KR20210000855A - Method and apparatus for correcting nerve position in dental image - Google Patents

Abstract

The present invention relates to a method and a device for correcting a nerve position. According to an embodiment of the present invention, as a method which corrects the nerve position detected in a three-dimensional image, the method for correcting a nerve position comprises: an image acquisition step of acquiring a curved planar reformation (CPR) image based on a nerve position; a display step of respectively displaying the nerve position and a reference line on the CPR image; and a correction step of correcting the nerve position on the CPR image based on the reference line.

Description

Method and apparatus for correcting nerve position in dental image}

The present invention relates to a method and apparatus for correcting nerve position, and more particularly, to a method and apparatus for correcting the position of a nerve detected in various ways.

In general, an implant refers to a substitute that can replace human tissue when the original human tissue is lost, and in particular, in dentistry, including a fixture, an abutment, and a crown. It means implanting an artificial tooth using a prosthesis at the actual tooth position.

In the dental implant procedure, a perforation is formed in the alveolar bone and a fixture is placed in the perforation, and when the fixture is fused to the alveolar bone, the abutment and crown are combined with the fixture.

In order to prevent problems such as nerve damage, these implant procedures determine the placement position, angle, and depth of the implant based on a cone beam computed tomography (CBCT) image before the implant procedure. Prior to confirmation.

In particular, drilling for alveolar bone perforation during an implant procedure requires maintaining a safe distance between the fixture and the inferior alveolar nerve, which is difficult due to various nerve paths and conditions for each patient. In addition, conventionally, the inferior alveolar nerve has been detected using a panoramic image of a tomography image, but there is an opaque and unclear inferior alveolar nerve in the panoramic image, which may cause damage to the nerve during implantation.

In order to solve this problem, a technique for detecting the position of a nerve based on an anatomical characteristic in which the nerve is distributed has been developed. For example, these nerve position detection techniques can automatically detect nerves faster and more accurately using various machine learning techniques.

However, such a nerve position detection technique has a problem in that the accuracy of the detected nerve position may be degraded.

In order to solve such a conventional problem, an object of the present invention is to provide a method and apparatus for calibrating a nerve position capable of simply and more accurately correcting a nerve position detected through various nerve position detection techniques.

The technical problems to be achieved in the present invention are not limited to the technical problems mentioned above, and other technical problems that are not mentioned can be clearly understood by those of ordinary skill in the technical field to which the present invention belongs from the following description. There will be.

In order to solve the above problems, a nerve position correction method according to an embodiment of the present invention is a method of correcting a nerve position detected in a 3D image, and (1) a curved surface reconstruction (CPR, Curved) based on the nerve position Planar Reformation) An image acquisition step of acquiring an image, (2) a display step of respectively displaying a nerve position and a reference line on the curved surface reconstructed image, (3) a correction of correcting the nerve position on the curved surface reconstructed image based on the reference line Includes steps.

In the nerve position correction method according to an embodiment of the present invention, the image acquisition step, the display step, and the correction step are repeatedly performed, but when repeatedly performed, the detected nerve position is updated to the nerve position corrected in the previous correction step. Thus, each step can be performed based on this.

The step of acquiring the image may include acquiring a curved reconstructed image having an angle different from that of the previously acquired curved reconstructed image during the repetition.

The displaying may include simultaneously displaying the 3D image or the processed image and the curved reconstruction image, and displaying the region of the nerve location on the 3D image or the processed image.

The step of obtaining the image may include receiving a range and an angle of the curved reconstructed image and obtaining a curved reconstructed image reflecting the range and angle.

The displaying may include displaying the reconstructed curved image reflecting the range and angle in real time.

The range or the angle can be input and selected through an input unit.

The step of obtaining the image may include receiving a selection for a specific nerve among a plurality of nerves from the reference image and obtaining the curved reconstruction image based on the nerve position for the input specific nerve.

The coordinate value of the reference line may include at least one of coordinate values of the nerve position.

The reference line may be a line passing through a portion of a neural region displayed on the curved reconstructed image.

The reference line may have a straight line or a curved shape.

A nerve position correction apparatus according to an embodiment of the present invention includes (1) a storage unit storing a nerve position detected in a three-dimensional image and a three-dimensional image, and (2) a control unit controlling to correct the nerve position, The controller acquires a CPR (Curved Planar Reformation) image based on the nerve position, displays a reference line on the curved surface reconstruction image, and corrects the nerve position based on the reference line.

