KR20200083822A - Computing device for analyzing dental image for orthodontic daignosis and dental image analyzing method - Google Patents

Abstract

According to one embodiment of the present invention, provided are a dental image analysis method for correction diagnosis and an apparatus using the same. The apparatus includes: at least one memory storing a program for dental image analysis; a communication part obtaining a dental image of a patient; and at least one processor preprocessing the dental image by resizing the dental image, detecting at least some of a plurality of landmarks for correction diagnosis from the preprocessed dental image by using a landmark detection module, and controlling the detected landmarks such that the landmarks overlap the dental image and is displayed. The landmarks are anatomical reference points indicating at least one of a facial skeleton, teeth and a facial contour which are required for correction diagnosis and the landmark detection module can include a machine learning module based on an artificial neural network.

Description

A computing device that supports dental image analysis for orthodontic diagnosis and a method for analyzing dental images{COMPUTING DEVICE FOR ANALYZING DENTAL IMAGE FOR ORTHODONTIC DAIGNOSIS AND DENTAL IMAGE ANALYZING METHOD}

The present invention relates to a dental image analysis method for corrective diagnosis and a device using the same, and more specifically, to a dental image analysis method capable of accurately and quickly detecting a plurality of measurement points for corrective diagnosis from a dental image and a device using the same It is an invention.

In general, occlusion refers to a state in which the teeth of the maxilla and mandible are engaged with each other when the mouth is closed. And, the malocclusion refers to an incorrect occlusal relationship in which the arrangement of the teeth is not aligned due to any cause or the upper and lower mandibular engagement state is out of the normal position, which is a functional and aesthetic problem.

Here, the cause of the malocclusion is known to have a large genetic effect, but may be caused by various causes, such as a problem in the shape or size of the teeth, environmental effects, bad habits, wrong posture, and congenital disorders such as dental caries.

When malocclusion occurs, the teeth are not uniform and food residue is likely to remain between the teeth. In addition, since it is not easy to clean it with an accurate brushing, the plaque in the oral cavity increases, so it is easy to progress to gum disease such as dental caries or gum inflammation. Moreover, if there is a tooth that deviates from the normal dentition or the position of the jaw is abnormal, there is a high possibility of damage to the tooth, such as a tooth fracture when an external impact is applied.

Thus, orthodontic treatment is performed to treat malocclusion. Here, orthodontic treatment uses the property that the tooth moves when it receives an external force. The orthodontic treatment may use various devices and methods according to the cause or treatment time, and may be classified into, for example, a device that suppresses or enhances the growth of the upper and lower jaw bones, or a device that gradually moves a tooth to a desired position.

In order to perform such orthodontic treatment appropriately for a patient, judgment on a patient's face type must be given priority. The cephalometric analysis method illustrated in FIG. 1 is mainly used for the face shape determination (ie, corrective diagnosis).

This cephaloanalysis is a method of determining a face shape for orthodontic treatment using an anatomical point reference point indicating a relative position of a facial skeleton, a tooth, or a facial contour. Conventionally, a head radiograph (cephalogram) of a patient in need of orthodontic treatment Looking at ), the orthodontist manually marked the required reference point by hand and judged the patient's face shape based on the relative angles of the straight lines connecting the reference point.

However, since such a conventional method is a method in which the proofreader randomly marks the required reference points based on his or her academic academy, standardization and sharing of the reference points is difficult because the reference points used for face-type judgment are different for each correction. Since it is necessary for the clinician to manually mark the reference point, it takes a lot of time, and has a problem in that deviations in accuracy occur depending on the skill of the clinician.

Therefore, there is a need for an orthodontic image analysis method for orthodontic diagnosis that can solve these problems.

The present invention was devised to solve the above problems, and the present invention aims to provide a dental image analysis method and an apparatus using the same for corrective diagnosis capable of accurately and quickly detecting a plurality of measurement points for corrective diagnosis from a dental image. Is done.

The technical problems of the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the following descriptions.

