KR102103281B1 - Ai based assistance diagnosis system for diagnosing cerebrovascular disease - Google Patents

Abstract

The present invention relates to an AI-based auxiliary diagnostic system for diagnosis of cerebrovascular diseases. According to the present invention, the AI-based auxiliary diagnostic system comprises: a learning database in which a plurality of MRA images labeled with cerebrovascular diseases are stored; a coordinate system registration module which registers a spatial coordinate system in each of the MRA images stored in the learning database; and an AI module which learns, through a preregistered deep learning-based learning model, each of the MRA images in which the spatial coordinate system is registered by the coordinate system registration module. When cerebrovascular diseases are read through a deep learning-based learning model, the same coordinate system is registered in learning images to be learned by the learning model so as to increase accuracy of learning and reading.

Description

AI-based assistive diagnosis system for diagnosis of cerebrovascular disease {AI BASED ASSISTANCE DIAGNOSIS SYSTEM FOR DIAGNOSING CEREBROVASCULAR DISEASE}

The present invention relates to an AI-based auxiliary diagnostic system for diagnosis of cerebrovascular disease, and more specifically, to increase the accuracy of learning and reading in reading cerebrovascular diseases such as cerebral aneurysms through a deep learning-based learning model. AI-based assisted diagnostic system for the diagnosis of possible cerebrovascular disease.

In modern medicine, medical imaging is a very important tool for effective disease diagnosis and patient treatment. In addition, the development of imaging technology makes it possible to acquire more sophisticated medical imaging data. In exchange for this sophistication, the amount of data is gradually increasing, making it difficult to analyze medical image data depending on human vision. Thus, in recent decades, clinical decision support systems and computer-aided reading systems have played an essential role in the automatic analysis of medical images.

A conventional clinical decision support system or computer-aided reading system detects and displays the lesion area or presents the reading information to a medical staff or medical personnel (hereinafter referred to as a user).

For example, in the 'medical image-based disease diagnosis information calculation method and apparatus' disclosed in Korean Patent Publication No. 10-2017-0017614, a region of interest in which an object to be analyzed is photographed is detected, a variation coefficient is calculated, and a variation coefficient is calculated. It includes the steps of creating an image and comparing it with a reference sample, and refers to the effect of diagnosing a patient's disease level by using medical images obtained through CT, MRI, MRA, and ultrasound imaging devices.

In particular, artificial intelligence (AI) technology based on machine learning, such as deep learning, has recently been the basis for bringing about a quantum leap in reading patients' diseases using medical images. .

Deep learning refers to a machine learning method based on an artificial neural network that simulates a human neuron and allows a machine to learn. Recently, deep learning technology has been rapidly developed in the field of image recognition, and is widely used in the field of reading using medical images.

In the deep learning technology in medical images, a machine learning model (hereinafter referred to as a 'learning model') is generated by performing machine learning using a plurality of medical images including diseases and the disease as learning data, and a medical image to be read When it is input to this learning model, it diagnoses whether or not there is a lesion.

As described above, the machine learning method based on deep learning collects learning data used to generate a learning model, for example, a plurality of lesion images and a plurality of normal images, and a learning model generated by learning the collected lesion images and the normal images. Since the newly inputted medical image to be read is read, it is possible to increase the accuracy of the read result when a variety of learning data are used for learning. In addition, when the medical image to be read is similar to the learning data, the accuracy of the read result may be increased.

The deep learning-based assistive diagnosis system as described above is also required to be applied to diagnose cerebrovascular diseases, such as cerebral aneurysms. 1 is a view showing an example of a 3D image of a cerebral blood vessel obtained based on an MRA image.

As shown in FIG. 1, in the case of a cerebral blood vessel, its shape is different depending on the viewing direction, and in the case of a cerebral aneurysm, it may appear normal according to the viewing direction, thereby limiting the accuracy of reading.

As a method of improving the accuracy of reading in image reading using a deep learning-based learning model, a method of inputting coordinate values at each location in the image together has been proposed. This is a principle in which a learning model trained using a learning image having a coordinate value uses the coordinate value for reading to increase accuracy when an image to be read having the same coordinate value is input.

However, as shown in FIG. 1, in the case of an MRA image, it may vary according to a shooting location or a distance, and may vary depending on a person, so if a coordinate system is set around a specific location in the MRA image, the same coordinates are used for each image. The problem arises in that different locations are specified.

