A method for segmenting tooth images, a storage medium, and an electronic device.

Abstract

This invention provides a method for tooth image segmentation, a storage medium, and an electronic device. The method includes the following steps: S1: acquiring a dental model image; S2: inputting the dental model image into a trained first neural network, outputting vector information corresponding to the feature points of the teeth in the dental model image; S3: performing cluster analysis on all the vector information, filtering out several vector information sets, and determining the corresponding tooth center point coordinates based on the vector information in each set; S4: segmenting the dental model image based on the coordinates of each tooth center point to obtain a segmented tooth image. This solution, by first obtaining the vector information corresponding to the feature points on the teeth, then performing cluster analysis based on all the vector information to obtain the center point coordinates of each tooth in the dental model image, and finally completing the identification and segmentation of a single tooth based on the center point coordinates, can effectively improve the accuracy of tooth image segmentation.

Description

Technical Field

[0001] This invention relates to the field of image processing, and more particularly to a method for segmenting tooth images, a storage medium, and an electronic device. Background Technology

[0002] With the continuous development of digital oral technology, oral auxiliary diagnostic and treatment systems based on three-dimensional digital technology all rely on the establishment of three-dimensional digital models of the patient's teeth and gums. Traditional methods include obtaining three-dimensional digital tooth and gum models by using silicone rubber to mold plaster casts and scanning the casts, or by directly obtaining the patient's three-dimensional digital tooth and gum models using a digital intraoral scanner. The resulting tooth model data has high accuracy.

[0003] Dental image segmentation involves dividing a dental model image containing numerous dental details into several images containing information about individual teeth. Due to the crisscrossing dental information on the dental model image, existing dental image segmentation methods suffer from problems such as low segmentation accuracy and low efficiency. Summary of the Invention

[0004] Therefore, a technical solution for tooth image segmentation is needed to address the problem of low accuracy in existing tooth image segmentation methods.

[0005] In a first aspect, the present invention provides a method for segmenting a tooth image, comprising the following steps:

[0006] S1: Obtain a dental model image; the dental model image includes information about at least one tooth;

[0007] S2: Input the dental model image into the trained first neural network, and output the vector information corresponding to the feature points where the tooth information is located on the dental model image; the vector information points from the feature point to the tooth center point that is closest to the feature point.

[0008] S3: Perform cluster analysis on all the vector information, filter out several vector information sets, and determine the corresponding tooth center point coordinates based on the vector information in each vector information set; the distance between the endpoints of vector information in the same vector information set is less than a preset error range;

[0009] S4: Segment the dental model image according to the coordinates of the center points of each tooth to obtain segmented tooth images.

[0010] In some embodiments, the dental model image is a three-dimensional image;

[0011] The steps following step S1 and before step S2 include:

[0012] S11: Adjust the size of each voxel on the dental model image to a first size; the first size is the voxel size on the dental model image input into the first neural network during training.

[0013] Inputting the dental model image into the trained first neural network includes: inputting the dental model image adjusted in step S11 into the trained first neural network.

[0014] In some embodiments, inputting the dental model image into the trained first neural network further includes:

[0015] Key regions are extracted from the dental model image to obtain key region images, and these key region images are input into the trained first neural network; the key region images contain all tooth information from the dental model image.

[0016] In some embodiments,

[0017] The first neural network is trained in the following manner:

[0018] Acquire several dental sample images, each dental sample image containing at least one tooth information, and each tooth information corresponding to a tooth center point;

[0019] The dental sample image is input into the first neural network for iterative training, and the vector information corresponding to the feature points in each tooth information is calculated and output.

[0020] In some embodiments, step S4 includes:

[0021] S41: Input the determined coordinates of several tooth center points and the dental model image into the trained second neural network, and output the tooth image segmented based on the coordinates of each tooth center point.

[0022] In some embodiments, the dental model image is a three-dimensional image;

[0023] The steps following step S3 and before step S41 include:

[0024] S31: Adjust the size of each voxel on the dental model image to the second size; the second size is the size of the voxels on the dental model image input into the second neural network during training.

