Introduction

Dental implants have revolutionized dental treatment by providing an effective solution for replacing missing teeth and restoring oral function and aesthetics1. One study estimated the implant prevalence among adults missing at least one tooth in the U.S. was 5.7% in 2016 and could reach up to 23% by 20262. Long term success of implants requires, among many factors, mechanical (primary) stability, which is strongly influenced by local bone quality and quantity1,2,3, to facilitate implant osseointegration with the surrounding bone structure1. Bone mineral density (BMD) is used clinically to evaluate the bone quality prior to surgery, to select a suitable implant site and to predict the primary stability3,4,5. Low BMD often results in low primary stability and delays on the final prosthetic rehabilitation.

Volumetric BMD can be determined by quantitative computed tomography (qCT)6 and dual energy CT (DECT)7, or assessed directly using the CT Hounsfield Unit (HU) value which scales with the BMD. Several studies have shown that there is a statistically significant correlation between the BMD/HU and the implant primary stability estimated by the insertion torque value (ITV) for implant placement5,8,9,10. However, CT, in general, requires high radiation exposure, high cost and large footprint and is not accessible for most dental clinics. Cone beam CT (CBCT) is the current 3D imaging modality of choice in dentistry11,12 and is used in presurgical dental implant planning13. Potential implant sites are evaluated based on their CBCT derived HU and are stratified into several categories following, for example, the Misch classification14, based on which clinical decisions including suitability for implant placement and surgical approach are made.

The current CBCT, however, is highly inaccurate and unreliable for quantifying the HU, with errors in the order of several hundred HU’s15,16,17. This has been attributed to the presence of strong x-ray photon scatter, beam hardening and incomplete data sampling that are intrinsic to the current CBCT design. Many commercial CBCT scanners provide only the image Grayscale Value (GV) instead of the HU. A recent study of two clinical CBCT units using multiple commercial implant planning software concluded that it was not possible to measure the HU reliably due to its large variations18. Attempts to correlate CBCT-derived HU/GV with implant primary stability produced mixed results varying from no19, weak20,21, to strong22 correlations. The inaccuracy leads to mischaracterization of bone types, unreliable prediction of implant stability, potentially incorrect clinical decisions, and an increased risk of implant failure.

The motivation of this study was to evaluate whether the accuracy of predicting implant primary stability can be improved by evaluating the BMD using a new multisource CBCT (ms-CBCT)23,24, which was designed with the aim to overcome the intrinsic limitations of the current CBCT. It replaces the conventional single wide cone angle x-ray tube with a carbon nanotube (CNT) based field emission x-ray source array25,26 that generates a stack of narrowly collimated x-ray beams23,27,28. Phantom imaging studies have shown the ms-CBCT significantly improves the HU accuracy, enhances the soft tissue contrast resolution, and reduces cone beam artifacts without increasing the radiation exposure, compared to multiple clinical dental CBCT scanners23,27,29. In this study, the relation between the ms-CBCT derived HU and the ITV for placing dental implants to porcine femurs was evaluated and compared to that of a CBCT. The results show a statistically significant correlation between the ms-CBCT derived HU and the ITV and a weak correlation between the conventional CBCT derived HU and the ITV.

Materials and methods

Specimen Preparation

Four commercially obtained frozen porcine femurs (from My Pet Carnivore, Indianapolis, IN 46201) were prepared for implant placement by first cleaning the surface soft tissue along the mid-femoral region. The femurs were initially stored at − 20 °C, then thawed at room temperature for 12–24 h prior to implant placement. Twelve identical titanium dental implants (NobelActive NP, 3.5 × 10 mm, Nobel Biocare) was utilized across the femurs to ensure consistency. The osteotomy for each implant site was prepared in stages, with sequential drilling from a precision drill (∅ ≈ 1.5 mm) to identify the entry point on the cortical surface and confirm appropriate alignment with the femur’s long axis. Drilling diameters were then increased gradually following the manufacturer’s guidelines until the ∅3.2 mm twist drill. The implant was inserted with a hand piece at a consistent speed and angulation maintained uniformly across all trials. Each femur accommodated three distinct osteotomy sites in the mid-diaphyseal cortical region, spaced at least 1 cm apart, to reduce the chance of structural overlap or bone weakening from previous placements. IACUC approval was not required because this study did not involve living vertebrate animals.

