Method, program and apparatus for multi-stage AI tooth segmentation based on CBCT data
Through the multi -stage AI tooth segmentation method, the problem of confusion and time of dental category during CBCT tooth segmentation is solved, and higher segmentation accuracy and efficiency are achieved.
Patent Information
- Application Number
- CN202411481562.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-23
- Publication Date
- 2025-01-28
AI Technical Summary
The prior art has problems of tooth category confusion, too long time and poor generalization performance in CBCT tooth segmentation.
The multi -stage AI tooth segmentation method is adopted. The steps include data cleaning, gastrocity pitch heavy samples, preliminary rough segmentation, ROI determination, dental segmentation, grid data processing, target detection, and single teeth segmentation, and gradually improve the accuracy and efficiency of the segmentation.
It significantly improves the accuracy and efficiency of tooth segmentation, reduces the confusion of dental categories, and improves the efficiency of the segmentation process.
Smart Images
Figure CN119379715A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of oral image digital processing, and in particular to a method, program and device for multi-stage AI tooth segmentation based on CBCT data. Background Art
[0002] Accurately segmenting each tooth in CBCT is an important step in orthodontics. Manual or traditional methods of segmenting teeth are gradually unable to meet the needs of the times, and methods based on single-stage deep learning often face the problem of insufficient accuracy, and many CBCT tooth segmentation results are confused with each other.
[0003] Based on this, it is necessary to provide a tooth segmentation method that can improve the confusion, long time and poor generalization performance that may occur in previous AI tooth segmentation. Summary of the invention
[0004] The technical problem to be solved by the present invention is to design a method, program and device for multi-stage AI tooth segmentation based on CBCT data. The method can use multi-stage AI to accurately segment each tooth in CBCT, thereby solving the existing technical problems.
[0005] In order to solve the above technical problems, the present invention provides a multi-stage AI tooth segmentation method based on CBCT data, comprising the following steps: Step S1: Obtain CBCT image data. Obtain CBCT images through CBCT imaging equipment, and the images can be in dcm format or nifti format.
[0006] Step S2: resample the voxel spacing of the CBCT image, and use the trained AI segmentation tooth model to perform preliminary rough segmentation on the resampled CBCT at the same voxel spacing.
[0007] Step S3: Determine the tooth ROI (Region Of Interest) of the original CBCT image.
[0008] Step S4: Segment the teeth from the ROI.
[0009] Step S5: Map all teeth obtained back to their original CBCT positions and convert them into mesh data.
[0010] Step S6: Perform target detection on the tooth grid data to determine the local tooth area corresponding to each tooth position.
[0011] Step S7: Determine the mesh range of each tooth based on the detection result of the tooth mesh data.
[0012] Step S8: Perform single tooth segmentation on the teeth within the range.
[0013] Step S9: Map the segmented single tooth back to the mesh data obtained in step S5 to obtain all teeth after instance segmentation.
[0014] Step S10: Convert the mesh data obtained in step S9 back to CBCT to obtain the tooth segmentation result of CBCT.
[0015] Furthermore, step S1 also includes data cleaning of the acquired CBCT data, using a third-party library such as ITK or VTK to check the CBCT data, removing multi-sequence CBCT, single CBCT and CBCT containing useless information, retaining the remaining CBCT and converting it into nifti format.
[0016] Furthermore, in step S2, the CBCT images are uniformly resampled to a voxel spacing of 1 mm×1 mm×1 mm, and the AI segmented tooth model is a nnUNet model trained at all 1 mm×1 mm×1 mm voxel spacings.
[0017] Furthermore, in step S3, a boundary is determined for the segmentation result after the preliminary rough segmentation, and the boundary is mapped back to the original CBCT range, and the tooth range in the original CBCT is screened out to obtain the tooth ROI of the original CBCT image.
