KR102242114B1 - Oct medical image based artificial intelligence computer aided diagnosis system and its method - Google Patents

Abstract

The present invention relates to an OCT medical image-based artificial intelligence computer-assisted diagnosis system and method thereof, and more specifically, as an OCT medical image-based artificial intelligence computer-assisted diagnosis system, from an IVOCT image taken by intravascular optical coherence tomography (IVOCT). A blood vessel lumen automatic detection unit that automatically detects the shape of the lumen of the blood vessel; A shape feature extraction unit for extracting a shape feature of the blood vessel lumen from the shape of the blood vessel lumen detected by the automatic blood vessel lumen detection unit; A vascular lumen classifying unit for classifying a normal vascular lumen and an abnormal vascular lumen according to the shape characteristics of the vascular lumen extracted from the shape feature extraction unit; And a display providing unit configured to provide a normal or abnormal vascular lumen classified according to the vascular lumen classification unit through a display.
The artificial intelligence computer-assisted diagnosis system and method based on OCT medical images proposed in the present invention automatically detects the inner diameter of a blood vessel from an IVOCT image taken by intravascular optical coherence tomography (IVOCT), and uses a pattern recognition technology to detect the lumen of the blood vessel. After extracting the shape features of, the extracted shape specification can be classified into normal vascular lumen and abnormal vascular lumen through SVM (Support Vector Machine) and provided.
In addition, according to the present invention, when an IVOCT image automatically classified into a normal vascular lumen and an abnormal vascular lumen based on the shape characteristics of the vascular lumen is provided through the display providing unit, normal or abnormal is further displayed on the IVOCT image and provided, By assisting the doctor's IVOCT image reading and diagnosis process, the probability of a false diagnosis can be reduced.

Description

OCT medical image-based artificial intelligence computer-assisted diagnosis system and its method {OCT MEDICAL IMAGE BASED ARTIFICIAL INTELLIGENCE COMPUTER AIDED DIAGNOSIS SYSTEM AND ITS METHOD}

The present invention relates to an artificial intelligence computer-assisted diagnostic system based on OCT (Optical Coherence Tomography) medical image and a method thereof, and more specifically, to a normal vascular lumen and abnormal vascular lumen in IVOCT images taken by intravascular optical coherence tomography (IVOCT). The present invention relates to an artificial intelligence computer-assisted diagnosis system based on OCT medical images and a method thereof that can be classified into and provided.

Currently, in the medical field, a method of photographing and observing internal structures such as biological tissues in a human body is widely used. For example, various transmission and tomography images of biological tissues in the human body are taken using an X-ray device, computerized tomography (CT), magnetic resonance image (MRI), and ultrasonography (Ultrasound). do. This method can grasp the internal structure of a living body tissue requiring diagnosis without directly cutting the human body. Therefore, this observation method is very important in the field of diagnostic medicine, since it can diagnose various diseases and grasp the progression of the disease. However, since the X-ray apparatus uses radiation, it can have a detrimental effect on the human body. In addition, CT is limitedly used only in some large hospitals because the size of the imaging device is very large and the price is high. In addition, the MRI and ultrasound apparatuses have a disadvantage in that the resolution of the captured image is deteriorated.

As an alternative to this, optical coherence tomography (OCT) or intravascular optical coherence tomography (IVOCT), which is used as an alternative, can photograph the inside of blood vessels in the human body in high resolution using the phenomenon of light interference. More specifically, in optical coherence tomography (IVOCT), an interference light source with a short wavelength is irradiated through a catheter, and reflected and collected light is analyzed, thereby sub-micron (Sub-micron) tomography of a minute portion of a vascular tissue. micron) area can be photographed in high resolution. These optical coherence tomography (OCT) or intravascular optical coherence tomography (IVOCT) are harmless to the human body and can acquire high-resolution (10-20㎛) images to observe the vascular tissue in the human body, so it is useful for diagnosing vascular diseases, etc. It is mainly used.

Radiology doctors can read high-resolution images of intravascular optical coherence tomography (IVOCT) to determine the shape and area of the lumen of the vessel, the microstructure of the vessel wall, the coronary artery, and the stent installed in the vessel. However, even if a doctor reads IVOCT images, the shape of the vascular lumen is very irregular, and it is difficult to clearly determine the normal lumen and abnormal lumen due to the shadow of the stand, so it is difficult to make a misdiagnosis during the IVOCT image reading and diagnosis process. There was a possibility. Korean Patent Application Publication No. 10-2014-0133372 has been disclosed as a prior art document.

