A post-processing method and system for AI segmentation of teeth based on CBCT data

By performing the connectivity domain analysis and merging of the AI ​​segmented teeth results of CBCT data, the problems of label confusion and crossover are solved, and automated post-processing is realized, which improves the accuracy of tooth segmentation and the accuracy of invisible orthodontic treatment.

CN114549559BActive Publication Date: 2024-12-03BONDENT TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210196096.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-01
Publication Date
2024-12-03
Estimated Expiration
2042-03-01

AI Technical Summary

Technical Problem

The existing CBCT data shows that tag confusion and crossover problems exist in the AI-segmented teeth results, resulting in inefficiency and affecting the accuracy of invisible orthodontic treatment.

Method used

Using the AI ​​segmentation tooth results post-processing method based on CBCT data, the intersecting connective domain is automatically judged and merged to generate optimized tooth segmentation results through the steps of connecting domain analysis, surrounding box calculation and morphological expansion.

Benefits of technology

It improves the accuracy of AI segmentation results, reduces the need for manual correction, significantly improves processing efficiency, and improves the accuracy of invisible orthodontic treatment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114549559B_ABST
    Figure CN114549559B_ABST
Patent Text Reader

Abstract

The present invention relates to a post-processing method for the result of AI segmentation of teeth based on CBCT data, comprising: S1 loading CBCT data; S2 performing AI segmentation on the CBCT data loaded in step S1; S3 performing connected domain analysis on the result of AI segmentation in step S2; S4 searching for the connected domains of other labels intersecting with each tooth based on step S3; S5 judging {R j} i Whether the connected domain in can be merged with tooth i; S6 generates the result. The post-processing method based on the AI ​​segmentation tooth result of CBCT data of the present invention can optimize the AI ​​segmentation tooth result of CBCT data, and does not require the user to manually check and correct the label confusion and intersection problems in the AI ​​segmentation result. The segmentation result can be automatically post-processed in a short time, and the chaotic and crossed labels can be merged to eliminate most of the label confusion, which greatly improves the efficiency, improves the AI ​​segmentation result, provides more information to orthodontists, and improves the accuracy of invisible orthodontic treatment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of oral medical image processing, and in particular to a post-processing method and system based on CBCT data AI segmentation of teeth results. Background Art

[0002] Cone Beam Computer Tomography (CBCT) is currently widely used in the diagnosis and treatment of dental diseases. With the improvement of people's living standards and the update of oral concepts, more and more people choose orthodontics. Orthodontic treatment requires segmentation of teeth based on the impact data collected by CBCT. Currently, automatic segmentation of teeth by AI will result in confusion and intersection of the labels of each tooth, which requires manual modification of the labels later, which is inefficient and affects the accuracy of invisible orthodontic treatment. Summary of the invention

[0003] The technical problem to be solved by the present invention is to design a processing method that can optimize the AI ​​segmentation results of CBCT data and improve the segmentation effect, so as to solve the existing technical problems.

[0004] In order to solve the above technical problems, the post-processing method of the tooth segmentation result based on CBCT data AI of the present invention comprises the following steps:

[0005] S1: Load CBCT data;

[0006] S2: Perform AI segmentation on the CBCT data loaded in step S1;

[0007] S3: Perform connected domain analysis on the result of AI segmentation in step S2;

[0008] S4: Based on step S3, for each tooth, find the connected domains of other labels that intersect with it;

[0009] S5: Based on step S4, determine {R j} i Whether the connected domain in can be merged with tooth i;

[0010] S6: Generate results.

[0011] Furthermore, in step S1, the CBCT data format is hundreds of files with a suffix of .dcm format.

[0012] Furthermore, in step S2, a conventional AI segmentation network framework is used to segment teeth on the CBCT dicom data, such as a V-Net network, a U-Net3D network, or a nn-UNet network framework.

[0013] Furthermore, the method for connected domain analysis in step S3 includes the following steps:

[0014] S31: Calculate the bounding box position p of all connected domains of each tooth label and size v label ;

[0015] S32: Calculate the volume of the connected domain in each bounding box and its label, and mark them from large to small. For example, the largest connected domain of tooth 1 is recorded as "0101".

[0016] S33: The information obtained in steps S31 and S32 is stored as the following dictionary structure D:

[0017] D={{″0101″:{p 1 , v 1}}, ...};

[0018] To facilitate subsequent indexing, the traversal operation on the connected domain is limited to the corresponding bounding box instead of the entire image.

