A Multi-Granularity CBCT Tooth Segmentation Method and System Based on SAM

By combining the SAM model with object detection and key point detection networks, multi-granularity tooth segmentation is achieved, which solves the problem of insufficient generalization ability of tooth image segmentation models on images from different patients and imaging devices, improves the accuracy and robustness of segmentation, and is suitable for automated segmentation of CBCT tooth images.

CN123118680APending Publication Date: 2026-11-17SHANDONG UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611059577.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-16
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing tooth segmentation models lack generalization ability when faced with images from different patients or imaging devices, making it difficult to accurately segment tooth images with complex structures and varied textures, especially in the multi-level detail segmentation of CBCT images.

Method used

By combining the Segment Anything Model (SAM) with object detection models and keypoint detection networks, and by introducing bounding box and point prompts, a multi-level feature extraction and fusion mechanism is adopted to achieve multi-granularity tooth segmentation, thereby improving the accuracy and robustness of segmentation.

Benefits of technology

It significantly improves the automation level and generalization ability of tooth image segmentation, and can maintain good segmentation performance on image data of different types and qualities, reducing the workload of doctors and improving the efficiency of clinical workflow.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118680A_ABST
    Figure CN123118680A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of image processing technology. It proposes a multi-granularity CBCT tooth segmentation method and system based on SAM (Signal-Assisted Image Processing). The method employs an image encoder based on the SAM model to process edge-enhanced X-ray CBCT tooth images, obtaining image embeddings. Based on a target detection model, bounding box cues are obtained, and a cue encoder encodes these cues. The encoded results and the image embeddings are input into a first decoder to obtain a coarse mask. Keypoint detection is performed using a keypoint detection network, and point cues are obtained based on the keypoint detection results and a point cue encoder. The coarse mask is convolved to obtain dense embeddings, which are then concatenated with the image embeddings. The concatenated result and the point cues are input into a second decoder to obtain the final tooth segmentation result. This invention improves the accuracy and robustness of segmentation and enhances the automation level of tooth image segmentation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and specifically to a multi-granularity CBCT tooth segmentation method and system based on SAM. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In dental care, accurate segmentation of tooth images is crucial for dental disease diagnosis, treatment planning, and postoperative evaluation. The complex structure of teeth and the significant differences in morphology, quality, and texture between patients' dental radiographs make automated tooth segmentation a challenging task. Medical procedures such as root canal treatment and orthodontic treatment rely on precise image segmentation to enable dentists to make scientific judgments and perform accurate procedures.

[0004] However, existing segmentation models often lack sufficient generalization ability when faced with images from different patients or imaging devices. This means that the models are prone to performance degradation on images outside the training data, leading to inaccurate segmentation or even failure. Furthermore, the complex structures, varied textures, and blurred or overlapping edge regions in dental images also increase the difficulty of segmentation. Traditional methods based on convolutional neural networks (CNNs) or other deep learning models, while achieving some success under specific conditions, still have many shortcomings in multi-level detail segmentation of CBCT images (i.e., cone-beam computed tomography of the oral and maxillofacial region).

[0005] In recent years, the Segment Anything Model (SAM) has demonstrated superior capabilities as a general-purpose image segmentation model. SAM can accurately segment various complex scenes without requiring domain-specific training and supports flexible input prompts. However, SAM still has limitations when handling fully automated segmentation tasks. For example, its segmentation performance may significantly degrade when input prompts are inaccurate or boundaries are blurred. Furthermore, the original architecture of SAM cannot provide sufficient support for tasks requiring multi-level, multi-granularity analysis. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a multi-granularity CBCT tooth segmentation method and system based on SAM (Signal-Assisted Aspect-Oriented Model). Combining the general segmentation advantages of SAM with an adaptive strategy, it achieves accurate segmentation of CBCT tooth images. By introducing an approximate bounding box (i.e., box cue) as a cue input for SAM through an object detection model, and utilizing a multi-level feature extraction and fusion mechanism, it can achieve comprehensive analysis and segmentation of teeth at different resolutions. The multi-scale feature fusion technology achieves a balance between details and overall structure, thereby significantly improving the accuracy and robustness of segmentation. This not only improves the automation level of tooth image segmentation but also expands the generalization ability of the SAM model, enabling it to maintain good performance on image data of different types and qualities.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a multi-granularity CBCT tooth segmentation method based on SAM.

