Accuracy of linear measurements performed with two imaging software in cone-beam computed tomography scans of dry human mandibles

The present study aimed to assess the accuracy of linear measurements performed with two software packages in multislice (MSCT) and cone beam (CBCT) computed tomography (CT) images. The sample consisted of 10 human mandibles marked standardly 8 times with metallic orthodontic wires. The mandibles underwent both MSCT and CBCT scanning and were measured linearly and digitally with two software packages, namely Studio 3.1 (Anne Solutions, SP, Brazil) (ST) and Implant Viewer 2.817 (Anne Solutions, SP, Brazil) (IV). The linear measurements were the mandibular height (MH) and width (MW), and the distance from the alveolar ridge to the mandibular canal (AC). To validate the measurements, the mandibles were sectioned in the marked regions and measured with a digital caliper. Statistically significant differences were not observed between the measurements (MH, MW, and AC) taken digitally with the two software packages from images obtained from the two scanning modalities (MSCT and CBCT) and the measurements taken manually with a digital caliper (p>0.05). MSCT and CBCT images measured digitally with ST and IV software packages resulted in reliable outcomes when compared to measurements obtained manually. Both software and imaging modalities are reliably useful for planning surgical procedure in the dental practice.


INTRODUCTION
Success of dental implant surgeries is partially assured by guidelines for the selection and application of maxillofacial imaging.Currently, computed tomography (CT) is the method most used for planning dental implant surgeries, as the limitations of conventional radiographs include the lack of detailed information on bone thickness ALESSANDRO A.L.S. SILVA et al. (Parnia et al. 2010).In CT imaging, maxillofacial hard tissues are captured by a sensor and manipulated three-dimensionally to avoid overlapping structures and to facilitate image interpretation (Frederiksen 1995).In the field of CT, multislice (MSCT) and cone-beam (CBCT) modalities figure more often in Medicine and Dentistry, respectively.Specifically, MSCT may be used in Medicine from the integral exam of the living (Deniz et al. 2017, Peng et al. 2016) to the cadaveric investigations on the cause of death and human identification (Puentes et al. 2009, Bolliger et al. 2005, 2007).On the other hand, CBCT is used in Dentistry for oral diagnosis (Larheim et al. 2015), investigation of dental and facial trauma (Palomo and Palomo 2009), guided surgeries (Deguchi et al. 2006, Garcia et al. 2008) pathological dissemination of dental lesions (Selden 1999, Obayashi et al. 2004), and planning of prosthetic rehabilitation (Dreiseidler et al. 2012, Tejaswi andPeriya 2013).
Previous studies highlight the advantages of CBCT imaging in face of MSCT, such as the use for less radiation and potentially higher resolution for images on dental and maxillofacial structures (Moura et al. 2009).Additionally, MSCT is designed mainly for full-body scans and requires larger and more expensive systems and facilities.Several studies have compared the accuracy of linear measurements (Pinsky et al. 2006, Gulobovic et al. 2012) between different CT modalities (Lund et al. 2009, Primo et al. 2012, Kim et al. 2012).In general, CBCT and MSCT perform similarly and accurately for linear measuring (Freire-Maia et al. 2017).However, these studies focused on comparing measurements taken digitally and manually without scoping deeper on the potential bias associated with the software packages used for image manipulation and analysis.
The present study aimed to assess the accuracy of linear measurements comparing two software packages in MSCT and CBCT scans of dry human mandibles.

MATERIALS AND METHODS
The present study was approved by the local Committee of Ethics in Research (protocol: #679.794).

STUDY DESIGN AND SAMPLING
A retrospective, cross-sectional and observational study was conducted.The sample consisted of 10 edentulous dry human mandibles.Each mandible was fixed in a parallelometer (B2-Bio-Art ® , Bio-Art Equipamentos Odontológicos Ltd. ® , São Paulo, SP, Brazil) with dental wax (Wilson Polidental ® , Cotia, SP, Brazil).The mandible bodies were marked 8 times with vertical linear drawings (4 marks in the posterior region and 4 marks in the anterior region) of 0.5mm thickness.The marks were performed with a prosthetic delineator crossing entirely the mandible border to simulate the parallelism of CT and MSCT scans.Orthodontic wire (Orthodontic Strong Elastic Floss CrNi floss ® , Morelli ® , Sorocaba, SP, Brazil) of 0.3mm thickness was bonded over each vertical mark using cyanoacrylate ester (Loctite 495 Super Bonder Instant Adhesive ® , Düsseldorf, Germany) (Figure 1).Small grooves were made on the orthodontic wire with diamond burs (FKG 1011 HL ® , Pdx Point Diamond Ind. Com.Ferram.Odontológ.Ltd. ® , Terra Preta, SP, Brazil) and a high-speed handpiece (Silent MRS 400 ® , Dabi Atlante®, Ribeirão Preto, SP, Brazil), without irrigation, to mark 4 points distributed in the vestibular/buccal, lingual, upper and lower surfaces of the mandible.These grooves were considered reference points for further measurements directly in the mandible and in CBCT/MSCT images.

