JPWO2020054019A1 - Image processing equipment, image processing methods, and programs - Google Patents

Abstract

The image processing device 3 sets the target tissue region based on the region estimation unit 304 that estimates the target tissue region by artificial intelligence from the captured image of the organ and the target tissue region estimated by the region estimation unit 304. It has a guide line determination unit 306 that determines the included guide lines, and an image extraction unit that extracts an image region corresponding to the target tissue from the captured image based on the guide lines determined by the guide line determination unit 306. Further, the target tissue is a tubular tissue, the region estimation unit 304 estimates the region of the tubular tissue, and the guide line determination unit 306 determines a guide line indicating at least one of the direction and range in which the tubular tissue spreads. The image extraction unit extracts an image region corresponding to the tubular tissue.

Description

The present invention relates to an image processing apparatus, an image processing method, and a program.

For example, in Non-Patent Document 1, the start point of the coronary artery (the start point extending from the ascending aorta to the end of the coronary artery) is searched for as Auto-Seeding from the diagnostic image data, and the coronary artery is extracted (separated) from the start point by the virtual control injection method. , Skeletalization) is disclosed.

Further, Patent Document 2 describes a system for segmenting a diagnostic image at present, in which a volume of interest is segmented in a previously generated diagnostic image of a selected volume of interest generated from a plurality of patients1. Alternatively, aligning the previously generated and segmented images with multiple workstations and merging the previously generated and segmented images (1) the probability that each voxel represents the volume of interest. One or more processors programmed to provide a probability map showing (2) the probability that each voxel represents the background, and (3) the mean segmentation boundary, and the probability map is the current volume of interest of the patient. A system is disclosed having a segmentation processor that aligns with a diagnostic image and produces a transformed probability map.

Further, Patent Document 3 describes a medical image processing apparatus that generates a medical image using three-dimensional volume data indicating the inside of a living body, and includes a region including a cross section of a predetermined tubular tissue indicated by the three-dimensional volume data. An imaging unit that images the cross section of the tubular structure included in the cross-sectional area by using a plurality of designated cross-sectional area designating parts and the three-dimensional volume data included in each cross-sectional area designated by the cross-sectional area designating part. And the three-dimensional volume data indicating the three-dimensional volume data indicating the center position of the image showing the cross section of the tubular structure and the three-dimensional volume data indicating the center position of the tube cross section in each cross section region specified by the center specifying portion. Based on this, a medical image processing apparatus is disclosed, which comprises a center line specifying portion for specifying the three-dimensional volume data indicating the long direction center line of the tubular tissue.

An Automatic Seding Method For Coronary Artery Segmentation and Skeletonization in CTA Special Table 2013-521844 Japanese Unexamined Patent Publication No. 2004-313736

An object of the present invention is to provide an image processing device that facilitates identification of a target tissue in a photographed image of an organ.

The image processing apparatus according to the present invention has a region estimation unit that estimates the region of the target tissue by artificial intelligence from a photographed image of an organ, and a region of the target tissue based on the region of the target tissue estimated by the region estimation unit. It has a guide line determination unit that determines the guide line included in the image, and an image extraction unit that extracts an image region corresponding to the target tissue from the captured image based on the guide line determined by the guide line determination unit. ..

The image processing apparatus according to the present invention has a region estimation unit that estimates the region of the target tissue by artificial intelligence from a photographed image of an organ, and a region of the target tissue based on the region of the target tissue estimated by the region estimation unit. It has a guide line determining unit for determining the guide line included in the above, and an image display unit for displaying a guide image along the guide line determined by the guide line determining unit on the captured image.

Preferably, the target tissue is a tubular tissue, the region estimation unit estimates the region of the tubular tissue, and the guide line determination unit provides a guide line indicating at least one of the direction and range in which the tubular tissue spreads. After determining, the image extraction unit extracts an image region corresponding to the tubular tissue.