According to an embodiment of the present invention, a computing device supporting dental image analysis for orthodontic diagnosis is provided. The device is provided with a dental image analysis method for corrective diagnosis and an apparatus using the same according to an embodiment of the present invention. The device includes at least one memory for storing a program for dental image analysis; Communication unit for obtaining the dental image of the examinee; And pre-processing the dental image by resizing the dental image size, detecting at least a portion of a plurality of landmarks for orthodontic diagnosis using a measurement point detection module from the pre-processed dental image, and detecting the detected multiples. And at least one processor that controls to display the measurement point of the superimposed on the dental image, wherein the measurement point is an anatomical point reference point indicating a relative position of at least one of a facial skeleton, a tooth, and a facial contour required for orthodontic diagnosis. , The measurement point detection module may include a machine learning module based on an artificial neural network.

According to the present invention, by using a machine learning module based on an artificial neural network, automatically and consistently automatically presenting 80 or more landmarks from a dental image of a trainee to a work level of a skilled major, so that accuracy, convenience, and speed of corrective diagnosis Can be increased.

BRIEF DESCRIPTION OF THE DRAWINGS In order to better understand the drawings cited in the detailed description of the present invention, a brief description of each drawing is provided.
1 schematically shows a conventional Cephalo analysis method.
2 illustrates an exemplary configuration of a computing device that performs a dental image analysis method for orthodontic diagnosis according to an embodiment of the present invention.
3 exemplarily illustrates a hardware and software architecture of a computing device performing a dental image analysis method for orthodontic diagnosis according to the present invention.
4 shows a method of analyzing a dental image for orthodontic diagnosis according to an embodiment of the present invention.
5 illustrates an embodiment of step S420 of FIG. 4.
6 illustrates a method of analyzing dental images for orthodontic diagnosis according to an embodiment of the present invention.
FIG. 7 shows an embodiment of step S630 of FIG. 6.
8 exemplarily shows a boundary box for detecting a measurement point in a dental image analysis room for orthodontic diagnosis according to an embodiment of the present invention.
9 and 10 in the dental image analysis method for orthodontic diagnosis according to an embodiment of the present invention, illustratively shows the process of generating a standard dental image.
11 exemplarily shows a process of performing a dental image analysis method for orthodontic diagnosis according to an embodiment of the present invention.
Throughout the specification, when a part is "connected" to another part, this includes not only "directly connected" but also "indirectly connected" with another element in between. . Throughout the specification, when a part “includes” a certain component, it means that the component may further include other components, not to exclude other components, unless otherwise stated. In addition, in describing the components of the embodiments of the present invention, terms such as first, second, A, B, (a), (b), and the like can be used. These terms are only for distinguishing the component from other components, and the nature, order, or order of the component is not limited by the term.
2 illustrates an exemplary configuration of a computing device that performs a dental image analysis method for orthodontic diagnosis according to an embodiment of the present invention.
The computing device 100 according to an embodiment of the present invention includes a communication unit 110, a processor 120, and a memory 130, and directly or indirectly with an external computing device (not shown) through the communication unit 110. Can communicate.
Specifically, computing device 100 includes typical computer hardware (eg, a computer processor, memory, storage, input device and output device, other devices that may include components of existing computing devices; electronics such as routers, switches, etc.) A desired system using a combination of communication devices; electronic information storage systems such as network attached storage (NAS) and storage area networks (SAN)) and computer software (i.e. instructions that cause a computing device to function in a particular way). It may be to achieve performance.
The communication unit 110 of the computing device 100 may transmit and receive requests and responses with other computing devices that are linked, for example, such requests and responses may be made by the same TCP session, but are not limited thereto. , And may be transmitted and received as a UDP datagram, for example. In addition, in a broad sense, the communication unit 110 may include a keyboard, a mouse, and other external input devices for receiving commands or instructions.
In addition, the processor 120 of the computing device 100 may include hardware configurations such as a micro processing unit (MPU) or a central processing unit (CPU), a cache memory, and a data bus. Also, it may further include a software configuration of an operating system and an application performing a specific purpose.