METHODOLOGY
The mandibles underwent CBCT scanning with an i-CAT 17-19 system ® (Imaging Sciences International LLC ® , Hatfield, PA, USA) at 120 kV, 8 mA, rotation of 360°, 0.25mm voxel size, and acquisition time of 8.9 seconds.MSCT scanning was performed using a Philips Mx8000 IDT 16 CT ® device (Philips Medical System ® , Amsterdam, The Netherlands) with 120 kV, 150 mA, collimation of 16 ´ 0.6mm, thickness of 0.5mm, resolution of 0.4 ´ 0.4mm and acquisition time of 5 seconds.The mandibles were scanned considering their basal surface as standard parameter.Stabilization during the image acquisition was achieved using a 50mm Styrofoam plate.
To establish a reference measurement, the mandibles were sectioned after CT scanning allowing gold-standard measurements.The section were performed adjacent to the orthodontic wire marks with a low-speed metallic saw (EXTEC Labcut 1010 ® , Emfield, CT, USA) mounted with a 76 ´ 0.2mm blade (11-10066 IsoMet Wafering Blades 15 HC ® , Buehler ITW ® , Lake Bluff, IL, USA) under constant irrigation and general rotation of 500rpm (limit of the machine).The sectioned regions underwent polishing (EcoMet 3000 Grinder ® , Buehler ITW ® , Lake Bluff, IL, USA) until reaching the exact position of the orthodontic wire and their corresponding marks.

DATA ANALYSIS
In both software packages, the linear measurements were performed over the parasagittal images.In the posterior region, these measurements consisted of the height and width of the mandible, and distance from the alveolar ridge to the mandibular canal.In the anterior region, the height and width of the mandible were considered.The high of the mandible  consisted of the distance between the grooves on the upper and the lower surfaces of the mandible, while the width consisted of the distance between the grooves on the vestibular/buccal and lingual surfaces.The distance from the alveolar ridge to the mandibular canal considered the uppermost cortical surface of both the ridge and the canal.

STATISTICS
Intra and inter-examiner agreement was performed repeating the measurements 3 times by the main examiner (intra) and by a second examiner included in the research (inter) at this stage.The measurements were repeated using the 2 software packages and with the digital caliper within an interval of 1 week.Intraclass Correlation Coefficient (ICC) was applied to assess intra-and inter-examiner reliability.
Data normality was assessed with Shapiro-Wilk test and the homogeneity of variance with Levene's test.Student's t-test was used to compare measurements between MSCT and CBCT, and between MSCT/CBCT and the digital caliper.Analysis of Variance (ANOVA) followed by post-hoc Dunnet's test was used to compare the measurements obtained with MSCT/CBCT with the 2 software packages and the measurements taken with the digital caliper.All the statistical tests were performed using SPSS ® (IBM ® , New York, NY, USA) software package with a significance rate 0.05.

RESULTS
ICC test indicated excellent examiner agreement, with outcomes ranging from 0.82 to 0.98 (Hallgren 2012).
The mean values for mandible height, width and distance from the alveolar ridge to the mandible canal were 20.43, 10.98, 8.65mm.Student's t-test did not reveal differences statistically significant (p>0.05)comparing the measurements obtained with CBCT and MSCT with the reference measurement approach (digital caliper) (Table I).
Lack of differences statistically significant (p>0.05) was also found applying Student's t-test to compare the measurements obtained with each software package and the measurements obtained with the digital caliper (Table II).
The Analysis of Variance did not show differences statistically significant (p>0.05)comparing the software packages separately according to each measuring approach (Table III).The convergence of measurements obtained with the different software and measuring approaches was confirmed by Dunnett's test (Table III).

