http://informahealthcare.com/jmt ISSN: 0309-1902 (print), 1464-522X (electronic) J Med Eng Technol, 2014; 38(5): 251–259 ! 2014 Informa UK Ltd. DOI: 10.3109/03091902.2014.913079
INNOVATION
Automated image-guided surgery for common and complex dental implants Xiaoyan Sun*1, Yongki Yoon2, Jiang Li3, and Frederic D. McKenzie4 J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
1
School of Information Science and Engineering, Hangzhou Normal University, Hangzhou, PR China, 2Whirlpool Corporation, Benton Harbor, MI, USA, 3Department of Electrical and Computer Engineering, Old Dominion University, Norfolk, VA, USA, and 4Department of Modeling, Simulation, and Visualization Engineering, Old Dominion University, Norfolk, VA, USA Abstract
Keywords
Dental implantation is now recognized as a standard of care for replacing missing teeth. Pre-operative planning with patient-specific images provides the basis for precise surgery, but such accuracy is hampered to some degree because of the manual drilling procedures performed by the surgeon. In this paper, a robotic system for automated site preparation for dental implants is presented in order to provide high accuracy drilling. The results of some experiments are given to validate the system. Additionally, results are shown for jaw surgery that utilize natural-root-formed implants which can contain multiple roots. This more complex type of implant is impossible to drill manually but can provide better long-term stability and success. With this robotic system, controlled and accurate drilling was achieved, which made more advanced implant designs possible.
Dental implantation, novel implants, robotic, volume decomposition
Introduction Dental implantation is now recognized as a standard of care for replacing missing teeth [1,2]. The loss of teeth affects not only one’s appearance but also abilities of eating, alimentation and speech, thus harming the patient’s long-term qualityof-life. The dental implant market worldwide expects strong growth through 2015 to reach $4.2 billion [3], although longterm studies (45 years) using Albrektsson et al. [4] criteria have shown success rates as low as 41.9% [3,5]. A human tooth is comprised of two components: a crown, and a root which embeds in the jawbone to support the crown. A dental implant is an artificial piece which substitutes the root portion of a missing tooth, and is usually in the shape of screw-typed cylinder. A corresponding shape in the jawbone needs to be surgically removed in order to allow the insertion and osseointegration of the implant. This bone removal procedure is called the ‘site preparation’ for the implant. Subsequently, the implant is inserted into the prepared site manually by the dentist. Traditionally, the site preparation is done with free-hand drilling by a surgeon, whose results lack accuracy and vary a lot with the surgeon’s experience [6,7]. An ideal dental implantation requires high anatomical accuracy. Appropriate spacing between an implant and a neighbouring implant, or natural tooth, is required to provide sufficient blood supply and avoid over-heating during drilling which can result *Corresponding author: Email: [email protected]
History Received 14 November 2013 Revised 1 April 2014 Accepted 3 April 2014
in the subsequent death of the bone cells. The minimum space between an implant and a neighbouring natural tooth and the space between two adjacent implants should be no less than 3 mm and range from 3–5 mm, respectively [8]. Anatomical structures around the implantation site for each patient also need to be considered, for example, bone density, neighbouring teeth, the schneiderian membrane of the maxillary sinus in the upper jaw and the mandibular nerve in the lower jaw. 3D medical imaging visualizes these structures, so that an appropriate surgical plan could be generated accordingly. Technologies such as image-guided navigation and surgical guides for site preparation were developed and proven to provide better accuracy than free-hand operations [9]. However, both methods still require manual drilling by the surgeon, with which errors caused by human factors like limited identification accuracy, trembling and fatigue, etc. are hardly avoidable. Robots were introduced into the field of surgery in the 1980s and remained popular ever since because of their unique advantages, including precision, flexibility, stability, etc. [10]. Robotic operation has been proven to reduce issues caused by human factors and provide superior accuracy, such as for needle insertion during prostate radiotherapy procedures [11] and for biopsy [12]. Likewise, robotic site preparation provides the possibility for ensuring more precise execution of the pre-operative plan in dental implantation. Although there were several publications on the topic of robotic dental implantation, most of them utilized a robot just for assistance and the site preparation was still performed
252
X. Sun et al.
J Med Eng Technol, 2014; 38(5): 251–259
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
by the surgeon. For example, Boesecke et al. [13] reported a dental implantation system with an assisting medical robot for holding the drill template and Chiarelli et al. [14] developed a 5-axes robotic system to create the surgical template. As proposed in our earlier publication [15], naturalroot-formed dental implants mimic the structures of the root portions of human teeth providing superior stress distribution than current cylinder-shape implants. However, such novel implants have more complicated geometry than the cylinder-shape implants and manual site preparation would be impossible. Robotic surgery for site preparation not only has the potential to provide high accuracy for emplacing any type of implant, but also makes the use of novel natural-root-formed implants possible.
Methods System architecture A wholly integrated system aiming at automated dental implantation utilizing image-guided robotics was designed and implemented in our lab. With a dental drill-bit and handpiece attached to its end-effector, a six degree-of-freedom (DOF) robot from Mitsubishi (MELFA RV-3S) functioned as a high accuracy milling machine. The pre-operative plan was generated based on patient-specific cone-beam CT (CBCT) images, and the registration between the pre-operative plan and the intra-operative workspace of the robot was performed utilizing a two-step registration procedure, with the help of a co-ordinate measurement machine (CMM). The CMM acted as the reference co-ordinate system (CS). The co-ordinate systems of the CBCT images and the robot were registered to this reference CS, respectively, as described in our earlier
publications [16,17]. As a result, the spacial relationship between the images and the robot was established, thus transferring the pre-operative plan to the surgical operation of the robot. The architecture of our image-guided robotic dental implantation system is given in Figure 1. At the first visit to the dental office, the patient will have fiducials glued to her natural teeth and then be subjected to a CBCT scan (sub-block 1 in Figure 1). Subsequently, a 3D model of the patients jaw with fiducials will be reconstructed from the CBCT images (sub-block 2 in Figure 1) and utilized by the dentist to plan the implant procedure (sub-block 3 in Figure 1). The surgical plan defines the type of implant going to be used and the position and orientation for the site preparation in the co-ordinate system of the planning software. The implant type determines the milling sequence of the robotic operation with pre-defined positions and orientations, while the planning co-ordinates along with the registration result determines the actual position and orientation of the robot milling during the surgery. On the next visit, the dentist will utilize the fiducials on the patient’s teeth to register the patient with her reconstructed 3D jaw model and the operational space of the robot. This will be done after restraining the patients head and mouth and then touching the fiducials with the tip of the CMM device. Many head fixation systems have been developed by researchers and applied for radiotherapy; for example, the thermoplastic mask-like fixation system from CIVCO [18], which could be adapted and customized for our application. Finally, the dentist identifies ‘no go’ areas with the CMM device and the robot can commence drilling. After the robotic site preparation is completed, the dentist manually emplaces the dental implant into the robotically drilled hole in the patients jaw bone.