The present invention can easily and more accurately correct the nerve position detected through various nerve position detection techniques, and through this, there is an advantage of enabling a safe implant procedure by more accurately grasping the position and state of different neural tubes for each patient.

The effects obtainable in the present invention are not limited to the above-mentioned effects, and other effects not mentioned can be clearly understood by those of ordinary skill in the art from the following description. will be.

1 is a schematic block diagram of a nerve position correction apparatus 100 according to an embodiment of the present invention.
2 is a flowchart of a method for correcting a nerve position according to an embodiment of the present invention.
3 shows an example of a 3D image displaying a nerve position.
4 shows an example of a curved reconstructed image obtained in FIG. 3.
5 shows a process of obtaining a curved reconstructed image.
6 shows an example in which nerve positions are displayed on the curved reconstructed image of FIG. 4.
7 shows an example in which nerve positions and a reference line are displayed together in the curved reconstructed image of FIG. 4.
8 shows an example of correcting a nerve position displayed in the curved reconstructed image of FIG. 4 along a reference line.
9 shows an example of a curved reconstructed image after correction according to FIG. 8 is performed.
10 to 13 show various examples of a user interface (UI) that displays a curved reconstruction image.
14 shows various ranges of curved reconstructed images.
15 is a view showing a comparison between a conventional correction method and a nerve position correction method according to an embodiment of the present invention.
16 is a schematic block diagram of an electronic device for detecting a nerve in a target image according to an embodiment of the present invention.
17 is a detailed block diagram of the controller 250 of FIG. 16.
18 is a flowchart of a method of detecting a nerve in a target image according to an embodiment of the present invention.
19 is a diagram illustrating a cross-sectional image of a tooth that varies according to the direction of the patient's head.
20 is a diagram illustrating detection targets in a tooth cross-sectional image in the coronal direction.
21 is a diagram illustrating a method of generating learning data according to an embodiment of the present invention.
22 is a flowchart of a step of detecting a nerve in the target image of FIG. 21.
FIG. 23 is a diagram for describing a step of detecting a nerve in the target image of FIG. 21.
24 is a detailed flowchart of a step of detecting a nerve in the target image of FIG. 22.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The detailed description to be disclosed hereinafter together with the accompanying drawings is intended to describe exemplary embodiments of the present invention, and is not intended to represent the only embodiments in which the present invention may be practiced. In the drawings, parts irrelevant to the description may be omitted in order to clearly describe the present invention, and the same reference numerals may be used for the same or similar components throughout the specification.

In an embodiment of the present invention, expressions such as "or" and "at least one" may represent one of words listed together, or a combination of two or more. For example, “A or B” and “at least one of A and B” may include only one of A or B, and may include both A and B.

1 is a schematic block diagram of a nerve position correction apparatus 100 according to an embodiment of the present invention.

Referring to FIG. 1, a nerve position correction apparatus 100 according to an embodiment of the present invention is an electronic device that corrects a nerve position previously detected in a 3D image, and includes a communication unit 110, an input unit 120, and a display unit ( 130), a memory 140, and a control unit 150 may be included.

The communication unit 110 communicates with external devices such as an image acquisition device (not shown) and a server (not shown). For example, the communication unit 110 5G (5 ^th generation communication), LTE-A (long term evolution-advanced), LTE (long term evolution), Bluetooth, BLE (bluetooth low energy), NFC (near field communication) Wireless communication such as, etc. can be performed, and wired communication such as cable communication can be performed.

The input unit 120 generates input data in response to a user's input. The input unit 120 includes at least one input means. For example, the input unit 120 may include a keyboard (key board), a keypad (key pad), a dome switch (dome switch), a touch panel (touch panel), a touch key (touch key), a mouse (mouse), etc. I can.

The display unit 130 displays display data according to an operation of the nerve position correction apparatus 100. For example, the display unit 130 includes a liquid crystal display (LCD), a light emitting diode (LED) display, an organic light emitting diode (OLED) display, and a micro electromechanical system (MEMS). mechanical systems) displays, electronic paper displays, and the like. In addition, the display unit 130 may be combined with the input unit 120 to be implemented as a touch screen.

The memory 140 stores operation programs of the apparatus 100 for correcting nerve position. In addition, the memory 140 may store information on a 2D or 3D image of a part of the human body, a previously detected nerve position, and the like. In this case, the nerve position includes a plurality of coordinate values corresponding to the nerve region in the 3D image. For example, these coordinate values may include center coordinates, edge coordinates, and the like of an area including a nerve.

