KR102030533B1 - Image processing apparatus for adopting human body morphometric based on artificial neural network for sarcopenia and image processing method using the same - Google Patents

Abstract

According to an embodiment, an image processing apparatus may receive an image acquisition unit configured to acquire a plurality of tomography images and a plurality of images acquired by the image acquisition unit to form a 3D image of an object by being stacked. And a lumbar level selection unit for selecting an image including a region corresponding to the L3 lumbar level through a first artificial neural network model previously trained by a deep learning method, and a lumbar level corresponding to the L3 lumbar level. Receive at least one of subcutaneous fat, visceral fat and muscle in the image including the region corresponding to the L3 lumbar level through a second artificial neural network model that is previously trained by a deep learning method as an input including an image including a region It comprises a labeling unit to display.

Description

IMAGE PROCESSING APPARATUS FOR ADOPTING HUMAN BODY MORPHOMETRIC BASED ON ARTIFICIAL NEURAL NETWORK FOR SARCOPENIA AND IMAGE PROCESSING METHOD USING THE SAME}

The present invention relates to an image processing apparatus employing an artificial neural network-based human shape analysis method for the analysis of muscle reduction and an image processing method using the same.

Sarcopenia is one of the changes in body composition that accompanies aging. Recent epidemiological studies have revealed the effects of these myopathy on body metabolism, such as systemic weakness or increased incidence of various diseases. Thus, many studies for diagnosing myopathy have been proposed.

In the past, a specialist such as a doctor diagnosed myotropia by visually analyzing an image obtained from a medical imaging device such as a computerized tomography (CT) device or a magnetic resonance imaging (MRI) device. However, as the number of patients increases, the number of images that are the target of the work increases rapidly. Accordingly, there is a need for a technology capable of diagnosing muscular dystrophy more quickly, efficiently and accurately.

United States Patent Application Publication No. 2017-0046837 (published Feb. 16, 2017)

The problem to be solved by the present invention is to provide a technique for distinguishing muscle, subcutaneous fat or visceral fat from images obtained from a medical imaging device such as a CT device, through which the basic data for the diagnosis of myotropia is automatically To produce or provide.

However, the problem to be solved of the present invention is not limited to those mentioned above, another problem to be solved is not mentioned can be clearly understood by those skilled in the art from the following description. will be.

According to an embodiment, an image processing apparatus may receive an image acquisition unit configured to acquire a plurality of tomography images and a plurality of images acquired by the image acquisition unit to form a 3D image of an object by being stacked. And a lumbar level selection unit for selecting an image including a region corresponding to the L3 lumbar level through a first artificial neural network model previously learned by a deep learning method, and the L3 lumbar spine The subcutaneous fat, visceral fat, and muscles of the subcutaneous fat, visceral fat, and muscles are input to the image including the region corresponding to the L3 lumbar level through a second artificial neural network model that is previously trained by deep learning. It includes a labeling unit for distinguishing at least one display.

The image processing apparatus may further include a preprocessor configured to preprocess by applying at least one of an anisotropic diffusion filter and a Gaussian smoothing filter to each of the plurality of images obtained by the image acquisition unit. The lumbar level selection unit may receive a plurality of images pre-processed by the preprocessing unit as inputs, and then input the L3 lumbar level through a first artificial neural network model trained in advance by the deep learning method. An image including a corresponding area may be selected.

The image processing apparatus may further include a preprocessor configured to preprocess by applying at least one of a gradient magnitude, a sigmoid filter, and a normalization to each of the plurality of images acquired by the image acquirer. The lumbar level selector receives a plurality of images preprocessed by the preprocessor, and receives an image including an area corresponding to the L3 lumbar level through a first artificial neural network model previously learned by the deep learning method. You can choose.

The first artificial neural network model may include the first image including any one of the L3 lumbar level and the L4 lumbar level, and the L4 lumbar level without being the L3 lumbar level. Or a convolutional neural network (CNN) classifier previously trained to classify the second image.

The image processing apparatus may further include the L3 lumbar level and the L4 lumbar level based on a relative position of a first image classified by the convolutional neural network classifier in the entirety of the plurality of images. It may further include a lumbar level post-processing unit for determining which one of the.