[0018] Furthermore, in step S4, a semantic segmentation network based on deep learning is used on the obtained tooth ROI, all teeth are regarded as the same foreground, and the teeth are segmented from the ROI.
[0019] Furthermore, in step S6, a point cloud target detection algorithm is used, and the detection result includes a classification result and a detection frame. The classification result is the tooth position number of each tooth, and the detection frame is the bounding box of each tooth.
[0020] Furthermore, in step S7, the grid data obtained in step S5 is extracted for the obtained tooth position and tooth local area, and the range of each tooth is obtained from the grid data.
[0021] Furthermore, in step S8, semantic segmentation of different teeth is performed within the range of each tooth, the tooth corresponding to the tooth position frame is regarded as the foreground, and other grids are regarded as the background, and each tooth is segmented.
[0022] The present invention also provides a device for multi-stage AI tooth segmentation based on CBCT data, comprising: at least one processor; and at least one memory in communication with the processor; The memory stores instructions executable by the processor, and the instructions are executed by the processor to enable the device to perform the aforementioned multi-stage AI tooth segmentation method based on CBCT data.
[0023] The present invention also provides a computer program product, comprising computer instructions, which, when executed by a processor, enable a computer device to perform the aforementioned method for multi-stage AI tooth segmentation based on CBCT data.
[0024] Beneficial effects of the present invention: The multi-stage AI tooth segmentation method based on CBCT data of the present invention has greatly improved the problems of confusion of tooth categories and long segmentation time that frequently occurred in the previous segmentation of CBCT teeth. After the improvement of the method of the present invention, the accuracy and efficiency of tooth segmentation are improved when segmenting CBCT teeth. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] The specific implementation of the present invention will be further explained below in conjunction with the accompanying drawings.
[0026] Figure 1 It is a flow chart of the method for multi-stage AI tooth segmentation based on CBCT data of the present invention; Figure 2 It is an original CBCT image (without segmentation processing); Figure 3 For Figure 2 Schematic diagram of the segmentation results using the original instance segmentation network; Figure 4 For Figure 2 Schematic diagram of segmentation results using the segmentation method of the present invention. DETAILED DESCRIPTION Example 1
[0027] Combination Figure 1 The multi-stage AI tooth segmentation method based on CBCT data of this embodiment includes the following steps: Step S1: Obtain CBCT image data. Obtain CBCT images through CBCT imaging equipment, and the images can be in dcm format or nifti format.
[0028] Preferably, in this embodiment, step S1 also includes cleaning the acquired CBCT data, checking the CBCT data using a third-party library such as ITK or VTK, removing multi-sequence CBCT, single CBCT and CBCT containing useless information, retaining the remaining CBCT and converting it into nifti format.
[0029] Step S2: resample the voxel spacing of the CBCT image, and use the trained AI segmentation tooth model to perform preliminary rough segmentation on the resampled CBCT at the same voxel spacing.
[0030] In this embodiment, preferably, in step S2, the CBCT images are uniformly resampled to a voxel spacing of 1 mm×1 mm×1 mm, that is, the previous voxel spacing is uniformly resampled to 1 mm×1 mm×1 mm, and the AI segmented tooth model is a nnUNet model trained at all 1 mm×1 mm×1 mm voxel spacings. The voxel spacing of the original CBCT is often more than 1 mm×1 mm×1 mm, and the step of segmenting teeth on the CBCT resampled to 1 mm×1 mm×1 mm is called preliminary coarse segmentation.
[0031] Step S3: Determine the tooth ROI (Region Of Interest) of the original CBCT image.
[0032] In this embodiment, preferably, in step S3, a boundary is determined for the segmentation result after the preliminary rough segmentation, the boundary is mapped back to the original CBCT range, the tooth range in the original CBCT is screened out, and the tooth ROI of the original CBCT image is obtained.