For example, a two-dimensional or three-dimensional image is a tooth structure including a mental tubercle, a mental nerve, an inferior alveolar nerve, an alveolar bone, and a mandibular angle. It may be a related image, a CT image, an oral scan image, an MRI image, or the like, or an image processed from these images.

The controller 150 controls an operation of correcting a coordinate value of a nerve position previously detected in a 3D image. In addition, the control unit 150 may control operations of the communication unit 110, the input unit 120, the display unit 130, and the memory 140.

Hereinafter, a method of correcting a nerve position according to an embodiment of the present invention controlled by the controller 150 will be described.

2 is a flowchart of a method for correcting a nerve position according to an embodiment of the present invention.

Referring to FIG. 2, a method of correcting a nerve position according to an embodiment of the present invention is a method of correcting a nerve position previously detected in a 3D image, and may include S10 to S30. Hereinafter, for convenience of explanation, a 3D image is an image related to a tooth and its surrounding skeletal structure, and the drawing will be illustrated and described assuming that the nerve previously detected therefrom is an alveolar nerve (especially, inferior alveolar nerve). , The present invention is not limited thereto.

S10 is an image acquisition step, a step of obtaining a Curved Planar Reformation (CPR) image based on a previously detected nerve position.

3 shows an example of a 3D image displaying nerve positions, in which case, nerve positions are indicated in red. In addition, FIG. 4 shows an example of the curved reconstructed image obtained in FIG. 3.

Referring to FIG. 3, the nerve location may be displayed in a separate color (eg, red) in a 3D image. As illustrated in FIG. 4, a curved reconstructed image may be obtained through a process of obtaining a curved reconstructed image to be described later based on the nerve position.

5 shows a process of obtaining a curved reconstructed image. Referring to FIG. 5, a process of obtaining a curved reconstruction image will be described as follows.

First, it may include a nerve stomach 1) for the alveolar nerve (C) in a 3D image of the tooth and its surrounding skeletal structure (see FIG. 5(a)). That is, C _i is coordinate values constituting C, and are N coordinate values for points of the center curve. Accordingly, C may be formed as a spatial area extended by a predetermined length R from the center line where each Ci is connected. That is, C = {C _i ∈R ³ | It may be i1}.

At this time, normal vectors n _i (indicated by a green arrow) at each C _i and a plane P _i perpendicular to each n _i can be derived (see FIG. 5(b)). That is, n _i is vectors representing the tangent direction at each C _i , and has a coordinate value of n _i = <n _x , n _y , n _z >, and can be derived through calculation of C _i+1 -C _i have.

)가 도출될 수 있다(도 5(c) 참고). 즉,

는 r이 P_i에 투영된 벡터이다. 또한, P_i에서 수직 평면 상에서 Ci를 중심점으로 하고

를 반지름으로 하는 영역을 갖는 평면(이하, “정사영 평면”이라 지칭함)이 도출될 수도 있다. 이때, r은 곡면 재구성 영상을 계산하기 위한 그 범위에 대한 파라미터 값으로서, 기 설정되거나 입력될 수 있다. In addition, the orthogonal projection vector according to the reference vector (r) in each C _i (

) Can be derived (see Fig. 5(c)). In other words,

Is the vector r is projected onto P _i . Also, let Ci as the center point on the vertical plane in P _i

A plane (hereinafter referred to as "orthogonal projection plane") having a region with a radius of may be derived. In this case, r is a parameter value for the range for calculating the curved reconstruction image, and may be preset or input.

When such an orthogonal projection plane is collected, a sample area S, which is a curved area for a curved reconstructed image, may be obtained (see FIG. 5(f)). In this case, a curved reconstruction (CPR) image may be finally obtained according to the cross-sectional direction θ in each P _i in S (see FIGS. 5(d) and 5(e)). That is, θ is a parameter value for the angle of CPR, and may be preset or input.