In one embodiment, the processor 120 pre-processes the dental image by resizing the dental image size, and uses at least some of the plurality of landmarks for orthodontic diagnosis using a measurement point detection module from the pre-processed dental image. It is possible to detect and control a plurality of detected measurement points to be superimposed and displayed on the dental image.
In one embodiment, the processor 120 may control to train the machine learning module from learning data including a plurality of accumulated comparative dental images.
In one embodiment, the processor 120 may control the measurement point detection module to detect the plurality of measurement points based on a single convolution network.
In one embodiment, the processor 120 detects a plurality of bounding boxes predicted to exist at least a portion of individual anatomical features corresponding to each of the plurality of measuring points, and at least one of the detected bounding boxes For each of the parts, a predetermined point included inside may be controlled to be determined as the measurement point.
In one embodiment, the processor 120 calculates the probability of existence of the individual anatomical features for each of the bounding boxes, and when a plurality of bounding boxes are detected for one individual anatomical feature, based on the probability of existence By filtering one of the plurality of bounding boxes corresponding to the individual anatomical features, it can be controlled to determine a predetermined point included in the filtered bounding box as the measurement point.
In one embodiment, the processor 120 may control to determine a center coordinate as the measurement point for at least a portion of the detected bounding box.
In one embodiment, the processor 120 identifies the missing measurement point by comparing the detected measurement point with the preset plurality of measurement points, and based on standard landmark information, at least one of the detected measurement points A standard dental image having a standard measurement point corresponding to a part may be searched and controlled to determine the location of the missing measurement point using the searched standard dental image and the standard measurement point of the searched standard dental image. In this case, the standard measurement point information may include a plurality of the standard dental images and a plurality of the standard dental images read for each of the standard dental images.
In one embodiment, the processor 120 is disposed adjacent to the missing measurement point among the detected measurement points based on information on a plurality of adjacent measurement points disposed adjacent to each of the standard measurement points included in the standard measurement point information. It may be controlled to search for the standard dental image having the standard measurement point corresponding to a plurality of measurement points.
In one embodiment, the processor 120 extracts a region of the detected measurement point from the dental image, and normalizes the extracted region to the same scale as the standard dental image to normalize the relative coordinates of the detected measurement point It can be controlled to search the standard dental image based on the relative coordinates of the detected measurement point and the relative coordinates of the standard measurement point, and determine the location of the missing measurement point.
In one embodiment, the processor 120 may control to determine the face shape of the examinee for orthodontic treatment by performing a cephalometric analysis based on the detected measurement point.
Also, various data associated with the operation of the computing device 100 may be stored in the memory 130 of the computing device 100. In particular, a program for performing analysis on a dental image may be stored in the memory 130. The memory 130, as is known to those skilled in the art, includes a hard disk drive (HDD), read only memory (ROM), random access memory (RAM), electrically erasable and programmable read only memory (EEPROM), and flash memory (flash). Memory, CF (Compact Flash) card, SD (Secure Digital) card, SM (Smart Media) card, MMC (Multimedia) card or memory stick (Memory Stick), etc. It may be provided inside the computing device 100, or may be provided in a separate device.
3 exemplarily illustrates a hardware and software architecture of a computing device performing a dental image analysis method for orthodontic diagnosis according to the present invention.
Referring to FIG. 3, the processor 120 of the computing device 100 according to an embodiment of the present invention includes an image acquisition module 210, a measurement point detection module 220, a measurement point correction module 230, and a face shape determination module 240 and a storage and transmission module 250. For example, each module may be implemented by the processor 120 and the communicator 110 and/or the memory 130, or may be implemented to operate in conjunction therewith.
The image acquisition module 210 may acquire a dental image of the examinee from another external computing device or another device (such as a dental imaging device) that is connected to the external computing device or the computing device 100 through the communication unit 110. Here, the dental image may be a head cephalogram obtained by X-ray of the head side of the examinee.
The measurement point detection module 220 may detect a plurality of landmarks required for orthodontic diagnosis from a dental image. Here, the plurality of measurement points refers to an anatomical point reference point indicating relative positions of at least one of the facial skeleton, teeth, and facial contour required for orthodontic diagnosis, and consists of N pieces determined by a user's setting or set as a default. It may be, preferably, it may be composed of 80.