Figure 1. System architecture of our image-guided robotic dental implantation system.
DOI: 10.3109/03091902.2014.913079
Detailed descriptions for the milling sequence generation and the robotic site preparation are given in the following sections.
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
Milling sequence generation Cylinder-shaped implants are currently the standard for dental implants, mainly because their site preparation is easy to perform and they approximate the shape of single-root human teeth. However, this basic design causes instability, especially for large molar cases. Natural-root-formed dental implants with single- and double-roots were designed as described in our earlier publication, and were proven to offer better stress distribution via Finite Element Method (FEM) analysis [15]. The shapes of such novel implants are given in Figure 2. These shapes are much more complex, thus their site preparation cannot be accomplished by utilizing manual surgical drilling, as is done for the current cylinder-shaped implants. Since our robot has very high accuracy (position repeatability of ± 0.02 mm), it is possible to remove the complex shapes in a point-by-point basis by 6-DOF milling. However, there are more than 20 000 voxels even for the volume of the single-root implant (resolution ¼ 0.2 mm), so the operating time of point-based milling would be too long and lead to possible over-heating during the surgery, which is considered a main reason for the early failure of implantation [19–21]. Besides, the large amount of data might easily cause a memory overflow in the robot, leading to unexpected errors. Alternatively, the shapes as shown in Figure 2 could be considered as the combination of some basic geometry. For instance, a root may be approximated by a cone shape and the top portion of the double-root implant could be approximated by a cylinder. Therefore, we propose a method to generate the milling sequences for our natural-root-formed implants based on volume decomposition, which decomposes the volume of a certain implant into the combination of basic geometry (sub-volumes) and discrete point-sequences. The main geometries include cylinder, cone, elliptic-cone, etc. and they can be defined using a uniform expression called an elliptic sub-volume with six parameters, as illustrated in Figure 3.
Automated image-guided surgery for dental implants
253
A Matlab algorithm was developed to generate the milling sequence (Figure 4). According for the width of the drill-bit attached to the robot (radius ¼ 1.0 mm), the voxelized implant model Voriginal was first ‘shrunk’ by 1.0 mm to get Vdiscretized. The shrunk model of the double-root implant was first segmented into three segments: one top and two roots; while the single-root model was considered as one single segment. Each segment was loaded as the current ‘target volume’ (Vi, as in Figure 4). A volume decomposition following the flowchart in Figure 4 was performed. The algorithm extracted elliptic sub-volumes with maximum volume from the current target volume and updated the target volume with the remaining segments then repeated the procedure until the latest volume was small enough to be recorded as a discrete point-sequence for milling. The result from this milling sequence generation algorithm for a double-root implant is given in Figure 5(b). The whole volume was decomposed into six components: an ellipticfrustum and a point-sequence set for each segment. The denser points located at the middle and bottom of the root are the discrete point milling sequences. Here, the threshold of error between the approximating elliptic-frustum and the original model was set as 0.8 mm, which makes sure the milled out elliptic-frustum shape approximates the original boundary with an error less than this threshold. It can be seen that the generated milling sequences enclose a volume which is smaller than the original model. This is because the width of the drill bit is taken into consideration when the milling
Figure 3. Uniform expression for sub-volumes using elliptic-frustum.
Figure 2. Shapes of our natural-root-formed implants with (a) single- and (b) double-root.
254
X. Sun et al.
J Med Eng Technol, 2014; 38(5): 251–259
sequence is generated. The actual volume removed by these milling sequences is shown in Figure 5(c), which gives a close approximation to the original model.
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
Robotic site preparation
Figure 4. Flowchart for volume decomposition algorithm.
The site preparation for the implant was performed by the robot, which has six DOF and a position repeatability of ± 0.02 mm. It comes with a controller (CR1B-571) which commands the robot operation using MELFA-BASIC IV programming language. A dental drill-bit with a diameter of 2 mm is attached to a dental hand piece from Aseptico (Woodinville, WA), which is driven by a high performance motor. With this hand piece rigidly connected to its endeffector, the robot functions as a high accuracy milling machine. It is commanded to complete the site preparation for the implantation automatically when appropriate parameters are given. There are two sets of information needed for the robot to perform the site preparation automatically: the type of implant and the position and orientation of the implant within the jaw bone. As mentioned earlier, this information was provided by the surgeon in the pre-operative planning stage. The implant type (single- or double-root, size of the implant) corresponds to a certain milling sequence generated from our volume decomposition algorithm, which comprises some elliptic-frustums and sets of discrete point-sequences. To coincide with the elliptic frustum, a sub-routine was
Figure 5. Milling sequence generation for a double-root implant: (a) the double-root model, (b) generated milling sequences, (c) volume removed by the milling sequences.
Automated image-guided surgery for dental implants
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
DOI: 10.3109/03091902.2014.913079
implemented for this geometry type in MELFA-BASIC IV language, within which the volume removed is determined by only several parameters instead of thousands of co-ordinates which can overflow the memory of the robot. The start point of robotic milling is either the top centre of the ellipticfrustum if it is the first component in the milling sequence or the first point if the milling sequence begins with a point sequence set. The orientation of the site preparation was actually realized by the orientation of the drill bit. The orientation differs for the two types of sub-volumes resulting from our volume decomposition algorithm: elliptic-frustum and point-sequence. The orientation of the former is that the drill bit points from the top centre to the bottom centre of the elliptic-frustum. The orientation of point-sequence milling remains unchanged if it is right after an elliptic-frustum milling; otherwise it is set as the same as the orientation of the first elliptic-frustum for a given implant type. A MELFA main function was defined for each implant type, within which sub-routines corresponding to each component decomposed from this implant were called in sequence. By commanding the robot with this main function from a PC, the site preparation for a certain implant can be executed automatically and accurately by the robot. Experiments The robotic site preparation (i.e. robotic milling of the root shape into the jaw bone) of the natural-root-formed implants with single- and double-root was tested with hard plaster (milling base) created by mixing plaster powder with water. The ratio between the powder and water was tested to find a suitable value with which the hardness of the plaster bases was suitable for the robot milling operation. Milling sequences for each implant were generated utilizing the algorithm based on volume decomposition (VD), as described in Section Methods. An arbitrary point on the surface of the milling base was set as the start point. In order to assess the accuracy and efficiency of the volumedecomposition-based algorithm for each implant type, the milling sequence was also generated without trying to extract any elliptic-frustums. Therefore, four different volumes were milled out using the robot: (1) S1: single-root shape from VD-based method;
255
(2) S2: single-root shape from point-based strategy; (3) D1: double-root shape from VD-based method; and (4) D2: double-root shape from point-based strategy. To assess these methods, the actual volumes removed and the milling times were recorded. The time of the milling operation for each of the implant designs and milling strategies was recorded by the robot and the experiment was performed five times for each type. Due to the fact that the volume removed for the site preparation is very small and in an irregular shape, the accurate shape of the removed volume is very hard to measure. Attempts were taken by cast moulding the shape of the volume removed, which helped to approximate and visualize the site preparation results. A dental material was filled into each of the removed volumes and remained overnight to set. Moulds were recovered by breaking down the milling base.
