-Fi), second generation mobile communication technology (2G), third generation mobile communication technology (3G), fourth generation mobile communication technology (4G) / long-term evolution of universal mobile communication technology (LTE), fifth generation mobile communication technology (5G), or a combination thereof. In one exemplary embodiment, communication component 816 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one illustrative example, the communication component 816 further includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, Infrared Communication Association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology and other technologies. .

In an exemplary embodiment, electronic device 800 includes one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable It may be implemented by a logic device (PLD), field programmable gate array (FPGA), controller, microcontroller, microprocessor or other device.

The exemplary embodiment further provides a non-volatile computer readable storage medium, memory 804 containing computer program instructions executable by the processor 820 of the electronic device 800 to complete the method described above.

FIG. 9 is a block diagram illustrating electronic device 1900 according to an embodiment of the disclosure. For example, electronic device 1900 may be provided as a server. Referring to FIG. 9, the electronic device 1900 includes a processing component 1922 which further includes one or more processors and a memory 1932 for storing instructions executable by the processing component 922, such as application programs. and memory resources represented by An application program stored in memory 1932 may include one or more portions, each portion corresponding to a group of instructions. Also, the processing component 1922 is configured to execute instructions for carrying out the methods described above.

The electronic device 1900 includes a power component 1926 configured to perform power management of the electronic device 1900; a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network; O) an interface 1958; The electronic device 1900 may support an operating system stored in memory 1932, such as the Microsoft Server Operating System (Windows ServerTM), the graphical user interface operating system (Mac OS XTM) introduced by Apple Inc., multi-user and multi-user. Process computer operating system (Unix), free and open source code UNIX-like operating system (Linux), open source code Unix-like operating system system (FreeBSD™) or similar.

The exemplary embodiment further provides a non-volatile computer-readable storage medium, such as memory 1932, containing computer program instructions executable by processor 1922 of electronic device 1900 to complete the method described above.

The present disclosure may be systems, methods and/or computer program products. The computer program product may include a computer-readable storage medium loaded with computer-readable program instructions for enabling a processor to implement various aspects of the present disclosure.

A computer-readable storage medium may be a tangible device capable of holding and storing instructions for use by an instruction-executing device. A computer-readable storage medium may be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (non-exhaustive list) of computer readable storage media include portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable random access memory ( EPROM), flash memory, static random access memory (SRAMy), portable compact disc read-only memory (CD-ROM: Compact Disc Read-Only Memory), digital versatile disc (DVD: Digital Video Disc), memory stick, floppy (registered (trademark) discs, mechanical coding devices such as punched cards or recessed structures storing instructions, and any suitable combination of the above. Computer-readable storage media, as used herein, includes radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical wires. It should not be construed as a transitory signal per se, such as an electrical signal transmitted via.

The computer readable program instructions described herein may be downloaded to each computing/processing device from a computer readable storage medium or to an external computer via networks such as the Internet, local area networks, wide area networks and/or wireless networks. Or it may be downloaded to an external storage device. A network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers and/or edge servers. A network adapter card or network interface in each computing/processing device receives computer-readable program instructions from the network and transfers the computer-readable program instructions for storage on a computer-readable storage medium in each computing/processing device. .

Computer program instructions for carrying out operations of the present disclosure may be assembler instructions, Instruction Set Architecture (ISA) instructions, machine instructions, instrument-related instructions, microcode, firmware instructions, state setting data, or one or more programming languages. The programming language may be an object-oriented programming language such as Smalltalk, C++, and a conventional procedural programming language such as the "C" language or similar Including programming languages. Computer readable program code may be executed entirely on the user's computer, partially executed on the user's computer, executed as a separate software package, partially executed on the user's computer, or remote computer. , or fully executed on a remote computer or server. When referring to a remote computer, the remote computer may be connected to the user computer via any type of network, including a local area network (LAN) or wide area network (WAN), or may be connected to an external computer. may be (eg, connected via the Internet by an Internet Service Provider). In some embodiments, by personalizing and customizing an electronic circuit, such as a programmable logic circuit, field programmable gate array (FPGA), or programmable logic array (PLA), with state information in computer readable program instructions, the electronic circuit , can execute computer-readable program instructions to implement various aspects of the present disclosure.

Aspects of the present disclosure are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products of embodiments of the disclosure. It is to be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer readable program instructions.