In addition, the first artificial neural network model may be any one of a convolutional neural network detection (CNN detector) model or a convolutional neural network segmentation (CNN segmentation) model.

The image processing apparatus may further include a level-set technique, a rolling-ball algorithm, and a connected element analysis on an image in which at least one of the subcutaneous fat, visceral fat, and muscle is divided and displayed. The method may further include a labeling post-processing unit applying at least one of -component analysis) techniques.

The image processing method according to an embodiment of the present invention is performed by an image processing apparatus, and obtaining a plurality of tomographic images, constituting a 3D image of an object by being stacked, and inputting the obtained plurality of images as inputs. And selecting an image including an area corresponding to the L3 lumbar level by using a first artificial neural network model previously trained by the deep learning method, and selecting an area corresponding to the L3 lumbar level. At least one of subcutaneous fat, visceral fat, and muscle is distinguished and displayed on an image including a region corresponding to the L3 lumbar level through a second artificial neural network model that is previously trained by deep learning. It includes a step.

The image processing method may further include preprocessing at least one of an anisotropic diffusion filter and a Gaussian smoothing filter to each of the plurality of images obtained in the acquiring step. The selecting may include receiving a plurality of images pre-processed in the preprocessing step, and corresponding to L3 lumbar levels through a first artificial neural network model previously trained by the deep learning method. An image including an area to be selected may be selected.

The image processing method may further include preprocessing each of the plurality of images obtained in the obtaining by applying at least one of a gradient magnitude, a sigmoid filter, and a normalization. The selecting may include receiving a plurality of images pre-processed in the preprocessing step as inputs, and include a region corresponding to the L3 lumbar level through a first artificial neural network model previously learned in the deep learning method. You can select the video to play.

A computer-readable recording medium according to an embodiment stores a computer program programmed to perform each step included in the image processing method.

A computer program stored in a computer readable recording medium according to an embodiment is programmed to perform each step included in the image processing method.

According to an embodiment of the present invention, a process of selecting a portion corresponding to the L3 lumbar level in the medical image, and a process of distinguishing and displaying subcutaneous fat / visceral fat / muscle on the portion corresponding to the L3 lumbar level are deep learning in advance. It is performed by the trained model. Therefore, the above-described processes can be performed more quickly and efficiently than in the case where the process is performed by a person, and the accuracy of the output according to the process can be improved. Furthermore, the diagnosis of myopathy based on these outputs can also be made more efficiently, quickly and accurately.

1 conceptually illustrates a configuration of an image processing apparatus according to an exemplary embodiment.
FIG. 2 illustrates a result of preprocessing an image by the preprocessor illustrated in FIG. 1.
FIG. 3 illustrates a process in which an image is preprocessed by the preprocessor illustrated in FIG. 1.
FIG. 4 illustrates the result of selecting the lumbar level by the lumbar level selection unit shown in FIG. 1.
FIG. 5 illustrates a process in which an image is preprocessed by the preprocessor illustrated in FIG. 1.
FIG. 6 illustrates training data for a second artificial neural network model included in the labeling unit illustrated in FIG. 1.
FIG. 7 illustrates training data for a second artificial neural network model included in the labeling unit illustrated in FIG. 1.
8 illustrates an output of an image processing apparatus according to an exemplary embodiment.
9 is a flowchart of an image processing method, according to an exemplary embodiment.

Advantages and features of the present invention, and methods of achieving them will be apparent with reference to the embodiments described below in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but can be implemented in various forms, and only the embodiments are to make the disclosure of the present invention complete, and common knowledge in the art to which the present invention pertains. It is provided to fully inform the person having the scope of the invention, which is defined only by the scope of the claims.

Terms used herein will be briefly described and the present invention will be described in detail.

The terms used in the present invention have been selected as widely used general terms as possible in consideration of the functions in the present invention, but this may vary according to the intention or precedent of the person skilled in the art, the emergence of new technologies, and the like. In addition, in certain cases, there is also a term arbitrarily selected by the applicant, in which case the meaning will be described in detail in the description of the invention. Therefore, the terms used in the present invention should be defined based on the meanings of the terms and the contents throughout the present invention, rather than the names of the simple terms.

When a part of the specification 'includes' a certain component, this means that it may further include other components, except to exclude other components unless otherwise stated.

