KR20240044045A - Method and device for processing medical image - Google Patents

Abstract

Obtaining a plurality of two-dimensional cross-sectional images of the oral cavity of the subject, generating a plurality of pre-processing images in which tooth areas are extracted from the plurality of two-dimensional cross-sectional images using a data recognition model, and performing the plurality of pre-processing A medical image processing method including storing an image is disclosed.

Description

Medical image processing method and device {METHOD AND DEVICE FOR PROCESSING MEDICAL IMAGE}

The present invention relates to a medical image processing method and device, and more specifically, to a technology for selecting teeth from images taken of the oral cavity. This can be used for implant installation guides, dental surgery guides, dental braces, dental crowns, etc.

Accurately modeling the shape of teeth is important for providing restorative dentistry and related services. However, the shape of the teeth and gums varies greatly for each dental patient, and the characteristics cannot be generalized, and it is not easy to distinguish the boundary between the teeth and gums, especially in the periapical area below the gums, so there have been limitations in establishing orthodontic treatment plans. In order to implement precision medicine, the shape of the teeth must be obtained. Conventionally, medical staff extracted the tooth area by manually marking the outline of the tooth in an image taken of the oral cavity. Therefore, it is easy for human error to intervene in the acquisition of tooth shape, and overall accuracy can often vary depending on the relative experience of the practitioner. In addition, a tooth image consisting of about 400 slides has the disadvantage that it takes a long time to obtain the tooth shape because a person must manually mark the tooth boundaries on each slide.

Accordingly, there is a demand for technology to accurately model the shape of teeth through artificial intelligence. For example, Republic of Korea Patent Registration No. 10-21048899 discloses a computer-implemented method and system for generating 3D model data based on a plurality of two-dimensional cross-sectional images of a three-dimensional object.

Republic of Korea Patent Registration No. 10-21048899

The present invention seeks to provide a medical image processing method and device that can accurately extract tooth areas using artificial intelligence.

As a technical means for achieving the above-described problem, the first aspect of the present invention includes obtaining a plurality of two-dimensional cross-sectional images of the oral cavity of a subject, using a data recognition model, from the plurality of two-dimensional cross-sectional images. Generating a plurality of preprocessed images from which tooth regions are extracted; and storing the plurality of pre-processed images.

In addition, the medical image processing method includes distinguishing the tooth area and the gum area in a plurality of two-dimensional cross-sectional images, emphasizing the tooth area and removing the gum area, extracting the outline of the tooth area, and It may include filling the color, and generating a plurality of pre-processed images by removing the outside of the outline from the plurality of two-dimensional cross-sectional images.

In addition, the medical image processing method may include loading a plurality of pre-processed images, setting different colors inside the outline in the plurality of pre-processed images, and generating a three-dimensional tooth model based on the plurality of pre-processed images. You can.

Additionally, the medical image processing method may further include storing each of the differently colored teeth in a 3D tooth model.

Additionally, the plurality of two-dimensional cross-sectional images may be at least one of Cone Beam CT (CBCT) and Multi Detector CT (MDCT) images.

Additionally, a second aspect of the present invention includes a memory in which computer-executable instructions are stored; and a control unit that generates a plurality of pre-processed images using the method of the first aspect by executing the computer-executable instructions.

A third aspect of the present invention can provide a computer-readable storage medium on which a program for executing the method of the first aspect is recorded on a computer.

The present invention has the advantage that tooth shapes can be manufactured more precisely by eliminating operator errors through artificial intelligence. In addition, there is a technical effect of more accurately extracting the boundaries of teeth through the unique tooth boundary determination algorithm of the present invention. Additionally, the tooth area can be quickly determined through artificial intelligence, and by linking it with 3D printing technology, dental procedures can be performed on the same day, eliminating the inconvenience of dental patients and reducing various opportunity costs associated with it.

FIG. 1 is a diagram illustrating an example of generating a 3D tooth model from a plurality of 2D cross-sectional images according to an embodiment.
Figure 2 is a flowchart illustrating a medical image processing method according to an embodiment.
Figure 3 is a diagram illustrating an example of processing a medical image according to an embodiment.
Figures 4A to 4H are diagrams illustrating examples of medical image processing according to another embodiment.
Figure 5 is a diagram illustrating an example of filling a pre-processed image with color for each tooth, according to an embodiment.
Figures 6 and 7 are diagrams showing examples of storing teeth in a 3D tooth model.
Figure 8 is a block diagram showing the configuration of a medical image processing device according to an embodiment.
Figure 9 is a detailed block diagram of a control unit according to an embodiment.
Figure 10 is a block diagram showing in detail a data learning unit according to an embodiment.
Figure 11 is a block diagram showing in detail a data recognition unit according to an embodiment.