·v_i = |

|·|v_i| cosθ 이다. 이에 따라, CPR을 구성하는 좌표 값들은 C_i + kv_i 와 같다. 이때, k는 y축 범위 지정 스칼라 값(실수, K∈R³)이다. 이에 따라, CPR 영상의 해상도(Sampling 수)는 C_i의 개수(N 값)와, k 값에 의해 결정될 수 있다.That is, the CPR image is composed of a curved plane constituting C, and can be obtained by C _i , r and θ. In particular, n _i ·v _i = 0,

·V _i = |

|·|v _i | is cosθ. Accordingly, the coordinate values constituting the CPR are equal to C _i + kv _i . In this case, k is the y-axis range designation scalar value (real number, K∈R ³ ). Accordingly, the resolution (Sampling number) of the CPR image may be determined by the number of C _i (N value) and k value.

6 shows an example in which nerve positions are displayed on the curved reconstructed image of FIG. 4, and FIG. 7 shows an example in which nerve positions and a reference line are displayed together on the curved reconstructed image of FIG. 4.

Thereafter, S20 is a display step, as shown in FIG. 4, a step of displaying the CPR image acquired in S10. In particular, in S20, as shown in FIGS. 6 and 7, the nerve location (green dotted line) and the reference line (red dotted line) may be displayed together on the CPR image. At this time, points indicating C _i may be displayed as the nerve location.

In particular, the baseline is a guide line for correcting the nerve position. In other words, the alveolar nerve region (C _d1 ) before correction (the black region between the upper and lower white regions) and the line of the nerve location (green dotted line) displayed on the CPR image should be displayed in a straight line due to the characteristics of the alveolar nerve. . However, referring to FIG. 6, C _d1 and its green dotted line appear in a curved shape of'~'. This indicates that a preset nerve position has been established.

Accordingly, the baseline provides a reference for correcting the alveolar nerve region C _d1 before correction displayed in the CPR image on the CPR based on the nerve position that is set incorrectly as described above. That is, the shape of the baseline displayed on the CPR may vary depending on the type and shape of the target nerve, and may be a straight or curved shape. In particular, when the target nerve is an alveolar nerve, the baseline may be displayed on the CPR image in a straight line by reflecting the shape of the alveolar nerve.

However, in most cases, since only a part of the nerve position will be corrected, it may be preferable that the coordinate value of the reference line includes at least one of the coordinate values of the nerve position. That is, the baseline may be a line passing through a portion of the nerve region displayed on the CPR image. Since various types of the reference line are stored, the user may select the reference line through the input unit 120, and a type designated as a default may be presented.

Meanwhile, the nerve position and the reference line may be displayed on the CPR image in different types of lines, shapes, thicknesses, colors, etc. to distinguish them.

8 shows an example of correcting a nerve position displayed in the curved reconstructed image of FIG. 4 along a reference line.

Thereafter, S30 is a correction step, in which the nerve position displayed on the CPR image is corrected on the CPR image based on the baseline displayed on the CPR image. That is, the point or line of the nerve position displayed on the CPR image may be changed on the CPR image (a portion to be changed in FIG. 8 is indicated by a red arrow) to correspond to the reference line. Accordingly, it is possible to correct by changing the coordinate value of the nerve position of the changed part to the coordinate value of the corresponding part in real time. At this time, the selection of the part to be changed among the nerve positions displayed in the CPR image and the degree of change are input from the user through the input unit 120 and are performed according to the input data, or various algorithms such as machine learning are used. It can be done through.

9 shows an example of a curved reconstructed image after correction according to FIG. 8 is performed.

According to S30, if the nerve position displayed on the CPR image is corrected on the CPR image based on the baseline, as shown in FIG. 9, the alveolar nerve region (C _d2 ) after correction displayed on the CPR image (between the upper and lower white regions Black area) is displayed almost in a straight line. This is the C _d2 corresponding to the alveolar nerve, which is the target nerve, indicating that C _d1 has been properly corrected.

Meanwhile, S10 to S30 may be repeatedly performed. In this case, by updating the previously detected nerve position to the nerve position corrected in the previous S30, S10 to S30 may be repeatedly performed based on this. That is, if the nerve position is corrected by performing S10 to S30 as a previous step, when performing S10 to S30 again afterwards, S10 to S30 may be performed based on the nerve position corrected in the previous step. Through such repetitive performance, the nerve position can be more accurately corrected.

In particular, when repeatedly performed, in S10, a CPR image having an angle θ ₂ different from the angle θ ₁ of the previously acquired CPR image may be obtained. However, the selection of θ ₂ is a predetermined angle change value selected or preset according to the input data received from the user through the input unit 120 within a certain range (eg, 90˚ to -90˚). Can be selected according to.

10 to 13 show various examples of a user interface (UI) that displays a curved reconstruction image.