The present invention has been proposed to solve the above problems of the previously proposed methods, and automatically detects the inner diameter of a blood vessel from an IVOCT image taken by IVOCT, and uses a pattern recognition technology to detect the lumen of the blood vessel. After extracting the shape features of the OCT medical image-based artificial intelligence computer-assisted diagnosis system and its method, the extracted shape specification can be classified into normal and abnormal vascular lumens through SVM (Support Vector Machine). It aims to provide.

In addition, in the present invention, when an IVOCT image automatically classified into a normal vascular lumen and an abnormal vascular lumen based on the shape characteristics of the vascular lumen is provided through the display providing unit, the normal or abnormal is further displayed on the IVOCT image. Another object of the present invention is to provide an OCT medical image-based artificial intelligence computer-assisted diagnosis system and method that assists in the process of reading and diagnosing an IVOCT image so that the probability of a false diagnosis can be reduced.

An artificial intelligence computer-assisted diagnosis system based on an OCT medical image according to a feature of the present invention for achieving the above object,

As an artificial intelligence computer-assisted diagnosis system based on OCT medical images,

An automatic vascular lumen detection unit for automatically detecting a lumen shape of a blood vessel from an IVOCT image taken by intravascular optical coherence tomography (IVOCT);

A shape feature extraction unit for extracting a shape feature of the blood vessel lumen from the shape of the blood vessel lumen detected by the automatic blood vessel lumen detection unit;

A vascular lumen classification unit for classifying a normal vascular lumen and an abnormal vascular lumen according to the shape characteristics of the vascular lumen extracted from the shape feature extraction unit; And

It is characterized in that the configuration comprises a display providing unit for providing through a display the normal or abnormal vascular lumen classified according to the vascular lumen classification unit.

Preferably, the blood vessel lumen automatic detection unit,

It may include a rectangular coordinate image conversion unit for converting and processing IVOCT images captured and provided by intravascular optical coherence tomography (IVOCT) into rectangular coordinate images.

More preferably, the blood vessel lumen automatic detection unit,

It may further include a boundary extracting unit configured to receive an IVOCT image converted into a rectangular coordinate image from the rectangular coordinate image conversion unit and perform convolution with an asymmetric Gaussian filter to extract a boundary of a blood vessel lumen.

Even more preferably, the blood vessel lumen automatic detection unit,

A shadow area detection unit for detecting a guide wire shadow area of a corresponding pattern at the boundary of the blood vessel lumen extracted by the boundary extraction unit may be further included.

Even more preferably, the guide wire shadow area,

The row length may be 220 pixels or more, and the column width may be 105 pixels or 175 pixels pattern.

Even more preferably, the blood vessel lumen automatic detection unit,

A shadow area removal unit for removing a shadow area of a guide wire of a corresponding pattern detected by the shadow area detection unit at a boundary of the blood vessel lumen extracted by the boundary extraction unit may be further included.

Even more preferably, the shadow area removal unit,

An interpolation process may be further performed so that the image processing is not affected with respect to the area from which the shadow area has been removed.

Even more preferably, the vascular lumen automatic detection unit,

Further comprising a stent strut processing unit that removes the shadow area of the guide wire in the shadow area removal unit and detects the shadow area of the stent strut at the boundary of the blood vessel lumen detected through an asymmetric Gaussian filter subjected to interpolation, and fills it with a preset high pixel value. Configurable.

Even more preferably, the vascular lumen automatic detection unit,

It may be configured to further include a blood vessel inner diameter shape detection unit that precisely detects the irregular shape of the lumen of the blood vessel using an active contour model (Active Contour Model).

Preferably, the shape feature extraction unit,

Shape features and texture features may be extracted from the lumen shape of the blood vessel detected by the automatic blood vessel lumen detection unit.

More preferably, the shape feature extraction unit,

Shape features and texture features may be extracted for the lumen shape of the blood vessel detected by the automatic blood vessel lumen detection unit, but 38 shape features and 54 texture features may be extracted.