Implant primary stability test

The maximum insertion torque value (ITV) for placing an implant, a commonly used technique to evaluate the implant mechanical stability, was recorded as each implant was placed into its prepared osteotomy. A digital torque wrench (Vpoer Digital Torque Wrench 1/4-inch Drive) equipped with a compatible implant driver was set to peak mode to capture the highest torque value attained during insertion. The wrench’s threshold was configured to not exceed the expected maximum torque. Implants were aligned with the osteotomy and inserted in a clockwise direction at a consistent speed until stopped. The maximal ITV displayed on the wrench was taken as the outcome measurement. The implant placement procedure is shown in Fig. 1.

Fig. 1
Fig. 1
Full size image

Titanium dental implant was placed into a porcine femur bone following manufacture recommended procedure. The maximum insertion torque value (ITV) for placing the implant was recorded.

Imaging using ms-CBCT and CBCT

The benchtop ms-CBCT scanner23 consists of a CNT x-ray source array (NuRay Technologies, ChangZhou, China) and a flat panel detector (Xineos-1511, Teledyne, CA) on a rotating gantry, as illustrated in Fig. 2. The system geometry is similar to that of a commercial dental CBCT30 (CS9300, Carestream Dental, USA). The x-ray source array contains 8 focal spots (“sources”) with a 12 mm inter-spot spacing, each with the output power and the spatial resolution of a typical fixed-anode dental x-ray tube30. The radiation from each focal spot is confined to a narrow cone angle by an external multisource collimator to expose only a small section of the object in the axial direction and the full detector width. The 8 collimated beams collectively cover the entire FOV. They are activated sequentially to cycle 361 times in one full gantry rotation, producing 2880 image frames.

Fig. 2
Fig. 2
Full size image

A photo of the benchtop ms-CBCT with a CNT x-ray source array and a multisource collimator that generates a stack of narrowly collimated x-ray beams in the axial direction.

To avoid complications from other factors such as variations in the system configuration, exposure parameters, and the reconstruction method, the ms-CBCT was compared to the same benchtop device operating in the conventional CBCT configuration using only one of the eight CNT x-ray sources and a separate wide cone-angle collimator under otherwise the same imaging conditions (referred to as “CBCT-1”). The benchtop ms-CBCT system is shown in Fig. 2. Previous studies have shown the imaging performance of CBCT-1 is similar to the clinical dental CBCT scanners investigated23,29. The femur bone with the implants was placed inside a cylindrical plastic container filled with water to simulate scatter from soft tissue. The exposure parameters used by the two scanners are shown in Table 1, which are similar to those used for CBCT imaging of an adult patient under the standard protocol in our institution.

Table 1 Imaging protocols of the ms-CBCT and CBCT-1.
Full size table

Image reconstruction and analysis

The projection images from the ms-CBCT were first processed using an adjacent scatter ratio subtraction (ASRS) method enabled by this imaging configuration27. The processed projections from all sources were then used for volumetric CT reconstruction using the 3D Simultaneous Iterative Reconstruction Technique (SIRT) algorithm based on the ASTRA Toolbox31. The reconstruction was performed using a single system matrix, including the source/detector rays from all sources23. The total variation (TV) algorithm was incorporated for image noise reduction based on the Tigre Toolbox32,33. The CT numbers of all reconstructions were calibrated according to Gammex CT ACR 464 phantom (Sun Nuclear) testing instruction. The projection images from CBCT-1 were reconstructed using the same algorithm without ASRS processing. Reconstructions were generated using voxel sizes of 0.25 mm × 0.25 mm × 0.25 mm.