10 to 13, for user convenience, a CPR image may be displayed together with a reference image. In this case, the reference image refers to a 3D image from which a CPR image is derived or an image obtained by processing a corresponding 3D image. In particular, when such a UI is provided, the CPR image and the reference image are simultaneously displayed, but the nerve location region may be displayed in the reference image (indicated by red regions in FIGS. 10 to 13 ).

In addition, when the above-described UI is provided, in S10, a selection of a specific nerve among a plurality of nerves is input from the reference image through the input unit 120, and a CPR image may be obtained based on the input nerve position. For example, the selection for the left or the right of the inferior alveolar nerve may be input through the input unit 120, and S10 is performed according to the input data, and the corresponding nerve position (left inferior alveolar nerve position or right inferior alveolar nerve position ) Based on the CPR image may be obtained.

14 shows various ranges of curved reconstructed images.

In addition, in S10, in addition to the angle θ of the CPR image, a range for the CPR image may be input and a CPR image reflecting the range may be obtained. In this case, the range for the CPR image is a value that can be input from the user through the input unit 120 and may be a value for r. As a result, as shown in FIG. 14, various ranges of CPR images can be obtained. Thereafter, in S200, a CPR image reflecting the corresponding range and angle θ may be displayed in real time.

Meanwhile, when the above-described UI is provided, the angle θ or the range for the CPR image may be selected through the input unit 120. For example, the angle θ or range may be selected through a mouse wheel manipulation, a mouse drag input, a keyboard direction key input, a keyboard value input for providing a scroll bar on the display unit 130, and the like. In particular, the angle (θ) or range of the CPR image may increase or decrease according to the operation of the mouse wheel or the input of the direction keys on the keyboard (in combination with other keys on the keyboard). Such an input means has the advantage of providing greater convenience to the user on the above-described UI.

15 is a view showing a comparison between a conventional correction method and a nerve position correction method according to an embodiment of the present invention.

Meanwhile, in the conventional correction method, as shown in Fig. 15(a), the cross-sectional image is moved and corrected for the cross-sectional image of the cylinder. On the other hand, in the neural position correction method according to an embodiment of the present invention, as shown in FIG. 15(b), cross-sectional rotation and correction are performed on a CPR image based on the center point of the cylinder.

16 is a schematic block diagram of a nerve detection apparatus 200 according to an embodiment of the present invention.

Meanwhile, the apparatus 100 for correcting a nerve position according to an embodiment of the present invention corrects a previously detected nerve position. At this time, the device for detecting such a nerve position, that is, the nerve detection device 200 may be configured as follows, and the nerve position correction device 100 according to an embodiment of the present invention may be the nerve detection device 200. have.

That is, the nerve detection device 200 according to an embodiment of the present invention is an electronic device that detects a nerve position in a 3D image, and includes a communication unit 210, an input unit 220, a display unit 230, a memory 240, and It may include a control unit 250.

The communication unit 210 communicates with external devices such as an image acquisition device (not shown) and a server (not shown). For example, the communication unit 110 5G (5 ^th generation communication), LTE-A (long term evolution-advanced), LTE (long term evolution), Bluetooth, BLE (bluetooth low energy), NFC (near field communication) Wireless communication such as, etc. can be performed, and wired communication such as cable communication can be performed.

The input unit 220 generates input data in response to a user input of the electronic device 200. The input unit 220 includes at least one input means. For example, the input unit 220 includes a keyboard (key board), a keypad (key pad), a dome switch (dome switch), a touch panel (touch panel), a touch key (touch key), a mouse (mouse), a menu button ( menu button), etc.

The display unit 230 displays display data according to an operation of the electronic device 200. For example, the display unit 230 includes a liquid crystal display (LCD), a light emitting diode (LED) display, an organic light emitting diode (OLED) display, a micro electromechanical system (MEMS). mechanical systems) displays and electronic paper displays. The display unit 130 may be combined with the input unit 120 to be implemented as a touch screen.

The memory 240 stores operation programs of the electronic device 200. In addition, the memory 240 may store an algorithm related to a convolutional neural network (CNN) such as U-Net and RCNN. Also, the memory 240 may store a plurality of training images received from an image acquisition device or the like.

The controller 250 learns the training image to generate training data, and detects a nerve in the target image based on the training data. In particular, the control unit 250, based on the anatomical characteristic that the nerve is located between the mandibular nodule and the mandibular angle, first detects the mandibular nodule and the mandibular angle, and then detects the nerve located between the mandibular nodule and the mandibular angle. It can accurately detect nerves.