Accordingly, the present invention was devised to solve the above problems, and in reading cerebrovascular diseases such as cerebral aneurysms through a deep learning-based learning model, the same model registers the same coordinate system in the learning images to be learned. The aim is to provide an AI-based assisted diagnostic system for diagnosis of cerebrovascular disease that can improve the accuracy of learning and reading by giving.

In accordance with the present invention, in the AI-based auxiliary diagnosis system for diagnosis of cerebrovascular disease, a plurality of MRA images labeled with cerebrovascular disease are stored, and each of the MRA images stored in the learning database. A coordinate system registration module for registering a spatial coordinate system in and an AI module for learning the MRA image in which the spatial coordinate system is registered by the coordinate system registration module through a pre-registered deep learning based learning model; The coordinate system registration module includes (a) extracting a vascular region from the MRA image, (b) extracting three pre-set anatomical first points from the vascular region, and (c) three first Determining a normal vector for a first plane formed by the point as a first coordinate vector, and (d) extracting two predetermined anatomical second points located on both sides with the first plane interposed therebetween. And, (e) extracting a location where the first plane meets a straight line connecting the two second points as a coordinate origin, and (f) extracting a predetermined anatomical third point, (g ) Extracting a second plane formed by the two second points and the third point, and extracting an intersection line intersecting the first plane and the second plane, and (h) based on the intersection line Determining a second coordinate vector by (i) Determining a first coordinate axis and a second coordinate axis by moving the first coordinate vector and the second coordinate vector to the coordinate origin, respectively, and (j) a third orthogonal to the first coordinate axis and the second coordinate axis, respectively. It is achieved by an AI-based auxiliary diagnostic system for the diagnosis of cerebrovascular disease characterized by performing the step of determining the coordinate axis and registering the spatial coordinate system.

Here, the step (b) comprises: (b1) extracting a pair of intersection points of the posterior carotid artery and the internal carotid artery with the two first points; (b2) extracting a pair of intersection points of the posterior carotid artery and the posterior cerebral artery; (b3) extracting the center point of the pair of intersections of the posterior carotid artery and the posterior cerebral artery as one of the first points.

In addition, the step (d) comprises: (d1) extracting the central point of the anterior transit artery as one second point; (d2) extracting a boundary point between the brain basal artery and the posterior cerebral artery as another second point.

And, in step (f), the boundary point between the brain base artery and the spinal artery may be extracted as the third point.

And, further comprising an image interface unit for inputting a new MRA image to be diagnosed; The coordinate system registration module registers the spatial coordinate system for the new MRA image and delivers it to the AI module; The AI module may diagnose whether or not a disease in the new MRA image in which the spatial coordinate system is registered through the learning model.

According to the configuration as described above, according to the present invention, by generating a spatial coordinate system using a plurality of points obtained based on the anatomical structure of the individual blood vessels constituting the brain blood vessels, a plurality of brain blood vessel learning images used for learning By providing a unified spatial coordinate system, an AI-based auxiliary diagnostic system for diagnosis of cerebrovascular lesions capable of generating a learning model with high reading accuracy is provided.

1 is a view showing an example of a 3D image of a cerebral blood vessel obtained based on an MRA image,
2 is a view showing the configuration of an AI-based auxiliary diagnostic system for the diagnosis of cerebrovascular lesions according to the present invention,
3 is a diagram showing a process of registering a spatial coordinate system in an AI-based auxiliary diagnostic system for diagnosis of cerebrovascular lesions according to the present invention,
4 to 17 are views for explaining a process of registering the spatial coordinate system shown in FIG. 3.

The present invention can be applied to various changes and may have various embodiments, and specific embodiments will be illustrated in the drawings and described in detail.

However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention.

The terms used in this application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "include" or "have" are intended to indicate the presence of features, numbers, steps, actions, components, parts or combinations thereof described herein, one or more other features. It should be understood that the existence or addition possibilities of fields or numbers, steps, operations, components, parts or combinations thereof are not excluded in advance.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person skilled in the art to which the present invention pertains. Terms such as those defined in a commonly used dictionary should be interpreted as having meanings consistent with meanings in the context of related technologies, and should not be interpreted as ideal or excessively formal meanings unless explicitly defined in the present application. Does not.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

2 is a view showing the configuration of an AI-based auxiliary diagnostic system 100 for the diagnosis of cerebrovascular lesions according to the present invention. Referring to FIG. 2, the AI-based auxiliary diagnosis system 100 according to the present invention includes a learning database 120, a coordinate system registration module 140 and an AI module 130. In addition, the AI-based auxiliary diagnostic system 100 according to the present invention may include an image interface unit 150 and a main processor 110.