Hereinafter, the invention will be described in detail by exemplary embodiments with reference to the attached drawings. The following examples are only intended to embody the invention and do not limit or limit the scope of the invention. Anything that can be easily inferred by an expert in the technical field to which the invention belongs from the detailed description and examples will be interpreted as falling within the scope of the invention's rights.

Terms such as 'consists of' or 'includes' used in this specification should not be construed as necessarily including all of several components or steps, and some of the components or steps may not be included. It may be possible, or it should be interpreted as being able to further include additional components or steps.

The terms used in this specification are described as general terms currently used in consideration of the functions mentioned in this specification, but they may mean various other terms depending on the intention or precedents of those skilled in the art, the emergence of new technologies, etc. You can. Therefore, the terms used in this specification should not be interpreted only by the name of the term, but should be interpreted based on the meaning of the term and the overall content of this specification. Additionally, singular expressions include plural meanings, unless the context clearly indicates singularity.

As used herein (particularly in the claims), “the” and similar referents may refer to both the singular and the plural. Additionally, in the absence of any description explicitly specifying the order of steps describing the method according to the present specification, the steps described may be performed in any suitable order. The present invention is not limited by the order of description of the steps described.

Phrases such as “in one embodiment” that appear in various places in this specification do not necessarily all refer to the same embodiment.

Some embodiments herein may be represented by functional block configurations and various processing steps. Some or all of these functional blocks may be implemented in various numbers of hardware and/or software configurations that perform specific functions.

Additionally, connection lines or connection members between components shown in the drawings merely exemplify functional connections and/or physical or circuit connections. In an actual device, connections between components may be represented by various replaceable or additional functional connections, physical connections, or circuit connections.

These embodiments relate to medical image processing methods and devices, and detailed descriptions of matters widely known to those skilled in the art to which the following embodiments belong will be omitted. Hereinafter, the present invention will be described in detail with reference to the attached drawings.

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

FIG. 1 is a diagram illustrating an example of generating a 3D tooth model from a plurality of 2D cross-sectional images according to an embodiment.

The subject is a person who undergoes an examination to receive a dental procedure, and the method according to the present invention uses a plurality of two images of the subject's oral region to create a three-dimensional model of the subject's individual tooth shape so that the subject receives the optimal treatment. This relates to technology for extracting tooth areas from dimensional cross-sectional images. To do this, you can first take a two-dimensional (2D) cross-sectional image (e.g., computed tomography (CT) scan image or magnetic resonance imaging (MRI) scan image) of the subject's oral region, or load the captured image. there is. For example, the 2D cross-sectional image may be a Cone Beam CT (CBCT) and a Multi Detector CT (MDCT) image. As shown in Figure 1, an exemplary CBCT or MDCT image of the oral cavity of a human subject shows, for example, the temporomandibular joint, vertebral foramen, spine, and teeth. Example objects may include bones, teeth, or body parts to determine the object's medical condition.

Additionally, the method according to the present invention can generate and store a plurality of preprocessed images in which tooth areas are extracted from a plurality of two-dimensional cross-sectional images by applying a tooth boundary determination algorithm. A specific method for generating a plurality of preprocessed images will be described later with reference to FIGS. 2 to 4F. Afterwards, a 3D tooth model of the subject's entire teeth can be created based on the plurality of preprocessed images. Here, individual teeth can be labeled with a different color. And in the 3D tooth model, individual teeth can be saved as STL files. This is a file that is input to a three-dimensional (3D) printer, and the 3D printer can produce it by 3D printing individual tooth shapes through a 3D tooth modeling file. In the past, after converting an STL file, the file was corrected for output, but in the present invention, image capture, 3D modeling, and output are possible without a separate correction process, making it possible to quickly and accurately produce a tooth replica that accurately reflects the structure of an actual tooth. . 3D tooth models make it easier to understand teeth with complex shapes. The shape and arrangement of teeth can be visually expressed from various perspectives through a virtual 3D model generated by a computer. Additionally, the 3D model can be modified to simulate changes in the number or position of teeth, and the 3D tooth model can be implemented as a tangible replica through a 3D printer.

Figure 2 is a flowchart illustrating a medical image processing method according to an embodiment.