These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer or other programmable data processing apparatus such that these instructions are executed by the processor of the computer or other programmable data processing apparatus. In some cases, the machines are manufactured to produce devices that perform the functions/acts specified in one or more blocks of the flowcharts and/or block diagrams. These computer readable program instructions can be stored on a computer readable storage medium and cause computers, programmable data processing apparatus and/or other devices to operate in a particular manner, thereby storing the instructions. The computer-readable media included in the flowcharts and/or block diagrams include articles of manufacture that include instructions for implementing aspects of the functionality/operations specified in one or more blocks of the flowcharts and/or block diagrams.

The computer readable program instructions may also be loaded into the computer, other programmable data processing apparatus, or other device to cause the computer, other programmable data processing apparatus, or other device to perform a series of operational steps, causing the computer to by instructions executed by a computer, other programmable data processing apparatus, or other device to perform the functions/functions defined in one or more blocks of the flowchart and/or block diagrams. Action is realized.

The flowcharts and block diagrams in the drawings illustrate possible architectures, functionality, and operation of systems, methods and computer program products according to embodiments of the present disclosure. In this regard, each block of a flowchart or block diagram can represent a portion, program segment or portion of an instruction, said portion, program segment or portion of an instruction being used to implement a given logical function. contains one or more executable instructions for In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two consecutive blocks may actually be executed essentially in parallel, or they may be executed in reverse order depending on the functionality involved. It should be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, are implemented in a dedicated, hardware-based system to perform the specified function or operation. or be implemented in a combination of dedicated hardware and computer instructions.

The computer program product may be specifically implemented in hardware, software or a combination thereof. In one alternative embodiment, said computer program product is specifically embodied as a computer storage medium, and in another alternative embodiment, the computer program product is a Software Development Kit (SDK). ) are concretely embodied as software products.

While embodiments of the present disclosure have been described above, the above description is illustrative, not exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the above-described embodiments. The choice of terminology used herein is such that it best describes the principle, practical application, or improvement over the marketed technology of each embodiment, or that a person of ordinary skill in the art would understand each embodiment disclosed herein. The purpose is to make it possible to understand.

In an embodiment of the present disclosure, predicting pixels belonging to a target object in an image to be processed, obtaining a rudimentary divided image corresponding to the image to be processed, based on edge information of the target object in the image to be processed, Adjusting pixel values of predicted pixels not belonging to the target object in closed regions included in edges of the target object in the preliminary segmented image, and obtaining a first segmentation result corresponding to the pending image. to get In this way, if the target object is a human or animal body, the pixels inside the organs of the target object can also be segmented into those belonging to the target object, thereby obtaining a more accurate and robust segmentation result. can be done.