In the learning database 120, a plurality of MRA images labeled with cerebrovascular diseases are stored as learning data. Then, the coordinate system registration module 140 registers the spatial coordinate system in each MRA image stored in the learning database 120. In addition, the AI module 130 learns the MRA image in which the spatial coordinate system is registered by the coordinate system registration module 140 through a pre-registered deep learning based learning model.

A new MRA image to be read is input to the image interface unit 150. When a new MRA image is input, the coordinate system registration module 140 registers a spatial coordinate system for the new MRA image and transmits it to the AI module 130, The AI module 130 diagnoses the disease in the new MRA image in which the spatial coordinate system is registered through a learning model.

The main processor 110 includes a hardware configuration such as a CPU and memory for the operation of the AI-based auxiliary diagnostic system 100 according to the present invention, and a software configuration such as an operating system (OS).

Hereinafter, the process of registering the spatial coordinate system in the MAR image in the AI-based auxiliary diagnostic system 100 according to the present invention will be described in detail with reference to FIG. 3.

First, the vascular region is extracted from the MRA image. As a method of extracting only the vascular region in the MRA image, various known methods can be applied. When the vascular region is extracted, an extraction process of the first plane PL1 is performed (S30). Then, the normal vector with respect to the first plane PL1 is calculated, and as illustrated in FIG. 4A, the normal vector is determined as the first coordinate vector (S31).

For the extraction of the first plane PL1, as shown in FIG. 4 (a), three anatomical first points P1_L, P1_R, and P2_C that are preset in the blood vessel region are extracted. First, as shown in FIG. 5, a pair of intersections of posterior communicating artery (PCom) and internal carotid artery (ICA) is extracted as two first points (P1_L, P1_R).

And, as shown in Figure 6, a posterior communicating artery (PCom, Posterior communicating artery) and a pair of intersections (P2_L, P2_R) of posterior cerebral artery (PCA, Posterior cerebral artery) are extracted. In addition, as shown in FIG. 7, the center point of the pair of intersections P2_L, P2_R of the posterior communicating artery (PCom) and posterior cerebral artery (PCA) is the remaining one first point (P2_C) ).

When three first points P1_L, P1_R, and P2_C are extracted from the anatomical structure of individual blood vessels constituting the cerebral blood vessel as described above, as illustrated in FIG. 8, the first plane formed by the three first points (PL1) is formed in the vascular region, and as illustrated in FIG. 9, the normal vector of the first plane PL1 is determined as the first coordinate vector. Here, the first coordinate vector is applied to determine one axis constituting the spatial coordinate system, for example, the z-axis, which will be described later.

When the first coordinate vector is determined, a process of extracting the coordinate origin O of the spatial coordinate system proceeds (S32). In the present invention, the coordinate origin is, as shown in FIG. 4 (b), using two predetermined anatomical second points P3_C and P4 located on both sides with the first plane PL1 interposed therebetween. Is extracted.

First, as illustrated in FIG. 10, a central point of an anterior communicating artery (ACom) is extracted as one second point P3_C. Then, as illustrated in FIG. 11, a boundary point between a basal arter (BA) and a posterior cerebral artery (PCA) is extracted as a second point P4 having a different boundary point.

When the two second points P3_C and P4 are extracted as described above, as shown in FIGS. 4B and 12, a straight line connecting the two (P3_C, P4) and a first plane PL1 The coordinate origin (O) is extracted from the meeting location (S32).

When the coordinate origin O is extracted, a process of extracting the second plane proceeds (S33). First, a predetermined anatomical third point P5 is extracted. In the present invention, as shown in Figure 13, the brain base artery (BA, Basilar artery) and the vertebral artery (VA, Vertebral artery) as an example of the extraction point of the third point (P5).