17 is a detailed block diagram of the controller 250 of FIG. 16.

Referring to FIG. 17, the controller 250 may include an image acquisition unit 251, a learning unit 252, and a nerve detection unit 253.

As such, the control unit 150 has an anatomical site such as a mental tubercle, a mental nerve, an inferior alveolar nerve, an alveolar bone, and a mandibular angle. It is characterized by generating learning data based on a tooth cross-sectional image in the coronal direction, and detecting a nerve based on this.

The image acquisition unit 251 receives a 3D image from an external image acquisition device and converts it into a plurality of 2D cross-sectional images (hereinafter, slices) arranged in a coronal direction. That is, the image acquisition unit 251 acquires a learning image including a plurality of slices arranged in the coronal direction.

The learning unit 252 includes nerves (chintip nerves and inferior alveolar nerves) in a plurality of learning images acquired by the image acquisition unit 251, a jaw nodule and mandibular angle, and inferior alveolar nerves that are a reference for detecting the nerves. At least one of the alveolar bones useful for detecting is generated training data set, and a learning model is generated by learning the training data. Here, the learning unit 252 may learn the training image using an algorithm related to a convolutional neural network (CNN) such as U-Net and RCNN stored in the memory 140.

The nerve detection unit 253 inputs a target image to the learning model generated by the learning unit 252, and detects a nerve in the target image based on the training data. Here, the target image may be a 3D image including a plurality of slices (2D cross-sectional images) arranged in the coronal direction of the patient.

Specifically, the nerve detection unit 253 detects the jaw nodule and the mandibular angle in the target image based on the learning data, sets the section between the detected jaw nodule and the mandibular angle as a nerve effective section, and searches for the set nerve effective section. Nerves (tip nerve and inferior alveolar nerve) are detected. Here, the nerve detection unit 253 may detect the alveolar bone and use it to detect the inferior alveolar nerve. That is, since the inferior alveolar nerve is located inside the alveolar bone, detection accuracy can be improved by detecting the alveolar bone first and then detecting the inferior alveolar nerve from the detected alveolar bone.

Meanwhile, the meaning that the nerve detection unit 253 detects the jaw nodule, the mandibular angle, the alveolar bone, and the nerve may be interpreted as including the meaning of detecting a slice including these detection targets.

The nerve detection unit 253 may set a first search start point based on statistics of a plurality of training images, and detect a jaw nodule by searching in an outward direction of the target image from the set first search start point.

Likewise, the nerve detection unit 253 may set a second search start point based on statistics of the plurality of training images, and detect the mandibular angle by searching in an outward direction of the target image from the second search start point.

Here, the statistical value of the plurality of training images is a value (eg, 0 to 1) obtained by normalizing the position of an anatomical region included in each image, focusing on the difference in resolution of the captured image for each image acquisition device. Statistically, there is little likelihood of finding a jaw nodule above the first search start point (eg, 0.323), and little possibility of finding the mandibular angle below the second search start point (eg, 0.715). Therefore, if the chin nodule and the mandibular angle are searched based on the first search starting point and the second search starting point, the chin tip nodule and the mandibular angle can be searched more quickly and accurately.

The nerve detection unit 253 sets a target region that is expected to include a jaw nodule, a mandibular angle, and a nerve to be detected in the target image. Then, position and size information of the target region is calculated, and a probability value in which the detection target is included in the target region is calculated. In addition, a target region having a calculated probability value equal to or greater than the reference value is detected as a detection region including the detection target, and a slice including the region is labeled so that the detection target is displayed.

The nerve detection unit 253 extracts coordinates (eg, center coordinates) of an area including a nerve among the detection areas, and detects a nerve based on the extracted coordinates (eg, center coordinates). That is, by collecting coordinates (eg, center coordinates) of a region including a nerve from a plurality of slices, a neural tube may be finally extracted.

Hereinafter, a nerve detection method according to an embodiment of the present invention controlled by the controller 250 will be described.

18 is a flowchart of a nerve detection method according to an embodiment of the present invention.

Referring to FIG. 18, in the neural detection method according to an embodiment of the present invention that can be performed by the controller 250, acquiring a training image (S100), generating training data and a learning model (S200) ), inputting a target image to the learning model (S300), and detecting a nerve in the target image (S400).