It should be understood that the foregoing general description and the following detailed description are exemplary and interpretative only and are not restrictive of the present disclosure.
For example, the present application provides the following items.
(Item 1)
An image segmentation method comprising:
predicting pixels belonging to a target object in a pending image and obtaining a rudimentary segmented image corresponding to the pending image;
calculating predicted pixel values of pixels not belonging to the target object in a closed region included in the edge of the target object in the preliminary segmented image based on the edge information of the target object in the image to be processed; adjusting and obtaining a first segmentation result corresponding to the pending image.
(Item 2)
In the rudimentary segmented image, predicted pixel values of pixels belonging to the target object are first preset values, and predicted pixel values of pixels not belonging to the target object are: a second preset value;
calculating predicted pixel values of pixels not belonging to the target object in a closed region included in the edge of the target object in the preliminary segmented image based on the edge information of the target object in the image to be processed; The step of adjusting and obtaining a first segmentation result corresponding to the pending image includes:
adjusting pixel values of closed regions whose pixel values in said preliminary segmented image are said second preset value to said first preset value to obtain a filled preliminary segmented image; When,
adjusting pixel values of the filled rudimentary segmented image based on the edge information of the target object in the pending image to obtain the first segmentation result corresponding to the pending image; characterized by comprising
The image segmentation method according to item 1.
(Item 3)
adjusting pixel values of closed regions whose pixel values in said preliminary segmented image are said second preset value to said first preset value to obtain a filled preliminary segmented image; teeth,
splicing a preset width edge around the rudimentary segmented image to obtain a spliced rudimentary segmented image, the pixels of the pixels on the spliced preset width edge; the value is the second preset value;
selecting pixels at image edges of the spliced primitive segmented image as seed points and performing a flooding filling operation on the spliced primitive segmented image to obtain the filled primitive segmented image; and
The image segmentation method according to item 2.
(Item 4)
adjusting pixel values of the filled rudimentary segmented image based on edge information of the target object in the pending image to obtain the first segmentation result corresponding to the pending image;
determining the maximum connected domain included in the edges of the target object in the filled rudimentary segmented image based on the edge information of the target object in the to-be-processed image;
adjusting pixel values of pixels other than the largest connected domain in the filled rudimentary segmented image to the second preset value to obtain the first segmentation result corresponding to the pending image; and
4. The image segmentation method according to item 2 or 3.
(Item 5)
After obtaining a first segmentation result corresponding to the pending image, the image segmentation method includes:
obtaining an image adjacent to the to-be-processed image and a second segmentation result corresponding to the adjacent image;
adjusting the first segmentation result based on the pixel value of the pixel at the same position in the process-waiting image and the adjacent image, and the second segmentation result, and generating a third segment corresponding to the process-waiting image; and obtaining a split result.
The image segmentation method according to any one of items 1-4.
(Item 6)
adjusting the first segmentation result based on the pixel value of the pixel at the same position in the process-waiting image and the adjacent image, and the second segmentation result, and generating a third segment corresponding to the process-waiting image; The step of obtaining the split result is
Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. , adjusting the first segmentation result to obtain the third segmentation result corresponding to the pending image.
6. An image segmentation method according to item 5.
(Item 7)
Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. , adjusting the first segmentation result to obtain the third segmentation result corresponding to the pending image;
obtaining a first set of pixels based on pixels whose co-located pixel value difference in the pending image and the adjacent image is less than or equal to a third preset value;
obtaining a second set of pixels based on pixels belonging to the target object in the second segmentation result from the first set of pixels;
adjusting pixels of the second set of pixels in the first segmentation result to belong to the target object to obtain the third segmentation result corresponding to the pending image. to be
An image segmentation method according to item 6.
(Item 8)
