Accuracy of linear measurements for implant planning based on low-dose cone beam CT protocols: a systematic review and meta-analysis

Abstract

Objectives

The aim of this systematic review was to verify the accuracy of linear measurements performed on low-dose CBCT protocols for implant planning, in comparison with those performed on standard and high-resolution CBCT protocols.

Methods

The literature search included four databases (Pubmed, Web of Science, Embase, and Scopus). Two reviewers independently screened titles/abstracts and full texts according to eligibility criteria, extracted the data, and examined the methodological quality. Risk of bias assessment was performed using the Quality Assessment Tool For In Vitro Studies. Random-effects meta-analysis was used for pooling measurement error data.

Results

The initial search yielded 4684 titles. In total, 13 studies were included in the systematic review, representing a total of 81 samples, while 9 studies were included in the meta-analysis. The risk of bias ranged from medium to low. The main results across the studies indicate a strong consistency in linear measurements performed on low-dose images in relation to the reference methods. The overall pooled planning measurement error from low-dose CBCT protocols was −0.24 mm (95% CI, −0.52 to 0.04) with a high level of heterogeneity, showing a tendency for underestimation of real values. Various studies found no significant differences in measurements across different protocols (eg, voxel sizes, mA settings, or dose levels), regions (incisor, premolar, molar) and types (height vs. width). Some studies, however, noted exceptions in measurements performed on the posterior mandible.

Conclusion

Low-dose CBCT protocols offer adequate precision and accuracy of linear measurements for implant planning. Nevertheless, diagnostic image quality needs must be taken into consideration when choosing a low-dose CBCT protocol.

Introduction

Cone beam CT (CBCT) is a dedicated imaging modality developed in the late nineties particularly for dentomaxillofacial imaging,¹ and has rapidly become the preferred imaging method for implant site assessment. Unlike traditional medical CT, CBCT uses a conical X-ray beam to capture a cylindrical or spherical volume of data, known as the field of view (FOV), which can be adjusted according to the area of interest, resulting in reduced patient exposure to harmful radiation.^2,3

The volumetric data acquired during CBCT rotation is reconstructed by a CT-based algorithm, resulting in a volume of data that can be assessed in axial, sagittal, and coronal planes or even in multiple variations of these plans.⁴ The reconstructed three-dimensional data comprises isotropic voxels (volume elements), providing high spatial resolution and enabling accurate measurements in all three dimensions.⁵

Cone beam CT has proven to be highly advantageous for implant planning since it offers reliable measurements that impact implant size selection, while also providing a comprehensive morphometric and skeletal analysis of the alveolar ridge, highlighting its proximity to vital anatomical structures.^6,7 However, the risk that radiation poses to the patient has been a constant concern in dentistry.⁸ The proper definition of the required diagnostic image quality for each indication is a grey area since fully balancing the radiation dose output and the image quality necessities requires considerable evidence.⁹ The variation in the diagnostic performance of the diverse modalities is directly linked with the radiation dose, and the ALADAIP (“As Low As Diagnostically Achievable being Indication-oriented and Patient-specific”) principle must always guide the challenging decision of what modality is to be used.⁹

The radiation dose associated with each tomographic acquisition is affected by different parameters selected by the operator, either manually or through predetermined exposure protocols. For most CBCT units, kVp (kilovoltage—peak) is fixed, whereas the tube current, ie, milliamperage (mA) and exposure time can be altered depending on desired image quality and patient size.¹⁰ Although reduction of the radiation dose is advantageous from a biological standpoint, extremely low dose levels may result in diagnostically useless images,¹¹ and the close relationship between radiation dose and image quality, particularly for implant planning, has been previously studied. Still, different devices, methodologies and protocols were used. To the best of the authors' knowledge, no study has conducted a systematic evaluation of the literature and meta-analysis regarding the use of low-dose protocols as the first choice for implant planning. Therefore, the aim of this study was to systematically investigate the literature regarding the accuracy of low-dose CBCT images when performing linear measurements for implant planning.

Methods

This systematic review was conducted following the preferred reporting items for systematic reviews (PRISMA) Statement,¹² and MOOSE guidelines.¹³ The research protocol was registered in the PROSPERO database under the number CRD42023431685. The following steps were defined: (i) review question, (ii) literature searches, (iii) study selection, and (iv) data extraction and synthesis.

