1
|
Hsu K, Yuh DY, Lin SC, Lyu PS, Pan GX, Zhuang YC, Chang CC, Peng HH, Lee TY, Juan CH, Juan CE, Liu YJ, Juan CJ. Improving performance of deep learning models using 3.5D U-Net via majority voting for tooth segmentation on cone beam computed tomography. Sci Rep 2022; 12:19809. [PMID: 36396696 PMCID: PMC9672125 DOI: 10.1038/s41598-022-23901-7] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/26/2022] [Accepted: 11/07/2022] [Indexed: 11/18/2022] Open
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.
Collapse
Affiliation(s)
- Kang Hsu
- grid.260565.20000 0004 0634 0356Department of Periodontology, School of Dentistry, Tri-Service General Hospital, National Defense Medical Center, Taipei, Taiwan, ROC ,grid.260565.20000 0004 0634 0356School of Dentistry and Graduate Institute of Dental Science, National Defense Medical Center, Taipei, Taiwan, ROC
| | - Da-Yo Yuh
- grid.260565.20000 0004 0634 0356Department of Periodontology, School of Dentistry, Tri-Service General Hospital, National Defense Medical Center, Taipei, Taiwan, ROC
| | - Shao-Chieh Lin
- Department of Medical Imaging, Xinglong Rd, China Medical University Hsinchu Hospital, 199, Sec. 1, Zhubei, 302 Hsinchu Taiwan, ROC ,grid.411298.70000 0001 2175 4846Ph.D. Program in Electrical and Communication Engineering, Feng Chia University, Taichung, Taiwan, ROC
| | - Pin-Sian Lyu
- Department of Medical Imaging, Xinglong Rd, China Medical University Hsinchu Hospital, 199, Sec. 1, Zhubei, 302 Hsinchu Taiwan, ROC ,grid.411298.70000 0001 2175 4846Department of Automatic Control Engineering, Feng Chia University, No. 100 Wenhwa Rd., Seatwen, 40724 Taichung Taiwan, ROC
| | - Guan-Xin Pan
- Department of Medical Imaging, Xinglong Rd, China Medical University Hsinchu Hospital, 199, Sec. 1, Zhubei, 302 Hsinchu Taiwan, ROC ,grid.411298.70000 0001 2175 4846Master’s Program of Biomedical Informatics and Biomedical Engineering, Feng Chia University, Taichung, Taiwan, ROC
| | - Yi-Chun Zhuang
- Department of Medical Imaging, Xinglong Rd, China Medical University Hsinchu Hospital, 199, Sec. 1, Zhubei, 302 Hsinchu Taiwan, ROC ,grid.411298.70000 0001 2175 4846Master’s Program of Biomedical Informatics and Biomedical Engineering, Feng Chia University, Taichung, Taiwan, ROC
| | - Chia-Ching Chang
- Department of Medical Imaging, Xinglong Rd, China Medical University Hsinchu Hospital, 199, Sec. 1, Zhubei, 302 Hsinchu Taiwan, ROC ,grid.260539.b0000 0001 2059 7017Department of Management Science, National Yang Ming Chiao Tung University, Taipei, Taiwan, ROC
| | - Hsu-Hsia Peng
- grid.38348.340000 0004 0532 0580Department of Biomedical Engineering and Environmental Sciences, National Tsing Hua University, Hsinchu, Taiwan, ROC
| | - Tung-Yang Lee
- grid.411298.70000 0001 2175 4846Master’s Program of Biomedical Informatics and Biomedical Engineering, Feng Chia University, Taichung, Taiwan, ROC ,grid.413844.e0000 0004 0638 8798Cheng Ching Hospital, Taichung, Taiwan, ROC
| | - Cheng-Hsuan Juan
- Department of Medical Imaging, Xinglong Rd, China Medical University Hsinchu Hospital, 199, Sec. 1, Zhubei, 302 Hsinchu Taiwan, ROC ,grid.411298.70000 0001 2175 4846Master’s Program of Biomedical Informatics and Biomedical Engineering, Feng Chia University, Taichung, Taiwan, ROC ,grid.413844.e0000 0004 0638 8798Cheng Ching Hospital, Taichung, Taiwan, ROC
| | - Cheng-En Juan
- grid.411298.70000 0001 2175 4846Department of Automatic Control Engineering, Feng Chia University, No. 100 Wenhwa Rd., Seatwen, 40724 Taichung Taiwan, ROC