Before the step of predicting pixels belonging to a target object in the pending image, the image segmentation method comprises training a neural network based on training images and labeled data of the training images, wherein: wherein the labeled data of includes true values of pixels belonging to the target object in the training images;
The step of predicting pixels belonging to a target object in the to-be-processed image and obtaining a rudimentary divided image corresponding to the to-be-processed image includes inputting the to-be-processed image to the neural network, and performing the processing by the neural network. predicting information of pixels belonging to the target object in an image; obtaining a preliminary segmented image corresponding to the to-be-processed image based on the information of pixels belonging to the target object in the to-be-processed image; characterized by comprising
The image segmentation method according to any one of items 1-7.
(Item 9)
the training images are computed tomography (CT) images;
The step of training a neural network based on training images and labeled data of the training images includes normalizing pixel values of the training images according to a preset CT value range to obtain normalized training images. and training the neural network based on the normalized training images and the labeled data of the training images.
9. An image segmentation method according to item 8.
(Item 10)
An image segmentation method comprising:
predicting pixels belonging to a target object in a pending image and obtaining a rudimentary segmented image corresponding to the pending image;
obtaining an image adjacent to the to-be-processed image and a second segmentation result corresponding to the adjacent image;
adjusting the preliminary divided image based on pixel values of pixels at the same position in the to-be-processed image and the adjacent image, and the result of the second division, and forming a fourth division corresponding to the to-be-processed image; and obtaining a result.
(Item 11)
adjusting the preliminary divided image based on pixel values of pixels at the same position in the to-be-processed image and the adjacent image, and the result of the second division, and forming a fourth division corresponding to the to-be-processed image; The step to get the result is
Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. and adjusting the rudimentary segmented image to obtain the fourth segmented result corresponding to the pending image.
11. An image segmentation method according to item 10.
(Item 12)
Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. , adjusting the rudimentary segmented image to obtain the fourth segmented result corresponding to the pending image;
obtaining a first set of pixels based on pixels whose co-located pixel value difference in the pending image and the adjacent image is less than or equal to a third preset value;
obtaining a second set of pixels based on pixels belonging to the target object in the second segmentation result from the first set of pixels;
adjusting pixels of the second set of pixels in the preliminary segmented image to belong to the target object to obtain the fourth segmented result corresponding to the pending image. do
12. An image segmentation method according to item 11.
(Item 13)
An image segmentation device,
a first segmentation unit configured to predict pixels belonging to a target object in a pending image and obtain a rudimentary segmented image corresponding to the pending image;
calculating predicted pixel values of pixels not belonging to the target object in a closed region included in the edge of the target object in the preliminary segmented image based on the edge information of the target object in the image to be processed; a first adjuster configured to adjust and obtain a first segmentation result corresponding to the pending image.
(Item 14)
An image segmentation device,
a second segmentation unit configured to predict pixels belonging to a target object in a pending image and obtain a rudimentary segmented image corresponding to the pending image;
a first acquisition unit configured to acquire an image adjacent to the to-be-processed image and a second division result corresponding to the adjacent image;
adjusting the preliminary divided image based on pixel values of pixels at the same position in the to-be-processed image and the adjacent image, and the result of the second division, and forming a fourth division corresponding to the to-be-processed image; and a second adjustment unit configured to obtain a result.
(Item 15)
a memory that stores executable instructions;
and one or more processors for calling executable instructions stored in said memory to perform the image segmentation method according to any one of items 1-12.
(Item 16)
A computer readable storage medium storing computer program instructions for, when executed by a processor, causing said processor to perform the image segmentation method of any one of items 1-12.
(Item 17)
A computer program product comprising computer readable code for, when run on an electronic device, causing a processor in said electronic device to perform the image segmentation method of any one of items 1-12.