The reconstructed dataset underwent additional processing to facilitate implant segmentation and peri-implant bone density assessment. First, the reconstructed volume was imported into the 3D Slicer software34 (version 5.6.1). A threshold of 2400 HU was applied to segment the titanium implant from the surrounding bone structure. A custom script was then used to identify the implant volume in the segmentation using connected-component labelling. The script also defined an annular bone region surrounding the implant, with an inner and outer surface respectively 0.5 mm and 1.5 mm away from the implant surface, as illustrated in Fig. 1. The mean HU value within this annulus volume was computed. Any voxels with the CT numbers < 1000 HU were excluded in the calculation to eliminate the trabecular bones and soft tissues, leaving only higher-density cortical bone for analysis. This procedure was repeated for each implant and for both the ms-CBCT and CBCT-1 scans, ensuring consistency in the ROI selection and HU computation. The thickness of the cortical bone supporting the implants was also measured using the Line tool in ImageJ by drawing a straight line across the cortical region, and the displayed length was recorded.

The mean HU values extracted from each peri-implant region were correlated with the corresponding ITV’s recorded during implant placement. A linear regression was performed for the data from each scanner, and the coefficient of determination (R2) was calculated to quantify the strength of each correlation.

Results

Fig. 3
Fig. 3
Full size image

A reconstructed ms-CBCT image showing two implants after (a) reconstruction, (b) segmentation, and (c) identifying the annular peri-implant bone regions (in red color). The clear hole through the cortical bone is the result from a failed implant placement where the implant was drilled through the bone.

Figure 3. shows an example of the reconstructed ms-CBCT image of a femur with two implants. The implant volumes determined from segmentation and the annular shells used to compute the averaged peri-implant bone densities are displayed in Fig. 3b) and Fig. 3c), respectively.

Figure 4 shows the measured maximum ITV versus the averaged HU of the cortical bones derived from the ms-CBCT (triangle) and a CBCT-1 (circle). Each dataset consists of data from two femur bones. The solid lines are results from linear regression analysis. A positive strong correlation is observed between the ITV and ms-CBCT derived HU, with ITV (N×cm) = 0.1565 × HU – 99.772. The coefficient of determinant R2 is 0.86 and the calculated p-value is 0.00014 indicating a statistically significant correlation between the ITV and the HU. For the HU value derived from CBCT-1, linear regression shows a weak correlation of ITV (N×cm) = 0.2844 × HU – 248.32, with a R2 of 0.55 and a p-value of 0.0056.

Fig. 4
Fig. 4
Full size image

A scatter plot of insertion torque value against measured cortical bone HU. Although the dataset is modest in size, the ms-CBCT regression curve yields a stronger coefficient of determination (\({R}^{2}\approx\:0.86\)) compared to the CBCT-1 (\({R}^{2}\approx\:0.55\)).

Discussion

The results from this study show there is a statistically significant correlation (R2 = 0.86, P = 0.00015) between the ms-CBCT derived cortical bone HU and the implant primary stability measured by its maximum insertion torque value. This is consistent with the conclusions from prior studies using CT and qCT5,8,9,10, and is a significant improvement compared to the conventional CBCT, which shows only a weak correlation in this study and mixed results in previous reports19,20,21,22. Contrary to a previous report using a conventional CBCT and porcine ilium model22, the result from this study shows no correlation between the cortical bone thickness and the ITV.

In an experiment using limited number of postmortem mandibles and a qCT, Homolka et al. found a strong correlation (R2 = 0.83 and P < 0.001) between the BMD and the implant ITV. Similarly, Ikumi and coworkers reported a strong correlation between CT derived HU values and the implant ITV (R2 = 0.77) in a pilot patient study9. In contrast, results from prior publications using CBCT are inconclusive. A study using porcine mandibular condyles and CBCT reported statistically insignificant correlation19. A low correlation was found for placing implants into the patient maxilla (r = 0.329)20. One study using pig ilium by Wada and coworkers showed a positive correlation (R2 = 0.63)22. Recently, Vautrin et al. reported that there was a weak correlation in an experiment using cadaveric mandibles (R2 = 0.39) and showed the correlation was improved by further finite element analysis21. The result obtained from the conventional CBCT-1 in this study (R2 = 0.55) is within the range of those from previously publications using CBCT.

