上顎All-on-4®植體配置須考量上顎竇與個別顎骨條件，但臨床規劃多仰賴醫師經驗。有限元素分析可量化生物力學表現，反覆建模與求解不利於大範圍搜尋。因此，本研究建立整合自動化有限元素分析(Finite Element Analysis, FEA)、多模態分支代理模型(Multimodal Branch Surrogate Model, MBSM)與粒子群最佳化(Particle Swarm Optimization, PSO)之植體配置流程。
本研究使用8組上顎骨模型，開發可依14項參數自動放置、旋轉植體並執行分析之FEA腳本，建立訓練資料。MBSM以顎骨體素與植體參數為雙分支輸入，分別經三維卷積神經網路與多層感知器擷取特徵，再以特徵式線性調制融合並採顎骨個別標準化，預測四支植體周圍骨平均von Mises應力。P1至P5用於模型建立，並採留一顎骨交叉驗證評估未見顎骨之泛化能力；完成訓練後，MBSM-PSO應用於P6至P8，以評估未知顎骨的泛化與預測穩定性。
結果顯示，以整組未見顎骨為測試資料的留一顎骨交叉驗證中，MBSM之R²為0.759±0.150。MBSM-PSO於8組顎骨皆找到低應力候選配置，且與FEA-PSO呈現相近的遠心端位置與植體尺寸趨勢。分位區間抽樣FEA驗證顯示，P1至P5有92.0%的區間平均相對誤差低於10%；P6至P8雖略高，各區間仍均不超過12.32%。整體77.5%的區間低於10%，且MBSM預測值與FEA結果之Kendall’s τ介於0.580至0.769，皆達統計顯著，顯示模型於不同應力區間仍能維持穩定之預測精度與排序一致性。
本研究建立的MBSM-PSO流程整合顎骨幾何與植體配置資訊，兼具跨顎骨應力預測與低應力配置搜尋能力，可為上顎All-on-4®個別化植體配置設計與最佳化提供具潛力的生物力學輔助工具。

Maxillary All-on-4® implant planning must account for anatomical constraints such as the maxillary sinus and patient-specific jawbone morphology, while clinical decisions still rely heavily on clinician experience. Although finite element analysis (FEA) can quantify biomechanical performance, repeated modeling and computation make large-scale configuration searches inefficient. Therefore, this study developed an integrated framework combining automated FEA, a multimodal branch surrogate model (MBSM), and particle swarm optimization (PSO) for patient-specific implant configuration optimization.
Eight maxillary models were included. An automated FEA workflow was developed to position and rotate implants according to 14 configuration parameters and to generate biomechanical training data. The MBSM used 128 × 128 × 128 jawbone voxels and implant parameters as two input branches. Jawbone features were extracted using a three-dimensional convolutional neural network, implant features using a multilayer perceptron, and both were fused through feature-wise linear modulation (FiLM). The model predicted the average von Mises stress in the peri-implant bone surrounding four implants.
MBSM-PSO identified low-stress configurations for all eight jawbones and showed optimization trends similar to FEA-PSO, particularly for distal implant positions and dimensions. Stratified FEA validation showed that 77.5% of jawbone–quantile intervals had mean relative errors below 10%, while all mean absolute errors were below 1 MPa. Kendall’s ranged from 0.580 to 0.769 and was statistically significant for all jawbones.
The proposed MBSM-PSO workflow integrates jawbone geometry and implant-configuration information and provides both cross-jawbone stress prediction and low-stress configuration searching. It therefore represents a potentially useful biomechanical support tool for patient-specific design and optimization of maxillary All-on-4® implant configurations.

[1]World Health Organization. Global oral health status report: towards universal health coverage for oral health by 2030. Geneva: World Health Organization; 2022.
[2] Borg-Bartolo R, Roccuzzo A, Molinero-Mourelle P, Schimmel M, Gambetta-Tessini K, Chaurasia A, et al. Global prevalence of edentulism and dental caries in middle-aged and elderly persons: a systematic review and meta-analysis. J Dent. 2022;127:104335. doi:10.1016/j.jdent.2022.104335.
[3] Soto-Peñaloza D, Zaragozí-Alonso R, Peñarrocha-Diago M, Peñarrocha-Diago M. The All-on-Four treatment concept: systematic review. J Clin Exp Dent. 2017;9(3):e474–e488. doi:10.4317/jced.53613.
[4] Patzelt SBM, Bahat O, Reynolds MA, Strub JR. The All-on-Four treatment concept: a systematic review. Clin Implant Dent Relat Res. 2014;16(6):836–855. doi:10.1111/cid.12068.