Focused question and study eligibility

The following research question was created based on the PIRD framework (Population, Index test, Reference test, and Diagnosis):¹⁴ Are linear measurements (D) for implant planning (P) performed on low-dose CBCT (I) as accurate as those performed on standard CBCT/real measurements (R)?

Inclusion criteria for this systematic review were (a) studies comparing linear measurements on low-dose protocols to standard or high-resolution protocols; (b) studies comparing linear measurements on low-dose protocols to physical measurements; (c) studies assessing voxel size/milliamperage change in acquisition protocol, compared to a gold standard; and (d) studies published in English language.

Criteria for excluding studies were (a) studies qualitatively assessing only image quality; (b) studies assessing virtual implant planning/CAD-CAM; (c) studies that didn’t compare low-dose to a reference standard (real measurements or standard/high-resolution protocols); (d) linear measurements performed around installed implants or unrelated to dental implant treatment (e.g., maxillofacial surgery, orthodontics, periodontology); and (e) systematic reviews or literature reviews.

Search strategy

An electronic search was performed in April 2023 in the following databases: MEDLINE^® (PubMed), Web of Science, Embase, and Scopus, according to the strategies presented in Table 1. The search in MEDLINE was based on MeSH (Medical Subject Headings) terms and free-text terms. The searches in Web of Science, Embase, and Scopus were performed using free-text terms.

Table 1.

Search strategies used in this study.

Database	Indexing terms	Publications (n)
MEDLINE (Pubmed)	#1 ((dental implantation[MeSH Terms]) OR (Dental implants[MeSH Terms])) OR (Implant planning) #2 ((((((((low dose) OR (ultra-low dose)) OR (dose)) OR (voxel size)) OR (milliamperage)) OR (exposure time)) OR (Radiation dosage[MeSH Terms])) OR (cone beam computed tomography[MeSH Terms])) OR (tomography, X-ray computed tomography[MeSH Terms]) #3 (((Linear measurement*) OR (Measurement*)) OR (Height)) OR (Width) #1 AND #2 AND #3	1524
Web of Science	#1 (ALL=(Dental implant*)) OR ALL=(Implant planning) #2 (((((((ALL=(low dose)) OR ALL=(ultra-low dose)) OR ALL=(voxel size)) OR ALL=(dose)) OR ALL=(milliamperage)) OR ALL=(exposure time*)) OR ALL=(cone beam computed tomography)) OR ALL=(CBCT) #3 (((ALL=(Linear measurement*)) OR ALL=(Measurement*)) OR ALL=(Height)) OR ALL=(Width) #1 AND #2 AND #3	1536
Embase	(‘dental implant*’ OR ‘implant planning’) AND (‘low dose’ OR ‘ultra-low dose’ OR dose OR ‘voxel size’ OR milliamperage OR ‘exposure time’ OR ‘cone beam computed tomography’ OR cbct) AND (‘linear measurement*’ OR measurement* OR height OR width)	638
Scopus	TITLE-ABS-KEY ((“Dental Implant*” OR “Implant planning”) AND (“low dose” OR “ultra-low dose” OR dose OR “voxel size” OR milliamperage OR “exposure time” OR “cone beam computed tomography” OR cbct) AND (“Linear measurement*” OR measurement* OR height OR width)	986

After removing duplicates, two reviewers (ALEC and INRDR) performed an initial analysis independently. Article titles and abstracts were screened according to the selection criteria. After examining the abstracts, potential articles were obtained in full text and checked for inclusion criteria. In case of disagreement, a third reviewer (RSN) was subsequently consulted for the decision. Reviewers were not blinded to the authors and institutions of the records during the study selection process. Rayyan QCRI software (Qatar Computing Research Institute, Doha, Qatar)¹⁵ was used to delete duplicates.

Data extraction and outcome of interest

Full texts of the eligible studies were analysed and the data was extracted for the following information: study identification (author, year, country, type of study, sample characteristics, reference method, compared methods, and acquisition protocols); linear measurements details (performed measurements and methods, region of the measurements, CBCT slices used for the measurements, standardization method, number of examiners, viewing software); main results (intra- and inter-examiner agreement, statistical relevance of compared methods to reference method, conclusions). Both authors (ALEC and INRDR) independently screened all the studies and conducted data extraction individually. In cases of disagreement regarding data extraction, the authors reached a consensus through discussion.