[0008] A multi-granularity CBCT tooth segmentation method based on SAM includes the following procedures: Obtain the original X-ray CBCT tooth image to be segmented, and perform edge enhancement processing on the original X-ray CBCT tooth image; An image encoder using the SAM model is used to process the edge-enhanced X-ray CBCT dental images to obtain image embeddings; Based on the original X-ray tooth image and the pre-trained target detection model, a bounding box cue is obtained. The bounding box cue is input into the cue encoder of the SAM model for encoding. The encoding result and the image are embedded into the first decoder to obtain a coarse mask. Keypoint detection is performed based on the original X-ray tooth image and a pre-trained keypoint detection network. Point cues are then obtained based on the keypoint detection results and the point cues encoder. After convolution processing of the coarse mask, dense embedding is obtained. The dense embedding is concatenated with the image embedding. The concatenation result and point cue are input into the second decoder to obtain the final tooth segmentation result.

[0009] Secondly, the present invention provides a multi-grained CBCT tooth segmentation system based on SAM.

[0010] A multi-granularity CBCT tooth segmentation system based on SAM includes: The image preprocessing unit is configured to: acquire the original X-ray CBCT tooth image to be segmented, and perform edge enhancement processing on the original X-ray CBCT tooth image; The image embedding unit is configured to: process the edge-enhanced X-ray CBCT dental image using an image encoder based on the SAM model to obtain the image embedding; The coarse mask generation unit is configured to: obtain bounding box cues based on the original X-ray tooth image and the pre-trained target detection model; input the bounding box cues into the cues encoder of the SAM model for encoding processing; and embed the encoding result and the image into the first decoder to obtain the coarse mask. The dot prompt generation unit is configured to: perform keypoint detection based on the original X-ray tooth image and a pre-trained keypoint detection network, and obtain dot prompts based on the keypoint detection results and the dot prompt encoder; The segmentation result generation unit is configured to: perform convolution processing on the coarse mask to obtain dense embedding, concatenate the dense embedding with the image embedding, and input the concatenation result and point cue into the second decoder to obtain the final tooth segmentation result.

[0011] Thirdly, the present invention provides a computer device, comprising: a processor and a computer-readable storage medium; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program that, when executed by the processor, implements the SAM-based multi-granularity CBCT tooth segmentation method as described in the first aspect of the present invention.

[0012] Fourthly, the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and executed as described in the first aspect of the present invention for multi-granularity CBCT tooth segmentation.

[0013] Fifthly, the present invention provides a computer program product comprising a computer program that, when executed by a processor, implements the SAM-based multi-granularity CBCT tooth segmentation method as described in the first aspect of the present invention.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention combines the high-quality segmentation capabilities of SAM with a prior knowledge enhancement strategy, demonstrating a significant improvement in accuracy when processing complex medical images. By generating a fully covered segmentation mask through prompts and introducing segmentation prior maps and boundary prior maps, it can effectively identify complex structures and blurred boundaries. This multi-level feature fusion improves the robustness of the segmentation model on different medical images, enabling it to adapt to various image noise and device differences.

[0015] This invention employs a fully automated input and result fusion strategy, such as grid prompts and prior information from SAM (Sectional Aspect-Oriented Analysis), eliminating the need for manual annotation or complex preprocessing steps in the segmentation process. This significantly improves the efficiency of clinical workflows and reduces the workload of physicians. Particularly in root canal treatment or complex procedures, this system can rapidly generate high-quality segmentation results, providing more reliable data support for physicians' diagnosis and treatment.

[0016] This invention combines the strong generalization ability of SAM with the cue boxes generated by YOLOv8 and the cue points generated by the key point detection model to achieve efficient segmentation and adaptation of CBCT. In particular, under the SAM architecture, the use of a multi-level feature extraction module further reduces the performance degradation caused by inaccurate cueing, enabling the model to maintain excellent performance on unseen data.