The measurement point detection module 220 may include a machine learning module 222, a filtering module 224, and a measurement point determination module 226.
The machine learning module 222 is implemented to simultaneously detect a plurality of objects from an image or image, and an artificial neural network, in particular, a convolution neural network (CNN) or the like It can be implemented based on a modified/improved artificial neural network.
In one embodiment, the machine learning module 222 may be implemented as a single convolution network to enable fast simultaneous detection of multiple objects. For example, an artificial neural network implemented by a YOLO (you only look once) algorithm may be applied, but is not limited thereto, according to an embodiment to which the present invention is applied, for detecting a plurality of objects such as SSD, R-CNN, etc. Various suitable algorithms or artificial neural networks can be applied.
The machine learning module 222 may include a plurality of convolutional layers and a fully connected layer. Here, the plurality of convolutional layers abstract the image to extract features, and the fully-connected layer may be implemented to predict the output probability of the detection object and the coordinates of the bounding box detecting it.
In the present invention, the machine learning module 222 may identify (or detect) individual anatomical features corresponding to a plurality of measurement points from a dental image through a boundary box. For example, the machine learning module 222 may divide the dental image into a plurality of cells and allocate a predetermined number of bounding boxes for each cell, and if individual anatomical features exist in a specific cell, the corresponding cell The bounding box assigned to can be implemented to identify it.
Through this, the machine learning module 222 is a boundary box in which individual anatomical features corresponding to a plurality of measurement points from a dental image are present, and coordinates, sizes, and respective anatomical features in the bounding box. You can predict the likelihood of existence.
The filtering module 224 may filter the bounding box detected by the machine learning module 222 based on the probability of existence of individual anatomical features. Specifically, when two or more boundary boxes are detected for one individual anatomical feature, the filtering module 224 converts one of the plurality of boundary boxes into a boundary box in which the respective anatomical features exist based on the probability of existence. You can choose.
The measurement point determination module 226 may reflect a filtering result and determine a predetermined point included therein for each of the last selected boundary boxes as a measurement point. For example, the measurement point determination module 226 may be implemented to determine the center coordinate of each boundary box as a measurement point.
The measurement point correction module 230 may identify whether a measurement point missing by the measurement point detection module 220 exists, and predict the location (or coordinate) of the missing measurement point using standard landmark information. Here, the standard measurement point information may include information regarding a plurality of the standard dental images, a plurality of the standard measurement points read for each of the plurality of standard dental images, and/or a plurality of adjacent measurement points disposed adjacent to each of the standard measurement points. Can.
The face shape determination module 240 may classify or determine the face shape of the examinee for orthodontic treatment by performing a cephalometric analysis based on at least some of the finally detected measurement points. Based on this face type, the diagnostician can establish a corrective treatment plan for the patient in the future.
The storage and transmission module 250 stores learning data for training of the machine learning module 222 (eg, a comparative dental image), a dental image of a patient, and a measurement point detection result in the memory 130, This may be transmitted to an external computing device, a display device, etc. through the communication unit 110.
FIG. 4 shows a method of analyzing a dental image for orthodontic diagnosis according to an embodiment of the present invention, and FIG. 5 shows an embodiment of step S420 of FIG. 4.
In step S410, the image acquisition module 210, through the communication unit 110, to obtain a dental image of the examinee from another device (such as a dental imaging device) interlocked with another computing device or the computing device 100 outside Can. As described above, the dental image may be a head radiograph of the examinee.
In step S420, the measurement point detection module 220 may detect at least some of a plurality of measurement points for orthodontic diagnosis from the dental image of the examinee. In one embodiment, step S420 may include steps S510 to S540, as shown in FIG. 5.
In step S510, the measurement point detection module 220 may pre-process the dental image by determining whether a predetermined quality is satisfied for the dental image, and resizing the dental image that satisfies the quality. That is, the measurement point detection module 220 may enlarge or reduce the dental image of the examinee at the same scale or ratio as the dental image previously learned by the machine learning module 222. Through this, the method 400 according to the present invention can further improve the detection accuracy of the machine learning module 222. The dental image may preferably be implemented to be resized to 416X640 pixels.