Even more preferably, the vascular lumen classification unit,

A feature vector may be defined through shape features and texture features of the blood vessel lumen extracted from the shape feature extraction unit.

Even more preferably, the vascular lumen classification unit,

The defined feature vector may be classified into a normal IVOCT image and an abnormal IVOCT image using SVM (Support Vector Machine) technology.

Even more preferably, the display providing unit,

An IVOCT image classified as normal or abnormal may be displayed so as to assist a doctor's IVOCT image reading and diagnosis process, but the normal or abnormal may be further displayed and provided on the displayed classified IVOCT image.

An artificial intelligence computer-assisted diagnosis method based on an OCT medical image according to a feature of the present invention for achieving the above object,

As an artificial intelligence computer-assisted diagnosis method based on OCT medical image,

(1) a step of automatically detecting the lumen shape of a blood vessel from an IVOCT image taken by an intravascular optical coherence tomography (IVOCT), by an automatic vascular lumen detection unit;

(2) extracting, by the shape feature extraction unit, shape features of the blood vessel lumen from the lumen shape of the blood vessel detected by the automatic blood vessel lumen detection unit in step (1);

(3) classifying the vascular lumen classification unit into normal vascular lumen and abnormal vascular lumen according to the shape characteristics of the vascular lumen extracted from the shape feature extraction unit of step (2); And

(4) The display providing unit is characterized in that it comprises the step of providing through a display the normal or abnormal vascular lumen classified according to the vascular lumen classification unit of step (3).

Preferably, in step (1),

(1-1) The vascular lumen automatic detection unit converts the IVOCT images captured and provided by intravascular optical coherence tomography (IVOCT) to rectangular coordinate images, and then convolves with an asymmetric Gaussian filter to extract the boundary of the vascular lumen. It may include steps.

More preferably, in the step (1),

(1-2) The blood vessel lumen automatic detection unit detects and removes a guide wire shadow region of a corresponding pattern at the boundary of the blood vessel lumen extracted in step (1-1), and the shadow region is removed. It may further include performing interpolation processing so that the image processing is not affected.

Even more preferably, the guide wire shadow area,

The row length may be 220 pixels or more, and the column width may be 105 pixels or 175 pixels pattern.

Even more preferably, in the step (1),

(1-3) The automatic blood vessel lumen detection unit detects the shadow region of the stent strut in the blood vessel lumen detected through the asymmetric Gaussian filter interpolated after removing the shadow region of the guide wire in step (1-2), Filling with a high pixel value may be further included.

Even more preferably, in the step (1),

(1-4) further comprising the step of accurately detecting the shape of the irregular vascular lumen using an active contour model for the vascular lumen processed in step (1-3) by the automatic vascular lumen detection unit It can be done by doing.

Preferably, in step (2),

The shape feature extraction unit may extract 38 shape features and 54 texture features from the luminal shape of the blood vessel detected by the automatic blood vessel lumen detection unit.

More preferably, in the step (3),

The vascular lumen classification unit defines a feature vector based on 38 shape features and 54 texture features of the vascular lumen extracted from the shape feature extraction unit in step (2), and normal IVOCT using SVM (Support Vector Machine) technology. It can be classified into images and abnormal IVOCT images.

Even more preferably, in the step (4),

The display providing unit displays IVOCT images classified as normal or abnormal so as to assist the doctor's IVOCT image reading and diagnosis process, but the displayed classified IVOCT images may be further displayed and provided with normal or abnormal.

The artificial intelligence computer-assisted diagnosis system and method based on OCT medical images proposed in the present invention automatically detects the inner diameter of a blood vessel from an IVOCT image taken by intravascular optical coherence tomography (IVOCT), and uses a pattern recognition technology to detect the lumen of the blood vessel. After extracting the shape features of, the extracted shape specification can be classified into normal vascular lumen and abnormal vascular lumen through SVM (Support Vector Machine) and provided.

In addition, according to the present invention, when an IVOCT image automatically classified into a normal vascular lumen and an abnormal vascular lumen based on the shape characteristics of the vascular lumen is provided through the display providing unit, normal or abnormal is further displayed on the IVOCT image and provided, It is possible to reduce the probability of a false diagnosis by assisting the doctor's IVOCT image reading and diagnosis process.