Preferably, the captured image is an MRI image of a region including the coronary arteries, the region estimation unit estimates the region of the coronary arteries, and the guide line determination unit is at least one of the direction and range in which the coronary arteries spread. The guide line indicating the above is determined, and the image extraction unit extracts an image region corresponding to the coronary artery from the MRI image.

Preferably, the guide line determination unit determines the center line passing through the central portion of the tubular tissue as the guide line, and the image extraction unit determines the brightness information of the position corresponding to the center line and the brightness around the center line. The image area corresponding to the tubular tissue is extracted by comparing with the information.

Preferably, the region estimation unit uses artificial intelligence configured with a set of a plurality of MRI images of the region including the coronary arteries and an image region of the coronary arteries extracted from these MRI images as training data. , Estimate the area of the coronary arteries.

Preferably, the guide line determination unit connects the regions of the estimated target tissue by a predetermined interpolation method to determine a guide line composed of a curved line.

In the image processing method according to the present invention, the step of estimating the region of the target tissue by artificial intelligence from the photographed image of the organ and the guide line included in the region of the target tissue are determined based on the estimated region of the target tissue. It has a step of extracting an image region corresponding to a target tissue from the captured image based on a determined guide line.

In the program according to the present invention, a step of estimating the region of the target tissue by artificial intelligence from a photographed image of the organ and a step of determining a guide line included in the region of the target tissue based on the estimated region of the target tissue. The computer is made to execute the step of superimposing the guide image along the determined guide line on the captured image and displaying the image.

It is possible to easily identify the target tissue in the photographed image of the organ.

It is a figure which illustrates the system structure of the image processing system 1. It is a figure which illustrates the hardware composition of the image processing apparatus 3. It is a figure which illustrates the functional structure of the image processing apparatus 3. It is a flowchart explaining the coronary artery extraction process (S10). It is a flowchart explaining the guide line determination process (S20). (A) is a diagram illustrating the scattered center lines, and (b) is a diagram illustrating how to connect the scattered center lines. It is a figure which illustrates the image which superposed the guide image 8 and the MRI image 9.

[background]
The background of the present invention will be described.
In cardiovascular diagnosis, a computer tomographic diagnostic device is used for diagnosis. The diagnostic procedure is, for example, (1) imaging with a computer tomographic diagnostic device, (2) extracting (trimming) the region of interest (VOI) from the captured image, (3) extracting the coronary artery, and (4) cardiac VR (three-dimensional three-dimensional). Image) Data is extracted, and (5) digitized based on this information and displayed in a format that is easy for doctors to see.
The mainstream imaging method is radiation CT using a contrast medium, but there are side effects of the contrast medium and radiation risk. These risk-free MRI (magnetic resonance imaging) methods are expected to be introduced in the future, but due to the roughness of the resolution and artifacts (specific noise), the conventional image processing technology extracts (separates, skeletalizes) coronary arteries. It is difficult to automate the extraction, and a skilled image interpreter performs the extraction semi-manually. And such an interpreter is in short supply. Therefore, in the field of medical care, coronary arteries are automatically extracted from MRI images without relying on a skilled interpreting doctor, and cardiac VR data effective for patient explanation is extracted and displayed in an appropriate format that is easy for the doctor to see. Realization is desired.

In general, MRI images, especially those in which the origin of the coronary artery is difficult to identify, have a low success rate for automatic detection of the coronary artery by computer processing. Factors with a low success rate include method factors and image characteristics factors. As a method factor, conventionally, the method of recognizing which blood vessel is in the entire image is to search from the start point of the coronary artery, so that the coronary artery cannot be detected unless the start point of the coronary artery is found. And, for the same reason, if pixel information cannot be recognized in a part of the coronary artery due to stenosis or other causes, even if the coronary artery at the point beyond it exists as a bright spot, it will not be searched. Can be mentioned.
In addition, as a factor of image characteristics, the resolution (1.2 mm) of the MRI image is not sufficient for the coronary artery diameter (about 3 mm) (CT resolution is 0.6 mm, and the success rate is relatively high). Then, due to the problem of artifacts, the coronary arteries are often buried in other tissues such as adipocytes and appear, and in particular, the starting point of the coronary arteries is often buried in other tissues.