In addition, the term 'part' as used herein refers to software or a hardware component such as an FPGA or an ASIC, and 'part' plays a role. But wealth is not limited to software or hardware. The 'unit' may be configured to be in an addressable storage medium or may be configured to play one or more processors. Thus, as an example, a 'part' may include components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, procedures, Subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays and variables. The functionality provided within the components and 'parts' may be combined into a smaller number of components and 'parts' or further separated into additional components and 'parts'.

In addition, in the present specification, the expression 'first', 'second' or 'first-1', etc. is an exemplary term for referring to different components, objects, images, pixels, or patches. Accordingly, the expressions such as 'first', 'second' or 'first-1' do not indicate the order between the components or indicate the priority.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention. In the drawings, parts irrelevant to the description are omitted in order to clearly describe the present invention.

As used herein, an 'object' includes a person or an animal, or a part of a person or an animal. For example, the subject may include organs such as the liver, heart, uterus, brain, breast, abdomen, blood vessels, bones, or the like.

In addition, in the present specification, 'image' refers to multi-dimensional data composed of discrete image elements (eg, pixels in a two-dimensional image and voxels in a three-dimensional image). Accordingly, the image includes, for example, organs, blood vessels, bones, and the like of the liver, heart, uterus, brain, breast, and abdomen of the aforementioned object.

In addition, in the present specification, the medical imaging apparatus includes a CT apparatus, an MRI apparatus, an X-ray apparatus, an ultrasound diagnostic apparatus, and the like, and may output the above-described image.

In addition, in the present specification, the 'user' may be a doctor, a nurse, a clinical pathologist, a medical imaging expert, or the like as a medical expert, but may be a technician repairing a medical device, but is not limited thereto.

FIG. 1 illustrates a configuration of an image processing apparatus 100 according to an exemplary embodiment, but the example illustrated in FIG. 1 is merely exemplary. The image processing apparatus 100 may be implemented in a PC or a server.

Referring to FIG. 1, the image processing apparatus 100 includes an image obtaining unit 110, a lumbar level selecting unit 130, and a labeling unit 150, and according to an embodiment, the preprocessing unit 120 and the lumbar level after The processor 140 may further include a labeling post-processing unit 160 and a display unit 170. Here, the image acquisition unit 110 may be implemented as a port or a wired / wireless communication module for receiving data. In addition, the preprocessor 120, the lumbar level selector 130, the lumbar level postprocessor 140, the labeling unit 150, and the labeling postprocessor 160 store instructions programmed to perform the functions described below. Memory and a microprocessor that executes these instructions. In addition, the display unit 170 may be implemented by a display device such as a monitor.

Referring to the image acquisition unit 110, the image acquisition unit 110 receives a plurality of tomography images at predetermined intervals with respect to the object. The image acquisition unit 110 may directly receive such an image from the above-described medical imaging apparatus.

The preprocessor 120 performs a pre-processing process on the plurality of images input by the image acquirer 110. When the preprocessing process is performed, noise may be removed from the image as shown in FIG. 2, and a boundary area such as an edge may be clearer. In addition, as will be described later, the preprocessor 120 may perform the above-described preprocessing on the image selected by the lumbar level selector 130, according to an embodiment.

In the pretreatment process, various things will be described below. First, in order to remove noise included in each of the plurality of images, an anisotropic diffusion filter or a Gaussian smoothing filter is repeatedly applied. The second is to apply a gradient magnitude (sigma magnitude) or sigmoid filter or normalization to produce an edge image. Thirdly, region growing is performed by considering each point of the edge image generated in the second as a seed. Fourth, the morphology operation is repeatedly applied to the third result. At this time, according to the embodiment, some or all of the first to fourth pretreatment processes may be performed, or none of the above-described pretreatment processes may be performed.

This pretreatment process is shown in FIG. 3. Referring to FIG. 3, the leftmost side of FIG. 3 illustrates a plurality of images stacked by the image acquisition unit 110. The various filters and histograms described above, which may be applied to the preprocessing process while moving to the right side from the left side of FIG. And the like are shown.