Data analysis and synthesis

Narrative synthesis and statistical analysis were conducted considering the error obtained from low-dose CBCT protocols. The data underwent a rigorous examination by grouping and comparing data reported by the included studies.

Random effects model meta-analyses were conducted using the DerSimonian & Laird method¹⁶ to evaluate the error in planning measurements obtained from low-dose CBCT protocols. A conservative approach, in which the lowest-dose protocol and the highest measurement error observed was employed. The overall estimate error was calculated by pooling study-specific estimates (mean and standard error) and was expressed with their 95% CI. Negative mean values represented that CBCT underestimated the true measurements, while positive mean values represented overestimation. Statistical heterogeneity was estimated by the I-squared index (I²) among the effect sizes of individual studies. Values of the I² over 75% were classified as high, 50%-75% as moderate and less than 50% as low.¹⁷ Data analyses were performed using Stata 14.2 Software with the metan package.¹⁸

Quality of evidence

The methodologic quality of the included studies was independently assessed by the same researchers using the Quality Assessment Tool For In Vitro Studies (the QUIN) developed by Sheth et al.¹⁹ Briefly, the following criteria were critically assessed in each study: aims and objectives, sampling technique, comparison group details, detailed explanation of methodology, operator details, randomization, outcome measurement, blinding, statistical analysis and presentation of results. Scores were assigned to each study based on the following criteria: adequately specified = 2; inadequately specified = 1; and not specified = 0.

Reporting bias assessment

A single-arm forest plot was generated considering the inherent characteristics of the data obtained from the included studies, specifically the error in planning measurements obtained from low-dose CBCT protocols. However, a funnel plot for analysing publication bias was not viable. Consequently, a comprehensive literature search was undertaken to mitigate the risk of publication bias.

Results

Study selection

The search yielded a total of 4684 references from four databases: PubMed produced 1524 papers, Scopus produced 986 studies, Embase produced 638 studies, and Web of Science produced 1,536 studies. After removing duplicates, 2754 results remained for analysis. A careful reading of the titles and abstracts excluded 2737 results. The 17 remaining results were fully read, of which four were excluded due to the following reasons: (i) did not use a reference method,^20,21 (ii) did not perform different acquisition protocols, only altered voxel size on the post-processing phase²² and (iii) was in the Chinese language.²³ After consensus among reviewers, 13 studies met the inclusion criteria and 9 studies were included in the meta-analysis.^24–36 Kappa scores regarding inter-examiner agreement for title/abstract review and full-text review were greater than 0.9. A PRISMA 2009 flow diagram³⁷ regarding study selection is shown in Figure 1.

Figure 1.

PRISMA 2020 flow diagram.

General characteristics of the studies

The 13 studies included in the review were published between 2012 and 2021 and were performed in 8 different countries: Brazil,^{24,25,33–35} Iran,²⁷ Lebanon,^30,32 Saudi Arabia,²⁹ Sweden,²⁸ Thailand,²⁶ Turkey³¹ and the USA.³⁶ Four studies used dry human full skulls as samples,^26–29 four studies used dry human mandibles,^25,31,34,35 one study used dry human hemi mandibles,²⁴ two studies used full skulls of fresh cadavers,^30,32 one study used synthetic mandibles made of polyurethane³³ and one study used embalmed cadaver heads with associated intact soft tissue.³⁶ Among the 13 included studies, twelve were ex vivo studies,^{24–32,34–36} and one study was in vitro.³³

Seven studies performed CBCT acquisitions with soft-tissue simulation, using different methodologies: three studies placed the mandibles in a Styrofoam/polystyrene box filled with water,^25,34,35 one study used acrylic resin to cover the full skull,²⁶ one study placed the skulls in a water-filled plastic bag and then placed the bag inside an acrylic container,²⁸ one study used wax and acrylic surrounding the full skulls²⁹ and one study placed the mandibles inside a cylindrical plexi glass block.³¹ Three studies performed the CBCT acquisitions with soft tissue preserved on the sample.^30,32,36

For the reference method, ten studies used direct measurements performed with a digital caliper,^{24–27,29,31,33–36} one study used a standard CBCT protocol,²⁸ and two studies used high-dose CBCT protocols.^30,32 For the compared (low-dose) methods, five studies used different voxel sizes,^24–27,36 four studies used different milliamperage settings,^30–32,35 two studies used low-dose/ultra-low dose protocols,^28,33 one study used different exposure times²⁹ and one study used two different scan modes (180°/360°).³⁴ Information regarding the aim of the studies, sample size and type, reference methods, CBCT parameters, and location of the studies is provided in Table 2.