In response to the above problems, the present invention estimates the area of the coronary artery by artificial intelligence without the intervention of an image interpreter, determines the centerline of the coronary artery in the estimated area, and captures an image of the original coronary artery along the centerline. This is a method of extracting a pixel portion of a blood vessel from (MRI image) based on brightness information and automatically extracting a coronary artery image.

Embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram illustrating the overall configuration of the image processing system 1.
As illustrated in FIG. 1, the image processing system 1 includes an image transmission device 2a, an image transmission device 2b, an image transmission device 2c, and an image processing device 3, and is connected to each other via a network 7. The image transmission device 2a, which is a combination of the image transmission device 2a, the image transmission device 2b, and the image transmission device 2c and is referred to as an image transmission device 2, is a computer terminal and transmits a photographed image of an organ to an image processing device. Specifically, the image transmission device 2 transmits tomographic image data (hereinafter referred to as image data) of the patient's body captured by an MRI (magnetic resonance imaging device) to the image processing device 3.
The image processing device 3 is a computer terminal, and extracts an image area of the target tissue based on the image data received by the image transmitting device 2. In this example, the target tissue is a tubular tissue, and in this example, a coronary artery. The image processing device 3 transmits the extracted image area to the image transmission device 2.

FIG. 2 is a diagram illustrating a hardware configuration of the image processing device 3.
As illustrated in FIG. 2, the image processing device 3 includes a CPU 200, a memory 202, an HDD 204, a network interface 206 (network IF206), a display device 208, and an input device 210, and these configurations are configured via the bus 212. Connected to each other.
The CPU 200 is, for example, a central arithmetic unit.
The memory 202 is, for example, a volatile memory and functions as a main storage device.
HDD 204 is, for example, a hard disk drive device, which stores computer programs and other data files as a non-volatile recording device.
The network IF206 is an interface for communicating by wire or wirelessly.
The display device 208 is, for example, a liquid crystal display.
The input device 210 is, for example, a keyboard and a mouse.

FIG. 3 is a diagram illustrating the functional configuration of the image processing device 3.
As illustrated in FIG. 3, an image processing program 30 is installed in the image processing device 3, and the image processing program 30 is stored in a recording medium such as a CD-ROM, and the image processing program 30 is stored in a recording medium such as a CD-ROM. , Is installed in the image processing device 3 and constitutes a learning data database 600 (learning data DB 600).
A part or all of the image processing program 30 may be realized by hardware such as an ASIC, or may be realized by borrowing a part of the functions of the OS (Operating System). In addition, the entire program may be installed on a single computer terminal or a virtual machine on the cloud.

The image processing program 30 includes an image acquisition unit 300, an area cutting unit 302, an area estimation unit 304, a guide line determination unit 306, an image extraction unit 308, and an image display unit 310.

In the image processing program 30, the image acquisition unit 300 acquires the MRI image data transmitted by the image transmission device 2.
The area cutting unit 302 extracts only the area (VOI) required for analysis from the image data.
The area estimation unit 304 estimates the area of the target tissue by artificial intelligence from the photographed image of the organ. Specifically, the target tissue is a tubular tissue, and more specifically, the region estimation unit 304 estimates the region of the coronary artery. The region estimation unit 304 uses artificial intelligence configured as teacher data as a set of a plurality of MRI images of the region including the coronary arteries and the image region of the coronary arteries extracted from these MRI images, and uses the region of the coronary arteries. To estimate.
In addition, artificial intelligence (AI: Artificial Intelligence) in the present invention learns teacher data by deep learning in a neural network. The teacher data is 3D data of coronary arteries extracted by a doctor associated with MRI image data. Artificial intelligence learns teacher data and estimates the area of the coronary arteries based on it. The learning data DB 600 holds the teacher data, and when new teacher data is input, the artificial intelligence may be relearned.