1 is a diagram showing a configuration of an artificial intelligence computer-assisted diagnosis system based on an OCT medical image according to an embodiment of the present invention as a function block.
2 is a diagram showing a configuration of an automatic vascular lumen detection unit applied to an artificial intelligence computer-assisted diagnosis system based on an OCT medical image according to an embodiment of the present invention as a functional block.
3 is a diagram illustrating a process of detecting a shadow area of a stent strut in an artificial intelligence computer-assisted diagnosis system based on an OCT medical image according to an embodiment of the present invention.
4 is a diagram illustrating a workflow of an artificial intelligence computer-assisted diagnosis system based on an OCT medical image according to an embodiment of the present invention.
5 is a diagram showing 38 shape features and 54 texture features for lumen classification of an artificial intelligence computer-assisted diagnosis system based on an OCT medical image according to an embodiment of the present invention.
6 is a diagram illustrating a flow of an artificial intelligence computer-assisted diagnosis method based on an OCT medical image according to an embodiment of the present invention.
7 is a diagram illustrating a flow of automatically detecting a lumen shape of a blood vessel in step S110 in an artificial intelligence computer-assisted diagnosis method based on an OCT medical image according to an embodiment of the present invention.

Hereinafter, preferred embodiments will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art may easily implement the present invention. However, in describing a preferred embodiment of the present invention in detail, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. In addition, the same reference numerals are used throughout the drawings for parts having similar functions and functions.

In addition, in the entire specification, when a part is said to be'connected' with another part, it is not only'directly connected', but also'indirectly connected' with another element in the middle. Includes. In addition, "including" certain components means that other components may be further included rather than excluding other components unless otherwise stated.

1 is a diagram showing the configuration of an OCT medical image-based artificial intelligence computer-assisted diagnosis system according to an embodiment of the present invention as a function block, and FIG. 2 is a OCT medical image-based artificial intelligence computer according to an embodiment of the present invention. A diagram showing a configuration of an automatic blood vessel lumen detection unit applied to an auxiliary diagnosis system as a functional block, and FIG. 3 is a process of detecting a shadow area of a stent strut of an artificial intelligence computer-assisted diagnosis system based on an OCT medical image according to an embodiment of the present invention FIG. 4 is a diagram illustrating a workflow of an artificial intelligence computer-assisted diagnosis system based on an OCT medical image according to an embodiment of the present invention, and FIG. 5 is an OCT according to an embodiment of the present invention. A diagram showing 38 shape features and 54 texture features for lumen classification in a medical image-based artificial intelligence computer-assisted diagnosis system. 1 to 5, the artificial intelligence computer-assisted diagnosis system 100 based on an OCT medical image according to an embodiment of the present invention includes an automatic vascular lumen detection unit 110, a shape feature extraction unit 120 , It may be configured to include a blood vessel lumen classification unit 130, and a display providing unit 140.

The automatic blood vessel lumen detection unit 110 is configured to automatically detect the lumen shape of a blood vessel from an IVOCT image photographed by intravascular optical coherence tomography (IVOCT). As shown in FIG. 2, the automatic blood vessel lumen detection unit 110 includes a rectangular coordinate image conversion unit 111, a boundary extraction unit 112, a shadow area detection unit 113, and a shadow area removal unit 114. Wow, it may be configured to include a stent strut processing unit 115, and a blood vessel inner diameter shape detection unit 116.

The rectangular coordinate image conversion unit 111 of the automatic blood vessel lumen detection unit 110 may convert IVOCT images captured and provided by intravascular optical coherence tomography (IVOCT) into a rectangular coordinate image. The rectangular coordinate image conversion unit 111 converts the IVOCT image into a rectangular coordinate image, thereby obtaining coordinate information for each pixel constituting the IVOCT image, and as a result, the IVOCT image can be converted into a matrix. have.

In addition, the boundary extraction unit 112 of the automatic vascular lumen detection unit 110 receives the IVOCT image converted into a rectangular coordinate image from the rectangular coordinate image conversion unit 111 and convolves with an asymmetric Gaussian filter to determine the boundary of the lumen of the blood vessel. Can be extracted. Here, the asymmetric Gaussian filter is a function that detects whether or not a feature exists in the data, and may be configured as a matrix.