[0025] Inputting the dental model image into the trained second neural network includes: inputting the dental model image adjusted in step S31 into the trained second neural network.

[0026] In some embodiments, inputting the determined coordinates of several tooth center points and the dental model image into the trained second neural network includes:

[0027] Centered on each tooth's center point, the dental model image is segmented into a corresponding number of images containing information about a single tooth. The coordinates of each tooth's center point and its corresponding image containing information about a single tooth are sequentially input into the trained second neural network.

[0028] In some embodiments, determining the coordinates of the corresponding tooth center point based on the vector information in each of the vector information sets includes:

[0029] The average value of the endpoint coordinates corresponding to the vector information in each of the vector information sets is used as the coordinates of the corresponding tooth center point.

[0030] Secondly, embodiments of this application also provide a storage medium storing a computer program, which, when executed by a processor, implements the method steps of the first aspect of this invention.

[0031] Thirdly, embodiments of this application also provide an electronic device, including a processor and a storage medium, wherein the storage medium is the storage medium as described in the second aspect;

[0032] The processor is used to execute a computer program stored in a storage medium to implement the method steps as described in the first aspect.

[0033] Unlike existing technologies, this invention provides a tooth image segmentation method, storage medium, and electronic device. The method includes the following steps: S1: acquiring a dental model image; S2: inputting the dental model image into a trained first neural network, outputting vector information corresponding to the feature points of the teeth in the dental model image; S3: performing cluster analysis on all the vector information, filtering out several vector information sets, and determining the corresponding tooth center point coordinates based on the vector information in each set; S4: segmenting the dental model image based on the center point coordinates of each tooth to obtain a segmented tooth image. This solution, by inputting the dental model image into a trained first neural network to obtain the vector information corresponding to the feature points of the teeth, and performing cluster analysis based on all the vector information to obtain the center point coordinates of each tooth in the dental model image, and then performing segmentation of a single tooth based on the center point coordinates, can effectively improve the accuracy of tooth image segmentation. Attached Figure Description

[0034] Figure 1 This is a flowchart of the tooth image segmentation method according to the first exemplary embodiment of this application;

[0035] Figure 2A flowchart of the tooth image segmentation method according to a second exemplary embodiment of the present invention;

[0036] Figure 3 A flowchart of the tooth image segmentation method according to the third exemplary embodiment of the present invention;

[0037] Figure 4 This is a flowchart of the tooth image segmentation method according to the fourth exemplary embodiment of the present invention;

[0038] Figure 5 This is a flowchart of the tooth image segmentation method according to the fifth exemplary embodiment of the present invention;

[0039] Figure 6 This is a schematic diagram illustrating the principle of the tooth image segmentation process described in this invention;

[0040] Figure 7 This is a schematic diagram of the preliminary segmentation result of a dental model image according to an exemplary embodiment of the present invention;

[0041] Figure 8 This is a schematic diagram of a tooth image obtained by segmenting a dental model image, as an exemplary embodiment of the present invention.

[0042] Figure 9 This is a schematic diagram of an electronic device according to an exemplary embodiment of the present invention;

[0043] Figure label:

[0044] 1. Dental model image;

[0045] 20. Images of key areas;

[0046] 30. Dental information;

[0047] 301. Feature points;

[0048] 302. Center point of the tooth;

[0049] 10. Electronic devices;

[0050] 101. Processor;

[0051] 102. Storage medium. Detailed Implementation

[0052] To illustrate the possible application scenarios, technical principles, implementable specific solutions, and achievable objectives and effects of this application in detail, the following description, in conjunction with the listed specific embodiments and accompanying drawings, provides a detailed explanation. The embodiments described herein are merely illustrative of the technical solutions of this application and are therefore intended to limit the scope of protection of this application.

[0053] In this document, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The term "embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment, nor does it specifically limit its independence or connection with other embodiments. In principle, in this application, as long as there are no technical contradictions or conflicts, the technical features mentioned in each embodiment can be combined in any way to form corresponding implementable technical solutions.