Claims (17)

An image segmentation method comprising:
predicting pixels belonging to a target object in a pending image and obtaining a rudimentary segmented image corresponding to the pending image;
calculating predicted pixel values of pixels not belonging to the target object in a closed region included in the edge of the target object in the preliminary segmented image based on the edge information of the target object in the image to be processed; adjusting and obtaining a first segmentation result corresponding to the pending image. In the rudimentary segmented image, predicted pixel values of pixels belonging to the target object are first preset values, and predicted pixel values of pixels not belonging to the target object are: a second preset value;
calculating predicted pixel values of pixels not belonging to the target object in a closed region included in the edge of the target object in the preliminary segmented image based on the edge information of the target object in the image to be processed; The step of adjusting and obtaining a first segmentation result corresponding to the pending image includes:
adjusting pixel values of closed regions whose pixel values in said preliminary segmented image are said second preset value to said first preset value to obtain a filled preliminary segmented image; When,
adjusting pixel values of the filled rudimentary segmented image based on the edge information of the target object in the pending image to obtain the first segmentation result corresponding to the pending image; The image segmentation method according to claim 1, characterized by comprising: adjusting pixel values of closed regions whose pixel values in said preliminary segmented image are said second preset value to said first preset value to obtain a filled preliminary segmented image; teeth,
splicing a preset width edge around the rudimentary segmented image to obtain a spliced rudimentary segmented image, the pixels of the pixels on the spliced preset width edge; the value is the second preset value;
selecting pixels at image edges of the spliced primitive segmented image as seed points and performing a flooding filling operation on the spliced primitive segmented image to obtain the filled primitive segmented image; 3. The image segmentation method according to claim 2, comprising: adjusting pixel values of the filled rudimentary segmented image based on edge information of the target object in the pending image to obtain the first segmentation result corresponding to the pending image;
determining the maximum connected domain included in the edges of the target object in the filled rudimentary segmented image based on the edge information of the target object in the to-be-processed image;
adjusting pixel values of pixels other than the largest connected domain in the filled rudimentary segmented image to the second preset value to obtain the first segmentation result corresponding to the pending image; 4. The image segmentation method according to claim 2 or 3, comprising: After obtaining a first segmentation result corresponding to the pending image, the image segmentation method includes:
obtaining an image adjacent to the to-be-processed image and a second segmentation result corresponding to the adjacent image;
adjusting the first segmentation result based on the pixel value of the pixel at the same position in the process-waiting image and the adjacent image, and the second segmentation result, and generating a third segment corresponding to the process-waiting image; and obtaining a segmentation result. adjusting the first segmentation result based on the pixel value of the pixel at the same position in the process-waiting image and the adjacent image, and the second segmentation result, and generating a third segment corresponding to the process-waiting image; The step of obtaining the split result is
Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. 6. The image segmentation method according to claim 5, further comprising: adjusting the first segmentation result to obtain the third segmentation result corresponding to the pending image. Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. , adjusting the first segmentation result to obtain the third segmentation result corresponding to the pending image;
obtaining a first set of pixels based on pixels whose co-located pixel value difference in the pending image and the adjacent image is less than or equal to a third preset value;
obtaining a second set of pixels based on pixels belonging to the target object in the second segmentation result from the first set of pixels;
adjusting pixels of the second set of pixels in the first segmentation result to belong to the target object to obtain the third segmentation result corresponding to the pending image. The image segmentation method according to claim 6. Before the step of predicting pixels belonging to a target object in the pending image, the image segmentation method comprises training a neural network based on training images and labeled data of the training images, wherein: wherein the labeled data of includes true values of pixels belonging to the target object in the training images;
The step of predicting pixels belonging to a target object in the to-be-processed image and obtaining a rudimentary divided image corresponding to the to-be-processed image includes inputting the to-be-processed image to the neural network, and performing the processing by the neural network. predicting information of pixels belonging to the target object in an image; obtaining a preliminary segmented image corresponding to the to-be-processed image based on the information of pixels belonging to the target object in the to-be-processed image; The image segmentation method according to any one of claims 1 to 7, characterized by comprising: the training images are computed tomography (CT) images;
The step of training a neural network based on training images and labeled data of the training images includes normalizing pixel values of the training images according to a preset CT value range to obtain normalized training images. and training the neural network based on the normalized training images and the labeled data of the training images. An image segmentation method comprising:
predicting pixels belonging to a target object in a pending image and obtaining a rudimentary segmented image corresponding to the pending image;
obtaining an image adjacent to the to-be-processed image and a second segmentation result corresponding to the adjacent image;
adjusting the preliminary divided image based on pixel values of pixels at the same position in the to-be-processed image and the adjacent image, and the result of the second division, and forming a fourth division corresponding to the to-be-processed image; and obtaining a result. adjusting the preliminary divided image based on pixel values of pixels at the same position in the to-be-processed image and the adjacent image, and the result of the second division, and forming a fourth division corresponding to the to-be-processed image; The step to get the result is
Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. 11. The image segmentation method of claim 10, further comprising adjusting the preliminary segmented image to obtain the fourth segmented result corresponding to the pending image. Based on pixels among the adjacent images that belong to the target object in the second division result and are located at the same position as the to-be-processed image and have a pixel value difference that is less than or equal to a third preset value. , adjusting the rudimentary segmented image to obtain the fourth segmented result corresponding to the pending image;
obtaining a first set of pixels based on pixels whose co-located pixel value difference in the pending image and the adjacent image is less than or equal to a third preset value;
obtaining a second set of pixels based on pixels belonging to the target object in the second segmentation result from the first set of pixels;
adjusting pixels of the second set of pixels in the preliminary segmented image to belong to the target object to obtain the fourth segmented result corresponding to the pending image. The image segmentation method according to claim 11. An image segmentation device,
a first segmentation unit configured to predict pixels belonging to a target object in a pending image and obtain a rudimentary segmented image corresponding to the pending image;
calculating predicted pixel values of pixels not belonging to the target object in a closed region included in the edge of the target object in the preliminary segmented image based on the edge information of the target object in the image to be processed; a first adjuster configured to adjust and obtain a first segmentation result corresponding to the pending image. An image segmentation device,
a second segmentation unit configured to predict pixels belonging to a target object in a pending image and obtain a rudimentary segmented image corresponding to the pending image;
a first acquisition unit configured to acquire an image adjacent to the to-be-processed image and a second division result corresponding to the adjacent image;
adjusting the preliminary divided image based on pixel values of pixels at the same position in the to-be-processed image and the adjacent image, and the result of the second division, and forming a fourth division corresponding to the to-be-processed image; and a second adjustment unit configured to obtain a result. a memory that stores executable instructions;
and one or more processors for calling executable instructions stored in the memory to perform the image segmentation method according to any one of claims 1-12. A computer readable storage medium storing computer program instructions for, when executed by a processor, causing said processor to perform the image segmentation method of any one of claims 1-12. A computer program product comprising computer readable code for, when run on an electronic device, causing a processor in said electronic device to perform the image segmentation method of any one of claims 1-12.

Images