Results The recorded milling time for each volume type are given in Table 1. A two-tailed t-test was performed to compare the milling time for the same model using different milling strategies. p values for both single-root model and double-root model comparison were less than 0.05. Experiments were also performed to compare the time required to remove a certain volume (basic geometry) with a sub-routine vs pointbased milling using the robot, and the results are given in Table 2. While monitoring the whole milling procedure, it was noticed that the top segment of D1 was affected when the elliptic-frustum decomposed from one of the roots was milled out. The body of the drill bit went outside the boundary of the top segment due to a relatively large orientation angle of the elliptic-frustum, thus a small piece of milling base was removed (area inside the red square in Figure 6). This problem was corrected by straightening the root which caused the errant milling due to the relatively large angle. The modified model was generated in Autodesk 3ds Max. Exactly the same parameters (depth, radii, entry point) were used for the milling of this root, but its orientation was set to be vertical down. The new model was labelled as D4. The difference between D1 and D4 is illustrated Figure 7.
Table 1. Milling time for each volume type (units in seconds, n ¼ 5). Volume Total milling time
S1 (single-root shape from VD-based method)
S2 (single-root shape from point-based strategy)
D1 (double-root shape from VD-based method)
D2 (double-root shape from point-based strategy)
177.003 0.165
391.008 0.265
941.924 0.283
1383.005 0.264
M SD
Table 2. Comparison between the speed of removing the same volume with direct milling and sub-routine. Milling time (s, M ± SD) Geometry type
Parameters (mm)
with point-based milling
with sub-routine
Cone Elliptic-cone Elliptic-frustum
r ¼ 3, d ¼ 6 a ¼ 4, b ¼ 3, d ¼ 6 a1 ¼ 4, b1 ¼ 3, a2 ¼ 2, b2 ¼ 1.5, d ¼ 3
311.32 ± 0.68 402.91 ± 0.40 329.65 ± 0.46
100.04 ± 0.74 130.69 ± 0.67 52.43 ± 0.44
256
X. Sun et al.
By monitoring the milling procedure, it was noticed that the elliptic-frustum removal of the straightened root did not affect the upper segment anymore, which confirmed the effectiveness of this modification.
J Med Eng Technol, 2014; 38(5): 251–259
For our natural-root-formed implants, methods with volume-decomposition were selected as the optimized milling strategy. S1 and D4 were chosen as the standard volumes which were to be milled out for single- and double-root implant design, respectively. Measurements were taken to check the repeatability of the robotic milling with casted moulds. For each volume, the milling was repeated five times on a single plaster (Figure 8) and parameters including the length of the root and the dimensions of the opening of the removed volume (see illustration in Figure 9) were recorded and the results are given in Table 3. The standard deviations of parameters for each volume type ranged between 0.06– 0.28 mm.
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
Discussion
Figure 6. Milling result for double-root implant using volume-decomposition algorithm (D1).
The performance of robotic system for the site preparation for dental implantation was evaluated by experiments, including the milling speed and the accuracy of the shape removed. The results of statistical test suggested that, for our specially designed implants, the volume-decomposition-based milling has a much shorter processing time than the simple pointbased strategy (p50.05), which was about half the amount of time shorter than the latter, mainly due to the reduction of discrete points that resulted from volume-decomposition-base method, as shown in Table 4. It was also confirmed that milling with the MELFA sub-routines has much shorter operation times than direct point-based milling (p50.05),
Figure 7. Milling sequence for (a) the original double-root model D1 and (b) the modified double-root model with a straightened root D4.
Figure 8. Repeating the milling procedure for (a) single-root implant (S1); (b) double-root implant (D4).
Automated image-guided surgery for dental implants
DOI: 10.3109/03091902.2014.913079
257
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
Figure 9. Casting the removed volumes. (a) site preparation results; (b) the extracted moulds.
Table 3. Measurements of milling results (mm).. S1
1 2 3 4 5 M SD
D4
d1s
d2s
ls
d1
d2
l1
l2
5.73 5.64 5.70 5.82 5.76 5.73 0.07
3.74 3.58 3.64 3.69 3.63 3.66 0.06
9.65 9.51 10.23 9.95 9.89 9.85 0.28
9.15 9.33 9.29 9.33 9.16 9.25 0.09
7.85 7.76 7.85 7.82 8.08 7.87 0.12
9.26 8.84 9.31 9.1 9.37 9.18 0.21
8.92 8.77 8.69 9.18 8.99 8.91 0.19
Table 4. Results of milling sequence generation (threshold ¼ 0.8 mm).
Volume S1 (single-root shape from VD-based method) S2 (single-root shape from point-based strategy) D1 (double-root shape from VD-based method) D2 (double-root shape from point-based strategy)
Elliptic-frustum 1 (d ¼ 9.5 mm) / 3 (d ¼ 2.0, 5.6, 6.5 mm) /
which is about one-third of the latter. This supports the results provided in Table 1. These results suggest that the volume-decomposition-based milling not only avoids memory overflow, but also significantly shortens the total milling time, thus decreasing the risk to the patient of bone over-heating during the operation. The top view of the removed volume of the double-root model with a straightened root is given in Figure 10.
Number of discrete points in each segment 91 1121 1224, 222, 109 813, 1391, 1404
Total number of discrete points 91 1121 1555 3608
Since this model utilized the volume-decomposition-based method for volume removal and has the same parameters for the elliptic-frustums as in D1, the milling time was the same as for D1, which is much shorter than the time for D2. Therefore, D4 provides good results on both the shape removed and the milling time. FEM analysis for the straightened model was performed, contour plots of von Mises stresses (axial load ¼ 200 N) recorded at the
258
X. Sun et al.
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
location of implant-bone interface are given in Figure 11. A lower stress indicated a better biomechanics of the model. The results suggested that the minor modification of the root provides similar biomechanics of the original double-root model. The modified version of our double-root implant with a straightened root solves the errant milling,
Figure 10. Milling result for straightened model.
J Med Eng Technol, 2014; 38(5): 251–259
while preserving good biomechanical features and retaining the same milling time as the original design. Although the cast moulding (Figure 9) was not perfect: top surfaces were preserved but with some degree of extension and the root areas were not complete, which we believe was due to the existence of air bubbles within the small spaces around those areas; still it indicated a close approximation to our designed natural-root forms. This preliminary result from phantom experiments was encouraging and suggested the feasibility of our robotic site preparation system for dental implantation. However, milling with patient’s jaw bone is expected to have different effects from milling with plaster due to their different material features. Cortical bones are extremely hard and may cause small tremor at the tip of the drill bit, thus introduce inaccuracy to the operation. Another possible factor which may hamper the actual clinical practice of our robotic site preparation system is the fixation for the patient’s head/jaw. The accuracy of the fixation could affect both the registration and also the milling results. Some commercially available head fixation systems could be adapted and applied for our application. For example, the thermoplastic mask-like fixation system from CIVCO [18], as mentioned earlier in section 2.1. Experiments are needed to be designed and carried out to investigate the final accuracy of our system before it could be clinically utilized. Standard deviations of parameters in Table 3 indicated that the milling method can provide the result we desired
Figure 11. FEM comparison for the original double-root model D1 (upper) and the modified double-root model with a straightened root D4 (lower).