In addition, the shadow area detection unit 113 of the blood vessel lumen automatic detection unit 110 may detect a guide wire shadow area of a corresponding pattern at the boundary of the blood vessel lumen extracted by the boundary extraction unit 112. Here, the guide wire shadow area may have a row length of 220 pixels or more and a column width of 105 pixels or a 175 pixel pattern. At this time, the shadow area of the guide wire refers to an area in which a black shadow is formed behind the guide wire as the guide wire absorbs all light. In the IVOCT image, there is a shadow area of this guide wire.

In addition, the shadow region removal unit 114 of the blood vessel lumen automatic detection unit 110 removes the guide wire shadow region of the corresponding pattern detected by the shadow region detection unit 113 from the boundary of the blood vessel lumen extracted by the boundary extraction unit 112 can do. In this case, the shadow area removal unit 114 may further perform interpolation processing so that the image processing is not affected with respect to the area from which the shadow area has been removed.

In addition, the stent strut processing unit 115 of the automatic blood vessel lumen detection unit 110 may detect the shadow area of the stent strut at the boundary of the lumen of the blood vessel detected through the asymmetric Gaussian filter, and fill it with a preset high pixel value. Here, the stent strut is a structure that supports the stent, and the shadow area of the stent strut is considered as an artifact in the image processing process, like the shadow area of the guide wire, and thus can be removed from the IVOCT image. 3 shows a process of detecting a shadow area of a tm tent strut in the stent strut processing unit 115.

In addition, the vessel inner diameter shape detection unit 116 of the vessel lumen automatic detection unit 110 may accurately detect the irregular shape of the vessel lumen using an active contour model. Here, the active contour model (ACM) is a method of actively extracting the contour of an object, also known as Snakes, and may extract a boundary in a direction that optimizes a globally defined boundary.

The shape feature extraction unit 120 is configured to extract shape features of the blood vessel lumen from the lumen shape of the blood vessel detected by the automatic blood vessel lumen detection unit 110. The shape feature extraction unit 120 may extract shape features and texture features from the luminal shape of the blood vessel detected by the automatic blood vessel lumen detection unit 110. That is, the shape feature extraction unit 120 extracts shape features and texture features for the lumen shape of the blood vessel detected by the automatic blood vessel lumen detection unit 110, but extracts 38 shape features and 54 texture features. do. FIG. 5 shows descriptions of 38 shape features and 54 texture features for lumen classification in an artificial intelligence computer-assisted diagnosis system based on an OCT medical image.

The vascular lumen classification unit 130 is configured to classify into a normal vascular lumen and an abnormal vascular lumen according to the shape characteristics of the vascular lumen extracted from the shape feature extraction unit 120. The vascular lumen classification unit 130 may define a feature vector through shape features and texture features of the vascular lumen extracted from the shape feature extraction unit 120. Here, the vascular lumen classifier 130 classifies the defined feature vector into a normal IVOCT image and an abnormal IVOCT image using a support vector machine (SVM) technology.

The display providing unit 140 is a configuration of a display output unit that provides a normal or abnormal vascular lumen classified according to the vascular lumen classification unit 130 through a display. The display providing unit 140 displays IVOCT images classified as normal or abnormal so as to assist the doctor's IVOCT image reading and diagnosis process, but may be further displayed and provided as normal or abnormal on the displayed classified IVOCT image. have. This assists the doctor's IVOCT image reading and diagnosis process, thereby reducing the possibility of a doctor's misdiagnosis.

6 is a diagram illustrating a flow of an artificial intelligence computer-assisted diagnosis method based on an OCT medical image according to an embodiment of the present invention, and FIG. 7 is a flowchart illustrating an artificial intelligence computer-assisted diagnosis method based on an OCT medical image according to an embodiment of the present invention. In the figure, in step S110, the flow of automatically detecting the shape of the lumen of the blood vessel is shown. As shown in Figure 6, the OCT medical image-based artificial intelligence computer-assisted diagnosis method according to an embodiment of the present invention, the step of automatically detecting a lumen shape of a blood vessel in an IVOCT image (S110), from the lumen shape of the blood vessel Extracting the shape features of the vascular lumen (S120), classifying them into a normal vascular lumen and an abnormal vascular lumen according to the shape characteristics of the vascular lumen (S130), and providing the classified normal or abnormal vascular lumen through a display. It may be implemented including step S140.