In step S210, the medical image processing method may acquire a plurality of two-dimensional cross-sectional images of the subject's oral region. For example, the plurality of two-dimensional cross-sectional images may be MRI images, CT images, Cone Beam CT (CBCT), or Multi Detector CT (MDCT) images. Additionally, the plurality of two-dimensional cross-sectional images may be images captured by the medical image processing device 1000 or may be images received from an external device or server.

In step S220, the medical image processing method may use a data recognition model to generate a plurality of preprocessed images in which tooth regions are extracted from a plurality of two-dimensional cross-sectional images. An example of generating a preprocessed image using a data recognition model will be described later with reference to FIG. 3.

In step S230, a plurality of preprocessed images may be stored. In one embodiment, the plurality of generated preprocessed images may be automatically stored in the medical image processing device 1000 without separate user input or command. Additionally, at this time, the preprocessed image may be saved in dcm or nifti format.

Figure 3 is a diagram illustrating an example of processing a medical image according to an embodiment.

Referring to FIG. 3, in order to extract a tooth area from a plurality of two-dimensional cross-sectional images, a plurality of two-dimensional cross-sectional images may be input into a data recognition model. For example, the input 2D cross-sectional images may be CBCT or MDCT images taken within 10 minutes of the oral cavity of a single subject. The data recognition model may be, for example, a model based on a neural network. For example, models such as Convolutional Neural Network (CNN), Deep Neural Network (DNN), Recurrent Neural Network (RNN), Bidirectional Recurrent Deep Neural Network (BRDNN), etc. may be used as data recognition models, but are not limited thereto. In one embodiment, identification of tooth areas in input images may be performed based on learning according to CNN technology. Through this, a plurality of preprocessed images in which only the subject's teeth area is filled with white can be generated as the output of the data recognition model. Here, the number of output preprocessed images will be equal to the number of input two-dimensional cross-sectional images. Previously, it was not easy to distinguish between the teeth and gum area at the root end of the tooth, but artificial intelligence has the advantage of being able to more accurately distinguish the tooth area at the root end of the tooth. How the data recognition model recognizes and learns data will be described in detail in FIGS. 9 to 11.

Figures 4A to 4F are diagrams illustrating examples of medical image processing according to another embodiment. 4A to 4F show an example of generating a preprocessed image from which a tooth region is extracted semi-automatically with human input involved in some steps, unlike in FIG. 3 .

To generate a plurality of preprocessed images, a plurality of two-dimensional cross-sectional images of the oral cavity of the subject in FIG. 4A can be loaded. In Figure 4b, the tooth area and the gum area can be distinguished from a plurality of two-dimensional cross-sectional images, and the gum area (soft tissue) can be automatically removed and the tooth area (bone tissue) can be emphasized. For this purpose, an auto-windowing image technique can be used. The windowing technique is a method of converting an area of interest to black or white and then increasing the color contrast of the area of interest to make the area of interest appear more distinct. For example, image arithmetic operations through a Gaussian filter, etc. can be performed to emphasize tooth areas. Afterwards, to separate the tooth area, a mask image can be created by optimizing the Gaussian filter and Otsu Thresholding technique. In addition, as a result of the AND operation between the image on which arithmetic operations have been performed through a Gaussian filter, etc., and the above-mentioned mask image, an image in which the gum area is removed and the tooth area is emphasized can be derived (see FIG. 4b). In Figure 4c, the outline of the tooth area can be extracted. In one embodiment, since the degree of exposure and brightness of each tooth is different, the brightness setting for each tooth can be set differently and then the boundary line of each tooth can be distinguished and processed through semi-automatic cumulative extraction of the outline. Figure 4d shows the process of filling color inside the outline of a tooth, and Figure 4e shows completing the setting of the tooth area by filling all the teeth of the subject with color. In order to fill color inside the outline of a tooth, when the user of the device 1000 manually selects a seed position inside the outline, the corresponding closed curve is automatically filled with color (auto hole filling). In Figure 4f, a preprocessed image can be created and stored by removing the area outside the tooth outline. For example, the device 1000 may automatically recognize the inside of the outline as a tooth area and automatically remove noise to generate a preprocessed image. Figure 4g shows re-extraction of the tooth outline portion through an automatic outline tracking algorithm. Figure 4h shows the user of the device 1000 finally confirming the boundary of the tooth by blending Figures 4a and 4g.

Figure 5 is a diagram illustrating an example of filling a pre-processed image with color for each tooth, according to an embodiment.