DOI: 10.3109/03091902.2014.913079
consistently. However, the depth of the milled out volume was shorter than its designed value, which we believe is due to the gathering of the powder being removed at the bottom of the volume. When the actual milling is taken in a clinical environment, there will be a small tube flushing away the removed debris, thus solving this problem occurring in the phantom experiments. We hope to prove this in future experiments.
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
Conclusion We presented an automated robotic dental implantation system. A 6-DOF robot with a dental drill-bit attached acted as a high accuracy milling machine. Milling sequences for the site preparation of novel natural-root-formed implants with single- and double-roots were generated by utilizing a volume-decomposition-based method. The implant volume was decomposed into the combination of some basic geometry and discrete point-sequences. For each basic geometry, a MELFA sub-routine was implemented to achieve its volume removal. By calling different sub-routines in sequence, the complete volume can be removed automatically by the robot. Phantom experiments proved that the complicated volumes of the natural-root-formed implants can be accomplished. With this robotic system, controlled and accurate drilling was achieved, which made more advanced implant designs possible. Although improvements such as convenient and affordable robots and testing different methods for head fixation will be necessary before it could be clinically utilized for dental implantation.
Acknowledgements The authors thank Krzysztof J. Rechowicz for his assistance with the model modification.
Declaration of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.
References 1. Barboza, E., Cau´la, A., and Carvalho, W., 2002, Crestal bone loss around submerged and exposed unloaded dental implants: a radiographic and microbiological descriptive study. Implant Dentistry, 11, 162–169. 2. Dental Implants - the tooth replacement solution. In International Congress of Oral Implantologists (ICOI). Available online at: http:// www.icoi.org/patient-education.php. Accessed 27 April 2010. 3. Schwartz-Arad, D., Kidron, N., and Dolev, E., 2005, A long-term study of implants supporting overdentures as a model for implant success. Journal of Periodontology, 76, 1431–1435. 4. Albrektsson, T., Zarb, G., Worthington, P., and Eriksson, A., 1986, The long-term efficacy of currently used dental implants: a review and proposed criteria of success. International Journal of Oral & Maxillofacial Implants, 1, 11–25. 5. Vogeler, M., Held, U., Gerds, T., and Strub, J.R., 2004, Long-term study using TPS-SteriOssÕ implants in partially edentulous patients. The IADR/AADR/CADR 82nd General Session. Hawaii.
Automated image-guided surgery for dental implants
259
6. Nickenig, H.J., Wichmann, M., Hamel, J., Schlegel, K.A., and Eitner, S., 2010, Evaluation of the difference in accuracy between implant placement by virtual planning data and surgical guide templates versus the conventional free-hand method-a combined in vivo-in vitro technique using cone-beam CT (Part II). Journal of Cranio-Maxillofacial Surgery, 38, 488–493. 7. Hoffmann, J., Westendorff, C., Gomez Roman, G., and Reinert, S., 2005, Accuracy of navigation guided socket drilling before implant installation compared to the conventional free hand method in a synthetic edentulous lower jaw model. Clinical Oral Implants Research, 16, 609–614. 8. Askary, A.S.E., Meffert, R.M., and Griffin, T., 1999, Why do dental implants fail? Part I. Implant Dentistry, 8, 173–185. 9. Kramer, F.J., Baethge, C., Swennen, G., and Rosahl, S., 2005, Navigated vs. conventional implant insertion for maxillary single tooth replacement. Clinical Oral Implants Research, 16, 60–68. 10. Korb, W., Marmulla, R., Raczkowsky, J., Mu¨hling, J., and Hassfeld, S., 2004, Robots in the operating theatre–chances and challenges. International Journal of Oral and Maxillofacial Surgery, 33, 721–732. 11. Podder, T.K., Sherman, J., Clark, D.P., Messing, E.M., Rubens, D.J., Strang, J.G., Liao, L., Brasacchio, R.A., Zhang, Y., Ng, W.S., and Yu, Y., 2005, Evaluation of robotic needle insertion in conjunction with in vivo manual insertion in the operating room. Robot and Human Interactive Communication, 2005. ROMAN 2005. IEEE International Workshop on. ROMAN 2005. pp. 66–72. 12. Boctor, E.M., Choti, M.A., Burdette, E.C., and Webster 3rd, R.J., 2008, Three-dimensional ultrasound-guided robotic needle placement: an experimental evaluation. The International Journal of Medical Robotics and Computer Assisted Surgery, 4, 180–191. 13. Boesecke, R., Brief, J., Raczkowsky, J., Schorr, O., Daueber, S., Krempien, R., Treiber, M., Wetter, T., and Habfeld, S., 2001, Robot assistant for dental implantology, Medical Image Computing and Computer-Assisted Intervention – MICCAI 2001 (Springer Berlin Heidelberg). pp. 1302–1303. 14. Chiarelli, T., Franchini, F., Lamma, A., Lamma, E., and Sansoni, T., 2012, From implant planning to surgical execution: an integrated approach for surgery in oral implantology. The International Journal of Medical Robotics and Computer Assisted Surgery, 8, 57–66. 15. Yoon, Y., Sun, X., Huang, J.-K., Hou, G., Rechowicz, K.D., and McKenzie, F., 2012, Designing natural-tooth-shaped dental implants based on soft-kill option optimization. Computer-Aided Design and Applications, 10, 59–72. 16. Sun, X., McKenzie, F.D., Bawab, S., Li, J., Yoon, Y., and Huang, J.K., 2011, Automated dental implantation using image-guided robotics: registration results. International Journal of Computer Assisted Radiology and Surgery, 6(5), 1–8. 17. Sun, X., Yoon, Y., Li, J., and McKenzie, F.D., 2011, An integrated computer-aided robotic system for dental implantation. MICCAI Workshop 2011: Systems and Architectures for Computer Assisted Interventions. Toronto, Canada. 18. Thermoplastics. Available online at: http://www.civco.com/ro/ products/Thermoplastics.htm. Accessed 2013.11.1. 19. Wagenberg, B., and Froum, S.J., 2006, A retrospective study of 1,925 consecutively placed immediate implants from 1988 to 2004. International Journal of Oral & Maxillofacial Implants, 21(1), 71–80. 20. Tawil, G., and Younan, R., 2003, Clinical evaluation of short, machined-surface implants followed for 12 to 92 months. International Journal of Oral & Maxillofacial Implants, 18(6), 894–901. 21. Park, H.-S., Jeong, S.-H., Kwon, O.-W., 2006, Factors affecting the clinical success of screw implants used as orthodontic anchorage. American Journal of Orthodontics and Dentofacial Orthopedics, 130, 18–25.