In step S110, the automatic blood vessel lumen detection unit 110 may automatically detect the lumen shape of a blood vessel from an IVOCT image photographed by intravascular optical coherence tomography (IVOCT). At this time, in step S110, the automatic blood vessel lumen detection unit 110 converts the IVOCT images photographed and provided by intravascular optical coherence tomography (IVOCT) to a rectangular coordinate image, and then convolves with an asymmetric Gaussian filter to determine the boundary of the blood vessel lumen. It may include the step of extracting (S111).

In addition, in step S110, the automatic blood vessel lumen detection unit 110 detects and removes a guide wire shadow region of a corresponding pattern at the boundary of the blood vessel lumen extracted in step S111, and the shadow region is removed from the region. For example, an interpolation process may be performed (S112) so as not to affect the image processing. In this case, the guide wire shadow area may have a row length of 220 pixels or more and a column width of 105 pixels or a 175 pixel pattern.

In addition, in step S110, the automatic blood vessel lumen detection unit 110 removes the shadow region of the guide wire in step S112 and detects the shadow region of the stent strut in the blood vessel lumen detected through the asymmetric Gaussian filter interpolated in step S112. A step of filling with a value (S113) may be further included.

In addition, in step S110, the automatic vascular lumen detection unit 110 accurately detects the shape of the irregular vascular lumen using an active contour model for the vascular lumen processed in step S113 (S114). It can be done by doing.

In step S120, the shape feature extraction unit 120 may extract shape features of the blood vessel lumen from the lumen shape of the blood vessel detected by the automatic blood vessel lumen detection unit 110 in step S110. Here, in step S120, the shape feature extraction unit 120 may extract 38 shape features and 54 texture features for the lumen shape of the blood vessel detected by the automatic blood vessel lumen detection unit 110.

In step S130, the vascular lumen classifying unit 130 may classify into a normal vascular lumen and an abnormal vascular lumen according to the shape characteristics of the vascular lumen extracted by the shape feature extraction unit 120 of step S120. Here, in step S130, the vascular lumen classification unit 130 defines a feature vector through 38 shape features and 54 texture features of the vascular lumen extracted by the shape feature extraction unit 120 in step S120, and Machine) technology is used to classify normal IVOCT images and abnormal IVOCT images.

In step S140, the display providing unit 140 may provide the normal or abnormal vascular lumen classified according to the vascular lumen classification unit of step S130 through the display. Here, in step S140, the display providing unit 140 displays an IVOCT image classified as normal or abnormal so that the display provider 140 can assist in the process of reading and diagnosing an IVOCT image of the doctor, but the normal or abnormal image is further displayed on the displayed classified IVOCT image. To be provided.

As described above, according to the OCT medical image-based artificial intelligence computer-assisted diagnosis system and method according to an embodiment of the present invention, the inner diameter of a blood vessel is automatically detected from an IVOCT image taken by intravascular optical coherence tomography (IVOCT), and , After extracting the shape features of the vascular lumen through pattern recognition technology, the extracted shape characteristics can be classified into normal vascular lumen and abnormal vascular lumen through SVM (Support Vector Machine) and provided. In particular, shape features of the vascular lumen When an IVOCT image that is automatically classified as a normal vascular lumen and an abnormal vascular lumen is provided through the display providing unit, the IVOCT image is further displayed and provided, thereby assisting the doctor's IVOCT image reading and diagnosis process. It is possible to allow the probability to be reduced.

The present invention described above can be modified or applied in various ways by those of ordinary skill in the technical field to which the present invention belongs, and the scope of the technical idea according to the present invention should be determined by the following claims.