In the step of acquiring the training image (S100), a 3D image is received from an external image acquisition device and converted into a plurality of 2D cross-sectional images (hereinafter, slices) arranged in a coronal direction. That is, a learning image including a plurality of slices arranged in the coronal direction is acquired.

19 is a diagram illustrating a tooth cross-sectional image that varies according to the direction of the patient's head, and FIG. 20 is a diagram illustrating detection targets in the tooth cross-sectional image in the coronal direction.

Referring to FIG. 19, the direction of the patient's head can be divided into a horizontal direction (a), a coronal (b), and a sagittal (c), depending on the direction of the head. It can be seen that the anatomical parts of the teeth appearing in the tooth cross-sectional image (slice) are different.

19 and 20, the nerve detection method according to an embodiment of the present invention, a chin nodule (mental tubercle), chin-tip nerve (Mental Nerve), inferior alveolar nerve (Inferior Alveolar Nerve), alveolar bone (Alveolar Bone) And it is characterized by generating learning data based on the tooth cross-sectional image in the coronal direction (b) where the anatomical parts such as the mandibular angle are well represented and the distinction is clearly indicated, and nerve detection based on this To do.

For reference, referring to FIG. 20, it can be seen that the inferior alveolar nerve is located inside the alveolar bone, and the jaw nerve is not located inside the alveolar bone, but it can be seen that it is in a form pierced outward.

In the step of generating the training data and the training model (S200), the nerves (chintip nerves and inferior alveolar nerves), the jaw nodules and mandibular angles, and inferior alveolar nerves, which are a reference for detecting the nerves, are At least one of the alveolar bones useful for detection is generated, and training data set is generated, and a learning model is generated by learning the training data.

Here, a learning model may be generated by learning training data using an algorithm related to a convolutional neural network (CNN) such as U-Net and RCNN stored in the memory 240.

21 is a diagram illustrating a method of generating learning data according to an embodiment of the present invention.

Referring to FIG. 21, the training data sets location and size information for an area (e.g., a box) including a chin nodule, mandibular angle, alveolar bone, and nerve (learning target) for each of a plurality of slices included in the training image. I can. Specifically, the x-axis and y-axis positions of the upper left of the box are set as the position coordinates of the box with the upper left of the training image as the origin, and the width and height of the box are determined based on the set position coordinates. Can be set.

Meanwhile, the method of setting the location and size information of the object to be learned in the training image as described above can be applied as it is to the method of setting the position and size information of the object to be detected in the target image to be described later.

In the step of detecting a nerve in the target image (S400), after inputting the target image to the learning model (S300), a nerve is detected in the target image based on the training data. Here, the target image may be a 3D image including a plurality of slices (2D cross-sectional images) arranged in the coronal direction of the patient.

FIG. 22 is a flowchart illustrating an operation of detecting a nerve in the target image of FIG. 21, and FIG. 23 is a diagram illustrating an operation of detecting a nerve in the target image of FIG. 21.

Referring to FIG. 22, in the step of detecting a nerve in a target image (S400), a chin nodule and a mandibular angle are detected in a target image based on the learning data (S410), and a section between the detected chin nodule and the mandibular angle is nerve effective. It is set as a section (S420), and the set nerve effective section is searched to detect the nerves (the jaw nerve and the inferior alveolar nerve) (S430).

Here, the detected alveolar bone can be used to detect the inferior alveolar nerve. That is, since the inferior alveolar nerve is located inside the alveolar bone, detection accuracy can be improved by detecting the alveolar bone first and then detecting the inferior alveolar nerve from the detected alveolar bone.

On the other hand, the meaning of detecting the jaw nodule, the mandibular angle, the alveolar bone, and the nerve may be interpreted as including the meaning of detecting a slice containing these detection targets.

Referring to FIG. 23, in the step of detecting a nerve in the target image (S400), a first search start point S1 is set based on the statistics of a plurality of training images, and the target image at the set first search start point S1. It is possible to detect the chin nodule (MT) by searching in the lateral direction of.

Likewise, a second search start point S2 may be set based on statistics of a plurality of training images, and the mandibular angle MA may be detected by searching in an outward direction of the target image at the second search starting point S2.

Here, a section between the detected jaw nodule MT and the mandibular angle MA is set as a nerve effective section, and nerves (chintip nerves and inferior alveolar nerves) can be detected by searching for a set nerve effective section.