INNOVATION
Automated image-guided surgery for common and complex dental implants Xiaoyan Sun*1, Yongki Yoon2, Jiang Li3, and Frederic D. McKenzie4 J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
1
School of Information Science and Engineering, Hangzhou Normal University, Hangzhou, PR China, 2Whirlpool Corporation, Benton Harbor, MI, USA, 3Department of Electrical and Computer Engineering, Old Dominion University, Norfolk, VA, USA, and 4Department of Modeling, Simulation, and Visualization Engineering, Old Dominion University, Norfolk, VA, USA Abstract
Keywords
Dental implantation is now recognized as a standard of care for replacing missing teeth. Pre-operative planning with patient-specific images provides the basis for precise surgery, but such accuracy is hampered to some degree because of the manual drilling procedures performed by the surgeon. In this paper, a robotic system for automated site preparation for dental implants is presented in order to provide high accuracy drilling. The results of some experiments are given to validate the system. Additionally, results are shown for jaw surgery that utilize natural-root-formed implants which can contain multiple roots. This more complex type of implant is impossible to drill manually but can provide better long-term stability and success. With this robotic system, controlled and accurate drilling was achieved, which made more advanced implant designs possible.
Dental implantation, novel implants, robotic, volume decomposition
Introduction Dental implantation is now recognized as a standard of care for replacing missing teeth [1,2]. The loss of teeth affects not only one’s appearance but also abilities of eating, alimentation and speech, thus harming the patient’s long-term qualityof-life. The dental implant market worldwide expects strong growth through 2015 to reach $4.2 billion [3], although longterm studies (45 years) using Albrektsson et al. [4] criteria have shown success rates as low as 41.9% [3,5]. A human tooth is comprised of two components: a crown, and a root which embeds in the jawbone to support the crown. A dental implant is an artificial piece which substitutes the root portion of a missing tooth, and is usually in the shape of screw-typed cylinder. A corresponding shape in the jawbone needs to be surgically removed in order to allow the insertion and osseointegration of the implant. This bone removal procedure is called the ‘site preparation’ for the implant. Subsequently, the implant is inserted into the prepared site manually by the dentist. Traditionally, the site preparation is done with free-hand drilling by a surgeon, whose results lack accuracy and vary a lot with the surgeon’s experience [6,7]. An ideal dental implantation requires high anatomical accuracy. Appropriate spacing between an implant and a neighbouring implant, or natural tooth, is required to provide sufficient blood supply and avoid over-heating during drilling which can result *Corresponding author: Email: [email protected]
History Received 14 November 2013 Revised 1 April 2014 Accepted 3 April 2014
in the subsequent death of the bone cells. The minimum space between an implant and a neighbouring natural tooth and the space between two adjacent implants should be no less than 3 mm and range from 3–5 mm, respectively [8]. Anatomical structures around the implantation site for each patient also need to be considered, for example, bone density, neighbouring teeth, the schneiderian membrane of the maxillary sinus in the upper jaw and the mandibular nerve in the lower jaw. 3D medical imaging visualizes these structures, so that an appropriate surgical plan could be generated accordingly. Technologies such as image-guided navigation and surgical guides for site preparation were developed and proven to provide better accuracy than free-hand operations [9]. However, both methods still require manual drilling by the surgeon, with which errors caused by human factors like limited identification accuracy, trembling and fatigue, etc. are hardly avoidable. Robots were introduced into the field of surgery in the 1980s and remained popular ever since because of their unique advantages, including precision, flexibility, stability, etc. [10]. Robotic operation has been proven to reduce issues caused by human factors and provide superior accuracy, such as for needle insertion during prostate radiotherapy procedures [11] and for biopsy [12]. Likewise, robotic site preparation provides the possibility for ensuring more precise execution of the pre-operative plan in dental implantation. Although there were several publications on the topic of robotic dental implantation, most of them utilized a robot just for assistance and the site preparation was still performed
252
X. Sun et al.
J Med Eng Technol, 2014; 38(5): 251–259
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
by the surgeon. For example, Boesecke et al. [13] reported a dental implantation system with an assisting medical robot for holding the drill template and Chiarelli et al. [14] developed a 5-axes robotic system to create the surgical template. As proposed in our earlier publication [15], naturalroot-formed dental implants mimic the structures of the root portions of human teeth providing superior stress distribution than current cylinder-shape implants. However, such novel implants have more complicated geometry than the cylinder-shape implants and manual site preparation would be impossible. Robotic surgery for site preparation not only has the potential to provide high accuracy for emplacing any type of implant, but also makes the use of novel natural-root-formed implants possible.
Methods System architecture A wholly integrated system aiming at automated dental implantation utilizing image-guided robotics was designed and implemented in our lab. With a dental drill-bit and handpiece attached to its end-effector, a six degree-of-freedom (DOF) robot from Mitsubishi (MELFA RV-3S) functioned as a high accuracy milling machine. The pre-operative plan was generated based on patient-specific cone-beam CT (CBCT) images, and the registration between the pre-operative plan and the intra-operative workspace of the robot was performed utilizing a two-step registration procedure, with the help of a co-ordinate measurement machine (CMM). The CMM acted as the reference co-ordinate system (CS). The co-ordinate systems of the CBCT images and the robot were registered to this reference CS, respectively, as described in our earlier
publications [16,17]. As a result, the spacial relationship between the images and the robot was established, thus transferring the pre-operative plan to the surgical operation of the robot. The architecture of our image-guided robotic dental implantation system is given in Figure 1. At the first visit to the dental office, the patient will have fiducials glued to her natural teeth and then be subjected to a CBCT scan (sub-block 1 in Figure 1). Subsequently, a 3D model of the patients jaw with fiducials will be reconstructed from the CBCT images (sub-block 2 in Figure 1) and utilized by the dentist to plan the implant procedure (sub-block 3 in Figure 1). The surgical plan defines the type of implant going to be used and the position and orientation for the site preparation in the co-ordinate system of the planning software. The implant type determines the milling sequence of the robotic operation with pre-defined positions and orientations, while the planning co-ordinates along with the registration result determines the actual position and orientation of the robot milling during the surgery. On the next visit, the dentist will utilize the fiducials on the patient’s teeth to register the patient with her reconstructed 3D jaw model and the operational space of the robot. This will be done after restraining the patients head and mouth and then touching the fiducials with the tip of the CMM device. Many head fixation systems have been developed by researchers and applied for radiotherapy; for example, the thermoplastic mask-like fixation system from CIVCO [18], which could be adapted and customized for our application. Finally, the dentist identifies ‘no go’ areas with the CMM device and the robot can commence drilling. After the robotic site preparation is completed, the dentist manually emplaces the dental implant into the robotically drilled hole in the patients jaw bone.
Figure 1. System architecture of our image-guided robotic dental implantation system.
DOI: 10.3109/03091902.2014.913079
Detailed descriptions for the milling sequence generation and the robotic site preparation are given in the following sections.