100: artificial intelligence computer assisted diagnosis system according to an embodiment of the present invention
110: automatic blood vessel lumen detection unit
111: Cartesian coordinate image conversion unit
112: boundary extraction unit
113: shadow area detection unit
114: shadow area removal unit
115: stent strut processing unit
116: vessel inner diameter shape detection unit
120: shape feature extraction unit
130: vascular lumen classification unit
140: display providing unit
S110: Automatically detecting the shape of the lumen of the blood vessel in the IVOCT image
S111: After converting the IVOCT images into rectangular coordinate images, extracting the boundary of the lumen of the blood vessel by convolving with an asymmetric Gaussian filter
S112: Step of detecting and removing a shadow area of a guide wire of a corresponding pattern at the boundary of the blood vessel lumen, and then performing interpolation processing
S113: Detecting the shadow area of the stent strut in the lumen of the blood vessel detected through the asymmetric Gaussian filter, and filling it with a preset high pixel value
S114: Precisely detecting the irregular shape of the vascular lumen using the active contour model
S120: Extracting shape features of the lumen of the blood vessel from the shape of the lumen of the blood vessel
S130: Classifying into a normal vascular lumen and an abnormal vascular lumen according to the shape characteristics of the vascular lumen
S140: providing the classified normal or abnormal vascular lumen through the display

Claims (23)