Then, the second plane PL2 formed by the two second points P3_C and P4 and one third point P5 is extracted as shown in FIG. 14 (S33). And, as shown in Figure 4 (c) and Figure 15, the first plane (PL1) and the second plane (PL2) intersecting the intersection line is extracted (S34), based on the intersection, the figure, As shown in 16, the second coordinate vector is determined. That is, since the intersection line is orthogonal to the normal vector of the plane, the second coordinate vector may be the other axial direction of the plane coordinate system.

Here, the direction of the second coordinate vector may be determined based on one of the first points P1_L, P1_R, and P2_C, the second points P3_C, and P4, or the third point P5. For example, a direction closer to P1_L among the first points may be determined as a direction of the second coordinate vector.

When the first coordinate vector and the second coordinate vector are determined as described above, as shown in FIG. 16, when the first coordinate vector and the second coordinate vector are moved to the coordinate origin O (S36), the second coordinate axis. And a second coordinate axis, for example, a z-axis and an x-axis are set, and a third coordinate axis orthogonal to the first coordinate axis and the second coordinate axis is determined (S37), so that the spatial coordinate system can be registered.

As described above, by determining the spatial coordinate system based on the anatomical structure between the anatomical individual blood vessels constituting the cerebral blood vessels, even if the MRA image is slightly different according to the cause of the direction of the captured image, the same coordinate system of the MRA image Setting becomes possible.

Therefore, the learning model trained through this provides an effect that the determination accuracy is higher than the learning model in which the spatial coordinate system is not registered.

Although some embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that the present invention may be modified without departing from the principles or spirit of the present invention. . The scope of the invention will be defined by the appended claims and their equivalents.

100: AI-based auxiliary diagnostic system
110: main processor 120: learning database
130: AI module 140: coordinate system registration module
150: video interface

Claims (5)

In the AI-based auxiliary diagnostic system for the diagnosis of cerebrovascular disease,
A learning database storing a plurality of MRA images labeled with cerebrovascular disease,
A coordinate system registration module for registering a spatial coordinate system to each of the MRA images stored in the learning database;
And an AI module for learning the MRA images in which the spatial coordinate systems are respectively registered by the coordinate system registration module through a pre-registered deep learning based learning model;
The coordinate system registration module
(a) extracting a vascular region from the MRA image,
(b) extracting three predetermined anatomical first points from the vascular region;
(c) determining a normal vector for a first plane formed by the three first points as a first coordinate vector,
(d) extracting two predetermined anatomical second points located on both sides with the first plane interposed therebetween;
(e) extracting a location where the straight line connecting the two second points and the first plane meet as a coordinate origin;
(f) extracting a predetermined anatomical third point,
(g) extracting a second plane formed by the two second points and the third point, and extracting an intersection line where the first plane and the second plane intersect;
(h) determining a second coordinate vector based on the intersection line;
(i) determining a first coordinate axis and a second coordinate axis by moving the first coordinate vector and the second coordinate vector to the coordinate origin, respectively;
(j) AI-based assisted diagnostic system for diagnosis of cerebrovascular disease characterized by performing a step of registering the spatial coordinate system by determining a third coordinate axis orthogonal to the first coordinate axis and the second coordinate axis, respectively. According to claim 1,
Step (b) is
(b1) extracting a pair of intersection points of the posterior carotid artery and the internal carotid artery with the two first points;
(b2) extracting a pair of intersection points of the posterior carotid artery and the posterior cerebral artery;
(b3) AI-based assisted diagnostic system for diagnosis of cerebrovascular disease, comprising extracting a center point of a pair of intersections of the posterior carotid artery and the posterior cerebral artery as one of the first points. According to claim 2,
Step (d) is
(d1) extracting the center point of the anterior transit artery as one of the second points;
(d2) AI-based assisted diagnostic system for diagnosis of cerebrovascular disease, comprising extracting a boundary point between the brain basal artery and the posterior cerebral artery as another second point. According to claim 3,
In step (f), an AI-based auxiliary diagnosis system for diagnosis of cerebrovascular disease characterized in that the boundary point between the brain base artery and the spinal artery is extracted as the third point. According to claim 1,
It further includes an image interface unit to which a new MRA image to be diagnosed is input;
The coordinate system registration module registers the spatial coordinate system for the new MRA image and delivers it to the AI module;
The AI module is an AI-based auxiliary diagnostic system, characterized in that the spatial coordinate system is registered to diagnose whether or not a disease in the new MRA image is acquired through the learning model.

Images