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
Milling sequence generation Cylinder-shaped implants are currently the standard for dental implants, mainly because their site preparation is easy to perform and they approximate the shape of single-root human teeth. However, this basic design causes instability, especially for large molar cases. Natural-root-formed dental implants with single- and double-roots were designed as described in our earlier publication, and were proven to offer better stress distribution via Finite Element Method (FEM) analysis [15]. The shapes of such novel implants are given in Figure 2. These shapes are much more complex, thus their site preparation cannot be accomplished by utilizing manual surgical drilling, as is done for the current cylinder-shaped implants. Since our robot has very high accuracy (position repeatability of ± 0.02 mm), it is possible to remove the complex shapes in a point-by-point basis by 6-DOF milling. However, there are more than 20 000 voxels even for the volume of the single-root implant (resolution ¼ 0.2 mm), so the operating time of point-based milling would be too long and lead to possible over-heating during the surgery, which is considered a main reason for the early failure of implantation [19–21]. Besides, the large amount of data might easily cause a memory overflow in the robot, leading to unexpected errors. Alternatively, the shapes as shown in Figure 2 could be considered as the combination of some basic geometry. For instance, a root may be approximated by a cone shape and the top portion of the double-root implant could be approximated by a cylinder. Therefore, we propose a method to generate the milling sequences for our natural-root-formed implants based on volume decomposition, which decomposes the volume of a certain implant into the combination of basic geometry (sub-volumes) and discrete point-sequences. The main geometries include cylinder, cone, elliptic-cone, etc. and they can be defined using a uniform expression called an elliptic sub-volume with six parameters, as illustrated in Figure 3.
Automated image-guided surgery for dental implants
253
A Matlab algorithm was developed to generate the milling sequence (Figure 4). According for the width of the drill-bit attached to the robot (radius ¼ 1.0 mm), the voxelized implant model Voriginal was first ‘shrunk’ by 1.0 mm to get Vdiscretized. The shrunk model of the double-root implant was first segmented into three segments: one top and two roots; while the single-root model was considered as one single segment. Each segment was loaded as the current ‘target volume’ (Vi, as in Figure 4). A volume decomposition following the flowchart in Figure 4 was performed. The algorithm extracted elliptic sub-volumes with maximum volume from the current target volume and updated the target volume with the remaining segments then repeated the procedure until the latest volume was small enough to be recorded as a discrete point-sequence for milling. The result from this milling sequence generation algorithm for a double-root implant is given in Figure 5(b). The whole volume was decomposed into six components: an ellipticfrustum and a point-sequence set for each segment. The denser points located at the middle and bottom of the root are the discrete point milling sequences. Here, the threshold of error between the approximating elliptic-frustum and the original model was set as 0.8 mm, which makes sure the milled out elliptic-frustum shape approximates the original boundary with an error less than this threshold. It can be seen that the generated milling sequences enclose a volume which is smaller than the original model. This is because the width of the drill bit is taken into consideration when the milling
Figure 3. Uniform expression for sub-volumes using elliptic-frustum.
Figure 2. Shapes of our natural-root-formed implants with (a) single- and (b) double-root.
254
X. Sun et al.
J Med Eng Technol, 2014; 38(5): 251–259
sequence is generated. The actual volume removed by these milling sequences is shown in Figure 5(c), which gives a close approximation to the original model.
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
Robotic site preparation
Figure 4. Flowchart for volume decomposition algorithm.
The site preparation for the implant was performed by the robot, which has six DOF and a position repeatability of ± 0.02 mm. It comes with a controller (CR1B-571) which commands the robot operation using MELFA-BASIC IV programming language. A dental drill-bit with a diameter of 2 mm is attached to a dental hand piece from Aseptico (Woodinville, WA), which is driven by a high performance motor. With this hand piece rigidly connected to its endeffector, the robot functions as a high accuracy milling machine. It is commanded to complete the site preparation for the implantation automatically when appropriate parameters are given. There are two sets of information needed for the robot to perform the site preparation automatically: the type of implant and the position and orientation of the implant within the jaw bone. As mentioned earlier, this information was provided by the surgeon in the pre-operative planning stage. The implant type (single- or double-root, size of the implant) corresponds to a certain milling sequence generated from our volume decomposition algorithm, which comprises some elliptic-frustums and sets of discrete point-sequences. To coincide with the elliptic frustum, a sub-routine was
Figure 5. Milling sequence generation for a double-root implant: (a) the double-root model, (b) generated milling sequences, (c) volume removed by the milling sequences.
Automated image-guided surgery for dental implants
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
DOI: 10.3109/03091902.2014.913079
implemented for this geometry type in MELFA-BASIC IV language, within which the volume removed is determined by only several parameters instead of thousands of co-ordinates which can overflow the memory of the robot. The start point of robotic milling is either the top centre of the ellipticfrustum if it is the first component in the milling sequence or the first point if the milling sequence begins with a point sequence set. The orientation of the site preparation was actually realized by the orientation of the drill bit. The orientation differs for the two types of sub-volumes resulting from our volume decomposition algorithm: elliptic-frustum and point-sequence. The orientation of the former is that the drill bit points from the top centre to the bottom centre of the elliptic-frustum. The orientation of point-sequence milling remains unchanged if it is right after an elliptic-frustum milling; otherwise it is set as the same as the orientation of the first elliptic-frustum for a given implant type. A MELFA main function was defined for each implant type, within which sub-routines corresponding to each component decomposed from this implant were called in sequence. By commanding the robot with this main function from a PC, the site preparation for a certain implant can be executed automatically and accurately by the robot. Experiments The robotic site preparation (i.e. robotic milling of the root shape into the jaw bone) of the natural-root-formed implants with single- and double-root was tested with hard plaster (milling base) created by mixing plaster powder with water. The ratio between the powder and water was tested to find a suitable value with which the hardness of the plaster bases was suitable for the robot milling operation. Milling sequences for each implant were generated utilizing the algorithm based on volume decomposition (VD), as described in Section Methods. An arbitrary point on the surface of the milling base was set as the start point. In order to assess the accuracy and efficiency of the volumedecomposition-based algorithm for each implant type, the milling sequence was also generated without trying to extract any elliptic-frustums. Therefore, four different volumes were milled out using the robot: (1) S1: single-root shape from VD-based method;
255
(2) S2: single-root shape from point-based strategy; (3) D1: double-root shape from VD-based method; and (4) D2: double-root shape from point-based strategy. To assess these methods, the actual volumes removed and the milling times were recorded. The time of the milling operation for each of the implant designs and milling strategies was recorded by the robot and the experiment was performed five times for each type. Due to the fact that the volume removed for the site preparation is very small and in an irregular shape, the accurate shape of the removed volume is very hard to measure. Attempts were taken by cast moulding the shape of the volume removed, which helped to approximate and visualize the site preparation results. A dental material was filled into each of the removed volumes and remained overnight to set. Moulds were recovered by breaking down the milling base.
