Improving performance of deep learning models using 3.5D U-Net via majority voting for tooth segmentation on cone beam computed tomography
Abstract Deep learning allows automatic segmentation of teeth on cone beam computed tomography (CBCT). However, the segmentation performance of deep learning varies among different training strategies. Our aim was to propose a 3.5D U-Net to improve the performance of the U-Net in segmenting teeth on...
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2022-11-01
|
| Series: | Scientific Reports |
| Online Access: | https://doi.org/10.1038/s41598-022-23901-7 |
| _version_ | 1798016901697241088 |
|---|---|
| author | Kang Hsu Da-Yo Yuh Shao-Chieh Lin Pin-Sian Lyu Guan-Xin Pan Yi-Chun Zhuang Chia-Ching Chang Hsu-Hsia Peng Tung-Yang Lee Cheng-Hsuan Juan Cheng-En Juan Yi-Jui Liu Chun-Jung Juan |
| author_facet | Kang Hsu Da-Yo Yuh Shao-Chieh Lin Pin-Sian Lyu Guan-Xin Pan Yi-Chun Zhuang Chia-Ching Chang Hsu-Hsia Peng Tung-Yang Lee Cheng-Hsuan Juan Cheng-En Juan Yi-Jui Liu Chun-Jung Juan |
| author_sort | Kang Hsu |
| collection | DOAJ |
| description | Abstract Deep learning allows automatic segmentation of teeth on cone beam computed tomography (CBCT). However, the segmentation performance of deep learning varies among different training strategies. Our aim was to propose a 3.5D U-Net to improve the performance of the U-Net in segmenting teeth on CBCT. This study retrospectively enrolled 24 patients who received CBCT. Five U-Nets, including 2Da U-Net, 2Dc U-Net, 2Ds U-Net, 2.5Da U-Net, 3D U-Net, were trained to segment the teeth. Four additional U-Nets, including 2.5Dv U-Net, 3.5Dv5 U-Net, 3.5Dv4 U-Net, and 3.5Dv3 U-Net, were obtained using majority voting. Mathematical morphology operations including erosion and dilation (E&D) were applied to remove diminutive noise speckles. Segmentation performance was evaluated by fourfold cross validation using Dice similarity coefficient (DSC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV). Kruskal–Wallis test with post hoc analysis using Bonferroni correction was used for group comparison. P < 0.05 was considered statistically significant. Performance of U-Nets significantly varies among different training strategies for teeth segmentation on CBCT (P < 0.05). The 3.5Dv5 U-Net and 2.5Dv U-Net showed DSC and PPV significantly higher than any of five originally trained U-Nets (all P < 0.05). E&D significantly improved the DSC, accuracy, specificity, and PPV (all P < 0.005). The 3.5Dv5 U-Net achieved highest DSC and accuracy among all U-Nets. The segmentation performance of the U-Net can be improved by majority voting and E&D. Overall speaking, the 3.5Dv5 U-Net achieved the best segmentation performance among all U-Nets. |
| first_indexed | 2024-04-11T15:56:54Z |
| format | Article |
| id | doaj.art-80966284a2854f8eb1de156a8110ed9e |
| institution | Directory Open Access Journal |
| issn | 2045-2322 |
| language | English |
| last_indexed | 2024-04-11T15:56:54Z |
| publishDate | 2022-11-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj.art-80966284a2854f8eb1de156a8110ed9e2022-12-22T04:15:07ZengNature PortfolioScientific Reports2045-23222022-11-0112111510.1038/s41598-022-23901-7Improving performance of deep learning models using 3.5D U-Net via majority voting for tooth segmentation on cone beam computed tomographyKang Hsu0Da-Yo Yuh1Shao-Chieh Lin2Pin-Sian Lyu3Guan-Xin Pan4Yi-Chun Zhuang5Chia-Ching Chang6Hsu-Hsia Peng7Tung-Yang Lee8Cheng-Hsuan Juan9Cheng-En Juan10Yi-Jui Liu11Chun-Jung Juan12Department of Periodontology, School of Dentistry, Tri-Service General Hospital, National Defense Medical CenterDepartment of Periodontology, School of Dentistry, Tri-Service General Hospital, National Defense Medical CenterDepartment of Medical Imaging, Xinglong Rd, China Medical University Hsinchu HospitalDepartment of Medical Imaging, Xinglong Rd, China Medical University Hsinchu HospitalDepartment of Medical Imaging, Xinglong Rd, China Medical University Hsinchu HospitalDepartment of Medical Imaging, Xinglong Rd, China Medical University Hsinchu HospitalDepartment of Medical Imaging, Xinglong Rd, China Medical University Hsinchu HospitalDepartment of Biomedical Engineering and Environmental Sciences, National Tsing Hua UniversityMaster’s Program of Biomedical Informatics and Biomedical Engineering, Feng Chia