The lumbar level selector 130 selects an image including a region corresponding to the L3 lumbar level from the plurality of images. In this case, the plurality of images that are the targets of the lumbar level selector 130 may be images preprocessed by the preprocessor 120. Referring to FIG. 4, an image including a region corresponding to the L3 lumbar level among the plurality of images is displayed in red.

The lumbar level selector 130 uses a first artificial neural network model to select an image including a region corresponding to the L3 lumbar level. The first artificial neural network model refers to a model that has been previously learned to output whether the image includes the region of the L3 lumbar level when an image is input. In this case, the learning method may be deep learning. This first artificial neural network model can be implemented in various ways as follows.

First, the first artificial neural network model may be implemented as a convolutional neural network (CNN) classifier. The convolutional neural classifier is trained by deep learning in advance, and if the target image includes the region corresponding to the L3 lumbar level or includes the region corresponding to the L4 lumbar level, the image is classified as the first image. If none of the L3 lumbar level and the L4 lumbar level, the corresponding image is classified as a second image. Since the convolutional neural classifier itself is a known technique, a detailed description thereof will be omitted.

When the first artificial neural network model is implemented as a convolutional neural network classifier, the image processing apparatus 100 may further include a lumbar level post-processing unit 140. The lumbar level post-processing unit 140 classifies whether the above-described first image corresponds to one of the L3 lumbar level and the L4 lumbar level. For example, if the classification method, the relative position (height) of the first image in the plurality of images obtained by the image acquisition unit 110 corresponds to the position of a predefined L3 lumbar level or L4 lumbar level The classification may be performed according to whether the position corresponds to. Alternatively, the anatomical features of the first region may be analyzed to classify whether the first image is an image corresponding to the L3 lumbar level or an image corresponding to the L4 lumbar level, but is not limited thereto.

Second, the first artificial neural network model may be implemented as either a convolutional neural network detector (CNN detector) or a convolutional segmentation (CNN segmentation) model. When implemented with either a CNN detector or a CNN segmentation model, the preprocessor 120 shown in FIG. 1 may perform the following preprocessing in addition to the following preprocessing process, that is, in addition to the preprocessing process described above. . For example, the preprocessor 120 obtains a coronal view, a sagittal view and a axial view from a three-dimensional image in which a plurality of images are stacked. Subsequently, the preprocessing unit 120 generates a bounding box for each of the coronal plane, sagittal plane and horizontal plane by applying a region of interest segmentation technique to each of the coronal plane, sagittal plane, and horizontal plane. Find the center of gravity. FIG. 5 exemplarily shows that a bounding box is generated for each of the coronal plane, sagittal plane and horizontal plane, and the center of gravity thereof is indicated. Here, since the ROI division technique itself is a known technique, a detailed description thereof will be omitted.

The CNN detector model or the CNN segmentation model is pre-trained by deep learning to receive or display the location of the L3 lumbar level in the coronal, sagittal, and horizontal planes with the center of gravity of the bounding box. . Therefore, the 'input' image as the training data is the image in the coronal plane, sagittal plane and horizontal plane, the 'correct' image is the image in which the position of the L3 lumbar level is displayed. 6A illustrates training data of the CNN detector model, and FIG. 6B illustrates training data of the CNN segmentation model. In this case, the training data or the data input during the actual driving may be a whole region or a region of interest (ROI) of each image in the coronal plane, the sagittal plane, and the horizontal plane. , Which is shown in FIG. 7.

Looking at the CNN detector model and the training process of the model described below, in the training process, a plurality of artificial neurons included in the middle layer of each model are interconnected with the input and the output by an interconnection network. Each of these networks can be weighted. This weight may be determined or changed to maximize the similarity between the input and the correct answer every time a plurality of training data is input and processed. Here, the process itself or the principle of learning as described above is already known technology, so further description thereof will be omitted.

Third, the first artificial neural network model may be implemented as a model using a maximum intensity difference accumulation (MIDA) and a maximum intensity projection (MIP) image. In the MIDA image or the MIP image, the object is displayed around the bone. Therefore, the first artificial neural network model applies a CNN detector model or a CNN segmentation model to the L3 region of the actual spine based on anatomical information.

That is, according to an embodiment, the selection process of the image including the region corresponding to the L3 lumbar level is performed by a model that has been previously learned in a deep learning manner. Thus, the accuracy, speed and efficiency of the selection process can be improved.