In step S520, the machine learning module 222 predicts that at least some of the individual anatomical features corresponding to each of the plurality of measurement points are present in the dental image of the examinee, based on the learning results of the accumulated plurality of comparative dental images. A plurality of bounding boxes to be detected can be detected, and the probability of existence of each individual anatomical feature in each bounding box can be calculated.
In one embodiment, step S520 may be performed through three steps of detection according to the degree of abstraction of the image. That is, the dental image of the examinee is abstracted to different levels while passing through a plurality of convolution layers included in the machine learning module 222, and the machine learning module 222 is individually anatomized at three different levels of abstraction. It can be implemented to perform detection for a bounding box containing features and calculation of the probability of existence of individual anatomical features.
As a result of performing step S520, information on the center coordinates, size, and probability of existence of each individual anatomical feature for each boundary box may be generated as output values.
In operation S530, the filtering module 224 may perform filtering on the bounding box based on the probability of existence of individual anatomical features. For example, when two or more bounding boxes are detected for one individual anatomical feature by applying the detection of step 3 in step S520, the filtering module 224 selects one of the plurality of bounding boxes based on the probability of existence. The individual anatomical features can be selected as a bounding box. In one embodiment, the filtering module 224 may be implemented to select one of the plurality of bounding boxes having the highest probability of existence of the respective anatomical feature.
In step S540, the measurement point determination module 226 may determine a point in the filtered bounding box as the coordinates of the measurement point. For example, the measurement point determination module 226 may determine the center coordinates of the detected bounding box as coordinates of the measurement points in correspondence with each individual anatomical feature.
Subsequently, in step S430, the face shape determination module 240 may classify or determine the face shape of the examinee for orthodontic treatment by performing cephalo analysis based on at least some of the detected measurement points. For example, if the face shape determination module 240 selects a part for calculating a significant straight line or angle required for the cephalo analysis from the detected measurement point, the cephalo analysis is automatically performed based on the selected measurement point, thereby performing orthodontic treatment. The face type of the examinee will be classified or judged. Here, the face shape for orthodontic treatment may include a hyperdivergent pattern, normodivergent pattern, hypodivergent pattern, etc., but this is exemplary, and according to an embodiment applied to the present invention, the face shape is more varied according to the relative position protruding degree of the maxilla and mandible. Can be classified. As described above, when the face type is determined, the diagnostician can establish an overall plan for corrective treatment based on this.
Meanwhile, although not shown in FIGS. 4 and 5, in one embodiment, the method 400 may further include learning the machine learning module 222. For example, such learning may be performed using a plurality of cumulative comparative dental images. That is, it can be implemented to train the machine learning module 222 by accumulating and collecting dental images of other examinees whose measurement points have been read by a specialist, and inputting this into the machine learning module 222 as training data. In this case, the size of the comparative dental image to be learned may be, for example, 416 X640 pixels.
Further, in one embodiment, the method 400 may further include displaying the detected measurement point. That is, when detection of the measurement point is completed through step S420, the storage and transmission module 250 transmits information on at least some of the detected measurement points to the display device or another computing device to which it is coupled through the communication unit 110. It can be displayed to a diagnostician or the like. In one embodiment, this indication may be performed based on the diagnostic preferences of the diagnostician. Here, the preferred measurement point information may include information on at least one of the area of the diagnosed person, the school of origin, the preferred school style related to corrective diagnosis, and the area of the examinee. For example, the measurement point detection module 220 or the storage and transmission module 250 selects some measurement points based on the diagnostic point preference information of the diagnostician, and transmits only the information related to the measurement points to a display device or the like, or selects some measurement points selected from the display device. It can be embodied to be highlighted in a certain way.
FIG. 6 shows a method of analyzing a dental image for orthodontic diagnosis according to an embodiment of the present invention, and FIG. 7 shows an embodiment of step S630 of FIG. 6.