First, a plurality of stored preprocessed images can be loaded as shown in Figure 4f. In the pre-processed image, the area inside the tooth outline will already be equally filled with white. To distinguish individual teeth more intuitively, the colors inside the outline can be set differently. If color labeling is applied to individual teeth, each tooth can be painted a different color. For example, all of a subject's teeth may be filled with different colors, and some teeth may be filled with the same color. By going through the STlizing (DICOM TO STL) process on a 2D preprocessed image with different colors for each tooth, individual teeth that can distinguish colors can be extracted as 3D STL.

Figures 6 and 7 are diagrams showing examples of storing teeth in a 3D tooth model.

If a 3D tooth model is created based on a plurality of 2D preprocessed images, a 3D tooth model for all of the subject's teeth will be created. When a STlizing (DICOM TO STL) process is performed to store differently colored teeth in a 3D tooth model, a grid-like staircase phenomenon is prominently displayed on the teeth, as shown in FIG. 6 . In one embodiment, when two adjacent teeth are separated into individual teeth, a crack (hole) may form in the previously connected area. This technology automatically groups multiple cracks (multi hole grouping), and when the cut surface or crack is convex, the crack can be automatically filled (auto repair) through a simple filling algorithm. In the right picture of FIG. 6 and the left picture of FIG. 7, a tooth whose crack has been automatically repaired is shown. However, due to the staircase phenomenon, it was not expressed as smoothly as a round tooth. By applying the Laplacian Smoothing algorithm, it is possible to eliminate the staircase phenomenon caused by layered modeling that occurs due to the characteristics of the STlizing (DICOM TO STL) process for 3D printing. The individual tooth model, which is smoothly expressed like a round tooth after removing the staircase phenomenon, can be seen in the picture on the right of FIG. 7.

Figure 8 is a block diagram showing the configuration of a medical image processing device according to an embodiment. As shown in FIG. 8, the medical image processing device 1000 according to an embodiment of the present invention may include a control unit 1300 and a memory 1700. However, not all of the illustrated components are essential components. The device 1000 may be implemented with more components than the illustrated components, or the device 1000 may be implemented with fewer components. Below, we look at the above components in turn.

The control unit 1300 typically controls the overall operation of the device 1000. The control unit 1300 may include at least one processor. Depending on its function and role, the control unit 1300 may include a plurality of processors or may include one processor in an integrated form. A processor may mainly refer to a central processing unit (CPU), an application processor (AP), or a graphics processing unit (GPU). Additionally, the CPU, AP, or GPU may include one or more cores therein, and the CPU, AP, or GPU may operate using an operating voltage and clock signal. However, while a CPU or AP consists of a few cores optimized for serial processing, a GPU can consist of thousands of smaller, more efficient cores designed for parallel processing.

For example, the control unit 1300 executes programs stored in the memory 1700, such as a user input unit (not shown), an output unit (not shown), a sensing unit (not shown), a communication unit (not shown), and A/ The V input unit (not shown) can be controlled overall. Additionally, the control unit 1300 may cause the device 1000 to generate a plurality of preprocessed images from which tooth regions are extracted.

The memory 1700 stores data supporting various functions of the control unit 1300. The memory 1700 may store a number of application programs (application programs or applications) running on the control unit 1300, data for operating the control unit 1300, and commands. At least some of these applications may be downloaded from an external server via wireless communication. Additionally, the application program may be stored in the memory 1700, installed in the control unit 1300, and driven by a processor to perform the operation (or function) of the control unit 1300.

The memory 1700 is a flash memory type, hard disk type, solid state disk type, SDD type (Silicon Disk Drive type), and multimedia card micro type. ), card-type memory (e.g. SD or XD memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), EEPROM (electrically erasable programmable read) -only memory), PROM (programmable read-only memory), magnetic memory, magnetic disk, and optical disk may include at least one type of storage medium. Additionally, the memory 520 may include web storage that performs a storage function on the Internet.

The above-described present invention can be modeled as computer-readable code on a program-recorded medium. Computer-readable media includes all types of recording devices that store data that can be read by a computer system. Examples of computer-readable media include HDD (Hard Disk Drive), SSD (Solid State Disk), SDD (Silicon Disk Drive), ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc. and also includes those modeled in the form of carrier waves (e.g., transmission over the Internet).