Table 2.

General characteristics of the included studies.

References	Sample size and type	Reference method and acquisition protocol (if applicable)	Compared parameter (details)	CBCT unit(s)	Acquisition protocol
Waltrick et al24	12 dry human hemimandibles	Direct measurements using a digital caliper	Voxel size (0.4, 0.3, and 0.2 mm)	i-CATa	FOV: 8 × 8 cm kVp: 120 mA: 3-8 Voxel size: 0.4, 0.3, and 0.2 mm Exposure time: 20 s (voxel sizes 0.4 and 0.3 mm)/40 s (voxel size 0.2 mm)
Torres et al25	8 dry human mandibles	Direct measurements using a digital caliper	Voxel size (0.4, 0.3, 0.25, and 0.2 mm)	i-CATa	FOV: 6 × 6 cm kVp: 120 mA: 46.72 and 23.87 Voxel size: 0.4, 0.3, 0.25, and 0.2 mm Exposure time: 20 s (voxel sizes 0.4 and 0.3 mm)/40 s (voxel sizes 0.2 and 0.25 mm)
Luangchana et al26	6 dry human full skulls	Direct measurements using a digital caliper	Voxel size (0.125, 0.16, 0.25 mm for unit 3D Accuitomo 170 and 0.2, 0.3 mm for unit CS 9500)	3D Accuitomo 170b and CS 9500c	Not reported
Moshfeghi et al27	4 dry human full skulls	Direct measurements using a digital caliper	Voxel size (0.3 and 0.15 mm)	Newtom VGd	FOV: Not reported kVp: 110 mA: 5.56 and 8.3 Voxel size: 0.4, 0.3, 0.25, 0.2 mm Exposure time: 20 s/40 s
Liljeholm et al28	8 dry human full skulls	Standard CBCT protocol Unit: ProMax 3D Classic FOV: 8 × 8 cm kVp: 90 mA: 8 Voxel size: 0.2 mm Exposure time: 12 s Rotation: single 200-degree	Protocol: ultra-low high definition (UL-HD), ultra-low normal definition (UL-MD) and ultra-low low definition (UL-LD)	ProMax 3D Classice	FOV: 8 × 8 cm kVp: 90 mA: 5.6 (UL-HD and UL-MD) and 2.5 (UL-LD) Voxel size: 0.15 (UL-HD), 0.2 (UL-MD) and 0.4 mm (UL-LD) Exposure time: 5 s (UL-HD), 4 s (UL-MD) and 3 s (UL-LD) Rotation: single 200-degree rotation
Al-Ekrish et al29	4 dry human full skulls and 1 mandible	Direct measurements using a digital caliper	Exposure time (40, 20, and 7 s)	Ilumaf	FOV: Large kVp: 120 mA: 3.8 Voxel size: 0.29 Exposure time: 40, 20, and 7 s
El Sahili et al30	2 full Skulls of fresh cadavers	High dose CBCT Unit: Carestream CS 9300 kV: 78 mA: 6.3 Voxel size: 0.09 Exposure time: 20 s	mA (2)	Carestream CS 9300c	FOV: not reported kVp: 78 mA: 2 Voxel size: 0.09 Exposure time: 20 s
Sener et al31	10 dry human mandibles	Direct measurements using a digital caliper	mA (5.2, 3.2, 6.3, 10, and 15)	Kodak 9000g	FOV: 7.5 × 3.7 kVp: 70 mA: 2, 3.2, 6.3, 10, and 15 Voxel size: 0.2 Exposure time: 32.4 s
El Sahili et al32	2 full Skulls of fresh cadavers	High dose CBCT Unit: Carestream CS 9300 kV: 78 mA: 6.3 Voxel size: 0.09 Exposure time: 20 s	mA (2)	Carestream CS 9300c	FOV: not reported kVp: 78 mA: 2 Voxel size: 0.09 Exposure time: 20 s
de Castro et al33	5 synthetic polyurethane mandibles	Direct measurements using a digital caliper	Exposure time (24 s: low-dose and 15 s: Ultra-low dose)	PaX-i3Dh	FOV: 5 × 5 cm kVp: 50 mA: 4 Voxel Size: 0.2 mm Exposure time: (24 s for low-dose and 15 s for Ultra-low dose)
Neves et al34	8 dry human mandibles	Direct measurements using a digital caliper	Rotation: half and full scan	Next Generation i-CATa	kVp: 120 mA: 20.27 (half scan) and 37.07 (full scan) Voxel Size: 0.2 mm Rotation: 180 (half scan) and 360 (full scan) Exposure time: 14.7 s (half scan) and 26.9 s (full scan)
Vasconcelos et al35	8 dry human mandibles	Direct measurements using a digital caliper	mA (2, 4, 6.3, 8, 10, 12, and 15)	Kodak 9000g	FOV 50 × 37 mm kVp: 60 mA: 2, 4, 6.3, 8, 10, 12, and 15 Voxel size: 0.2 mm Exposure time: 10.8 s
Ganguly et al36	4 cadaver heads	Direct measurements using a digital caliper	Voxel size (0.3, 0.2, and 0.16 mm)	i-CAT Classica and Planmeca Promax 3De	i-CAT Classic: FOV: 6 × 13 cm; voxel size: 0.3 and 0.2 mm; acquisition time: 20 s Promax 3D: FOV: 5 × 8 cm; voxel size: 0.16 mm; kVp: 84; mA: 14; acquisition time: 12 s