Steps S610, S620, and S640 in the method 600 are the same as steps S410 to S430 of the method 400 described above with reference to FIGS. 4 and 5, and a detailed description thereof will be omitted here.
In step S630, the measurement point correction module 230 may correct the missing measurement point based on the standard measurement point information when there is an undetected measurement point among the plurality of measurement points in step S620.
Here, the standard measurement point information includes information on a plurality of the standard dental images, a plurality of the standard measurement points read for each of the plurality of standard dental images, and/or a plurality of adjacent measurement points disposed adjacent to each of the standard measurement points. Can. The standard dental image may be generated, for example, by extracting the area of the measurement point from the original dental image at which the measurement point is determined by a specialist, and in this case, information regarding the standard measurement point may be obtained from the standard measurement point in each standard dental image. It may further include information about the relative coordinates.
In one embodiment, step S630 may include steps S710 to S740, as shown in FIG. 7.
In step S710, the measurement point correction module 230 may identify at least one measurement point that is missing detection in step S620. That is, the measurement point correction module 230 may identify a missing measurement point by contrasting a plurality of measurement points set by a user or set as default values and detected measurement points.
Meanwhile, in step S710, if the missing measurement point is not identified, steps S720 to S740 described below are not performed, and step S640 may be performed immediately.
In step S720, the measurement point correction module 230 may calculate relative coordinates of at least some of the detected measurement points. For example, step S720 may be performed by extracting a measurement point from the dental image of the patient whose measurement point has been detected and normalizing it to the same scale as a standard dental image. That is, hereinafter, as described above with reference to FIGS. 10 to 11, the measurement point correction module 230 extracts a measurement point presence area from the dental image, and then scales the area to convert at least one relative coordinate of the measurement point (0 , 0) to (1, 1).
In one embodiment, step S720 may be implemented to be performed on two or more measurement points disposed adjacent to the missing measurement point.
In step S730, the measurement point correction module 230 may search for a standard dental image having a standard measurement point corresponding to at least a part of the measurement points detected using the calculated relative coordinates. For example, the measurement point correction module 230 includes a plurality of peripheral points (preferably 5 to 7 measurement points) disposed adjacent to the missing measurement point, and a standard measurement point corresponding to the measurement point missing from each standard dental image. By comparing the relative coordinates of a plurality of adjacent measurement points arranged adjacently, a standard dental image having a standard measurement point closest to the peripheral measurement point of the missing measurement point can be searched.
In step S740, the measurement point correction module 230 may determine the location (or coordinate) of the missing measurement point using the standard measurement point of the found standard dental image. That is, the measurement point correction module 230 sets the relative coordinates of the standard measurement point corresponding to the missing measurement point in the searched standard dental image as the relative coordinates of the missing measurement point, and scales the relative coordinates to the original of the dental image, The location (or coordinate) of the missing measurement point can be determined.
On the other hand, although not shown in FIGS. 6 and 7, in one embodiment, the method 600 comprises re-learning the machine learning module 222 based on information about the final corrected measurement point and the dental image of the examinee. It may further include. As described above, by making the machine learning module 222 re-learn the result of the correction of the missing measurement point, it is possible to further improve the detection accuracy of the machine learning module 222.
8 exemplarily shows a boundary box for detecting a measurement point in a dental image analysis room for orthodontic diagnosis according to an embodiment of the present invention.
Referring to FIG. 8, the plurality of bounding boxes 810 may identify an area in which individual anatomical features corresponding to a plurality of measurement points are defined from a dental image. Here, the size of a label size defining each individual anatomical characteristic may be set to 30X30 pixels, in order to maximize detection accuracy.
In one embodiment, the machine learning module 222 may divide the dental image into a plurality of cells and allocate a predetermined number of bounding boxes for each cell, when individual anatomical features exist in a specific cell, corresponding The bounding box assigned to the cell can be implemented to detect it.
Accordingly, as described above, the machine learning module 222 provides information about the center coordinates (relative coordinates in each cell), size (width, height), and probability of existence of each individual anatomical feature for each bounding box. Will output