[0017] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0018] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0019] Figure 1 A flowchart illustrating a multi-granularity CBCT tooth segmentation method based on SAM, provided as an exemplary embodiment of the present invention; Figure 2 A schematic diagram illustrating the principle of a multi-granularity CBCT tooth segmentation method based on SAM, provided as an exemplary embodiment of the present invention; Figure 3 A schematic diagram of a SAM-based multi-granularity CBCT tooth segmentation system provided as an exemplary embodiment of the present invention; Figure 4 A schematic diagram of a computer device provided for an exemplary embodiment of the present invention. Detailed Implementation

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0021] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0022] This implementation proposes a multi-granularity CBCT tooth segmentation method based on SAM (Signal-Assisted Image Processing). This method combines the general segmentation advantages of SAM with an adaptive strategy to achieve accurate segmentation of CBCT tooth images. By introducing YOLOv8 to generate approximate bounding boxes as prompts for SAM, and utilizing a multi-level feature extraction and fusion mechanism, this method can achieve comprehensive tooth analysis and segmentation at different resolutions. Furthermore, this invention employs multi-scale feature fusion technology, achieving a balance between details and overall structure, thereby significantly improving the accuracy and robustness of segmentation. Finally, this invention not only improves the automation level of tooth image segmentation but also expands the model's generalization ability, enabling it to maintain good performance on image data of different types and qualities.

[0023] More specifically, such as Figure 1 As shown, the process includes the following: S101: Obtain the original X-ray CBCT tooth image to be segmented, and perform edge enhancement processing on the original X-ray CBCT tooth image; S102: The image encoder using the SAM model processes the edge-enhanced X-ray CBCT dental image to obtain the image embedding; S103: Based on the original X-ray tooth image and the pre-trained target detection model, obtain the bounding box cue. Input the bounding box cue into the cue encoder of the SAM model for encoding processing. Embed the encoding result and the image into the first decoder to obtain the coarse mask. S104: Perform key point detection based on the original X-ray tooth image and the pre-trained key point detection network, and obtain point cues based on the key point detection results and the point cues encoder; S105: After convolution processing of the coarse mask, dense embedding is obtained. The dense embedding is concatenated with the image embedding. The concatenation result and point cue are input into the second decoder to obtain the final tooth segmentation result.

[0024] In this implementation, CBCT imaging is a medical imaging device that uses cone-beam X-ray scanning technology combined with computer reconstruction algorithms to generate three-dimensional images. It features low radiation dose (50% to 90% lower than traditional CT), short scanning time, and high spatial resolution. Its applications cover fields such as orthodontics, dental implants, treatment of dental caries and pulp diseases, maxillofacial surgery, and diagnosis of tumors and trauma. It can accurately assess the mass and density of the jawbone, observe root lesions, and preoperative structural relationships.

[0025] In this implementation, during the training of the SAM model, the original X-ray dental image data of hospital patients are acquired to form a CBCT image dataset; more specifically, a certain number of X-ray CBCT image data are acquired, all of which come from patients with relatively complete dental structures (the X-ray CBCT image data here are all data authorized by the patients or publicly available datasets), while CBCT data with extremely poor exposure and poor quality are removed.

[0026] In this implementation, professional dentists annotate the teeth in the entire CBCT section with points and bounding boxes. The annotations are then reviewed by multiple professional dentists to ensure accuracy and consistency. More specifically, multiple professional dentists annotate all teeth with points and bounding boxes, and multiple dentists who did not participate in the annotation process evaluate the annotations to guarantee their authenticity. The bounding box annotation refers to the smallest outer bounding box of the tooth, which can completely enclose the tooth. This label is used to generate a coarse-grained mask. The point annotation refers to three points: the left and right sides of the tooth and the lowest point of the root. This label uses an optimized coarse-grained mask to generate a fine-grained mask.

[0027] In this implementation, an edge enhancement algorithm is used to increase the distinguishability of tooth edges. An edge algorithm (i.e., edge sharpening processing) is introduced to enhance the distinguishability of teeth from alveolar bone and other surrounding areas, so as to ensure the robustness of the segmentation model. More specifically, the segmentation output of SAM is combined with the original image input to generate a SAM-enhanced input image, which is used to construct the downstream medical image segmentation model. This process provides the model with rich contextual and structural information by introducing segmentation prior maps and boundary prior maps.