[0054] Unless otherwise defined, the technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the use of related terms herein is merely for the purpose of describing particular embodiments and is not intended to limit this application.

[0055] In the description of this application, the term "and / or" is used to describe the logical relationship between objects, indicating that three relationships can exist. For example, A and / or B means: A exists, B exists, and A and B exist simultaneously. Additionally, the character " / " in this document generally indicates that the preceding and following objects have an "or" logical relationship.

[0056] In this application, terms such as “first” and “second” are used only to distinguish one entity or operation from another, and do not necessarily require or imply any actual quantity, hierarchy or order relationship between these entities or operations.

[0057] Unless otherwise specified, the use of terms such as “comprising,” “including,” “having,” or other similar expressions in this application is intended to cover non-exclusive inclusion, which does not exclude the presence of additional elements in a process, method, or product that includes elements, such that a process, method, or product that includes a series of elements may include not only those defined elements but also other elements not expressly listed, or elements inherent to such process, method, or product.

[0058] Similar to the understanding in the Examination Guidelines, in this application, expressions such as "greater than," "less than," and "exceeding" are understood to exclude the stated number; expressions such as "above," "below," and "within" are understood to include the stated number. Furthermore, in the description of the embodiments in this application, "multiple" means two or more (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups" and "multiple times," unless otherwise explicitly specified.

[0059] like Figure 1 As shown, in a first aspect, the present invention provides a method for segmenting a tooth image, the method comprising the following steps:

[0060] S1: Obtain dental model image;

[0061] S2: Input the dental model image into the trained first neural network and output the vector information corresponding to the feature points where the tooth information is located on the dental model image;

[0062] S3: Perform cluster analysis on all the vector information, filter out several vector information sets, and determine the corresponding tooth center point coordinates based on the vector information in each of the vector information sets;

[0063] S4: Segment the dental model image according to the coordinates of the center points of each tooth to obtain the segmented tooth image.

[0064] In step S1, the dental model image is an oral scan dental model image, preferably a three-dimensional image, which is placed in a three-dimensional spatial coordinate system. The three-dimensional image can be generated by scanning the human oral cavity with an oral scanner, or by scanning the human oral cavity by running a computer program built into a mobile terminal or cloud server.

[0065] The dental model image includes information on at least one tooth. The dental model image contains information on multiple teeth, and the adjacent tooth information is arranged relatively closely. The dental model image needs to be processed to identify individual tooth images.

[0066] In step S2, the first neural network, or first neural network model, takes a dental model image as input and outputs vector information corresponding to the feature points of each tooth in the dental model image. The vector information points from the feature point to the tooth center point closest to that feature point. A feature point refers to a coordinate point within the region encompassed by the tooth information in the dental model image. Taking a 3D image as an example, the feature point is all the voxels corresponding to the tooth information; a voxel is the smallest unit for 3D image segmentation. The closest tooth center point is the tooth center point closest to the feature point in the 3D coordinate system. During model training, the coordinates of the center point of each tooth are synchronously input into the first neural network to be trained. After model training is complete, inputting any dental model image will yield the vector information corresponding to each feature point of the tooth information in that dental model image.

[0067] In step S3, the distance between the endpoints of vector information in the same vector information set is less than a preset error range. If the tooth information on the dental model image belongs to the same tooth, after the first neural network operation in step S2, the endpoints of the vector information corresponding to the feature points of that tooth information will mostly point to the same position. That is, the endpoint coordinates of the vector information belonging to the same tooth are very close. Therefore, by classifying the vectors whose endpoints are less than the preset error range, it is possible to analyze how many clustered vector information sets there are on the current dental model image, that is, how many tooth center points or how many teeth there are on the dental model image.

[0068] In step S4, once the center point coordinates of each tooth are confirmed, the image of each tooth can be derived by working backward from these coordinates. Specifically, this can be achieved by first determining a region containing the size of a single tooth based on its center point, then binarizing and filtering this region to obtain the tooth image corresponding to that center point; alternatively, the dental model image and the coordinates of each center point can be directly input into another trained neural network for calculation to obtain the tooth image corresponding to the dental model image.