[5]Lin GH, Strauss FJ, Brunello G, Stilwell C, Jung RE, Kopp I, Schwarz F, Wang HL. 1st Global Consensus for Clinical Guidelines for the Rehabilitation of the Edentulous Maxilla: Single-Round Survey on Implant-Supported Fixed and Removable Prostheses. Clin Oral Implants Res. 2026;37(Suppl 30):S121–S134. doi:10.1111/clr.70027.
[6]Schwarz F, Jung RE, Kopp I, Heitz-Mayfield L, Wang HL; participants of the First Global Consensus Conference Group for Implant Dentistry (GCCG). First Global Consensus for Clinical Guidelines: Structured Recommendations (S2k-Level Guideline Framework) for the Rehabilitation of the Edentulous Maxilla Based on Core Outcome Sets. Clin Oral Implants Res. 2026;37 Suppl 30(Suppl 30):S7-S27. doi:10.1111/clr.70084
[7] Donos, N., Ng, E., Pannuti, C.M., Romito, G.A., Francisco, H.C.O., Abou-Ayash, S., Avila-Ortiz, G., Carames, J.M.M., Casentini, P., Chackartchi, T., Chappuis, V., Chen, S., Fugazzotto, P., Giannobile, W.V., Hagiwara, Y., Hamilton, A., Ivanovski, S., Kahn, S., Kan, J., Lambert, F., Levine, R.A., Navarro, J.M., Ornekol, T., Payer, M., Schoenbaum, T., Singh, M.K., Thacker, S. and Alcoforado, G. (2026), Consensus Report of Group 1 of the 1st Global Consensus for Clinical Guidelines for the Rehabilitation of the Edentulous Maxilla: Number of Implants, Timing of Implant Placement and Loading. Clin Oral Impl Res, 37: S28-S48. https://doi.org/10.1111/clr.70063
[8] Pala, K., Neugebauer, J., Buser, D., Brunello, G., Chmielewski, K., Chowdhary, R., Cordaro, L., Hosokawa, R., Jensen, O., Jung, U.-W., Lin, Y., Mardas, N., Markovic, A., Mattheos, N., Mijiritsky, E., Milinkovic, I., Nunes, L., Polido, W.D., Sanz, M., Schwarz, F., Sculean, A., Stavropoulos, A., Strauss, F.J., Sumi, T., Thieringer, F.M., Thoma, D.S., Yi, Y. and Gallucci, G.O. (2026), Consensus Report of Group 2 of the 1st Global Consensus for Clinical Guidelines for the Rehabilitation of the Edentulous Maxilla: Zygomatic, Standard-Length, and Short Implant-Supported Prostheses. Clin Oral Impl Res, 37: S49-S67. https://doi.org/10.1111/clr.70081
[9]Güçlü, Z. B., et al. (2022). "Mechanical response of different frameworks for maxillary all-on-four implant-supported fixed dental prosthesis: 3D finite element analysis." Biomedical Engineering / Biomedizinische Technik 67(5): 419–428.
[10]Ayali, A., et al. (2020). "Biomechanical comparison of the All-on-4, M-4, and V-4 techniques in an atrophic maxilla: A 3D finite element analysis." Computers in Biology and Medicine 123: 103880.
[11]Sezer, Taygun et al. “Effect of Anterior Implant Position on Biomechanical Performance in the Maxillary All-on-Four Treatment: A 3-D Finite Element Analysis.” The Journal of oral implantology vol. 48,3 (2022): 177-186. doi:10.1563/aaid-joi-D-21-00022
[12]Tezerişener, H. A., et al. (2024). "Comparison of stress distribution around all-on-four implants of different angulations and zygoma implants: a 7-model finite element analysis." BMC Oral Health 24(1): 176.
[13]Phellan, R., Hachem, B., Clin, J., Mac-Thiong, J.-M. and Duong, L. (2021), Real-time biomechanics using the finite element method and machine learning: Review and perspective. Med. Phys., 48: 7-18. https://doi.org/10.1002/mp.14602
[14]Kriging surrogate model for resonance frequency analysis of dental implants by a Latin hypercube-based finite element method
[15]Akkad, K., et al. (2023). "A Machine-Learning-Based Approach for Predicting Mechanical Performance of Semi-Porous Hip Stems." Journal of Functional Biomaterials 14(3).
[16]Gandía-Sastre, T. and M. Prados-Privado (2025). "Integrating Finite Element Data with Neural Networks for Fatigue Prediction in Titanium Dental Implants: A Proof-of-Concept Study." Applied Sciences (Switzerland) 15(19).