The guide line determination unit 306 determines the guide line included in the region of the coronary artery based on the region of the coronary artery estimated by the region estimation unit 304. Specifically, the guide line determination unit 306 determines a guide line indicating at least one of the direction and range in which the coronary artery spreads. More specifically, the guide line determining unit 306 determines the center line passing through the central portion of the coronary artery as the guide line. As a method of determining the guide line, the guide line determination unit 306 connects the estimated coronary artery regions by a predetermined interpolation method to determine a guide line composed of a curved line. The guide line indicates a direction or a range, and the line width and the pattern are arbitrary. The guide line may be a dotted line, a broken line, a chain line, an arrow line, a hollow line, or the like.

The image extraction unit 308 extracts an image region corresponding to the coronary artery from the captured image based on the guide line determined by the guide line determination unit 306. Specifically, the image extraction unit 308 extracts an image region corresponding to the coronary artery from the MRI image. Further, the image extraction unit 308 compares the luminance information at the position corresponding to the center line with the luminance information around the center line to extract the image region corresponding to the coronary artery. More specifically, the image extraction unit 308 represents the extracted image area as an extraction diagram.

The image display unit 310 superimposes and displays a guide image along the guide line determined by the guide line determination unit 306 on the MRI image. Further, the image display unit 310 outputs a 3D image of the coronary artery, a coronary artery diameter value, or the like as a report together with the coronary artery extract image.

FIG. 4 is a flowchart illustrating the coronary artery extraction process (S10).
As illustrated in FIG. 4, in step 100 (S100), the image acquisition unit 300 acquires image data to be diagnosed, which is taken by MRI.
In step 105 (S105), the area cutting unit 302 extracts only the area (VOI) required for analysis from the image data. Then, the area cutting unit 302 performs preprocessing such as data file format conversion in order to input the extracted area to the neural network.
In step 110 (S110), the region estimation unit 304 estimates and extracts the coronary arteries in the extracted region using a neural network.
In step 115 (S115), the guide line determination unit 306 determines the center line passing through the central portion of the coronary artery as the guide line.
In step 120 (S120), the image extraction unit 308 connects the luminance signals of the original patient data (MRI image data to be diagnosed) based on the guide line determined by the guide line determination unit 306, and draws a coronary artery extraction diagram. create.
In step 125 (S125), the image display unit 310 reports and outputs a shape (for example, a 3D image) useful for a doctor's diagnosis, a coronary artery diameter value, and the like.

FIG. 5 is a flowchart illustrating the guide line determination process (S20).
As illustrated in FIG. 5, in step 200 (S200), the region estimation unit 304 is estimated to be a part of the target tissue by AI in the image of the coronary artery based on the teacher data and the learning of the blood vessel running. Identify pixels whose expected value is greater than or equal to the threshold.
In step 205 (S205), the guide line determination unit 306 thins the area containing the specified pixel to obtain the center line of the coronary artery. As illustrated in FIG. 6A, since the line structure is a sequence of points in space in the program, the center lines are scattered along the coronary arteries.
In step 210 (S210), when the front end and the rear end of the scattered center lines are close to each other, the guide line determination unit 306 shifts to S215, and when they are not close to each other, the guide line determination unit 306 shifts to S220.
In step 215 (S215), the guide line determination unit 306 regards the scattered point clouds as one row, as illustrated in FIG. 6 (b), and another row close to the point at the lower end of the row. Search for the point at the tip of, connect between them, and determine the guide line.
In step 220 (S220), the guide line determination unit 306 connects the scattered center lines and determines the guide line by using the blood vessel estimation technique. Specifically, the guide line determination unit 306 connects the center lines by a curve regression method (for example, nth-order spline interpolation) that passes through the extraction points.