Here, the statistical value of the plurality of training images is a value (eg, 0 to 1) obtained by normalizing the position of an anatomical region included in each image, focusing on the difference in resolution of the captured image for each image acquisition device. Statistically, there is little likelihood of finding a jaw nodule above the first search start point (eg, 0.323), and little possibility of finding the mandibular angle below the second search start point (eg, 0.715). Therefore, if the chin nodule and the mandibular angle are searched based on the first search starting point and the second search starting point, the chin tip nodule and the mandibular angle can be searched more quickly and accurately.

24 is a detailed flowchart of a step of detecting a nerve in the target image of FIG. 22.

Referring to FIG. 24, in the step of detecting a nerve in a target image (S400), a target region that is expected to include a jaw nodule, a mandibular angle, and a nerve to be detected in the target image is set based on the training data (S431). Then, information on the location and size of the target area is calculated (S432), and a probability value that the detection target is included in the target area is calculated (S433). In addition, a target region having a calculated probability value equal to or greater than a reference value may be detected as a detection region including the detection target (S434), and a slice including the region may be labeled so that the detection target is displayed.

Then, the coordinates (eg, center coordinates) of an area including a nerve among the detection areas are extracted (S435), and a nerve is detected based on the extracted coordinates (eg, center coordinates) (S436). That is, by collecting coordinates (eg, center coordinates) of a region including a nerve from a plurality of slices, a neural tube may be finally extracted.

As described above, in the nerve detection method according to an embodiment of the present invention, by detecting the nerve based on the anatomical characteristics of the nerve distribution, the nerve can be detected more quickly and accurately, and through this By accurately grasping the location and condition, you can safely perform the implant procedure.

The embodiments of the present invention disclosed in the present specification and drawings are only provided for specific examples to easily explain the technical content of the present invention and to aid understanding of the present invention, and are not intended to limit the scope of the present invention. Therefore, the scope of the present invention should be construed that all changes or modified forms derived based on the technical idea of the present invention in addition to the embodiments disclosed herein are included in the scope of the present invention.

100: nerve position correction device
110, 210: communication unit 120, 220: input unit
130, 230: display unit 140, 240: memory
150, 250: control unit
200: nerve detection device

Claims (11)

As a method of correcting a nerve position detected in a 3D image,
An image acquisition step of obtaining a Curved Planar Reformation (CPR) image based on the nerve position;
Displaying a nerve position and a reference line on the curved reconstructed image, respectively; And
A correction step of correcting the nerve position on the curved reconstructed image based on the reference line;
Nerve position correction method comprising a.
The method of claim 1,
The image acquisition step, the display step, and the correction step are repeatedly performed, but when repeatedly performed, the detected nerve position is updated to the nerve position corrected in the previous correction step, and each step is performed based on this. Nerve position correction method.
The method of claim 2,
And the step of acquiring the image includes acquiring a curved reconstructed image having an angle different from that of the previously acquired curved reconstructed image when the iteratively performed.
The method of claim 1,
The displaying step comprises simultaneously displaying the 3D image or the processed image and the curved reconstruction image, and displaying the region of the nerve location on the 3D image or the processed image. Nerve position correction method.
The method of claim 4,
The image acquisition step includes receiving a range and an angle for the curved surface reconstructed image and obtaining a curved surface reconstructed image reflecting the range and angle,
The displaying step includes displaying the reconstructed curved image reflecting the range and angle in real time.
The method of claim 5,
The nerve position correction method, characterized in that the range or the angle can be input and selected through an input unit.
The method of claim 4,
The image acquisition step includes receiving a selection for a specific nerve among a plurality of nerves from the reference image and obtaining the curved reconstruction image based on the nerve position for the input specific nerve. Calibration method.
The method of claim 1,
The coordinate value of the reference line comprises at least one of the coordinate values of the nerve position.
The method of claim 1,
The reference line is a nerve position correction method, characterized in that the line passing through a portion of the nerve region displayed on the curved reconstruction image.
The method of claim 1,
The reference line is a nerve position correction method, characterized in that the straight or curved shape.
A storage unit storing a 3D image and a nerve position detected in the 3D image; And
Includes; a control unit for controlling to correct the nerve position,
The control unit,
A nerve position correction apparatus, comprising: acquiring a CPR (Curved Planar Reformation) image based on the nerve position, displaying a reference line on the curved surface reconstruction image, and correcting the nerve position based on the reference line.

Images