[0069] The above scheme obtains the vector information corresponding to the feature points of the teeth by inputting the dental model image into the trained first neural network, and performs cluster analysis based on all the vector information to obtain the center point coordinates of each tooth on the dental model image. Then, the recognition and segmentation of a single tooth is completed based on the center point coordinates, which can effectively improve the accuracy of tooth image segmentation.

[0070] like Figure 2 As shown, in some embodiments, the following step is included after step S1 and before step S2:

[0071] S11: Adjust the size of each voxel on the dental model image to the first size;

[0072] Step S2 includes:

[0073] S21: Input the dental model image adjusted in step S11 into the first trained neural network, and output the vector information corresponding to the feature points where the tooth information is located on the dental model image.

[0074] In step S11, the first size is the voxel size of the dental model image input into the first neural network during training. The dental model images to be segmented vary in size, and their corresponding voxel sizes also differ (smaller voxel sizes result in higher accuracy). By first adjusting the dental model image to match the voxel size image used during the training of the first neural network, the first neural network can output the vector information corresponding to the teeth in the dental model image more quickly and accurately.

[0075] like Figure 3 As shown, in some embodiments, step S2 includes:

[0076] S22: Extract the key regions from the dental model image to obtain key region images, input the key region images into the trained first neural network, and output the vector information corresponding to the feature points where the tooth information is located on the dental model image.

[0077] In step S22, the key region image contains all tooth information from the dental model image. Typically, the acquired dental model image includes not only the tooth information portion but also the background portion. If the background portion occupies too large a proportion of the overall dental model image, it will affect the accuracy of tooth image segmentation. Therefore, it is necessary to first extract the key region image containing all tooth information, and then perform analysis based on the key region image. The key region image can be determined as follows: When the dental model image is acquired, feature points that are closest to the image edge in four directions (up, down, left, and right) and whose grayscale values ​​reach the tooth information can be obtained through grayscale judgment. Then, based on the coordinates of these four feature points, extend outward along the image edge by a preset number of coordinate units. The line connecting the extended coordinate points is taken as the key region image, ensuring that all tooth information is included in the key region. By extracting the key region image from the dental model image, the amount of data calculated by the first neural network can be effectively reduced, improving the calculation accuracy.

[0078] In some embodiments, the first neural network is trained in the following manner:

[0079] Acquire several dental sample images, each dental sample image containing at least one tooth information, and each tooth information corresponding to a tooth center point;

[0080] The dental sample image is input into the first neural network for iterative training, and the vector information corresponding to the feature points in each tooth information is calculated and output.

[0081] Preferably, the first neural network includes, but is not limited to, a three-dimensional segmentation network VNet. VNet is a type of neural network. During training, many parameters of this network can be continuously adjusted, and the difference between the output result of the neural network and the standard result can be continuously reduced iteratively to obtain the training weights that are paired with it. The training weights obtained at the end are used as the optimal parameter settings for the neural network to obtain the first neural network after training.

[0082] In some embodiments, such as Figure 4As shown, step S4 includes: S41: Inputting the determined coordinates of several tooth center points and the dental model image into the trained second neural network, and outputting a segmented tooth image based on the coordinates of each tooth center point. Obtaining segmented tooth information through the trained neural network achieves more accurate tooth image segmentation compared to directly using binarization.

[0083] It should be noted that, in this embodiment, the tooth information obtained by reconstructing the coordinates of each tooth center point can be marked with different identifiers. For example, if there are 20 tooth center points, the feature points of the reconstructed tooth information can be marked with numbers 1-20 respectively, so that they can be distinguished and highlighted when the segmented tooth image is obtained later.