9 and 10 in the dental image analysis method for orthodontic diagnosis according to an embodiment of the present invention, illustratively shows the process of generating a standard dental image.
9 and 10, a standard dental image may be generated based on a predetermined original dental image from which a measurement point is read by a specialist. Here, the original dental image may be at least a part of the comparative dental image provided as learning data of the machine learning module 222.
That is, for example, in the standard dental image, the presence area of the standard measurement point is extracted based on the outermost measurement points of 2 or more, and then the scale is converted into a coordinate area corresponding to (0, 0) to (1, 1). By doing so, a standard dental image can be generated. Accordingly, each standard measurement point in the transformed coordinate region may have relative coordinates between (0, 0) to (1, 1).
11 exemplarily shows a process of performing a dental image analysis method for orthodontic diagnosis according to an embodiment of the present invention.
Referring to FIG. 11, the process of performing a dental image analysis method for orthodontic diagnosis according to an embodiment of the present invention is summarized as follows.
When the dental image of the examinee acquired through the image acquisition module 210 and the communication unit 110 is input to the machine learning module 222, the machine learning module 222 abstracts the dental image through a plurality of convolutional layers, Depending on the degree of abstraction, it detects a bounding box that is predicted to have individual anatomical features corresponding to each measurement point at three levels.
Subsequently, when a plurality of bounding boxes for one individual anatomical feature is detected, the filtering module 224 filters the bounding box having the highest probability of existence based on the probability of existence of the corresponding anatomical feature, and a measurement point determination module In step 226, the center coordinate of the last detected bounding box is determined as a measurement point according to the filtering result.
Subsequently, the measurement point correction module 230 identifies whether there is a missing measurement point among a plurality of set measurement points, and determines the location (or coordinate) of the missing measurement point with reference to standard measurement point information, and finally set All measurement points can be detected, for example, superimposed on a dental image on a display device and output in the form of coordinates or points.
Meanwhile, various embodiments described in this specification may be implemented by hardware, software, and/or combinations thereof. For example, various embodiments may include one or more on-demand semiconductors (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGA) ), processors, controllers, microcontrollers, microprocessors, other electronic units designed to perform the functions presented herein, or combinations thereof.
Also, for example, various embodiments may be recorded or encoded in a computer-readable medium including instructions. Instructions stored or encoded on a computer-readable medium may cause a programmable processor or other processor to perform a method, such as when instructions are executed. Computer-readable media includes computer storage media. The storage medium may be any available medium that can be accessed by a computer. For example, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage medium, magnetic disk storage medium or other magnetic storage device, or instructions that are accessible by a computer to a desired program code, or And any other media that can be used to store data in the form of structures.
Such hardware, software, etc. may be implemented within the same device or within separate devices to support various operations and functions described herein. Additionally, components, units, modules, components, etc., described as “˜parts” in the present invention may be implemented individually together as separate or interoperable logic devices. The depiction of different features for modules, units, etc. is intended to highlight different functional embodiments, and does not necessarily mean that they must be realized by individual hardware or software components. Rather, the functionality associated with one or more modules or units may be performed by separate hardware or software components or incorporated within common or separate hardware or software components.
Although the operations are shown in the drawings in a specific order, it should not be understood that these operations are performed in a specific order, or in a sequential order, to achieve the desired result, or that all the illustrated actions need to be performed. . In any environment, multitasking and parallel processing can be advantageous. Moreover, the division of various components in the above-described embodiments should not be understood as requiring such division in all embodiments, and the described components are generally integrated together into a single software product or packaged into multiple software products. It should be understood that it can.
As described above, optimal embodiments have been disclosed in the drawings and specifications. Although specific terms have been used herein, they are only used for the purpose of describing the present invention and are not used to limit the scope of the present invention as defined in the claims or claims. Therefore, those skilled in the art will appreciate that various modifications and other equivalent embodiments are possible. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the appended claims.