[0028] like Figure 2 As shown, this invention uses the YOLOv8 model as the object detection model to detect target regions and generate coarse-grained masks. YOLOv8, as an efficient object detection model, automatically identifies regions of interest in images and generates corresponding bounding box hints. These hints provide the approximate location and boundary information of the object, guiding subsequent coarse mask generation. (1); in, This represents the YOLOv8 object detection model; Input image; Text prompts are provided to enhance detection effectiveness; The box indicates a hint.

[0029] In this implementation, the generated dialog box is used. This is then fed into the cue encoder of the SAM model for processing, and combined with image embedding and the first decoder to generate a coarse mask. : (2); in, The frame indicates the encoder after fine-tuning. As the first decoder, Image embeddings generated by the image encoder for the SAM model.

[0030] Building upon coarse-grained mask generation, a key point detection network is used to automatically generate point prompts to further improve the accuracy of edge recognition. These point prompts are used to capture complex edges and minute structures in the target area, avoiding edge blurring or mask merging errors between adjacent areas.

[0031] The formula for keypoint recognition using a keypoint detection network is as follows: (3); in, Represents a keypoint detection network. This represents the key points obtained from the detection.

[0032] Because there may be information conflicts between dot hints and box hints, to avoid such conflicts, the system converts the coarse mask into a masked hint. The masked hint generates a dense embedding through convolution operations, which is then combined with the dot hint and input into the decoder for further optimization to obtain a fine-grained mask. : (4); With the fine-grained mask described above As the final segmentation result, this invention achieves progressive optimization from coarse to fine through the above two-stage segmentation strategy. The coarse mask generated by YOLOv8 provides the ability to quickly cover the main target region, while the keypoint detection network further focuses on detail optimization to ensure the accuracy and integrity of the boundaries. This multi-level segmentation strategy not only improves the robustness of the segmentation but also effectively reduces misjudgments caused by blurred boundaries or overlapping regions. Through the coarse-to-fine segmentation strategy, this invention can progressively optimize the segmentation result. After quickly covering the main region in the coarse-grained stage, the fine-grained stage further improves the accuracy of the edges. This invention effectively avoids conflicts between prompts, enhances the fineness and robustness of the segmentation, and achieves significant performance improvements on multiple publicly available medical datasets.

[0033] After training the model, the input X-ray image needs to be fed into the trained model for processing. The model first performs preliminary mask generation on the image and merges it with the original image to form the input for the downstream segmentation task, clarifying the tooth boundaries. Then, it is fed into a multi-granularity SAM model to complete the segmentation.

[0034] Figure 3 A multi-granularity CBCT tooth segmentation system based on SAM is shown, comprising: The image preprocessing unit 301 is configured to: acquire the original X-ray CBCT tooth image to be segmented, and perform edge enhancement processing on the original X-ray CBCT tooth image; The image embedding unit 302 is configured to: process the edge-enhanced X-ray CBCT dental image using an image encoder based on the SAM model to obtain the image embedding; The coarse mask generation unit 303 is configured to: obtain bounding box cue based on the original X-ray tooth image and the pre-trained target detection model; input the bounding box cue to the cue encoder of the SAM model for encoding processing; and embed the encoding result and the image into the first decoder to obtain the coarse mask. The dot prompt generation unit 304 is configured to: perform key point detection based on the original X-ray tooth image and a pre-trained key point detection network, and obtain dot prompts based on the key point detection results and the dot prompt encoder; The segmentation result generation unit 305 is configured to: perform convolution processing on the coarse mask to obtain dense embedding, concatenate the dense embedding with the image embedding, and input the concatenation result and point cue into the second decoder to obtain the final tooth segmentation result.