Meanwhile, the configuration of the device 1000 shown in FIG. 8 is an example, and each component of the device 1000 may be integrated, added, or omitted depending on the specifications of the device 1000 to be implemented. That is, as needed, two or more components may be combined into one component, or one component may be subdivided into two or more components. In addition, the functions performed by each component (or module) are for explaining the embodiments, and the specific operations or devices do not limit the scope of the present invention.

Figure 9 is a block diagram of the control unit 1300 according to one embodiment.

Referring to FIG. 9, the control unit 1300 according to one embodiment may include a data learning unit 1310 and a data recognition unit 1320.

The data learning unit 1310 may analyze a plurality of two-dimensional cross-sectional images, identify the tooth area, and learn a standard for generating a plurality of preprocessed images from which the tooth area is extracted. The data learning unit 1310 can learn standards regarding what data to use to determine a predetermined tooth area and how to determine the tooth area using the data. The data learning unit 1310 acquires data to be used for learning, and applies the acquired data to a data recognition model to be described later, thereby learning standards for generating a plurality of preprocessed images from which tooth regions are extracted.

The data recognition unit 1320 may determine which area in the plurality of two-dimensional cross-sectional images is a tooth area based on the data. The data recognition unit 1320 can recognize the tooth area from predetermined data using a learned data recognition model. The data recognition unit 1320 acquires predetermined data according to a preset standard through learning and uses a data recognition model using the obtained data as an input value, thereby extracting the tooth area based on the predetermined data. . Additionally, the result value output by the data recognition model using the acquired data as an input value can be used to update the data recognition model.

At least one of the data learning unit 1310 and the data recognition unit 1320 may be manufactured in the form of at least one hardware chip and mounted on the device 1000. For example, at least one of the data learning unit 1310 and the data recognition unit 1320 may be manufactured in the form of a dedicated hardware chip for artificial intelligence (AI), or an existing general-purpose processor (e.g. CPU Alternatively, it may be manufactured as part of an application processor) or a graphics-specific processor (e.g., GPU) and may be mounted on the various devices described above.

In this case, the data learning unit 1310 and the data recognition unit 1320 may be mounted on one device, a server, or may be mounted on separate devices or servers. For example, one of the data learning unit 1310 and the data recognition unit 1320 may be included in the device 1000, and the other may be included in the server. In addition, the data learning unit 1310 and the data recognition unit 1320 may provide model information built by the data learning unit 1310 to the data recognition unit 1320 through wired or wireless communication, and the data recognition unit ( Data input through 1320) may be provided to the data learning unit 1310 as additional learning data.

Meanwhile, at least one of the data learning unit 1310 and the data recognition unit 1320 may be implemented as a software module. When at least one of the data learning unit 1310 and the data recognition unit 1320 is implemented as a software module (or a program module including instructions), the software module is a readable non-transitory program that can be read by a computer. It may be stored on a readable recording medium (non-transitory computer readable media). Additionally, in this case, at least one software module may be provided by an operating system (OS) or a predetermined application. Alternatively, part of at least one software module may be provided by an operating system (OS), and the remaining part may be provided by a predetermined application.

Figure 10 is a block diagram of the data learning unit 1310 according to one embodiment.

Referring to FIG. 10, the data learning unit 1310 according to one embodiment includes a data acquisition unit 1310-1, a preprocessing unit 1310-2, a learning data selection unit 1310-3, and a model learning unit 1310. -4) and a model evaluation unit (1310-5).

The data acquisition unit 1310-1 may acquire data necessary for generating a plurality of preprocessed images from which tooth regions are extracted. The data acquisition unit 1310-1 may acquire data necessary for learning to extract the tooth area.

For example, the data acquisition unit 1310-1 may acquire video data (eg, image, video), etc. For example, the data acquisition unit 1310-1 may receive data through an input device (eg, a microphone, camera, or sensor) of the device 1000. Alternatively, the data acquisition unit 1310-1 may acquire data through an external device that communicates with the device 1000.

Additionally, for example, the data acquisition unit 1310-1 may receive data input from a user, load data previously stored in the device 1000, or receive data from a server, but is not limited thereto. No. For example, necessary data may be obtained by combining data previously stored in the device 1000, data sensed by the device 1000, data input from the user, and data obtained from the server. The data may include at least one of an image, a video, and image metadata. For example, the data acquisition unit 1310-1 may receive an image as input. For example, the data acquisition unit 1310-1 is a camera of the device 1000 including the data learning unit 1310, or an external CT scanner capable of communicating with the device 1000 including the data learning unit 1310. You can input an image through .