[0084] In some embodiments, such as Figure 5 As shown, the dental model image is a three-dimensional image; the process includes the following steps after step S3 and before step S41:

[0085] S31: Adjust the size of each voxel on the dental model image to the second dimension;

[0086] Step S4 includes: inputting the determined coordinates of several tooth center points and the dental model image adjusted in step S31 into the trained second neural network, and outputting a tooth image segmented based on the coordinates of each tooth center point.

[0087] In step S31, the second size is the voxel size of the dental model image input into the second neural network during training. The dental model images to be segmented vary in size, and their corresponding voxel sizes also differ (smaller voxel sizes result in higher accuracy). Before the dental model image is input into the trained second neural network, by adjusting the dental model image to match the voxel size image used during the second neural network's training, the second neural network can output the corresponding tooth image (i.e., the segmented result) more quickly and accurately.

[0088] Preferably, inputting the determined coordinates of several tooth center points and the dental model image into the trained second neural network includes: dividing the dental model image into a corresponding number of images containing information of a single tooth, centered on each tooth center point; and sequentially inputting the coordinates of each tooth center point and its corresponding image containing information of a single tooth into the trained second neural network. The second neural network includes, but is not limited to, the UET++ model, the UET model, and the DeepLabv3 model.

[0089] In short, during the training of the second neural network, the input to the second neural network is a dental model image and the coordinates of the tooth center point. Each dental model image contains information about one or more teeth, and the output of the second neural network is each image containing information about a single tooth. During training, the second neural network continuously adjusts relevant parameters and iteratively reduces the difference between the neural network output and the label result (i.e., the standard image containing information about a single tooth), thereby obtaining paired training weights. The final training weights are used as the optimal parameter settings for the neural network to obtain the trained second neural network. Then, by inputting any dental model image containing tooth information and the coordinates of the tooth center point (obtained through cluster analysis of vector information) into the trained second neural network, all images containing information about a single tooth (i.e., tooth images) corresponding to that dental model image can be obtained. All the obtained tooth images can be displayed as single or multiple teeth.

[0090] In some embodiments, determining the corresponding tooth center point coordinates based on the vector information in each of the vector information sets includes: using the average of the endpoint coordinates corresponding to the vector information in each of the vector information sets as the corresponding tooth center point coordinates. This can improve the accuracy of the calculated tooth center point coordinates.

[0091] like Figure 6 As shown, the principle of the method involved in this embodiment will be further explained below. A dental model image 1 contains information 30 of four teeth, namely tooth information A, tooth information B, tooth information C, and tooth information D. After obtaining the dental model image 1, it is first processed to extract the key region image 20 containing all tooth information 30. Then, the key region image 20 can be input into the trained first neural network. The first neural network will output the vector information corresponding to all voxel points (i.e., feature points 301) in the tooth information AD. The vector information is then processed by... Figure 6 The tooth information is indicated by arrows within the circled area in AD. It should be noted that... Figure 6 The image only shows a portion of the feature points in the AD (Advanced Visual Aspect Rating) of the teeth. Figure 6 All voxel points contained within the circled area of ​​the AD should be considered as feature points corresponding to the tooth information AD.

[0092] After obtaining the vector information corresponding to the feature points of all tooth information, cluster analysis can be performed on these vector information. Figure 6It is easy to see that although the vector information of some tooth feature points points outwards, the vector information corresponding to feature points belonging to the same tooth tends to cluster towards the center of the tooth. Therefore, by analyzing the endpoint coordinates of these clustered vector information, the vector information with smaller endpoint distance differences is classified, and then the average of the endpoint coordinates of the classified vector information is taken to obtain the coordinates of each tooth center point 302. Then, the coordinates of each tooth center point 302 are input together with the dental model image into the trained second neural network to obtain the segmented tooth image.

[0093] The following section provides a further explanation of the method involved in this application, using specific application scenarios. This method specifically involves a training phase and a forward inference phase.

[0094] The training phase specifically includes the following stages:

[0095] (1) Roi detection stage, which is to detect which location area of ​​CBCT contains tooth information.