As an artificial intelligence computer-assisted diagnosis system 100 based on OCT medical image,
A blood vessel lumen automatic detection unit 110 for automatically detecting a lumen shape of a blood vessel from an IVOCT image taken by intravascular optical coherence tomography (IVOCT);
A shape feature extraction unit 120 for extracting shape features of the vascular lumen from the lumen shape of the blood vessel detected by the vascular lumen automatic detection unit 110;
A vascular lumen classification unit 130 for classifying a normal vascular lumen and an abnormal vascular lumen according to the shape characteristics of the vascular lumen extracted from the shape feature extraction unit 120; And
Including a display providing unit 140 for providing through a display the normal or abnormal vascular lumen classified according to the vascular lumen classification unit 130,
The shape feature extraction unit 120,
A shape feature and a texture feature are extracted from the luminal shape of the blood vessel detected by the automatic blood vessel lumen detection unit 110,
The shape feature extraction unit 120,
The shape features and texture features are extracted for the lumen shape of the blood vessel detected by the blood vessel lumen automatic detection unit 110, but 38 shape features and 54 texture features are extracted,
The blood vessel lumen classification unit 130,
A feature vector is defined through the shape features and texture features of the blood vessel lumen extracted by the shape feature extraction unit 120,
The blood vessel lumen classification unit 130,
Classify the defined feature vector into a normal IVOCT image and an abnormal IVOCT image using SVM (Support Vector Machine) technology,
The display providing unit 140,
OCT medical image, characterized in that an IVOCT image classified as normal or abnormal is displayed to aid in the process of reading and diagnosing an IVOCT image by a doctor, but is further displayed and provided with normal or abnormality on the displayed classified IVOCT image. Based artificial intelligence computer-assisted diagnostic system.
The method of claim 1, wherein the automatic blood vessel lumen detection unit 110,
An artificial intelligence computer-assisted diagnosis based on OCT medical images, characterized in that it comprises a rectangular coordinate image conversion unit 111 that converts and processes IVOCT images captured and provided by intravascular optical coherence tomography (IVOCT) into rectangular coordinate images. system.
The method of claim 2, wherein the automatic blood vessel lumen detection unit 110,
A boundary extraction unit 112 configured to receive an IVOCT image converted into a rectangular coordinate image from the rectangular coordinate image conversion unit 111 and convolve with an asymmetric Gaussian filter to extract the boundary of the blood vessel lumen. AI computer-assisted diagnosis system based on OCT medical image.
The method of claim 3, wherein the automatic blood vessel lumen detection unit 110,
OCT medical image, characterized in that it further comprises a shadow region detection unit 113 for detecting a guide wire shadow region of a corresponding pattern at the boundary of the blood vessel lumen extracted by the boundary extraction unit 112 Based artificial intelligence computer-assisted diagnostic system.
The method of claim 4, wherein the guide wire shadow area,
An artificial intelligence computer-assisted diagnosis system based on an OCT medical image, characterized in that a row length is 220 pixels or more and a column width is a 105 pixel or 175 pixel pattern.
The method of claim 4, wherein the automatic blood vessel lumen detection unit 110,
And a shadow area removal unit 114 that removes a shadow area of the guide wire of a corresponding pattern detected by the shadow area detection unit 113 at the boundary of the blood vessel lumen extracted by the boundary extraction unit 112. AI computer-assisted diagnosis system based on OCT medical image.
The method of claim 6, wherein the shadow area removal unit 114,
An artificial intelligence computer-assisted diagnosis system based on OCT medical images, characterized in that interpolation processing is further performed so that the image processing is not affected by the area from which the shadow area has been removed.
The method of claim 6, wherein the automatic blood vessel lumen detection unit 110,
A stent strut processing unit 115 that removes the shadow area of the guide wire in the shadow area removal unit 114 and detects the shadow area of the stent strut at the boundary of the blood vessel lumen detected through an asymmetric Gaussian filter interpolated and fills it with a preset pixel value. ) Further comprising an artificial intelligence computer-assisted diagnosis system based on an OCT medical image.
The method of claim 8, wherein the automatic blood vessel lumen detection unit 110,
An artificial intelligence computer-assisted diagnosis system based on OCT medical images, characterized in that it further comprises a blood vessel inner diameter shape detection unit 116 that precisely detects irregular blood vessel lumen shapes using an active contour model.
delete delete delete delete delete As an artificial intelligence computer-assisted judgment method based on OCT medical image,
(1) a step of automatically detecting the shape of the lumen of the blood vessel from the IVOCT image captured by the intravascular optical coherence tomography (IVOCT) by the automatic blood vessel lumen detection unit 110;
(2) extracting, by the shape feature extraction unit 120, the shape features of the blood vessel lumen from the lumen shape of the blood vessel detected by the automatic blood vessel lumen detection unit 110 in step (1);
(3) classifying, by the vascular lumen classification unit 130, into a normal vascular lumen and an abnormal vascular lumen according to the shape characteristics of the vascular lumen extracted by the shape feature extraction unit 120 of step (2); And
(4) including the step of providing, by the display providing unit 140, a normal or abnormal vascular lumen classified according to the vascular lumen classification unit of step (3) through a display,
In step (2),
The shape feature extraction unit 120 extracts shape features and texture features for the lumen shape of the blood vessel detected by the automatic blood vessel lumen detection unit 110, and extracts 38 shape features and 54 texture features,
In step (3),
The vascular lumen classification unit 130 defines a feature vector through 38 shape features and 54 texture features of the vascular lumen extracted by the shape feature extraction unit 120 of step (2), and supports vector machine (SVM). ) Using technology to classify into normal IVOCT images and abnormal IVOCT images,
In step (4),
The display providing unit 140 displays IVOCT images classified as normal or abnormal so that they can assist the doctor's IVOCT image reading and diagnosis process, but the normal or abnormal is further displayed on the displayed classified IVOCT image. Characterized in that, OCT medical image-based artificial intelligence computer-assisted determination method.
The method of claim 15, wherein in step (1),
(1-1) The blood vessel lumen automatic detection unit 110 converts the IVOCT images photographed and provided by intravascular optical coherence tomography (IVOCT) to a rectangular coordinate image, and then convolves with an asymmetric Gaussian filter to the boundary of the vessel lumen. It characterized in that it comprises the step of extracting, OCT medical image-based artificial intelligence computer-assisted determination method.
The method of claim 16, wherein in the step (1),
(1-2) The blood vessel lumen automatic detection unit 110 detects and removes a guide wire shadow region of a corresponding pattern at the boundary of the blood vessel lumen extracted in step (1-1), and the shadow region is An artificial intelligence computer-assisted determination method based on an OCT medical image, further comprising: performing interpolation processing so that the image processing is not affected by the removed region.
The method of claim 17, wherein the guide wire shadow area,
An artificial intelligence computer-assisted determination method based on an OCT medical image, characterized in that a row length is 220 pixels or more and a column width is a 105 pixel or 175 pixel pattern.
The method of claim 17, wherein in the step (1),
(1-3) The blood vessel lumen automatic detection unit 110 removes the guide wire shadow region in step (1-2) and detects the shadow region of the stent strut in the blood vessel lumen detected through an asymmetric Gaussian filter interpolated. , OCT medical image-based artificial intelligence computer-assisted determination method further comprising the step of filling with a preset pixel value.
The method of claim 19, wherein in step (1),
(1-4) A step in which the automatic vascular lumen detection unit 110 accurately detects the irregular shape of the vascular lumen using an active contour model for the vascular lumen processed in step (1-3). Characterized in that further comprises a, OCT medical image-based artificial intelligence computer-assisted determination method.
delete delete delete

Images