^aImaging Sciences International, Hatfield, PA, United States.

^bJ Morita Manufacturing, Tokyo, Japan.

^cCarestream Health Inc., Onex Corporation in Toronto, Canada.

^dQuantitative Radiology, Verona, Italy.

^ePlanmeca USA Inc., Roselle, IL, United States.

^fImtek Imaging, 3M Co., St Paul, MN, United States.

^gKodak Carestream Health, Trophy, France.

^hVatech, Hwaseong-si, Korea.

Regarding linear measurements, eight studies evaluated alveolar bone height and width,^{24,25,29–33,36} four studies evaluated bone height,^26,28,34,35 and one study evaluated linear distances between anatomical landmarks.²⁷ In the studies that used direct measurements as the gold standard, seven^{24,26,27,29,31,34,35} used gutta-percha for standardization of the region to be measured. One study used hollow spherical markers,²⁵ one study used carbide drill perforations,³³ and one study used stainless steel fiduciary markers.³⁶ Information regarding measurement region, CBCT slices used for measurement, standardization methods, examiners, and software details is presented in Table 3.

Table 3.

Measurement information and examiner details of the included studies.

References	Performed measurements	Measurement region	CBCT slices used for measurements	Standardization method	Number of examiners/measurement times	Software
Waltrick et al24	Total mandible height, alveolar bone height, and alveolar bone width	Region between the second premolar and the second molar	Cross-sectionals	Gutta-percha markers	2/twice	i-CAT Vision®a
Torres et al25	Bone height and width	Incisor, premolar and molar	Cross-sectionals	Hollow spherical markers	1/twice	Xoran CAT®b
Luangchana et al26	Bone height	Maxilla: incisor, premolar and molar Mandible: incisor, canine, premolar and molar	Cross-sectionals	Gutta-percha markers	2/twice	One Volume Viewer®c and CS 3D imaging®d
Moshfeghi et al27	Linear distances between anatomical landmarks	Mental foramen, mandibular foramen, lateral pterygoid plate, medial and lateral pterygoid plate, occipital condyle, greater palatine foramen, infraorbital foramen, foramen oval, foramen lacerum, anterior and posterior clinoid process	Axial and coronal	Gutta-percha markers	3/twice	NNT®e
Liljeholm et al28	Bone height	Maxila: premolar and molar Mandible: anterior, premolar and molar	Cross-sectionals	Three 2D cross-sectional sequences were constructed from each CBCT volume, oriented according to the inclination of the neighbouring roots in the region	7/twice for 3 examiners	Romexis®f
Al-Ekrish et al29	Bone height and width	Incisor, canine, premolar and molar	Cross-sectionals	Gutta-percha markers	2/twice for one examiner	IlumaVision 3-D®g
El Sahili et al30	Bone height and width	Not reported	Cross-sectionals	Not reported	10/once	CS 3D imaging®d
Sener et al31	Bone height and width	Anterior and posterior regions the mandible	Axial and sagittal	Gutta-percha markers	3/once	Kodak Dental Imaging Software®h
El Sahili et al32	Bone height and width	Not reported	Cross-sectionals	Not reported	5/once	CS 3D imaging®d
de Castro et al33	Bone height and width	Between the mental foramina	Cross-sectionals	Perforations with a carbide drill	1/twice on 50% of the sample	CS 3D imaging®d
Neves et al34	Bone height	Incisor, canine, premolar, first molar and second molar	Cross-sectionals	Gutta-percha cone	2/twice on 50% of the sample	OnDemand3®i
Vasconcelos et al35	Bone height	incisor, canine, premolar, first molar and second molar	Cross-sectionals	Gutta-percha cone	3/once	3D- KDIS®d
Ganguly et al36	Bone height and width	First and second premolars, first and second molars	Cross-sectionals	Stainless steel fiduciary markers	2/twice on 2 of the specimens	Xoran CAT®b and Planmeca Romexis®f