[0035] It is understood that the aforementioned units can be individually or entirely merged into one or more other units, or some of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effects of the embodiments of this application. The aforementioned units are based on logical functional division. In practical applications, the function of one unit can be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of this application, the system may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0036] According to another embodiment of this application, the system described in this embodiment can be constructed by running a computer program (including program code) capable of performing the steps involved in the corresponding method of the present invention on a general-purpose computing device, such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). The computer program can be recorded on, for example, a computer-readable recording medium, loaded into the aforementioned computing device through the computer-readable recording medium, and run therein.

[0037] Figure 4 A computer device is shown, which includes a processor 401, a communication interface 402, and a computer-readable storage medium 403. The processor 401, communication interface 402, and computer-readable storage medium 403 can be connected via a bus or other means.

[0038] The communication interface 402 is used to receive and send data. The computer-readable storage medium 403 can be stored in the memory of the electronic device. The computer-readable storage medium 403 is used to store computer programs, which include program instructions. The processor 401 is used to execute the program instructions stored in the computer-readable storage medium 403.

[0039] The processor 401 is the computing and control core of the electronic device. It is suitable for implementing one or more instructions, specifically for loading and executing one or more instructions to achieve the corresponding method flow or corresponding function.

[0040] The processor 401 is configured to perform the following process: Obtain the original X-ray CBCT tooth image to be segmented, and perform edge enhancement processing on the original X-ray CBCT tooth image; An image encoder using the SAM model is used to process the edge-enhanced X-ray CBCT dental images to obtain image embeddings; Based on the original X-ray tooth image and the pre-trained target detection model, a bounding box cue is obtained. The bounding box cue is input into the cue encoder of the SAM model for encoding. The encoding result and the image are embedded into the first decoder to obtain a coarse mask. Keypoint detection is performed based on the original X-ray tooth image and a pre-trained keypoint detection network. Point cues are then obtained based on the keypoint detection results and the point cues encoder. After convolution processing of the coarse mask, dense embedding is obtained. The dense embedding is concatenated with the image embedding. The concatenation result and point cue are input into the second decoder to obtain the final tooth segmentation result.

[0041] This invention also provides a computer-readable storage medium, which is a memory device in an electronic device for storing programs and data. It is understood that the computer-readable storage medium here may include both built-in storage media in the electronic device and extended storage media supported by the electronic device. The computer-readable storage medium provides storage space for storing the processing system of the electronic device.

[0042] Furthermore, this storage space also contains one or more instructions suitable for loading and execution by the processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM memory or unstable memory, such as at least one disk storage device; optionally, it can also be at least one computer-readable storage medium located remotely from the aforementioned processor.

[0043] In one embodiment, the computer-readable storage medium stores one or more instructions; the processor loads and executes the one or more instructions stored in the computer-readable storage medium to perform the following process: Obtain the original X-ray CBCT tooth image to be segmented, and perform edge enhancement processing on the original X-ray CBCT tooth image; An image encoder using the SAM model is used to process the edge-enhanced X-ray CBCT dental images to obtain image embeddings; Based on the original X-ray tooth image and the pre-trained target detection model, a bounding box cue is obtained. The bounding box cue is input into the cue encoder of the SAM model for encoding. The encoding result and the image are embedded into the first decoder to obtain a coarse mask. Keypoint detection is performed based on the original X-ray tooth image and a pre-trained keypoint detection network. Point cues are then obtained based on the keypoint detection results and the point cues encoder. After convolution processing of the coarse mask, dense embedding is obtained. The dense embedding is concatenated with the image embedding. The concatenation result and point cue are input into the second decoder to obtain the final tooth segmentation result.

[0044] The present invention also provides a computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the following process: Obtain the original X-ray CBCT tooth image to be segmented, and perform edge enhancement processing on the original X-ray CBCT tooth image; An image encoder using the SAM model is used to process the edge-enhanced X-ray CBCT dental images to obtain image embeddings; Based on the original X-ray tooth image and the pre-trained target detection model, a bounding box cue is obtained. The bounding box cue is input into the cue encoder of the SAM model for encoding. The encoding result and the image are embedded into the first decoder to obtain a coarse mask. Keypoint detection is performed based on the original X-ray tooth image and a pre-trained keypoint detection network. Point cues are then obtained based on the keypoint detection results and the point cues encoder. After convolution processing of the coarse mask, dense embedding is obtained. The dense embedding is concatenated with the image embedding. The concatenation result and point cue are input into the second decoder to obtain the final tooth segmentation result.