| | - Yi-Jui Liu
- grid.411298.70000 0001 2175 4846Department of Automatic Control Engineering, Feng Chia University, No. 100 Wenhwa Rd., Seatwen, 40724 Taichung Taiwan, ROC
| | - Chun-Jung Juan
- Department of Medical Imaging, Xinglong Rd, China Medical University Hsinchu Hospital, 199, Sec. 1, Zhubei, 302 Hsinchu Taiwan, ROC ,grid.38348.340000 0004 0532 0580Department of Biomedical Engineering and Environmental Sciences, National Tsing Hua University, Hsinchu, Taiwan, ROC ,grid.254145.30000 0001 0083 6092Department of Radiology, School of Medicine, College of Medicine, China Medical University, Taichung, Taiwan, ROC ,grid.411508.90000 0004 0572 9415Department of Medical Imaging, China Medical University Hospital, Taichung, Taiwan, ROC ,grid.260565.20000 0004 0634 0356Department of Biomedical Engineering, National Defense Medical Center, Taipei, Taiwan, ROC ,grid.19188.390000 0004 0546 0241Department of Computer Science and Information Engineering, National Taiwan University, Taipei, Taiwan, ROC
| |
Collapse
|
2
|
Pauwels R, Pittayapat P, Sinpitaksakul P, Panmekiate S. Scatter-to-primary ratio in dentomaxillofacial cone-beam CT: effect of field of view and beam energy. Dentomaxillofac Radiol 2021; 50:20200597. [PMID: 33882256 DOI: 10.1259/dmfr.20200597] [Citation(s) in RCA: 3] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/05/2022] Open
Abstract
OBJECTIVE The aim of this study was to evaluate the effect of field of view (FOV) and beam energy on the scatter-to-primary ratio (SPR) in dental cone-beam CT (CBCT). METHODS An anthropomorphic phantom representing an adult male (ATOM Max 711-HN, Norfolk, VA, USA) was scanned using the 3D Accuitomo 170 CBCT (J. Morita, Kyoto, Japan) using 11 FOVs. During each scan, half of the X-ray beam was blocked. Each scan was performed at three exposure settings with varying beam energy and equal radiation dose: 90 kV 5 mA, 77 kV 7.5 mA and 69 kV 10 mA. The SPR was estimated by measuring the grey values in the blocked and non-blocked regions of the RAW data. The effect of FOV on SPR was evaluated using Dunn's multiple comparison test, and the effect of the exposure settings was compared using a Wilcoxon signed rank test. RESULTS Larger FOVs showed increased scatter. FOVs with a shorter isocenter-detector distance showed a particularly high SPR. Most intercomparisons between FOVs were statistically significant. The largest difference was found between 17 × 12 cm and 6 × 6 cm (lower jaw), with the former showing a 4.9-fold higher SPR. The effect of beam energy was relatively small and varied between FOV sizes and positions. CONCLUSION While the choice of FOV size and position is determined by the diagnostic region of interest, the image quality deterioration for large FOVs due to scatter provides another incentive to limit the FOV size as much as possible.
Collapse
Affiliation(s)
- Ruben Pauwels
- Aarhus Institute of Advanced Studies, Aarhus University, Aarhus, Denmark.,Department of Radiology, Faculty of Dentistry, Chulalongkorn University, Bangkok, Thailand
| | - Pisha Pittayapat
- Department of Radiology, Faculty of Dentistry, Chulalongkorn University, Bangkok, Thailand
| | - Phonkit Sinpitaksakul
- Department of Radiology, Faculty of Dentistry, Chulalongkorn University, Bangkok, Thailand
| | - Soontra Panmekiate
- Department of Radiology, Faculty of Dentistry, Chulalongkorn University, Bangkok, Thailand
| |
Collapse
|