^aImaging Sciences International, Hatfield, PA, United States.

^bXoran technologies, Ann Arbor, MI, United States.

^cJ Morita Manufacturing, Tokyo, Japan.

^dCarestream Health, Onex Corporation in Toronto, Canada.

^eNewtom, Quantitative Radiology, Verona, Italy.

^fPlanmeca USA Inc., Roselle, IL, United States.

^gImtek Imaging, 3M Co., St Paul, MN, United States.

^hKodak Carestream Health, Trophy, France.

ⁱCyberMed Inc, Seoul, Korea.

Overall, there was no consistency in the use of statistical methods to compare low-dose CBCT protocols to the reference method, entailing a mix of parametric and non-parametric tests, correlation coefficients, and techniques for estimating reliability and agreement. Two studies^25,33 did not have multiple examiners performing the linear measurements. Therefore inter-examiners correlation was not assessed. The examiners of three studies^30–32 did not perform a duplicate assessment of linear measurements. Consequently, intra-examiner correlation was not evaluated.

In the analysis of measurements and inter-observer or intra-observer reliability, various statistical methods were employed across the studies. Pearson^24,31 and Intraclass correlation coefficients (ICC)^{26–28,33–36} were used to assess the difference among the measurements and inter-observer or intra-observer reliability, respectively. For comparing means of paired or single-sample data, authors resorted to paired t-tests^24,26,28 or 1-sample t-tests^29,33 in five studies. Furthermore, the Wilcoxon signed-rank test was a frequently used non-parametric method to test differences in measurements.^{25,29,30,32,34,36} To discern differences in linear measurements among multiple groups, researchers often turned to one-way ANOVA,^{24,29,31,33,35} and this was frequently followed by post hoc tests like Dunn²⁵ or Tukey.³⁵ Similarly, the Friedman test^25,29,36 and Kruskal-Wallis test²⁵ were used as non-parametric alternatives for comparing measurements across groups. Three studies^25,30,32 used Lin concordance to assess reproducibility or agreement between measurements. Additionally, error calculation^24,25,29 and Bland-Altman plots^31,36 were also used for measurement accuracy and agreement between methods or observers.

The results indicated a strong consistency in linear measurements performed on low-dose images and their correlation with measurements performed on the reference methods. Various studies found no significant differences in measurements across different protocols (eg, voxel sizes, mA settings, or dose levels), regions (incisor, premolar, molar), or types (height vs. width). Some studies, however, noted exceptions in measurements performed on the posterior mandible.^24,34,35 The mean error values between image-based and direct measurements were typically sub-millimetric, suggesting high precision. Inter-observer and intra-observer differences were generally small, with some larger discrepancies observed for larger measurements. Overall, these studies suggest that the accuracy, reliability, and reproducibility of radiographic measurements using different low-dose protocols are high. Details regarding statistical tests, main results and conclusions are provided in Table 4.

Table 4.

Statistical analysis, main results and conclusions of the included studies.