[0033] The result of the initial rough segmentation carries a label for each tooth. The label has coordinates in three directions: length, width, and height in the voxel space. The maximum and minimum values (edge contours) of the overall label in the three directions are taken to calculate the bounding box boundary. The result of the rough segmentation is converted into a bounding box that surrounds all teeth. This step is called determining the boundary.
[0034] Step S4: Segment the teeth from the ROI.
[0035] In this embodiment, preferably, in step S4, a semantic segmentation network based on deep learning is used for the obtained tooth ROI, all teeth are regarded as the same foreground, and the teeth are segmented from the ROI. The segmentation network used in this step is consistent with the AI segmentation tooth model method used in step S2, except that the voxel spacing of the segmented CBCT is the original voxel spacing.
[0036] Step S5: Map all teeth obtained back to their original CBCT positions and convert them into mesh data.
[0037] Specifically, in this embodiment, the teeth are mapped back to the original CBCT positions to obtain the real segmentation range of all teeth in the CBCT, and the teeth in the range are converted into grid data through the Marching Cube algorithm.
[0038] Step S6: Perform target detection on the tooth grid data to determine the local tooth area corresponding to each tooth position.
[0039] Preferably, in step S6 of this embodiment, a point cloud target detection algorithm is used, such as Point RCNN, 3DBoNet and PV-RCNN, and the detection result includes a classification result and a detection box. The classification result is the tooth position number of each tooth, and the detection box is the bounding box of each tooth.
[0040] Step S7: Determine the mesh range of each tooth based on the detection result of the tooth mesh data.
[0041] In this embodiment, preferably, in step S7, the grid data obtained in step S5 is extracted for the obtained tooth position and tooth local area, and the range of each tooth is obtained from the grid data.
[0042] Step S8: Perform single tooth segmentation on the teeth within the range.
[0043] In this embodiment, preferably, in step S8, semantic segmentation of different teeth is performed within the range of each tooth, the tooth corresponding to the tooth position frame is regarded as the foreground, and other grids are regarded as the background, and each tooth is segmented.
[0044] Specifically, in this embodiment, the PointNet++ point cloud segmentation algorithm is used to perform semantic segmentation on the grids within the bounding box corresponding to each tooth, that is, only the target teeth are segmented, and the remaining grids are treated as background.
[0045] Step S9: Map the segmented single tooth back to the mesh data obtained in step S5 to obtain all teeth after instance segmentation.
[0046] Specifically, in this embodiment, all teeth obtained after semantic segmentation are classified as the classification results in the respective target detection, and are mapped back to the grid data obtained in step S5 to obtain all teeth after instance segmentation.
[0047] Step S10: convert the grid data obtained in step S9 back to the original CBCT, and finally obtain the tooth segmentation result of CBCT.
[0048] Combination Figure 2 As shown in the figure, the original CBCT image is segmented using the original instance segmentation network and the multi-stage AI tooth segmentation method of this embodiment. The segmentation result using the original instance segmentation network is shown in Figure 3 As shown, the segmentation result using the method of this embodiment is as follows Figure 4 As shown, it can be seen that the results of tooth confusion between teeth 35 and 36 have been greatly improved ( Figure 3 and Figure 4The labels are changed in color, but they are fine as long as they are different, i.e. to distinguish different teeth). Example 2
[0049] The device for multi-stage AI tooth segmentation based on CBCT data of this embodiment includes: at least one processor; and at least one memory in communication with the processor; The memory stores instructions that can be executed by the processor, and the instructions are executed by the processor to enable the device to perform the method for multi-stage AI tooth segmentation based on CBCT data in Example 1. Example 3
[0050] The computer program product of this embodiment includes computer instructions, which, when executed by a processor, enable a computer device to perform the aforementioned method for multi-stage AI tooth segmentation based on CBCT data.