Claims (15)

As a dental imaging analysis method for orthodontic diagnosis,
Obtaining a dental image of the examinee; And
And detecting at least a part of a plurality of landmarks for orthodontic diagnosis using the measurement point detection module from the dental image,
The measurement point is an anatomical point reference point indicating a relative position of at least one of a facial skeleton, a tooth, and a face contour required for orthodontic diagnosis, and the measurement point detection module includes a machine learning module based on an artificial neural network. , Way. According to claim 1,
Further comprising the step of learning the machine learning module from the learning data including a plurality of cumulative comparison dental image,
The comparative dental image is a dental image of another examinee whose measurement point is read by a specialist. According to claim 1,
The dental image is a head radiograph (cephalogram), method. According to claim 1,
In the step of detecting at least some of the measurement points, the measurement point detection module detects the plurality of measurement points based on a single convolution network. According to claim 1,
The step of detecting at least a portion of the measurement point,
Detecting a plurality of bounding boxes predicted to exist at least part of individual anatomical features corresponding to each of the plurality of measuring points; And
And for each of at least some of the detected bounding boxes, determining a predetermined point included therein as the measurement point. The method of claim 5,
The step of detecting at least a portion of the measurement point,
Further comprising the step of resizing (resizing) the received dental image,
The detecting step is performed based on the resized dental image. The method of claim 5,
The step of detecting at least a portion of the measurement point,
Calculating the probability of existence of the individual anatomical feature for each of the bounding boxes,
The determining step,
If a plurality of bounding boxes are detected for one individual anatomical feature, filtering one of the plurality of bounding boxes corresponding to the one individual anatomical feature based on the existence probability; And
And determining a predetermined point included in the filtered bounding box as the measurement point. The method of claim 5,
In the determining step, a center coordinate is determined as the measurement point for at least a portion of the detected bounding box. According to claim 1,
Identifying a measurement point that is missing detection by comparing the detected measurement point with the plurality of predetermined measurement points; And
Searching for a standard dental image having a standard measurement point corresponding to at least a part of the detected measurement points based on standard landmark information-the standard measurement point information includes a plurality of the standard dental images and a plurality of the standards Contains information about the plurality of standard measurement points read for each dental image;
And determining the location of the missing measurement point using the searched standard dental image and the standard measurement point of the searched standard dental image. The method of claim 9,
The standard measurement point information further includes information about a plurality of adjacent measurement points disposed adjacent to each of the standard measurement points,
In the step of searching for the standard dental image, the standard dental image having the standard measurement point corresponding to a plurality of measurement points disposed adjacent to the missing measurement point among the detected measurement points is searched based on information about the adjacent measurement point. How to. The method of claim 9,
The standard dental image is generated by extracting the area of the standard measurement point from the original dental image, and the information on the standard measurement point includes information on the relative coordinates of the standard measurement point on the standard dental image,
The above method,
Further comprising the step of extracting the area of the detected measurement point from the dental image, and normalizing the extracted area to the same scale as the standard dental image to calculate the relative coordinates of the detected measurement point,
The step of searching for the standard dental image and determining the location of the missing measurement point is performed based on the relative coordinates of the detected measurement point and the relative coordinates of the standard measurement point. According to claim 1,
Receiving information of a preferred measurement point of a diagnostician; And
And highlighting and displaying a part of the detected measurement points corresponding to the preferred measurement point information. According to claim 1,
And determining a face shape of the examinee for corrective treatment by performing a cephalometric analysis based on the detected measurement point. A computer-readable recording medium in which a program for performing the method according to any one of claims 1 to 13 is recorded. A computing device supporting dental image analysis for orthodontic diagnosis,
Communication unit for obtaining the dental image of the examinee; And
And a processor including a measurement point detection module that detects at least a portion of a plurality of landmarks for corrective diagnosis from the dental image,
The measurement point is an anatomical point reference point indicating a relative position of at least one of a facial skeleton, a tooth, and a face contour required for orthodontic diagnosis, and the measurement point detection module includes a machine learning module based on an artificial neural network. , Device.

Images