References	Main statistical tests performed	Main results	Intra-examiner assessment	Inter-examiner assessment	Conclusions
Waltrick et al24	PCC (difference among the measurements and among examiners); Paired Student t-test (difference in measurement errors between the two examiners); One-way ANOVA (difference among measurement errors comparing the three voxel sizes)	Strong positive correlation between the measurements made on the images and the direct measurements (r > 0.99); Inter-examiner difference for alveolar ridge width, at the regions of the second molar (P = .0001) and first molar (P = .0028); Overall, no differences in the accuracy among voxel sizes, except for alveolar bone height at the first molar region for one of the examiners	PCC >0.99	Difference for alveolar ridge width, in the second molar region (P = .0001) and first molar region (P = .0028)	Voxel sizes of 0.2, 0.3, or 0.4 mm are adequate for linear measurements in the posterior region of the mandible, and 0.3 mm is a good compromise between image quality and low radiation dose
Torres et al25	ME was calculated (subtracting the value of direct measurement from that of tomographic images); Friedman, Kruskal-Wallis, and Wilcoxon tests followed by the Dunn test (ME comparison between protocols, for general evaluation, by site and by measurement); LCC (reproducibility of the measurements between the evaluations and the gold standard)	No difference between the ME for all protocols in the overall assessment (P = .606); No difference between the ME of the regions: incisor, premolar and molar for all protocols (P-value ranging from .1696 to .6516); Significant differences were found between the ME for vertical and horizontal measurements on all protocols (P-value ranging from .0066 to .0276); Mean value of the difference between measurements from the images and the dry mandible varied from 0.68 to 0.72 mm on all protocols	LCC ranging from 0.960 to 0.986	Not applicable	Accuracy of vertical and horizontal measurements was shown to be comparable with the measurements performed on the dry mandible. Protocols with voxel sizes of 0.3 and 0.4 mm must be preferably indicated in the evaluation of the linear measurements for planning of dental implants treatment, because the radiation dose is reduced
Luangchana et al26	Paired sample correlation (correlation between each protocol and the physical measurements); Paired difference t-test (difference between each protocol and the physical measurements); ICC confirmed by Cronbach alpha (intra- and inter-examiner reliability)	Paired sample correlation, ranging from 0.903 to 0.998, showed statistically significant correlations (P < .05); There were no differences between any of the radiographic protocols and the direct physical measurements of the bone	ICC ranging from 0.996 to 1.000	ICC ranging from 0.991 to 1.000	3D Accuitomo 170 and CS 9500 CBCT units have sufficient accuracy for linear measurement of alveolar bone height. Voxel size does not affect the accuracy of linear measurements
Moshfeghi et al27	ICC (accuracy of the radiographic measurements, inter-examiner and intra-examiner reliability)	ICC ranged from 0.9935 to 0.9944, showing that the radiographic measurements were accurate (P > .05); No difference was found between the radiographic measurements and real measurements (P > .05); ICC for each individual landmark ranged from 0.7407 to 0.9996, showing no difference between radiographic measurements and real measurements (P > .05)	ICC ranging from 0.9849 to 0.9998	ICC ranging from 0.9991 to 0.9996	Newtom VG unit is highly accurate and reproducible in linear measurements. A CBCT scan with a larger voxel size (0.3 mm in comparison with 0.15 mm) is recommended for the measurement of linear distances. This will result in lower patient radiation dose and faster scan time
Liljeholm et al28	Paired t-test (differences in examiner means between the four examined protocols); ICC (inter-rater reliability based on seven examiners and intra-rater reliability based on three examiners)	No difference between any of the ultra-low dose protocols and the standard protocol (Sig (2-tailed) P > .673)	ICC ranging from 0.942 to 0.982	ICC ranging from 0.978 to 0.989	Low-dose protocols may be applied for pre-implant radiographic assessment. However, image quality can be hampered if the radiation exposure is too low and the voxel size too large. The UL-HD and UL-MD protocols were preferred to the Standard protocol, due to a reduction of radiation dose
Al-Ekrish et al29	Error (CBCT image mean minus the mean direct bone measurement); 1-sample t-test (mean of the absolute errors of the overall measurements for each protocol); Wilcoxon signed rank test (mean of the absolute errors for height measurements at sites for each protocol); Repeated measures ANOVA (difference between the means of the absolute errors obtained by the different protocols for the entire data set); Friedman test (difference between the means of the absolute errors for height measurements at the posterior mandible); Correlation testing (Intra- and inter-examiner reliability)	Although the means of the absolute errors were sub millimetric for all protocols, they were statistically significant for the entire sample and for the height measurements at the posterior mandible; No significant difference (P = .539) between the means of the absolute errors obtained with the three examination protocols for the entire sample; No significant difference (P = .856) between the means of the absolute errors for the subset of height measurements at the posterior mandible	Cronbach Alpha ranging from 0.996 to 0.999	Inter-item correlation matrix ranging from 0.993 to 0.999	Lowering CBCT exposure time from 40 to 20 s does not adversely affect the reliability or accuracy of implant site measurements. The possibility of a 1-mm overestimation of measurements should be considered with all three examination protocols, and correction should be performed accordingly