[0051] Many specific details are described in the above description to facilitate a full understanding of the present invention. However, the above description is only a preferred embodiment of the present invention. The present invention can be implemented in many other ways different from those described herein, so the present invention is not limited to the specific implementation disclosed above. At the same time, any person familiar with the art can make many possible changes and modifications to the technical solution of the present invention using the methods and technical contents disclosed above without departing from the scope of the technical solution of the present invention, or modify it into an equivalent embodiment of equivalent changes. Any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of protection of the technical solution of the present invention.
Claims
1. A multi-stage AI tooth segmentation method based on CBCT data, characterized by: The steps include: Step S1: Acquire CBCT image data; Step S2: resample the voxel spacing of the CBCT image, and use the trained AI segmentation tooth model to perform preliminary rough segmentation on the resampled CBCT at the same voxel spacing; Step S3: determining the tooth ROI of the original CBCT image; Step S4: segment the teeth from the ROI; Step S5: Mapping all teeth obtained back to the original CBCT positions and converting them into mesh data; Step S6: performing target detection on the tooth grid data to determine the local tooth area corresponding to each tooth position; Step S7: determining the mesh range of each tooth through the detection result of the tooth mesh data; Step S8: performing single tooth segmentation on the teeth within the range; Step S9: Mapping the segmented single tooth back to the mesh data obtained in step S5 to obtain all teeth after instance segmentation; Step S10: Convert the mesh data obtained in step S9 back to CBCT to obtain the tooth segmentation result of CBCT.
2. The method for multi-stage AI tooth segmentation based on CBCT data according to claim 1, characterized in that: Step S1 also includes data cleaning of the acquired CBCT data, using ITK or VTK to check the CBCT data, removing multi-sequence CBCT, single CBCT and CBCT containing useless information, retaining the remaining CBCT and converting it into nifti format.
3. The method for multi-stage AI tooth segmentation based on CBCT data according to claim 1, characterized in that: In step S2, the CBCT images are uniformly resampled to a voxel spacing of 1 mm×1 mm×1 mm, and the AI segmented tooth model is a nnUNet model trained at all 1 mm×1 mm×1 mm voxel spacings.
4. The method for multi-stage AI tooth segmentation based on CBCT data according to claim 1, characterized in that: In step S3, a boundary is determined for the segmentation result after the preliminary rough segmentation, and the boundary is mapped back to the original CBCT range, and the tooth range in the original CBCT is screened out to obtain the tooth ROI of the original CBCT image.
5. The method for multi-stage AI tooth segmentation based on CBCT data according to claim 1, characterized in that: In step S4, a semantic segmentation network based on deep learning is used on the obtained tooth ROI, all teeth are regarded as the same foreground, and the teeth are segmented from the ROI.
6. The method for multi-stage AI tooth segmentation based on CBCT data according to claim 1, characterized in that: In step S6, a point cloud target detection algorithm is used, and the detection result includes a classification result and a detection frame. The classification result is the tooth position number of each tooth, and the detection frame is the bounding box of each tooth.
7. The method for multi-stage AI tooth segmentation based on CBCT data according to claim 1, characterized in that: In step S7, the grid data obtained in step S5 is extracted for the obtained tooth position and tooth local area, and the range of each tooth is obtained from the grid data.
8. The method for multi-stage AI tooth segmentation based on CBCT data according to claim 1, characterized in that: In step S8, semantic segmentation of different teeth is performed within the range of each tooth, the tooth corresponding to the tooth position frame is regarded as the foreground, and other grids are regarded as the background, and each tooth is segmented.
9. A device for multi-stage AI tooth segmentation based on CBCT data, characterized in that: include: at least one processor; as well as at least one memory in communication with the processor; The memory stores instructions executable by the processor, and the instructions are executed by the processor to enable the device to perform the method for multi-stage AI tooth segmentation based on CBCT data as described in any one of claims 1-8.
10. A computer program product, characterized in that: The method comprises computer instructions, which, when executed by a processor, enable a computer device to perform the method for multi-stage AI tooth segmentation based on CBCT data according to any one of claims 1 to 8.