Referring back to FIG. 1, the labeling unit 150 classifies subcutaneous fat, visceral fat, or muscle on an image selected as including the region of the L3 lumbar level by the lumbar level selection unit 130. At this time, according to an embodiment, the preprocessing process may be performed on the image selected by the lumbar level selection unit 130 by the preprocessor 120, and the labeling unit 150 may perform the preprocessing process as described above. Subcutaneous fat, visceral fat or muscles can be identified and displayed on the image.

The labeling unit 150 uses a second artificial neural network model to separately display subcutaneous fat, visceral fat, or muscle in the image. The second artificial neural network model refers to a model that has been previously learned in a deep learning manner so that subcutaneous fat, visceral fat, and muscle are separately displayed on the corresponding image when the image is input. The second artificial neural network model may be implemented by modifying some layers of the ResNet model, but is not limited thereto.

The labeling post-processing unit 160 is configured to post-process the result labeled by the labeling unit 150. There are a variety of post-processing steps to be described below. First, the change in the shape of the geometric curve shown in the image is corrected using a level-set technique. Secondly, there is a rolling ball algorithm to minimize false positive errors near the boundary. Third, there is a connected-component analysis technique. Fourth, the threshold value of a previously set hounsfield unit (HU) is applied based on actual pixel values (12-16 bits) of the image acquired by the image acquisition unit 110. According to an embodiment, some or all of the first to fourth post-treatment processes may be performed, or no post-treatment process may be performed. The result of this post-process is shown in FIG. 8.

As described above, according to an embodiment, the process of selecting a portion corresponding to the L3 lumbar level in the medical image, and the process of distinguishing and displaying subcutaneous fat / visceral fat / muscle on the portion corresponding to the L3 lumbar level It is performed by a previously trained model in a deep learning manner. Therefore, the above-described processes can be performed more quickly and efficiently than in the case where the process is performed by a person, and the accuracy of the output according to the process can be improved. Furthermore, the diagnosis of myopathy based on these outputs can also be made more efficiently, quickly and accurately.

9 conceptually illustrates a procedure of an image processing method, according to an exemplary embodiment. Such an image processing method may be performed by the image processing apparatus 100 described above.

On the other hand, since the procedure shown in Figure 9 is merely exemplary, the spirit of the present invention is not limited to the one shown in Figure 9, according to the embodiment the procedure is performed in a different order than that shown in Figure 9 At least one of the procedures shown in FIG. 9 may not be performed, and a procedure not shown in FIG. 9 may be additionally performed.

First, in operation S100, a plurality of tomography-photographed images constituting a 3D image of an object by stacking are obtained.

The method may further include selecting an image including an area corresponding to the L3 lumbar level through a first artificial neural network model trained in advance by using the deep learning method, by receiving the obtained plurality of images as an input ( S110) is performed.

At this time, before performing step S110, the above-described preprocessing process may be performed on the plurality of images acquired in step S100.

In addition, the subcutaneous fat in the image including the region corresponding to the L3 lumbar level through a second artificial neural network model that is previously trained by receiving the image including the region corresponding to the L3 lumbar level as an input, A step (S120) of distinguishing visceral fat and muscles is performed.

In this case, before performing step S120, the above-described preprocessing process may be performed on an image including a region corresponding to the L3 lumbar level in step S110.

Meanwhile, since details of the image processing method overlap with those of the image processing apparatus 100, a description of the image processing apparatus 100 will be used.

Meanwhile, each step included in the image processing method according to the above-described embodiment may be implemented in a computer readable recording medium for recording a computer program programmed to perform such steps.

In addition, each step included in the image processing method according to the above-described embodiment may be implemented in the form of a computer program stored in a computer-readable recording medium programmed to perform such steps.

The above description is merely illustrative of the technical idea of the present invention, and those skilled in the art to which the present invention pertains may make various modifications and changes without departing from the essential quality of the present invention. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention but to describe the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted by the following claims, and all technical ideas that fall within the scope of equivalents should be construed as being included in the scope of the present invention.