[17]Maag, C., et al. (2024). "Evaluation of machine learning techniques for real-time prediction of implanted lower limb mechanics." Front Bioeng Biotechnol 12: 1461768.
[18]Ardestani, M. M., et al. (2015). "A real-time topography of maximum contact pressure distribution at medial tibiofemoral knee implant during gait: Application to knee rehabilitation." Neurocomputing 154: 174–188.
[19] Kim, J. Y., et al. (2023). "Rapid Estimation of Contact Stresses in Imageless Total Knee Arthroplasty." Mathematics 11(16).
[20]Minku and R. Ghosh (2024). "A macro–micro FE and ANN framework to assess site-specific bone ingrowth around the porous beaded-coated implant: an example with BOX® tibial implant for total ankle replacement." Medical and Biological Engineering and Computing 62(6): 1639–1654.
[21] Murat, F., et al. (2025). "Predictive mathematical modeling of biomechanical behavior in all-on-4 implants design: effects of distal implant and occlusal load angulation using RSM based on FEA." Front Bioeng Biotechnol 13: 1644776.
[22] Kung, P. C., et al. (2023). "Prediction of Bone Healing around Dental Implants in Various Boundary Conditions by Deep Learning Network." International Journal of Molecular Sciences 24(3).
[23] Ghosh, R., et al. (2022). Neural Network (NN) Based Qualitative Assessment of Bone Growth Over Textured Implant Surfaces. Lecture Notes in Mechanical Engineering, Springer Science and Business Media Deutschland GmbH.
[24] Hsu, C. W., et al. (2021). "Engineer design process assisted by explainable deep learning network." Scientific reports 11(1): 22525.
[25] Shih, K. S., et al. (2025). Integration of Convolutional Neural Networks and Finite Element Analysis for Fast Evaluation of Implant Stress on Orthopedic Bone Plates. Mechanisms and Machine Science, Springer Science and Business Media B.V.
[26] Jia, Z., et al. (2024). "Designing three-dimensional lattice structures with anticipated properties through a deep learning method." Materials and Design 244.
[27] Yáñez González-Cuéllar, J. A., et al. (2026). "Dual-CNN-based AI framework for patient-specific design of short femoral stems." Comput Methods Programs Biomed 285: 109521.
[28] Al-Dirini, R. M. A., et al. (2020). "Computational efficient method for assessing the influence of surgical variability on primary stability of a contemporary femoral stem in a cohort of subjects." Biomechanics and Modeling in Mechanobiology 19(4): 1283–1295.
[29] Chanda, S., et al. (2015). "A combined neural network and genetic algorithm based approach for optimally designed femoral implant having improved primary stability." Applied Soft Computing Journal 38: 296–307.
[30] Chanda, S., et al. (2016). "Effects of interfacial conditions on shape optimization of cementless hip stem: an investigation based on a hybrid framework." Structural and Multidisciplinary Optimization 53(5): 1143–1155.
[31] Chatterjee, S., et al. (2019). "Computational intelligence based design of implant for varying bone conditions." International journal for numerical methods in biomedical engineering 35(6): e3191.
[32] Lee, C. H., et al. (2018). "An Optimization Study for the Bone-Implant Interface Performance of Lumbar Vertebral Body Cages Using a Neurogenetic Algorithm and Verification Experiment." Journal of Medical and Biological Engineering 38(1): 22–32.
[33] Lee, C. H., et al. (2014). "An optimization study of the screw orientation on the interfacial strength of the anterior lumbar plate system using neurogenetic algorithms and experimental validation." Journal of Biomechanical Engineering 136(11).
[34] Roy, S., et al. (2018). "Design of patient specific dental implant using FE analysis and computational intelligence techniques." Applied Soft Computing Journal 65: 272–279.
[35]Chen, Y. C., et al. (2025). "Real-time optimization of prosthetic design for complete arch implant-supported treatments using finite element-based machine learning." Journal of Prosthetic Dentistry 134(5): 1920–1928.
[36] Chen, Y. C., et al. (2024). "Personalized prosthesis design in all-on-4® treatment through deep learning-accelerated structural optimization." J Dent Sci 19(4): 2140–2149.
[37] Liu, C., et al. (2025). "Biomechanical design of titanium-PEEK combined fusion cage based on PLIF surgical model." Front Bioeng Biotechnol 13: 1671068.
[38] Beisekenov, N., et al. (2025). "Data-Driven Design and Additive Manufacturing of Patient-Specific Lattice Titanium Scaffolds for Mandibular Bone Reconstruction." Journal of Functional Biomaterials 16(9).