Results The recorded milling time for each volume type are given in Table 1. A two-tailed t-test was performed to compare the milling time for the same model using different milling strategies. p values for both single-root model and double-root model comparison were less than 0.05. Experiments were also performed to compare the time required to remove a certain volume (basic geometry) with a sub-routine vs pointbased milling using the robot, and the results are given in Table 2. While monitoring the whole milling procedure, it was noticed that the top segment of D1 was affected when the elliptic-frustum decomposed from one of the roots was milled out. The body of the drill bit went outside the boundary of the top segment due to a relatively large orientation angle of the elliptic-frustum, thus a small piece of milling base was removed (area inside the red square in Figure 6). This problem was corrected by straightening the root which caused the errant milling due to the relatively large angle. The modified model was generated in Autodesk 3ds Max. Exactly the same parameters (depth, radii, entry point) were used for the milling of this root, but its orientation was set to be vertical down. The new model was labelled as D4. The difference between D1 and D4 is illustrated Figure 7.
Table 1. Milling time for each volume type (units in seconds, n ¼ 5). Volume Total milling time
S1 (single-root shape from VD-based method)
S2 (single-root shape from point-based strategy)
D1 (double-root shape from VD-based method)
D2 (double-root shape from point-based strategy)
177.003 0.165
391.008 0.265
941.924 0.283
1383.005 0.264
M SD
Table 2. Comparison between the speed of removing the same volume with direct milling and sub-routine. Milling time (s, M ± SD) Geometry type
Parameters (mm)
with point-based milling
with sub-routine
Cone Elliptic-cone Elliptic-frustum
r ¼ 3, d ¼ 6 a ¼ 4, b ¼ 3, d ¼ 6 a1 ¼ 4, b1 ¼ 3, a2 ¼ 2, b2 ¼ 1.5, d ¼ 3
311.32 ± 0.68 402.91 ± 0.40 329.65 ± 0.46
100.04 ± 0.74 130.69 ± 0.67 52.43 ± 0.44
256
X. Sun et al.
By monitoring the milling procedure, it was noticed that the elliptic-frustum removal of the straightened root did not affect the upper segment anymore, which confirmed the effectiveness of this modification.
J Med Eng Technol, 2014; 38(5): 251–259
For our natural-root-formed implants, methods with volume-decomposition were selected as the optimized milling strategy. S1 and D4 were chosen as the standard volumes which were to be milled out for single- and double-root implant design, respectively. Measurements were taken to check the repeatability of the robotic milling with casted moulds. For each volume, the milling was repeated five times on a single plaster (Figure 8) and parameters including the length of the root and the dimensions of the opening of the removed volume (see illustration in Figure 9) were recorded and the results are given in Table 3. The standard deviations of parameters for each volume type ranged between 0.06– 0.28 mm.
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
Discussion
Figure 6. Milling result for double-root implant using volume-decomposition algorithm (D1).
The performance of robotic system for the site preparation for dental implantation was evaluated by experiments, including the milling speed and the accuracy of the shape removed. The results of statistical test suggested that, for our specially designed implants, the volume-decomposition-based milling has a much shorter processing time than the simple pointbased strategy (p50.05), which was about half the amount of time shorter than the latter, mainly due to the reduction of discrete points that resulted from volume-decomposition-base method, as shown in Table 4. It was also confirmed that milling with the MELFA sub-routines has much shorter operation times than direct point-based milling (p50.05),
Figure 7. Milling sequence for (a) the original double-root model D1 and (b) the modified double-root model with a straightened root D4.
Figure 8. Repeating the milling procedure for (a) single-root implant (S1); (b) double-root implant (D4).
Automated image-guided surgery for dental implants
DOI: 10.3109/03091902.2014.913079
257
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
Figure 9. Casting the removed volumes. (a) site preparation results; (b) the extracted moulds.
Table 3. Measurements of milling results (mm).. S1
1 2 3 4 5 M SD
D4
d1s
d2s
ls
d1
d2
l1
l2
5.73 5.64 5.70 5.82 5.76 5.73 0.07
3.74 3.58 3.64 3.69 3.63 3.66 0.06
9.65 9.51 10.23 9.95 9.89 9.85 0.28
9.15 9.33 9.29 9.33 9.16 9.25 0.09
7.85 7.76 7.85 7.82 8.08 7.87 0.12
9.26 8.84 9.31 9.1 9.37 9.18 0.21
8.92 8.77 8.69 9.18 8.99 8.91 0.19
Table 4. Results of milling sequence generation (threshold ¼ 0.8 mm).
Volume S1 (single-root shape from VD-based method) S2 (single-root shape from point-based strategy) D1 (double-root shape from VD-based method) D2 (double-root shape from point-based strategy)
Elliptic-frustum 1 (d ¼ 9.5 mm) / 3 (d ¼ 2.0, 5.6, 6.5 mm) /
which is about one-third of the latter. This supports the results provided in Table 1. These results suggest that the volume-decomposition-based milling not only avoids memory overflow, but also significantly shortens the total milling time, thus decreasing the risk to the patient of bone over-heating during the operation. The top view of the removed volume of the double-root model with a straightened root is given in Figure 10.
Number of discrete points in each segment 91 1121 1224, 222, 109 813, 1391, 1404
Total number of discrete points 91 1121 1555 3608
Since this model utilized the volume-decomposition-based method for volume removal and has the same parameters for the elliptic-frustums as in D1, the milling time was the same as for D1, which is much shorter than the time for D2. Therefore, D4 provides good results on both the shape removed and the milling time. FEM analysis for the straightened model was performed, contour plots of von Mises stresses (axial load ¼ 200 N) recorded at the
258
X. Sun et al.
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
location of implant-bone interface are given in Figure 11. A lower stress indicated a better biomechanics of the model. The results suggested that the minor modification of the root provides similar biomechanics of the original double-root model. The modified version of our double-root implant with a straightened root solves the errant milling,
Figure 10. Milling result for straightened model.
J Med Eng Technol, 2014; 38(5): 251–259
while preserving good biomechanical features and retaining the same milling time as the original design. Although the cast moulding (Figure 9) was not perfect: top surfaces were preserved but with some degree of extension and the root areas were not complete, which we believe was due to the existence of air bubbles within the small spaces around those areas; still it indicated a close approximation to our designed natural-root forms. This preliminary result from phantom experiments was encouraging and suggested the feasibility of our robotic site preparation system for dental implantation. However, milling with patient’s jaw bone is expected to have different effects from milling with plaster due to their different material features. Cortical bones are extremely hard and may cause small tremor at the tip of the drill bit, thus introduce inaccuracy to the operation. Another possible factor which may hamper the actual clinical practice of our robotic site preparation system is the fixation for the patient’s head/jaw. The accuracy of the fixation could affect both the registration and also the milling results. Some commercially available head fixation systems could be adapted and applied for our application. For example, the thermoplastic mask-like fixation system from CIVCO [18], as mentioned earlier in section 2.1. Experiments are needed to be designed and carried out to investigate the final accuracy of our system before it could be clinically utilized. Standard deviations of parameters in Table 3 indicated that the milling method can provide the result we desired
Figure 11. FEM comparison for the original double-root model D1 (upper) and the modified double-root model with a straightened root D4 (lower).