The preprocessor 1310-2 may preprocess the acquired data so that the acquired data can be used for learning to determine the degree of correlation between text and images. The pre-processing unit 1310-2 can process the acquired data into a preset format so that the model learning unit 1310-4, which will be described later, can use the acquired data for learning to determine the correlation between text and images. there is.

For example, the preprocessor 1310-2 may overlap at least a portion of the plurality of images based on a common area included in each of the plurality of input images (or frames). The common area may be an area containing the same or similar common objects (eg, teeth and gums, etc.) in each of the plurality of images. Alternatively, the common area may be an area in which the color, shade, RGB value, or CMYK value is the same or similar in each of the plurality of images.

The learning data selection unit 1310-3 may select data required for learning from preprocessed data. The selected data may be provided to the model learning unit 1310-4. The learning data selection unit 1310-3 may select data required for learning from preprocessed data according to preset standards for determining the tooth area. Additionally, the learning data selection unit 1310-3 may select data according to preset criteria through learning by the model learning unit 1310-4, which will be described later.

The model learning unit 1310-4 may learn standards for how to determine the tooth area based on the learning data. Additionally, the model learning unit 1310-4 may learn standards for what learning data should be used to determine the tooth area.

Additionally, the model learning unit 1310-4 can train a data recognition model used to determine the tooth area using training data. In this case, the data recognition model may be a pre-built model. For example, a data recognition model may be a pre-built model that receives basic training data (e.g., sample images, etc.) as input.

A data recognition model may be built considering the application field of the recognition model, the purpose of learning, or the computer performance of the device. The data recognition model may be, for example, a model based on a neural network. For example, models such as Convolutional Neural Network (CNN), Deep Neural Network (DNN), Recurrent Neural Network (RNN), Bidirectional Recurrent Deep Neural Network (BRDNN), etc. may be used as data recognition models, but are not limited thereto.

According to various embodiments, when there are a plurality of pre-built data recognition models, the model learning unit 1310-4 selects a data recognition model with a high correlation between the input training data and the basic training data as a data recognition model to be learned. You can decide. In this case, the basic learning data may be pre-classified by data type, and the data recognition model may be pre-built by data type. For example, the basic training data is pre-classified by various criteria such as the region where the training data was created, the time the training data was created, the size of the training data, the genre of the training data, the creator of the training data, the type of object in the training data, etc. It may be.

Additionally, the model learning unit 1310-4 may train a data recognition model using, for example, a learning algorithm including error back-propagation or gradient descent. .

Additionally, the model learning unit 1310-4 may learn a data recognition model through, for example, supervised learning using training data as input. In addition, the model learning unit 1310-4 discovers a standard for determining the correlation between text and images, for example, by learning the type of data needed to determine the correlation between text and images without any guidance. Through unsupervised learning, a data recognition model can be learned. In addition, the model learning unit (1310-4) is a data recognition model through Decision Tree (DT), Support Vector Machine (SVM), Semi-Supervised Learning (SSL), etc. can also be learned. Additionally, the model learning unit 1310-4 may learn a data recognition model through, for example, reinforcement learning that uses feedback regarding whether the result of tooth region extraction according to learning is correct.

Additionally, when the data recognition model is learned, the model learning unit 1310-4 may store the learned data recognition model. In this case, the model learning unit 1310-4 may store the learned data recognition model in the memory 1700 of the device 1000 including the data recognition unit 1320. Alternatively, the model learning unit 1310-4 may store the learned data recognition model in the memory of a server connected to the device 1000 through a wired or wireless network.

In this case, the memory where the learned data recognition model is stored may also store, for example, commands or data related to at least one other component of the device 1000. Additionally, memory may store software and/or programs. A program may include, for example, a kernel, middleware, an application programming interface (API), and/or an application program (or “application”).

The model evaluation unit 1310-5 inputs evaluation data into the data recognition model, and if the recognition result output from the evaluation data does not meet a predetermined standard, it can cause the model learning unit 1310-4 to learn again. there is. In this case, the evaluation data may be preset data for evaluating the data recognition model. For example, if the difference between the tooth area output by inputting the image selected as the tooth area by the user into the data recognition model and the tooth area selected by the user exceeds a predetermined value, the model learning unit 1310-4 can be evaluated as something that needs to be learned again.