[39] All-on-4 treatment concept manual US_79360 C
[40] Niedermaier, Robert et al. “Implant-Supported Immediately Loaded Fixed Full-Arch Dentures: Evaluation of Implant Survival Rates in a Case Cohort of up to 7 Years.” Clinical implant dentistry and related research vol. 19,1 (2017): 4-19. doi:10.1111/cid.12421
[41] Tong, H., et al. (2012). "Mesiodistal angulation and faciolingual inclination of each whole tooth in 3-dimensional space in patients with near-normal occlusion." American Journal of Orthodontics and Dentofacial Orthopedics 141(5): 604–617.
[42] Falcinelli, C., et al. (2023). "Finite element analysis in implant dentistry: State of the art and future directions." Dental Materials 39(6): 539–556.
[43] Adiguzel, Ozkan et al. “Maxillary Cortical Bone Thickness in a South-Eastern Anatolian Population: A Cone-Beam Computed Tomography Study.” Medical science monitor : international medical journal of experimental and clinical research vol. 23 5812-5817. 7 Dec. 2017, doi:10.12659/msm.906229
[44] Çelebi, A., et al. (2024). "Maxillary sinus detection on cone beam computed tomography images using ResNet and Swin Transformer-based UNet." Oral Surgery, Oral Medicine, Oral Pathology and Oral Radiology 138(1): 149–161.
[45] Namatevs, I.; Nikulins, A.; Edelmers, E.; Neimane, L.; Slaidina, A.; Radzins, O.; Sudars, K. Modular Neural Networks for Osteoporosis Detection in Mandibular Cone-Beam Computed Tomography Scans. Tomography 2023, 9, 1772-1786. https://doi.org/10.3390/tomography9050141
[46] Chen, Y.-C., et al. (2025). "Real-time optimization of prosthetic design for complete arch implant-supported treatments using finite element-based machine learning." The Journal of Prosthetic Dentistry 134(5): 1920–1928.
[47] Moriwaki, Hiroyoshi et al. “Influence of Implant Length and Diameter, Bicortical Anchorage, and Sinus Augmentation on Bone Stress Distribution: Three-Dimensional Finite Element Analysis.” The International journal of oral & maxillofacial implants vol. 31,4 (2016): e84-91. doi:10.11607/jomi.4217
[48] Masoomi F, Mahboub F. Stress distribution pattern in all-on-four maxillary restorations supported by porous tantalum and solid titanium implants using three-dimensional finite element analysis. Eur J Transl Myol [Internet]. 2024 Feb. 16 [cited 2026 Jul. 21];34(1). Available from: https://www.pagepressjournals.org/bam/article/view/12170
[49] Al-Matrafi, T., Ahmed, K., Basunbul, G. et al. Enhanced implant longevity: Finite element optimization of all-on-four implant designs to minimize jawbone stresses. Saudi Dent. J. 37, 90 (2025). https://doi.org/10.1007/s44445-025-00099-8
[50] Al-Matrafi, T., Ahmed, K., Basunbul, G. et al. Enhanced implant longevity: Finite element optimization of all-on-four implant designs to minimize jawbone stresses. Saudi Dent. J. 37, 90 (2025). https://doi.org/10.1007/s44445-025-00099-8
[51] Song L, Wang D, Liu X, Yin A, Long Z. Prediction of mechanical properties of composite materials using multimodal fusion learning. Sensors and Actuators A: Physical. 2023;358:114433. doi:10.1016/j.sna.2023.114433.
[52] Perez E, Strub F, de Vries H, Dumoulin V, Courville A. FiLM: Visual Reasoning with a General Conditioning Layer. Proceedings of the AAAI Conference on Artificial Intelligence. 2018;32(1):3942–3951. doi:10.1609/aaai.v32i1.11671.
[53] Huang HL, Chang CH, Hsu JT, Fallgatter AM, Ko CC. Comparison of implant body designs and threaded designs of dental implants: A 3-dimensional finite element analysis. International Journal of Oral & Maxillofacial Implants. 2007;22(4):551–562.
[54] Huang HL, Hsu JT, Fuh LJ, Tu MG, Ko CC, Shen YW. Bone stress and interfacial sliding analysis of implant designs on an immediately loaded maxillary implant: A non-linear finite element study. Journal of Dentistry. 2008;36:409–417. doi:10.1016/j.jdent.2008.02.015.
[55] Bozkaya, D., & Müftü, S. (2005). Mechanics of the taper integrated screwed-in (TIS) abutments used in dental implants. Journal of biomechanics, 38(1), 87–97.