100: image processing device

Claims (12)

An image acquisition unit configured to acquire a plurality of tomographic images, which are stacked to form a 3D image of the object;
A coronal view, a sagittal view and a axial view are obtained for each of the obtained plurality of images, and a bounding box is obtained from each of the obtained coronal, sagittal and horizontal planes. After generating the pre-processing unit for displaying the center of gravity of each of the generated bounding box,
The preprocessing unit receives a plurality of images in which the center of gravity is displayed as an input, and corresponds to an L3 lumbar level through a first artificial neural network model previously learned by a deep learning method. A lumbar level selector for selecting an image including an area;
Subcutaneous fat and visceral fat are received in an image including the region corresponding to the L3 lumbar level through a second artificial neural network model that is previously trained by a deep learning method as an input of an image including the region corresponding to the L3 lumbar level. And a labeling unit for distinguishing and displaying at least one of the muscles.
An image processing device employing artificial body network-based Human Body Morphometrics to support sarcopenia analysis. The method of claim 1,
The preprocessing unit,
Pre-processing by applying at least one of an anisotropic diffusion filter and a Gaussian smoothing filter to each of the plurality of images obtained by the image acquisition unit
An image processing device employing artificial body network-based Human Body Morphometrics to support sarcopenia analysis. The method of claim 1,
The preprocessing unit,
Preprocessing each of the plurality of images obtained by the image acquisition unit by applying at least one of a gradient magnitude, a sigmoid filter, and a normalization.
An image processing device employing artificial body network-based Human Body Morphometrics to support sarcopenia analysis. The method of claim 1,
The first artificial neural network model,
The input plurality of images may be classified into a first image including any one of an L3 lumbar level and an L4 lumbar level, and a second image that does not correspond to the L4 lumbar level but does not correspond to the L3 lumbar level. Learned convolutional neural network (CNN) classifier
An image processing device employing artificial body network-based Human Body Morphometrics to support sarcopenia analysis. The method of claim 4, wherein
Determining whether the classified first image corresponds to one of the L3 lumbar level and the L4 lumbar level based on a relative position of the first image classified by the convolutional neural network classifier in the entire plurality of images. Lumbar level post-processing further comprising
An image processing device employing artificial body network-based Human Body Morphometrics to support sarcopenia analysis. The method of claim 1,
The first artificial neural network model,
Either the convolutional neural network detection model or the convolutional neural network segmentation model
An image processing device employing artificial body network-based Human Body Morphometrics to support sarcopenia analysis. The method of claim 1,
At least one of a level-set technique, a rolling-ball algorithm, and a connected-component analysis technique for an image in which at least one of the subcutaneous fat, visceral fat, and muscle are divided and displayed. Further comprising a labeling post-processing portion applying one
An image processing device employing artificial body network-based Human Body Morphometrics to support sarcopenia analysis. An image processing method performed by an image processing apparatus employing an artificial neural network based human body morphometrics to support sarcopenia analysis,
Obtaining a plurality of tomography images, which are stacked to form a 3D image of the object;
Acquiring a coronal view, a sagittal view, and a axial view for each of the obtained plurality of images;
Generating a bounding box in each of the obtained coronal plane, sagittal plane and horizontal plane;
Displaying a center of gravity at each of the generated bounding boxes;
Selecting an image including an area corresponding to the L3 lumbar level through a first artificial neural network model previously trained by a deep learning method, by receiving a plurality of images displaying the center of gravity as an input; Wow,
Subcutaneous fat and visceral fat are received in an image including the region corresponding to the L3 lumbar level through a second artificial neural network model that is previously trained by a deep learning method as an input of an image including the region corresponding to the L3 lumbar level. And classifying and displaying at least one of muscles
Image processing method. The method of claim 8,
Pre-processing each of the plurality of images obtained in the obtaining step by applying at least one of an anisotropic diffusion filter and a Gaussian smoothing filter.
Image processing method. The method of claim 8,
And preprocessing each of the plurality of images obtained in the obtaining step by applying at least one of a gradient magnitude, a sigmoid filter, and a normalization.
Image processing method. A computer program stored in a computer readable recording medium programmed to perform each step included in the image processing method of any one of claims 8 to 10. A computer-readable recording medium storing a computer program programmed to perform each step included in the image processing method of any one of claims 8 to 10.

Images