[0096] The specific steps are as follows: Read the images (i.e., dental model images) and labels (i.e., label types), perform uniform normalization on the images and labels, normalize the values ​​of the images to a reasonable range to facilitate subsequent calculations, and scale the voxel size of all CBCT images to a uniform size, such as 0.2 or 0.25.

[0097] CBCT data is a three-dimensional array containing numbers of varying sizes. This is because X-rays, when penetrating the human head, encounter different materials, such as bones and simple skin tissue, resulting in varying transmittance. Labels, on the other hand, are markers that indicate which data in the images represents teeth, which represents jawbones, etc.

[0098] (2) CNT detection stage, namely the extraction of key areas in the dental model image and the training of the first neural network.

[0099] The CNT detection during the training phase first locates the range of tooth information in the input images using labels (i.e., which range in the dental model image contains tooth information), and then extracts the images and labels within this range. Based on this, the offset of the three-dimensional coordinates of each tooth relative to the center point of each tooth (i.e., vector information) is calculated according to the labels. (3) Single tooth segmentation, i.e., training of the second neural network

[0100] The primary objective of this training phase is to generate the necessary training data using labels and images. In the single-tooth segmentation step, this application first locates the 3D coordinates of the center point of each tooth using the labels. Then, it extracts voxel blocks containing the entire tooth (the size of the voxel block can be 64*64*64 or 128*128*128, etc.) based on the located 3D coordinates of the center points of each tooth. This extraction process is performed on the 3D data of both images and labels to ensure that the images and labels correspond. The paired data is then fed into the neural network used for ROI detection to obtain the single-tooth segmentation weights. The forward inference phase may include the following steps:

[0101] (1) Read CBCT images and normalize them so that the voxel sizes in the normalized images are consistent with those processed during the training phase. Feed the preprocessed images into the ROI detection network to obtain preliminary tooth segmentation results, such as... Figure 7 As shown.

[0102] (2) Scale the initial tooth segmentation results to the size of the images after data reading.

[0103] (3) Crop the scaled image to determine the specific tooth region location and obtain its three-dimensional coordinate range.

[0104] (4) The images are cropped according to the three-dimensional coordinate range and fed into the CNT point detection neural network according to the rule of step size of 2. The post-processing is performed based on the forward inference results, and the results are clustered to ensure that one tooth corresponds to one CNT point.

[0105] (5) Map the CNT point results back to the initial image coordinates. Based on this, extract a certain number of voxel blocks within the 3D space of the CNT points according to the CNT point coordinates (the size of the voxel blocks remains consistent with the training phase), and feed them into the single-tooth segmentation network. Then, based on the CNT point coordinates and the segmentation results, gradually map the segmentation results to empty label fields of the same size as the images. This result is the final tooth segmentation result. The effect is as follows: Figure 8 As shown.

[0106] In a second aspect, the present invention also provides a storage medium storing a computer program that, when executed by a processor, implements the method steps of the first aspect of the present invention.

[0107] like Figure 9As shown, in a third aspect, the present invention also provides an electronic device 10, including a processor 101 and a storage medium 102, the storage medium 102 being the storage medium as described in the second aspect; the processor 101 is configured to execute a computer program stored in the storage medium 102 to implement the method steps as described in the first aspect.

[0108] In this embodiment, the electronic device is a computer device, including but not limited to: an oral scanner, personal computer, server, general-purpose computer, special-purpose computer, network device, embedded device, programmable device, smart mobile terminal, smart home device, wearable smart device, and in-vehicle smart device. Storage media include but are not limited to: RAM, ROM, magnetic disk, magnetic tape, optical disk, flash memory, USB flash drive, portable hard drive, memory card, memory stick, network server storage, and network cloud storage. Processors include but are not limited to: CPU (Central Processing Unit), GPU (Graphics Processing Unit), and MCU (Microprocessor Unit).

[0109] Those skilled in the art will understand that the above embodiments can be provided as methods, apparatus, or computer program products. These embodiments may take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. All or part of the steps in the methods involved in the above embodiments can be implemented by a program instructing related hardware, and the program can be stored in a computer-readable storage medium for performing all or part of the steps of the methods of the above embodiments.