UniversityDepartment of Medical Imaging, Xinglong Rd, China Medical University Hsinchu HospitalDepartment of Automatic Control Engineering, Feng Chia UniversityDepartment of Automatic Control Engineering, Feng Chia UniversityDepartment of Medical Imaging, Xinglong Rd, China Medical University Hsinchu HospitalAbstract Deep learning allows automatic segmentation of teeth on cone beam computed tomography (CBCT). However, the segmentation performance of deep learning varies among different training strategies. Our aim was to propose a 3.5D U-Net to improve the performance of the U-Net in segmenting teeth on CBCT. This study retrospectively enrolled 24 patients who received CBCT. Five U-Nets, including 2Da U-Net, 2Dc U-Net, 2Ds U-Net, 2.5Da U-Net, 3D U-Net, were trained to segment the teeth. Four additional U-Nets, including 2.5Dv U-Net, 3.5Dv5 U-Net, 3.5Dv4 U-Net, and 3.5Dv3 U-Net, were obtained using majority voting. Mathematical morphology operations including erosion and dilation (E&D) were applied to remove diminutive noise speckles. Segmentation performance was evaluated by fourfold cross validation using Dice similarity coefficient (DSC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV). Kruskal–Wallis test with post hoc analysis using Bonferroni correction was used for group comparison. P < 0.05 was considered statistically significant. Performance of U-Nets significantly varies among different training strategies for teeth segmentation on CBCT (P < 0.05). The 3.5Dv5 U-Net and 2.5Dv U-Net showed DSC and PPV significantly higher than any of five originally trained U-Nets (all P < 0.05). E&D significantly improved the DSC, accuracy, specificity, and PPV (all P < 0.005). The 3.5Dv5 U-Net achieved highest DSC and accuracy among all U-Nets. The segmentation performance of the U-Net can be improved by majority voting and E&D. Overall speaking, the 3.5Dv5 U-Net achieved the best segmentation performance among all U-Nets.https://doi.org/10.1038/s41598-022-23901-7 |
| spellingShingle | Kang Hsu Da-Yo Yuh Shao-Chieh Lin Pin-Sian Lyu Guan-Xin Pan Yi-Chun Zhuang Chia-Ching Chang Hsu-Hsia Peng Tung-Yang Lee Cheng-Hsuan Juan Cheng-En Juan Yi-Jui Liu Chun-Jung Juan Improving performance of deep learning models using 3.5D U-Net via majority voting for tooth segmentation on cone beam computed tomography Scientific Reports |
| title | Improving performance of deep learning models using 3.5D U-Net via majority voting for tooth segmentation on cone beam computed tomography |
| title_full | Improving performance of deep learning models using 3.5D U-Net via majority voting for tooth segmentation on cone beam computed tomography |
| title_fullStr | Improving performance of deep learning models using 3.5D U-Net via majority voting for tooth segmentation on cone beam computed tomography |
| title_full_unstemmed | Improving performance of deep learning models using 3.5D U-Net via majority voting for tooth segmentation on cone beam computed tomography |
| title_short | Improving performance of deep learning models using 3.5D U-Net via majority voting for tooth segmentation on cone beam computed tomography |
| title_sort | improving performance of deep learning models using 3 5d u net via majority voting for tooth segmentation on cone beam computed tomography |
| url | https://doi.org/10.1038/s41598-022-23901-7 |
| work_keys_str_mv | AT kanghsu improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT dayoyuh improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT shaochiehlin improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT pinsianlyu improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT guanxinpan improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT yichunzhuang improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT chiachingchang improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT hsuhsiapeng improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT tungyanglee improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT chenghsuanjuan improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT chengenjuan improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT yijuiliu improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography AT chunjungjuan improvingperformanceofdeeplearningmodelsusing35dunetviamajorityvotingfortoothsegmentationonconebeamcomputedtomography |