As described above, according to the image processing system 1 of the present embodiment, even in an image having low resolution and noise such as an MRI image in which it is difficult to identify the start point of the coronary artery, the estimation of the coronary artery by AI and the guide line can be performed. Since the coronary artery extraction process can be performed based on the above, the coronary artery extraction process can be performed with higher accuracy and a high success rate. In addition, since the coronary artery extraction process can be performed automatically without the intervention of a doctor, it is possible to reduce the dependence on the skill of the doctor. Further, since the output by AI is used only for outputting the coronary center line, it is possible to reduce the variation due to the accuracy of AI based on the training data as compared with the coronary artery extraction using only AI, and finally based on the brightness information. By extracting the coronary arteries, it is possible to obtain information that can be used for diagnosis with the same reliability as before.

In the above embodiment, the image extraction unit 308 creates an extraction diagram of the coronary artery, but the present invention is not limited to this, and for example, as illustrated in FIG. 7, the image display unit 310 uses the guide line determination unit 306. The guide image 8 along the determined guide line may be superimposed on the MRI image 9 and displayed, and an extraction map of the coronary artery may be created by an image interpreter based on the displayed image.

1 ... Image processing system 2 ... Image transmission device 3 ... Image processing device 30 ... Image processing program

Claims (9)

A region estimation unit that estimates the region of the target tissue by artificial intelligence from the captured image of the organ,
A guide line determination unit that determines a guide line included in the area of the target organization based on the area of the target organization estimated by the area estimation unit.
An image processing apparatus including an image extraction unit that extracts an image region corresponding to a target tissue from the captured image based on a guide line determined by the guide line determination unit. A region estimation unit that estimates the region of the target tissue by artificial intelligence from the captured image of the organ,
A guide line determination unit that determines a guide line included in the area of the target organization based on the area of the target organization estimated by the area estimation unit.
An image processing device including an image display unit that superimposes and displays a guide image along the guide line determined by the guide line determination unit on the captured image. The target tissue is a tubular tissue and
The region estimation unit estimates the region of the tubular tissue and
The guide line determination unit determines a guide line indicating at least one of the direction and range in which the tubular tissue spreads.
The image processing apparatus according to claim 1 or 2, wherein the image extraction unit extracts an image region corresponding to a tubular tissue. The captured image is an MRI image of a region including a coronary artery.
The region estimation unit estimates the region of the coronary artery and
The guide line determination unit determines a guide line indicating at least one of the direction and range in which the coronary artery spreads.
The image processing apparatus according to claim 3, wherein the image extraction unit extracts an image region corresponding to a coronary artery from an MRI image. The guide line determining portion determines the center line passing through the central portion of the tubular tissue as the guide line.
The image processing apparatus according to claim 3, wherein the image extraction unit compares the brightness information at a position corresponding to the center line with the brightness information around the center line to extract an image area corresponding to a tubular tissue. The region estimation unit uses artificial intelligence configured with a set of a plurality of MRI images of the region including the coronary arteries and an image region of the coronary arteries extracted from these MRI images as training data, and the region of the coronary arteries. The image processing apparatus according to claim 4. The image processing apparatus according to claim 3, wherein the guide line determination unit connects the estimated target tissue regions by a predetermined interpolation method to determine a guide line composed of a curved line. The step of estimating the area of the target tissue by artificial intelligence from the photographed image of the organ,
Steps to determine the guidelines contained in the area of the target organization based on the estimated area of the target organization,
An image processing method including a step of extracting an image area corresponding to a target tissue from the captured image based on a determined guide line. The step of estimating the area of the target tissue by artificial intelligence from the photographed image of the organ,
Steps to determine the guidelines contained in the area of the target organization based on the estimated area of the target organization,
A program that causes a computer to perform a step of superimposing and displaying a guide image along a determined guide line on the captured image.

Images