[0110] The above embodiments are described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to the embodiments. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a computer device to produce a machine, such that the instructions, which execute via the processor of the computer device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0111] These computer program instructions may also be stored in a computer device-readable storage medium that can direct a computer device to operate in a particular manner, such that the instructions stored in the computer device-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0112] These computer program instructions can also be loaded onto a computer device, causing a series of operational steps to be performed on the computer device to produce a computer-implemented process, thereby providing instructions that execute on the computer device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0113] Although the above embodiments have been described, those skilled in the art, once they understand the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the above are merely embodiments of the present invention and do not limit the scope of patent protection of the present invention. Any equivalent structural or procedural transformations made using the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method of segmenting a dental image, the method comprising: The method comprises the following steps: S1: acquiring a dental mold image; the dental mold image comprises at least one tooth information; S2: inputting the dental mold image into a first neural network trained, and outputting vector information corresponding to feature points of the tooth information on the dental mold image; the vector information is directed to a tooth center point closest to the feature points; S3: performing cluster analysis on all the vector information, screening a plurality of vector information sets, and determining corresponding tooth center point coordinates according to the vector information in each of the vector information sets; the determination of the corresponding tooth center point coordinates according to the vector information in each of the vector information sets comprises: taking an average value of end point coordinates of the vector information in each of the vector information sets as the corresponding tooth center point coordinates, and the distance between the end points of the vector information in the same vector information set is less than a preset error range; S4: segmenting the dental mold image according to each tooth center point coordinate to obtain a segmented tooth image.

2. The dental image segmentation method of claim 1, wherein, The dental mold image is a three-dimensional image; After step S1 and before step S2, the method further comprises the following step: S11: adjusting the size of each voxel on the dental mold image to a first size; the first size is the size of the voxel on a dental mold sample image input into the first neural network during training of the first neural network; Inputting the dental mold image into the first neural network trained comprises inputting the dental mold image adjusted in step S11 into the first neural network trained.

3. The tooth image segmentation method of claim 1, wherein Inputting the dental mold image into the first neural network trained further comprises: extracting a key area in the dental mold image to obtain a key area image, and inputting the key area image into the first neural network trained; the key area image contains all tooth information on the dental mold image.

4. The tooth image segmentation method of claim 1 or 2, wherein The first neural network is trained according to the following manner: acquiring a plurality of dental mold sample images; the dental mold sample images contain at least one tooth information; each tooth information corresponds to a tooth center point; inputting the dental mold sample images into the first neural network for iterative training, calculating vector information corresponding to feature points in each tooth information, and outputting the vector information.

5. The dental image segmentation method of claim 1, wherein, Step S4 comprises: S41: inputting the determined plurality of tooth center point coordinates and the dental mold image into a second neural network trained, and outputting a tooth image segmented based on each tooth center point coordinate.

6. The dental image segmentation method of claim 5, wherein, The dental mold image is a three-dimensional image; After step S3 and before step S41, the method further comprises the following step: S31: adjusting the size of each voxel on the dental mold image to a second size; the second size is the size of the voxel on a dental mold sample image input into the second neural network during training of the second neural network; Inputting the dental mold image into the second neural network trained comprises inputting the dental mold image adjusted in step S31 into the second neural network trained.

7. The tooth image segmentation method of claim 5 or 6, wherein, inputting the determined plurality of tooth center point coordinates and the dental cast image into the trained second neural network comprises: segmenting the dental cast image into a corresponding number of images containing single tooth information, each image centered on a respective tooth center point, and inputting each tooth center point coordinate and its corresponding image containing single tooth information into the trained second neural network in sequence.

8. A storage medium, characterized by The storage medium stores a computer program, and the computer program is executed by the processor to implement the method steps of any one of claims 1-7.

9. An electronic device, comprising: The storage medium comprises a processor and a storage medium, and the storage medium is the storage medium of claim 8. The processor is configured to execute the computer program stored in the storage medium to implement the method steps of any one of claims 1-7.

Images