[0019] Furthermore, the method for searching for the connected domains of other labels intersecting with each tooth in step S4 includes the following steps:

[0020] S41: Traverse all teeth, take tooth i as an example, and select its maximum connected domain C i As a benchmark;

[0021] S42: Set the shape and size of the convolution kernel (e.g., a spherical convolution kernel with a radius of 3). i Perform morphological expansion and record the expansion result as D i ;

[0022] S43: Calculate D i The bounding box of , traverse the bounding box area, find all other connected domains that intersect with the expansion result, denoted as {R j} i , where j is the tooth to which other connected domains belong.

[0023] Furthermore, in step S5 {R j} i The method for judging whether the connected domain in can be merged with tooth i comprises the following steps:

[0024] S51: Set the upper bound U of the combined volume of tooth i i and the connected domain R j Volume lower bound L j , U i and L j It is obtained by calculating the statistical information of tooth data, and its mean is recorded as mean i, the variance is recorded as std i ,but:

[0025] U i =mean i +2*std i ;

[0026] L j =mean j -std j ;

[0027] S52: j} i The connected domain is traversed in R j For example, first calculate the sum of its volume and the volume of tooth i. If this value is less than U i , indicating that the volume after merging is within the normal range; then judge R j Is the volume less than L? j , if it is less than, it means that the connected domain is not the normal range of the tooth to which it belongs, and then it is merged, which can avoid merging the entire tooth into other teeth.

[0028] S53: Modify the position and size information of the affected bounding box after merging.

[0029] The present invention also provides a post-processing system for AI segmentation of teeth based on CBCT data, comprising:

[0030] one or more processors; and

[0031] One or more memories storing a computer executable program, when the computer executable program is executed by the processor, the post-processing method of the above-mentioned AI segmentation tooth result based on CBCT data is performed.

[0032] Beneficial effects of the present invention:

[0033] The post-processing method of the tooth segmentation results based on CBCT data AI of the present invention can optimize the tooth segmentation results of CBCT data AI, and does not require the user to manually check and correct the problems of chaotic and crossed labels in the AI ​​segmentation results. The segmentation results can be automatically post-processed in a short time, and the chaotic and crossed labels can be merged to eliminate most of the label confusion, which greatly improves the efficiency, improves the results of AI segmentation, provides more information to orthodontists, and improves the accuracy of invisible orthodontic treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The specific implementation of the present invention will be further explained below in conjunction with the accompanying drawings.

[0035] Figure 1It is a flowchart of the post-processing method of the present invention based on CBCT data AI segmentation tooth result;

[0036] Figure 2 This is a schematic diagram of the AI ​​segmentation results of CBCT data;

[0037] Figure 3 Schematic diagram of the post-processing AI segmentation results of CBCT data. DETAILED DESCRIPTION

[0038] Combination Figure 1 The post-processing method of the tooth segmentation result based on CBCT data AI in this embodiment includes the following steps:

[0039] S1: Loading CBCT data; in this embodiment, the CBCT data format is hundreds of files with the suffix .dcm.

[0040] S2: Perform AI segmentation on the CBCT data loaded in step S1; In this embodiment, a conventional AI segmentation network framework is used to segment teeth on the DICOM data of CBCT, such as a V-Net network, a U-Net3D network, or a nn-UNet network framework, and the segmentation result is as follows: Figure 2 shown.

[0041] S3: Performing connected domain analysis on the result of AI segmentation in step S2; the method for connected domain analysis in this embodiment includes the following steps:

[0042] S31: Calculate the bounding box position p of all connected domains of each tooth label and size v label ;

[0043] S32: Calculate the volume of the connected domain in each bounding box and its label, and mark them from large to small. For example, the largest connected domain of tooth 1 is recorded as "0101".

[0044] S33: The information obtained in steps S31 and S32 is stored as the following dictionary structure D:

[0045] D={{″0101″:{p 1 , v 1}}, ...};

[0046] To facilitate subsequent indexing, the traversal operation on the connected domain is limited to the corresponding bounding box instead of the entire image.

[0047] S4: Based on step S3, searching for connected domains of other labels intersecting with each tooth; in this embodiment, the method for searching for connected domains of other labels intersecting with each tooth includes the following steps:

[0048] S41: Traverse all teeth, take tooth i as an example, and select its maximum connected domain C i As a benchmark;

[0049] S42: Set the shape and size of the convolution kernel (e.g., a spherical convolution kernel with a radius of 3 mm). i Perform morphological expansion and record the expansion result as D i ;

[0050] S43: Calculate D i The bounding box of the object is traversed in the bounding box area to find all other connected domains that intersect with the expansion result, denoted as {R j} i , where j is the tooth to which other connected domains belong.

