| 摘要: | 當前在人工數位牙科的發展中，修復體透過牙科電腦輔助設計與電腦輔助製造 (CAD/CAM) 人工數位化系統做三維立體 (3D) 列印已發展相當成熟，再加上人工智慧 (Artificial intelligence, AI) 應用也是日常生活與牙科必然趨勢，因此傳統的手工修復體製作也跟著進步更新。然而，修復體的特異性、再現性、邊緣完整性、密合度與咬合真實型態複雜度添加了人工數位化 CAD/CAM 製作與 AI 應用製程修復體與牙科設計的挑戰。所以，目前透過人工數位口掃機印模與牙科用椎形狀電腦斷層 (CBCT) 影像的擴展，可克服 AI 與 CAD/CAM 在修復體的臨床應用。本實驗目的在驗證 AI 對牙科修復體治療的可能性。
本研究方法先分析口外 AI 製程、人工數位化 CAD/CAM 設計、傳統人工去蠟鑄造三種不同方式，並利用牙科標準模型與樹脂材料做 3D 列印成形製造的 11 Veneer、 16 Crown、 36 Onlay、 46 Inlay 後再觀察彼此間差異性。同時比較不同製程設計修復體的所需時間、 3D表面再現性、邊緣完整性、以及密合度與精密度等。
結果顯示，三種方式設計製作修復體所需時間依序是 AI 最快速 (57-60秒)，其次是人工數位 (229-450秒)，最後是傳統人工去蠟鑄造最耗費時間 (263-623秒)。基本上利用根均方誤差值 (RMSE) 的計算與比較得到初步結果是人工數位設計的修復體比 AI 設計具有較優的3D表面再現性。至於邊緣完整性在光學顯微鏡下， AI 修復體除了具有良好的密合度外， AI 設計的全牙冠修復體都將可大幅減少邊緣間隙。但不同方式製作的修復體邊緣間隙顯示邊緣缺陷範圍在 (2 ± 1 values) – (5 ± 2 values) ，沒有顯著差異。三種製作方式在 X 光與 CBCT 影像上顯示在不同修復體的精密度上，不論是那個面向空間都是在臨床可接受範圍內 (< 120 μm)。
同時利用口內 AI 與人工數位兩種方式製作下顎第一大臼齒全牙冠進行比對佐證，結果也顯示 RMSE 範圍從 189 μm 到 252 μm有顯著差異。 原則上 AI 設計的近心接觸與遠心接觸距離都大於人工數位製作，因此臨床上人工數位製作調整也較耗費時間。而不論在 AI 或人工數位製作全牙冠的四個面向的邊緣間隙總和數值，依測試案例依序是 332 μm、 376 μm、 366 μm 與 276 μm、 341 μm、 293 μm。平均每個面向邊緣間隙均小於 90 μm。
本研究的結果顯示 AI 設計相較於傳統人工去蠟鑄造與人工數位設計將節省更多的時間而提高製作效率。再現性數據得知 AI 設計與人工數位設計的修復體表面真實性相近，表明 AI 已接近人工數位設計的牙科修復體形態。同時 AI 將可精準地設計修復體邊緣以減少邊緣間隙。總結這些結果也跟本研究假設相符合。 1. 人工智慧製程比傳統人工去蠟鑄造製程更符合患者的美學與口腔構造、 2. 人工智慧設計可以將修復體在牙科臨床表現更完美，但不代表人工智慧一定能取代人工數位化 CAD/CAM 的牙科製作流程。
In the current development of digital dentistry, three-dimensional (3D) printing of restorations through dental computer-aided design and computer-aided manufacture (CAD/CAM) digital systems has become quite mature, coupled with artificial intelligence (AI) application is also an inevitable trend in daily life and dentistry, so traditional manual dental restoration has also been updated with progress. However, the specificity, reproducibility, marginal integrity, accuracy, and occlusal complexity of restorations have added challenges to digital CAD/CAM production and AI application to make restorations and dental designs. Therefore, the current expansion of digital oral scanner impressions and dental cone-beam computed tomography (CBCT) images can overcome some of the clinical applications of AI and CAD/CAM in restorations. The purpose of this study is to verify the possibility of AI treatment of dental restorations.
This research method first analyzes three different methods: extraoral AI manufacturing, manual digital CAD/CAM design, and traditional manual wax removal casting. Use dental standard models and resin materials to 3D print and make 11 veneer, 16 crown, 36 onlay, and 46 inlay, and then observe the differences between them. At the same time, the time required to design the restoration, 3D surface reproducibility, marginal completeness, and accuracy of different processes were compared.
The results show that among the three methods, the time required for designing restorations is in order: AI is the fastest (57-60 seconds), followed by manual digital (229-450 seconds), and finally manual wax-up craftsmanship is the most time-consuming (263-623 seconds). Basically, the calculation and comparison of the root mean square error (RMSE) are used to obtain preliminary results. The manual digitally designed restoration has better 3D surface reproducibility than the AI design. As for the marginal completeness under the optical microscope, in addition to the good tightness of AI restorations, both AI and manual digitally designed full crown restorations can significantly reduce the marginal gap. However, the marginal gaps of restorations made in different ways showed that the marginal defects ranged from (2 ± 1 values) – (5 ± 2 values), with no significant difference. The X-ray and CBCT images of the three production methods show that the accuracy of different restorations is within the clinically acceptable range (< 120 μm) regardless of which side is facing.
At the same time, AI and manual digital methods were used to make the full crown of the first molar of the mandible for comparison. The results also showed that there was a significant difference in the RMSE range from 189 μm to 252 μm. In principle, the mesial and distal contact distances of AI design are larger than those of manual digital production, so manual digital production adjustments are also more time-consuming in clinical practice. The total marginal clearance values of the four sides of the full crown produced by AI or manual digital methods are 332 μm, 376 μm, 366 μm and 276 μm, 341 μm, 293 μm in order according to the test cases. The average gap per side is less than 90 μm.
The results of this study show that AI design will save more time and improve production efficiency compared with traditional manual wax removal casting and manual digital design. The reproducibility data shows that the surface trueness of the restorations designed by AI and manual digital design is similar, indicating that AI is close to the morphology of dental restorations designed by manual digital design. At the same time, AI will accurately design the edges of the restoration to reduce marginal gaps. In summary, these results are also consistent with the hypotheses of this study. 1. AI manufacture is more in line with the patient's aesthetics and oral structure than the traditional technique of manual wax removal casting. 2. AI design can make the restoration more perfect in clinical dentistry, but it does not mean that AI can replace the dentistry production process of the manual digital CAD/CAM. |