[0045] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0046] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic cable, digital cable) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0047] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-grained CBCT tooth segmentation method based on SAM, characterized in that, The process includes the following: Obtain the original X-ray CBCT tooth image to be segmented, and perform edge enhancement processing on the original X-ray CBCT tooth image; An image encoder using the SAM model is used to process the edge-enhanced X-ray CBCT dental images to obtain image embeddings; Based on the original X-ray tooth image and the pre-trained target detection model, a bounding box cue is obtained. The bounding box cue is input into the cue encoder of the SAM model for encoding. The encoding result and the image are embedded into the first decoder to obtain a coarse mask. Keypoint detection is performed based on the original X-ray tooth image and a pre-trained keypoint detection network. Point cues are then obtained based on the keypoint detection results and the point cues encoder. After convolution processing of the coarse mask, dense embedding is obtained. The dense embedding is concatenated with the image embedding. The concatenation result and point cue are input into the second decoder to obtain the final tooth segmentation result.

2. The SAM-based multi-grained CBCT tooth segmentation method as described in claim 1, characterized in that, Edge enhancement processing is performed, including: using an edge enhancement algorithm to sharpen the edges of the original X-ray CBCT dental images to enhance the distinction between the teeth and the surrounding areas.

3. The SAM-based multi-grained CBCT tooth segmentation method as described in claim 1, characterized in that, The encoder receives prompts when the target detection result is entered into the input box, including: ; in, The representative box prompts, Represents the YOLOv8 object detection model. The input is the original X-ray CBCT image of the teeth, and T is the text prompt.

4. The SAM-based multi-grained CBCT tooth segmentation method as described in claim 1, characterized in that, The result of the encoding process and the image are embedded into the first decoder to obtain a coarse mask, including: ; in, Represents a coarse mask. The representative box prompts, The cue encoder representing the SAM model, Represents image embedding, Represents the first decoder.

5. The SAM-based multi-graininess CBCT tooth segmentation method as described in claim 1, characterized in that, Keypoint detection is performed based on the original X-ray dental images and a pre-trained keypoint detection network, including: ; in, This represents the original X-ray CBCT image of the teeth input. This represents a pre-trained keypoint detection network. This represents the detection results of key points.

6. The SAM-based multi-grained CBCT tooth segmentation method as described in claim 1, characterized in that, Dense embeddings and image embeddings are concatenated, and the concatenated result, along with point cues, is input into a second decoder to obtain the final tooth segmentation result, including: ; in, Represents a coarse mask. Represents image embedding, Represents the second decoder. Represents the key point detection results. Representative point prompt encoder.

7. A multi-grained CBCT tooth segmentation system based on SAM, characterized in that, include: The image preprocessing unit is configured to: acquire the original X-ray CBCT tooth image to be segmented, and perform edge enhancement processing on the original X-ray CBCT tooth image; The image embedding unit is configured to: process the edge-enhanced X-ray CBCT dental image using an image encoder based on the SAM model to obtain the image embedding; The coarse mask generation unit is configured to: obtain bounding box cues based on the original X-ray tooth image and the pre-trained target detection model; input the bounding box cues into the cues encoder of the SAM model for encoding processing; and embed the encoding result and the image into the first decoder to obtain the coarse mask. The dot prompt generation unit is configured to: perform keypoint detection based on the original X-ray tooth image and a pre-trained keypoint detection network, and obtain dot prompts based on the keypoint detection results and the dot prompt encoder; The segmentation result generation unit is configured to: perform convolution processing on the coarse mask to obtain dense embedding, concatenate the dense embedding with the image embedding, and input the concatenation result and point cue into the second decoder to obtain the final tooth segmentation result.

8. A computer device, characterized in that, include: Processor and computer-readable storage media; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program that, when executed by the processor, implements the SAM-based multi-granularity CBCT tooth segmentation method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1 to 6, the SAM-based multi-granularity CBCT tooth segmentation method.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the SAM-based multi-granularity CBCT tooth segmentation method as described in any one of claims 1 to 6.