The study was designed to evaluate the in-vitro accuracy of implant placement. Approval was obtained from our University Faculty of Dentistry Ethical Committee (approval no. 2018/157). Digital Imaging and Communication in Medicine (DICOM) data from a real clinical case were used to produce the base model. Materialise Mimics (Materialise Medical Software, Leuven, Belgium) was used to obtain standard tessellation language (STL) data. Input parameters were between 226 and 3,071 Hounsfield units. In the 3D models, the maxillary and mandibular regions were distinguished. The 3D printed models were made from Die and Model Resin ( Sprintray, Los Angeles, CA, USA). It has a 1700 mPa Flexural Modulus and 66,7 mPa Flexural Strength. The models are all designed to have a bone-like structure. The 3D printing process was performed with a layer height of 20 microns to ensure optimal model accuracy. In the Z direction, the outermost 4 mm is like cortical bone with 100% filling, and the inner part, the cancellous bone, is less with different filling patterns. All models are 3D printed in this way.

Meshmixer software (Autodesk, Mill Valley, CA, USA) was used to derive maxillary and mandibular models. The Cawood and Howell classification was used to differentiate among levels of atrophy [7]. Cawood and Howell III–V residual ridge types were designed for both the maxillary and mandibular models. Digital light processing technology (Moonray S 3D printer; Sprintray, Los Angeles, CA, USA) was used to print five copies of each design for the maxilla and mandible.

Initially, the gingival mask was constructed virtually on bone models with a 3 mm thickness for all bone surfaces. A pattern model was created to suit precisely these models containing the standard gingiva thickness. This model was designed to prevent sliding during slicon rubber weaving by fixing pin gaps at three distinct locations. On the same 3D printer, this gingival mask pattern mold was also produced. After affixing it to the bone models, RTV-2 silicone (Aydın Kompozit, Konya, Turkey) was poured in and allowed to cure for 24 h.

The sample size was calculated with G*Power software (ver. 3.1.3; Heinrich-Heine Universität, Düsseldorf, Germany) based on an alpha value of 0.05 and statistical power of 90%. A sample size of 120 implants was required (20 implants per group). Five copies of each Cawood and Howell model (III–V) were reproduced for the maxilla and mandible. In total, 30 models were produced (Fig. 1).

RTV-2 silicone rubber is used for the gum models, along with a surgical guide

Three 2.4-mm screws were placed to demarcate the anterior midline and posterior lateral lines on each side of the model. Computer-aided design/computer-aided manufacturing (CAD-CAM) wax (MarmoScan; Siladent, Goslar, Germany) was used to cover the screw heads and facilitate recognition during optical scanning (Fig. 2). We utilized a high-resolution cone-beam computed tomography machine (Planmeca Promax 3D Mid Dental Volumetric Tomography, Helsinki, Finland) to perform the CBCT scans. These scans were carried out using the subsequent parameters: 90 kV, 10 mA, and 36 s, with a Field of View (FOV) of 16 × 9 cm. White CAD-CAM spray (Dr. Mat, Istanbul, Turkey) was applied to the models to obtain higher-quality scans. The scanned gingival surface texture was transferred to the software of the NeWay optical 3D scanner (Open Technologies, Rezzato, Italy) (Fig. 3).

Radiopaque wax-covered screw heads are used to accurately superimpose STL data from the optical scan onto the DICOM data. STL, standard tessellation language; DICOM, Digital Imaging and Communication in Medicine

Superimposing STL data from the optical scan on the DICOM data in the coDiagnostiX software STL, standard tessellation language; DICOM, Digital Imaging and Communication in Medicine

coDiagnostiX software (Dental Wings Inc., Montreal, Canada) was used for implant planning and designing surgical guides. The radiopaque, wax-covered screw heads were used for accurate superposition of the STL and DICOM data. The gingival thickness of 3 mm on all surfaces and the cortical outer surface of 4 mm were measured and double-checked again on software, and models that did not have these features were not included in the study. After segmentation and marking of anatomical landmarks, virtual implants were positioned considering the available bone volume. Straumann (Basel, Switzerland) bone-level tapered implants (3.3 mm × 12 mm) were used in all regions. Four implants (two axial and two tilted) were planned for all models. In the mandibular models, the axial implants were located close to teeth #32 and #42. Two posteriorly angulated implants were positioned in front of the mental foramen at an approximately 30° angle. In the maxillary models, the axial implants were located close to teeth #12 and #22. Posterior angulated implants were placed in front of the anterior maxillary sinus wall at an approximately 30° angle. A constant anterior–posterior distance was maintained between the virtual implants. The surgical guide was designed using a sleeve of 5.0 mm in diameter and height. The guide design was sent to the laboratory and printed using the CARES P30 printer (Straumann). The entire process, from implant planning to surgical guide design, was overseen by a Straumann digital product consultant (Fig. 4).

The comparison module in the coDiagnostiX software is used for comparing the planned and actual implant positions

During the interventions, the models were firmly fastened in a vise and stationed on a table to ensure stability and diminish variability throughout the procedure. A single operator performed all interventions, an accomplished oral surgeon with a track record exceeding ten years in the field of implant placement. The aim behind this was to mitigate any operator-dependent factors that could possibly affect the precision of implant placement. Surgical guides were fixed to the models using three pins passed through the sleeves (1.3 × 28 mm). The implants were placed in accordance with the recommended surgical protocol. Following implant placement, CBCT scans of the models were obtained. DICOM data were used to assess deviations from the planned locations. Marker screws were used for superimposition. The comparison module of the coDiagnostiX software was used for the assessment. Positional accuracy was evaluated by comparing the virtually planned and actual implant positions (Fig. 4). The implant placement accuracy was assessed based on angular deviations at the base [angle (A), 3D offset (B3D), distal (BD), vestibular (BV), and apical (BA)] and tip [3D offset (A3D), distal (AD), vestibular (AV), and apical (AA)].

Statistical analysis was performed using SPSS Statistics software (version 23.0; IBM Corp., Armonk, NY, USA). A two-way analysis of variance was conducted for the analysis of jaw shape and region. Multiple comparisons were made using Duncan's multiple-range test. The quantitative data are presented as the mean ± standard deviation. Statistical significance was set at p < 0.05.