For example, the model evaluation unit 1310-5 applies a predetermined criterion when the number or ratio of evaluation data for which the recognition result is inaccurate among the recognition results of the data recognition model learned for the evaluation data exceeds a preset threshold. It can be evaluated as unsatisfactory. For example, when the predetermined standard is defined as a ratio of 2%, and the learned data recognition model outputs incorrect recognition results for more than 20 evaluation data out of a total of 1000 evaluation data, the model evaluation unit 1310-5 It can be evaluated that the learned data recognition model is not suitable.

Meanwhile, when there are a plurality of learned data recognition models, the model evaluation unit 1310-5 evaluates whether each learned video recognition model satisfies a predetermined standard, and recognizes the model that satisfies the predetermined standard as the final data. You can decide as a model. In this case, when there are multiple models that satisfy the predetermined criteria, the model evaluation unit 1310-5 may determine one or a predetermined number of models preset in descending order of evaluation scores as the final data recognition model.

Meanwhile, in the data learning unit 1310, the data acquisition unit 1310-1, preprocessing unit 1310-2, learning data selection unit 1310-3, model learning unit 1310-4, and model evaluation unit 1310 At least one of -5) may be manufactured in the form of at least one hardware chip and mounted on the device 1000. For example, at least one of the data acquisition unit 1310-1, the preprocessor 1310-2, the learning data selection unit 1310-3, the model learning unit 1310-4, and the model evaluation unit 1310-5. One may be manufactured in the form of a dedicated hardware chip for artificial intelligence (AI), or it may be manufactured as part of an existing general-purpose processor (e.g. CPU or application processor) or graphics-specific processor (e.g. GPU), as described above. It can be mounted on various devices.

In addition, the data acquisition unit 1310-1, preprocessing unit 1310-2, learning data selection unit 1310-3, model learning unit 1310-4, and model evaluation unit 1310-5 are one device. It may be mounted on a server, or it may be mounted on separate devices. For example, some of the data acquisition unit 1310-1, preprocessing unit 1310-2, learning data selection unit 1310-3, model learning unit 1310-4, and model evaluation unit 1310-5. may be included in the device 1000, and the remaining portion may be included in the server.

In addition, at least one of the data acquisition unit 1310-1, the preprocessor 1310-2, the learning data selection unit 1310-3, the model learning unit 1310-4, and the model evaluation unit 1310-5 It can be implemented as a software module. At least one of the data acquisition unit 1310-1, preprocessing unit 1310-2, learning data selection unit 1310-3, model learning unit 1310-4, and model evaluation unit 1310-5 is a software module. When implemented as a program module (or a program module including instructions), the software module may be stored in a non-transitory computer readable recording medium that can be read by a computer. Additionally, in this case, at least one software module may be provided by an operating system (OS) or a predetermined application. Alternatively, part of at least one software module may be provided by an operating system (OS), and the remaining part may be provided by a predetermined application.

Figure 11 is a block diagram of the data recognition unit 1320 according to one embodiment.

Referring to FIG. 11, the data recognition unit 1320 according to one embodiment includes a data acquisition unit 1320-1, a preprocessing unit 1320-2, a recognition data selection unit 1320-3, and a recognition result providing unit ( 1320-4) and a model update unit 1320-5.

The data acquisition unit 1320-1 may acquire data necessary for determining the tooth area, and the preprocessing unit 1320-2 may preprocess the acquired data so that the acquired data can be used to determine the tooth area. there is. The pre-processing unit 1320-2 may process the acquired data into a preset format so that the recognition result providing unit 1320-4, which will be described later, can use the acquired data to determine the tooth area.

The recognition data selection unit 1320-3 may select data necessary for determining the tooth area from the preprocessed data. The selected data may be provided to the recognition result provider 1320-4. The recognition data selection unit 1320-3 may select some or all of the preprocessed data according to preset criteria for determining the tooth area. Additionally, the recognition data selection unit 1320-3 may select data according to preset criteria through learning by the model learning unit 1310-4, which will be described later.

The recognition result provider 1320-4 may determine the tooth area by applying the selected data to the data recognition model. The recognition result providing unit 1320-4 can provide recognition results according to the recognition purpose of the data. The recognition result provider 1320-4 can apply the selected data to the data recognition model by using the data selected by the recognition data selection unit 1320-3 as an input value. Additionally, the recognition result may be determined by a data recognition model.

For example, the image recognition result may be provided as a video, image, or command (eg, application execution command, module function execution command, etc.). The recognition result provider 1320-4 may apply a composite image generated from a plurality of images to a data recognition model to provide an image recognition result. As an example, the recognition result providing unit 1320-4 may provide a recognition result of an object included in an image. The recognition result may, for example, provide information about teeth included in the oral image in the form of a 3D model, image, or command.