DISCUSSION
Over the last decades, CT scanning became one of the main complementary exams used to provide cross-sectional craniofacial images for diagnosis and treatment planning in Implantology (Frederiksen 1995, Parnia et al. 2010, Tyndall and Brooks 2000, Naitoh et al. 2010, Rubio Serrano et al. 2008).The vast use of CT in Dentistry led to need for constant assessment on the calibration of software tools for linear measurements.
To set up an adequate research design towards the investigation of linear measurements, mandible positioning was standardized for stabilization during the image acquisition.The concern with object positioning is justified on potential image alterations (due to head movement) that may occur on the spatial position in the focus point of the CT (El-Beialy et al. 2011).To overcome positioning errors, the mandibles used in the present study were fixed in the CT device over a Styrofoam plate.Clearly, an advantage of the present study relies on the use of dry human mandibles instead of images from living patients, as the first are not subjects to voluntary or involuntary movements when stabilized properly.
Once positioned, the mandibles were scanned in CBCT and MSCT.The comparison of linear measurement accuracy between CBCT and MSCT was previously performed and reported in the literature (Soumalainen et al. 2009, Freire-Maia et al. 2017).However, major attention is given to the modality of CT device itself in face of the software packages used for image analysis.In general, CBCT and MSCT perform similarly concerning linear measurements (Loubele et al. 2008).Similarly, the present study confirmed the performance of CBCT and MSCT based on the lack of differences statistically significant (p>0.05) between these CT modalities (Figure 1).However, the lack of differences statistically significant between CT does not guarantee reliability.Based on that, the comparison of CT measurements with a reference measurement (digital caliper) was performed.Accordingly, no differences statistically significant (p>0.05) were observed (Table I).The clinical significance of these findings remains on the reliable use of CBCT and MSCT scans for dental purposes -especially allowing linear measurements.
Once scanned, the mandibles are registered three-dimensionally as DICOM files, which may be analyzed with a variety of software available.Following the methodological set up, the present study also aimed to compare software packages.Kamburoğlu et al. 2009, addressed the important role of software packages and computer devices in the outcomes of radiographic linear measuring.Currently, the literature hampers the investigation of imaging software, as a great part of the studies published lack identification information (e.g., version, manufacturer and configuration) of the software packages used.Concerning on the software performance, Maloney et al. (2011) analyzed compared SimPlant 3D Pro ® software (Materialize ® , Leuven, Belgium) with the original i-Cat ® software (Imaging Sciences International ® , Hatfield, PA, USA) verifying no difference statistically significant.Considering the three distances measured in the present study, namely the mandible height, width and distance from the alveolar ridge to the mandible canal, none reached differences statistically significant (p>0.05) between software packages (Table II).Yet the measurements taken with the software were also not different statistically (p>0.05) from the measurements taken with the digital caliper (Table II).These outcomes validate the use of both software for image analysis and treatment planning.
Regarding the high inter-and intra-examiner reproducibility achieved, our results are similar to most studies in the literature, which show that images from CBCT and MSCT software packages lead to high reproducibility of results.The examiner agreement was especially higher for measuring long distances (MH), which reflect the application of software packages for planning dental implants in the posterior region of the mandible, which is a risk area for nerve injury during surgical approaches (Givol et al. 2013).
In light of these results, it is important to accredit both types of tomography used and the two software programs employed in this study for diagnosis and planning of dental implants.Future studies should take into account the alterations introduced when applying filters and when using other software packages.Additionally, large samples of both maxilla and mandible must be investigated and confronted with the existing literature.

CONCLUSIONS
Measurements of anatomic distances with both CBCT and MSCT software packages revealed high accuracy when compared with distances measured with a digital caliper.The present study points towards the effectiveness and reliability of CBCT and MSCT with the respective software packages tested in the present study for planning dental implants in the routine of Dentistry.

Figure 1 -
Figure 1 -Dry human mandible marked prior to scanning and measuring procedures.Edentulous mandible body marked 8 times with prosthetic delineator (a) and orthodontic wire (b).

Figure 2 -
Figure 2 -Dry human mandible manually measured.High of the mandible body measured with a digital caliper (a) in the region previously marked with orthodontic wire and cut with diamond burs (b).

Figure 3 -
Figure 3 -Dry human mandible manually measured.Operator interface of Imaging Studio® (Anne Solutions, São Paulo, SP, Brazil) software package showing the mandible in axial (a), panoramic (b) and parasagittal (c) views, and Implant Viewer® (Anne Solutions, São Paulo, SP, Brazil) software package showing the mandible in the parasagittal view through the highlighted control acquisition parameters (d).

TABLE I Student's t-test outcomes for the comparison of measuring approaches used in the present study.
CBCT: cone-Beam computed tomography; MSCT: multislice computed tomography; mm: millimeter; SD: standard deviation; p: p-value considering significance rate of 0.05.

TABLE II Student's t-test outcomes for the comparison of software packages used in the present study.
® (Anne Solutions ® , São Paulo, SP, Brazil); Implant Viewer ® (Anne Solutions ® , São Paulo, SP, Brazil); mm: millimeter; SD: standard deviation; p: p-value considering significance rate of 0.05.

TABLE III Analysis of Variance for the comparison of measuring approaches considering each software package separately.
Anne Solutions ® , São Paulo, SP, Brazil); Implant Viewer ® (Anne Solutions ® , São Paulo, SP, Brazil); mm: millimeter; SD: standard deviation; p: p-value considering significance rate of 0.05; *: Based on Dunnett's outcomes all the approaches converged to resemble the real value (digital caliper), except for three approaches (*) which resembled discretely less (0.95-0.99/1.00).