DOI: 10.3109/03091902.2014.913079
consistently. However, the depth of the milled out volume was shorter than its designed value, which we believe is due to the gathering of the powder being removed at the bottom of the volume. When the actual milling is taken in a clinical environment, there will be a small tube flushing away the removed debris, thus solving this problem occurring in the phantom experiments. We hope to prove this in future experiments.
J Med Eng Technol Downloaded from informahealthcare.com by McMaster University on 02/20/15 For personal use only.
Conclusion We presented an automated robotic dental implantation system. A 6-DOF robot with a dental drill-bit attached acted as a high accuracy milling machine. Milling sequences for the site preparation of novel natural-root-formed implants with single- and double-roots were generated by utilizing a volume-decomposition-based method. The implant volume was decomposed into the combination of some basic geometry and discrete point-sequences. For each basic geometry, a MELFA sub-routine was implemented to achieve its volume removal. By calling different sub-routines in sequence, the complete volume can be removed automatically by the robot. Phantom experiments proved that the complicated volumes of the natural-root-formed implants can be accomplished. With this robotic system, controlled and accurate drilling was achieved, which made more advanced implant designs possible. Although improvements such as convenient and affordable robots and testing different methods for head fixation will be necessary before it could be clinically utilized for dental implantation.
Acknowledgements The authors thank Krzysztof J. Rechowicz for his assistance with the model modification.
Declaration of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.
References 1. Barboza, E., Cau´la, A., and Carvalho, W., 2002, Crestal bone loss around submerged and exposed unloaded dental implants: a radiographic and microbiological descriptive study. Implant Dentistry, 11, 162–169. 2. Dental Implants - the tooth replacement solution. In International Congress of Oral Implantologists (ICOI). Available online at: http:// www.icoi.org/patient-education.php. Accessed 27 April 2010. 3. Schwartz-Arad, D., Kidron, N., and Dolev, E., 2005, A long-term study of implants supporting overdentures as a model for implant success. Journal of Periodontology, 76, 1431–1435. 4. Albrektsson, T., Zarb, G., Worthington, P., and Eriksson, A., 1986, The long-term efficacy of currently used dental implants: a review and proposed criteria of success. International Journal of Oral & Maxillofacial Implants, 1, 11–25. 5. Vogeler, M., Held, U., Gerds, T., and Strub, J.R., 2004, Long-term study using TPS-SteriOssÕ implants in partially edentulous patients. The IADR/AADR/CADR 82nd General Session. Hawaii.
Automated image-guided surgery for dental implants
259
6. Nickenig, H.J., Wichmann, M., Hamel, J., Schlegel, K.A., and Eitner, S., 2010, Evaluation of the difference in accuracy between implant placement by virtual planning data and surgical guide templates versus the conventional free-hand method-a combined in vivo-in vitro technique using cone-beam CT (Part II). Journal of Cranio-Maxillofacial Surgery, 38, 488–493. 7. Hoffmann, J., Westendorff, C., Gomez Roman, G., and Reinert, S., 2005, Accuracy of navigation guided socket drilling before implant installation compared to the conventional free hand method in a synthetic edentulous lower jaw model. Clinical Oral Implants Research, 16, 609–614. 8. Askary, A.S.E., Meffert, R.M., and Griffin, T., 1999, Why do dental implants fail? Part I. Implant Dentistry, 8, 173–185. 9. Kramer, F.J., Baethge, C., Swennen, G., and Rosahl, S., 2005, Navigated vs. conventional implant insertion for maxillary single tooth replacement. Clinical Oral Implants Research, 16, 60–68. 10. Korb, W., Marmulla, R., Raczkowsky, J., Mu¨hling, J., and Hassfeld, S., 2004, Robots in the operating theatre–chances and challenges. International Journal of Oral and Maxillofacial Surgery, 33, 721–732. 11. Podder, T.K., Sherman, J., Clark, D.P., Messing, E.M., Rubens, D.J., Strang, J.G., Liao, L., Brasacchio, R.A., Zhang, Y., Ng, W.S., and Yu, Y., 2005, Evaluation of robotic needle insertion in conjunction with in vivo manual insertion in the operating room. Robot and Human Interactive Communication, 2005. ROMAN 2005. IEEE International Workshop on. ROMAN 2005. pp. 66–72. 12. Boctor, E.M., Choti, M.A., Burdette, E.C., and Webster 3rd, R.J., 2008, Three-dimensional ultrasound-guided robotic needle placement: an experimental evaluation. The International Journal of Medical Robotics and Computer Assisted Surgery, 4, 180–191. 13. Boesecke, R., Brief, J., Raczkowsky, J., Schorr, O., Daueber, S., Krempien, R., Treiber, M., Wetter, T., and Habfeld, S., 2001, Robot assistant for dental implantology, Medical Image Computing and Computer-Assisted Intervention – MICCAI 2001 (Springer Berlin Heidelberg). pp. 1302–1303. 14. Chiarelli, T., Franchini, F., Lamma, A., Lamma, E., and Sansoni, T., 2012, From implant planning to surgical execution: an integrated approach for surgery in oral implantology. The International Journal of Medical Robotics and Computer Assisted Surgery, 8, 57–66. 15. Yoon, Y., Sun, X., Huang, J.-K., Hou, G., Rechowicz, K.D., and McKenzie, F., 2012, Designing natural-tooth-shaped dental implants based on soft-kill option optimization. Computer-Aided Design and Applications, 10, 59–72. 16. Sun, X., McKenzie, F.D., Bawab, S., Li, J., Yoon, Y., and Huang, J.K., 2011, Automated dental implantation using image-guided robotics: registration results. International Journal of Computer Assisted Radiology and Surgery, 6(5), 1–8. 17. Sun, X., Yoon, Y., Li, J., and McKenzie, F.D., 2011, An integrated computer-aided robotic system for dental implantation. MICCAI Workshop 2011: Systems and Architectures for Computer Assisted Interventions. Toronto, Canada. 18. Thermoplastics. Available online at: http://www.civco.com/ro/ products/Thermoplastics.htm. Accessed 2013.11.1. 19. Wagenberg, B., and Froum, S.J., 2006, A retrospective study of 1,925 consecutively placed immediate implants from 1988 to 2004. International Journal of Oral & Maxillofacial Implants, 21(1), 71–80. 20. Tawil, G., and Younan, R., 2003, Clinical evaluation of short, machined-surface implants followed for 12 to 92 months. International Journal of Oral & Maxillofacial Implants, 18(6), 894–901. 21. Park, H.-S., Jeong, S.-H., Kwon, O.-W., 2006, Factors affecting the clinical success of screw implants used as orthodontic anchorage. American Journal of Orthodontics and Dentofacial Orthopedics, 130, 18–25.