[0051] S5: Based on step S4, determine {R j} i Whether the connected domain in can be merged with tooth i; in this embodiment {R j} i The method for judging whether the connected domain in can be merged with tooth i comprises the following steps:

[0052] S51: Set the upper bound U of the combined volume of tooth i i and the connected domain R j Volume lower bound L j , U i and L j The mean is calculated by the statistical information of the tooth data (the statistical data in this embodiment is 100 cases), and its mean is recorded as mean i , the variance is recorded as std i ,but:

[0053] U i =mean i +2*std i ;

[0054] L j =mean j -std j ;

[0055] S52: j} i The connected domain is traversed in R j For example, first calculate the sum of its volume and the volume of tooth i. If this value is less than U i , indicating that the volume after merging is within the normal range; then judge R j Is the volume less than L? j, if it is less than, it means that the connected domain is not the normal range of the tooth to which it belongs, and then it is merged, which can avoid merging the entire tooth into other teeth.

[0056] S53: Modify the position and size information of the affected bounding box after merging.

[0057] S6: Generate results, such as Figure 3 shown.

[0058] This embodiment also provides a post-processing system for AI segmentation of teeth based on CBCT data, including:

[0059] one or more processors; and

[0060] One or more memories storing a computer executable program, when the computer executable program is executed by the processor, the post-processing method of the above-mentioned AI segmentation tooth result based on CBCT data is performed.

[0061] The post-processing method of the tooth segmentation results based on the AI ​​CBCT data of this embodiment can optimize the tooth segmentation results of the AI ​​CBCT data. The user does not need to manually check and correct the problems of label confusion and intersection in the AI ​​segmentation results. The segmentation results can be automatically post-processed in a short time, and the chaotic and crossed labels can be merged to eliminate most of the label confusion, which greatly improves the efficiency, improves the results of AI segmentation, provides more information to orthodontists, and improves the accuracy of invisible orthodontic treatment.

[0062] 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 post-processing method for AI segmentation of teeth based on CBCT data, characterized by: The steps include: S1: Load CBCT data; S2: Perform AI segmentation on the CBCT data loaded in step S1; S3: Perform connected domain analysis on the result of AI segmentation in step S2; S4: Based on step S3, for each tooth, find the connected domains of other labels that intersect with it; S5: Based on step S4, determine {R j } i Whether the connected domain in can be merged with tooth i; S6: Generate results; The method for each tooth in step S4 to search for connected domains of other labels intersecting with it includes the following steps: S41: Traverse all teeth, take tooth i as an example, and select its maximum connected domain C i As a benchmark; S42: Set the shape and size of the convolution kernel, for C i Perform morphological expansion and record the expansion result as D i ; S43: Calculate D i The bounding box of the object is traversed in the bounding box area to find all other connected domains that intersect with the expansion result, denoted as {R j } i , where j is the tooth to which other connected domains belong; In step S5, {R j } i The method for judging whether the connected domain in can be merged with tooth i comprises the following steps: S51: Set the upper bound U of the combined volume of tooth i i and the connected domain R j Volume lower bound L j , U i and L j It is obtained by calculating the statistical information of tooth data, and its mean is recorded as mean i , the variance is recorded as std i ,but: U i =mean i +2*std i ; L j =mean j -std j ; S52: j } i The connected domain is traversed in R j For example, first calculate the sum of its volume and the volume of tooth i. If this value is less than U i , indicating that the volume after merging is within the normal range; then judge R j Is the volume less than L? j , if it is smaller than , it means that the connected domain is not within the normal range of the tooth to which it belongs, and then it is merged.

2. The post-processing method of the AI ​​segmentation results of teeth based on CBCT data according to claim 1, characterized in that: In step S1, the CBCT data format is hundreds of files with the suffix .dcm.

3. The post-processing method of the AI ​​segmentation results of teeth based on CBCT data according to claim 1, characterized in that: In step S2, a conventional AI segmentation network framework is used to segment teeth on the CBCT DICOM data.

4. The post-processing method of the AI ​​segmentation results of teeth based on CBCT data according to claim 1, characterized in that: The method for connected domain analysis in step S3 includes the following steps: S31: Calculate the bounding box position p of all connected domains of each tooth label and size v label ; S32: Calculate the volume of the connected domain in each bounding box and its label, and mark them from large to small. The largest connected domain of tooth 1 is recorded as "0101". S33: The information obtained in steps S31 and S32 is stored as the following dictionary structure D: D={{″0101″:{p1, v1}},...}; To facilitate subsequent indexing, the traversal operation on the connected domain is restricted to the corresponding bounding box.

5. A post-processing system based on AI segmentation of teeth based on CBCT data, characterized by: include: one or more processors; as well as One or more memories storing a computer executable program, wherein when the computer executable program is executed by the processor, the post-processing method of the tooth segmentation result based on CBCT data AI according to any one of claims 1 to 4 is executed.