The model updating unit 1320-5 may update the data recognition model based on the evaluation of the recognition result provided by the recognition result providing unit 1320-4. For example, the model updating unit 1320-5 provides the recognition result provided by the recognition result providing unit 1320-4 to the model learning unit 1310-4, so that the model learning unit 1310-4 The data recognition model can be updated.

Meanwhile, in the data recognition unit 1320, a data acquisition unit 1320-1, a pre-processing unit 1320-2, a recognition data selection unit 1320-3, a recognition result providing unit 1320-4, and a model update unit ( At least one of 1320-5) may be manufactured in the form of at least one hardware chip and mounted on the device 1000. For example, among the data acquisition unit 1320-1, pre-processing unit 1320-2, recognition data selection unit 1320-3, recognition result providing unit 1320-4, and model update unit 1320-5. At least one may be manufactured in the form of a dedicated hardware chip for artificial intelligence (AI), or may be manufactured as part of an existing general-purpose processor (eg, CPU or application processor) or graphics-specific processor (eg, GPU). It can be mounted on various devices.

In addition, the data acquisition unit 1320-1, the preprocessor 1320-2, the recognition data selection unit 1320-3, the recognition result provision unit 1320-4, and the model update unit 1320-5 are one unit. It may be mounted on a device, or it may be mounted on separate devices. For example, among the data acquisition unit 1320-1, pre-processing unit 1320-2, recognition data selection unit 1320-3, recognition result providing unit 1320-4, and model update unit 1320-5. Some may be included in the device 1000, and others may be included in the server.

In addition, at least one of the data acquisition unit 1320-1, the preprocessor 1320-2, the recognition data selection unit 1320-3, the recognition result provision unit 1320-4, and the model update unit 1320-5. Can be implemented as a software module. At least one of the data acquisition unit 1320-1, the pre-processing unit 1320-2, the recognition data selection unit 1320-3, the recognition result providing unit 1320-4, and the model update unit 1320-5 is software. When implemented as a module (or a program module including instructions), the software module may be stored in a non-transitory computer readable recording medium that can be read by a computer. Additionally, in this case, at least one software module may be provided by an operating system (OS) or a predetermined application. Alternatively, part of at least one software module may be provided by an operating system (OS), and the remaining part may be provided by a predetermined application.

Meanwhile, the above-described embodiments of the present invention can be written as a program that can be executed on a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium.

The computer-readable recording media includes storage media such as magnetic storage media (eg, ROM, floppy disk, hard disk, etc.) and optical readable media (eg, CD-ROM, DVD, etc.).

Although embodiments of the present invention have been described with reference to the above and the attached drawings, those skilled in the art will understand that the present invention can be implemented in other specific forms without changing the technical idea or essential features. You will understand that it exists. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive.

1000: device
1300: Control unit
1700: Memory

Claims (7)

Obtaining a plurality of two-dimensional cross-sectional images of the oral region of the subject;
Using a data recognition model, generating a plurality of preprocessed images in which tooth regions are extracted from the plurality of two-dimensional cross-sectional images; and
storing the plurality of pre-processed images;
Including, a medical image processing method.
According to paragraph 1,
The step of generating the plurality of preprocessed images is,
distinguishing a tooth area and a gum area from the plurality of two-dimensional cross-sectional images;
emphasizing the tooth area and removing the gum area;
Extracting the outline of the tooth area;
filling the inside of the outline with color; and
generating a plurality of pre-processed images by removing the exterior of the outline from the plurality of two-dimensional cross-sectional images;
Including, a medical image processing method.
According to paragraph 2,
Loading the plurality of pre-processed images;
setting different colors inside the outline in the plurality of pre-processed images; and
Generating a 3D tooth model based on the plurality of preprocessed images;
A medical image processing method further comprising:
According to paragraph 3,
Saving each of the differently colored teeth in the three-dimensional tooth model;
A medical image processing method further comprising:
According to paragraph 1,
A medical image processing method, wherein the plurality of two-dimensional cross-sectional images are at least one of CBCT (Cone Beam CT) and MDCT (Multi Detector CT) images.
a memory in which computer executable instructions are stored; and
a control unit generating a plurality of pre-processed images using the method of any one of claims 1 to 5 by executing the computer-executable instructions;
A medical image processing device comprising:
A computer-readable storage medium on which a program for executing the method of any one of claims 1 to 5 on a computer is recorded.