The large variation and overall weak correlation between the BMD derived from the conventional CBCT and implant primary stability are attributed to the intrinsic limitations of the current CBCT design. They are known to be inaccurate and inconsistent in quantifying the HU. The presence of strong scatter from the large imaging volume causes underestimation of the linear attenuation coefficient and consequently the HU value. This is clearly shown in Fig. 4, where the HU value calculated from CBCT-1 is ~ 100–200 HU smaller than that from the ms-CBCT for the same ROI. In addition, for the same material, CBCT derived HU varies with position inside the object due to scatter and beam hardening, which is known as the “Cupping” artifact. The image quality and accuracy are further compromised by the presence of cone beam artifacts caused by the large divergency of the x-ray beams and the resulting undersampling in regions away from the central imaging plane. The HU values are also strongly affected by imaging protocol including the field of view16. As a result, CBCT derived HU/GV is not a reliable measure of the bone density/quality and cannot reliably predict the implant primary stability.

There has been considerable amount of research directed towards improving the accuracy of dental CBCT for quantifying HU/BMD16, primarily through software-based postprocessing35,36. The ms-CBCT was designed to overcome the root cause problem of the current clinical CBCT, the large imaging volume, by replacing the wide cone angle x-ray tube with multiple narrow cone angle x-ray beams (2.4 degrees per beam vs. 10.5 degrees in CBCT-1)23,27. This significantly reduces scatter and cone-beam artifact and consequently improves the HU accuracy. A study comparing the HU accuracy of the ms-CBCT and several clinical CBCT23 units reports that the median deviation from the clinical CT was reduced from − 166.7HU and 354.4 HU for the two clinical CBCT units (Accuitomo 170 from J. Morita and CS9300 from Carestream Dental) to 43.7HU for the ms-CBCT in the Bland-Altman analysis. A separate study reports the ms-CBCT improves the accuracy of jawbone BMD compared to the two commercial CBCT scanners evaluated29. The enhanced HU accuracy and reliability contribute to the improved correlation between the ms-CBCT derived HU and the implant ITV observed in this study.

This feasibility study has its limitations. The scope of this study was limited. Only twelve implants of the same diameter and length were evaluated. A larger size sample that also includes implants with different specifications would be more desirable and will be used in follow up studies to improve the statistics and to cover the different clinical scenarios. Porcine femurs instead of mandibles were tested because of their consistent availability and comparable bone structure. Their cortical bones have higher HU and ITV values when compared to those from a typical human jawbone. The range of the HU values of their cortical bones is relatively narrow and does not cover the entire range of the different grades of bones in the Misch classification. This is not expected to change the conclusion of this report but needs to be validated in future studies. In a previous study, for example, we show the ms-CBCT is significantly more accurate than CBCT-1 in measuring the HU value of a dried human skull for both cortical and cancellous bones in the entire range of 0-2000HU29. Multiple implants were placed in each femur. The metal artifacts caused by the implants in proximity could affect the HU values of cortical bones.

The new ms-CBCT improves the imaging performance without increasing the imaging Air Kerma compared to a conventional CBCT. In this study, the Air Kerma value for the ms-CBCT is slightly lower than that of a conventional CBCT (NT5G, New Tom, Italy) in the normal dose protocol for an adult patient in our institution. The ms-CBCT scanner with multiple narrowly collimated x-ray beams acquires more image frames and requires a faster detector readout to maintain the scanning time compared to a conventional CBCT. By replacing the half-width and laterally offset FPD used presently with a full width one and by utilizing region of interest readout to reduce the detector readout time, an estimated scanning time of ~ 12 s can be achieved, which is within the range of the current dental CBCT with a similar FOV17,30. The ms-CBCT technology is still in the development stage. The main cost difference between this and a conventional CBCT is the CNT x-ray source array. Based on the costs of the multisource CNT x-ray source array manufactured for an intraoral tomosynthesis device37 and the monoblock x-ray generator used in the conventional CBCT, the cost of the ms-CBCT will be slightly higher but not significantly. The device can be constructed by retrofitting a conventional CBCT with the CNT x-ray source array.

Conclusion

This preliminary study demonstrates an improved correlation between the Hounsfield Unit value measured from the new ms-CBCT and the maximum implant insertion torque, suggesting that the ms-CBCT technology can improve the accuracy and reliability of predicting implant stability in presurgical dental implant planning compared to the current clinical CBCT.