El Sahili et al30	Wilcoxon signed rank test (differences in measurements between the two image qualities for each examiner in each of the two panels); LCC (variability between examiners)	In the juniors panel almost all the differences were not significant (P = .034 on one measurement performed by one examiner); In the senior panel there were no differences between measurements	Not reported	LCC ranging from 0.5 to 0.9 (juniors panel) and >0.9 (senior panel)	There is significant potential for the reduction of patient radiation exposure by reducing milliamperage when using CBCT imaging for the planning of dental implant placement. Seniors and juniors alike showed similar performance on high- and low-dose images in recording linear quantitative measurements
Sener et al31	Bland-Altman plots (comparison of height and width measurements from each of the mA settings to the gold standard); Repeated measures ANOVA (overall comparison of the measurements); PCC (correlation between individual examiners)	No differences were found in the height and width measurements for both anterior (P > .05) and posterior (P > .05) segments of the mandible for all of the five different mA settings. Increase in mA caused a decrease, although not significant, in the geometric accuracy of CBCT images; Bland-Altman plots showed an equal distribution around the mean, with most points lying very close to the mean, within the 2 standard deviations of the mean difference, in most mA settings	Not reported	PCC ranging from 0.97 and 0.98, with no difference for increasing tube settings (P > .05)	It is possible to lower the tube current of the Kodak 9000 3D CBCT system as low as 2 mA and still maintain linear measurement accuracy
El Sahili et al32	Wilcoxon signed Rank test (differences in measurement between the two images qualities); LCC (variability between examiners)	No differences between measurements recorded from low dose images and those recorded from high dose images (P > .05)	Not reported	LCC >0.9	Patient radiation exposure may be significantly reduced without affecting diagnostic image quality when using CBCT imaging for implant planning, by reducing milliamperage. Planning implant placement in the posterior mandible may require higher image dosage, but still in the lower range of mA settings provided by CBCT machines
de Castro et al33	One-way ANOVA (comparison of the measurements of the three groups: reference standard, low dose [L] and ultralow dose [UL]); Student t-test (comparison of the measurements obtained with the different acquisition protocols to the reference standard); ICC (reproducibility of the analyses)	No differences for the measurements of bone height (P = .9986) and thickness (P = .7764); The L and UL protocols did not differ from the reference standard regarding bone height and thickness; There was no difference between the measurements obtained with the L and UL protocols for bone height and thickness	ICC =0.9878	Not applicable	Low-dose radiation protocols can be used for precise measurements of bone height and thickness in implantology, without compromising clinical planning
Neves et al34	Wilcoxon signed rank test (correlation between the measurements obtained in different scan mode with the gold standard); ICC (intra- and inter-examiner agreement)	No difference in the dental implant measurements compared to the gold standard, except in bone height of the second molar region in full scan mode (P = .02)	ICC ranging from 0.96 to 0.98	ICC =0.98	Both modes (half and full scan) provided real measures. Half scan mode uses smaller radiation doses, thus offering the best dose–effect relationship with less risk to the patient
Vasconcelos et al35	One-way ANOVA (effect of the different milliamperage settings on the linear measurements); Post hoc Tukey test (when ANOVA detected differences between the groups); ICC (intra- and inter-examiner reproducibility)	No differences in the accuracy of the measurements were observed between the milliamperage settings analysed (P > .05), except in the second molar region (P = .008); Tukey test showed that milliamperage settings did not influence their accuracy, however, images obtained with 6.3 mA were closest to the real measurements	ICC ranging from 0.81 to 0.99	ICC ranging from 0.91 to 0.98	The evaluated milliamperage settings did not influence the objective evaluation of the images using measurements and, although they influenced image quality, this influence was limited to values less than 6.3 mA. Higher milliamperage values generally gave images of similar quality. Thus, the use of higher milliamperage settings does not seem to be necessary
Ganguly et al36	Wilcoxon signed-rank test (determine whether the accuracy of measurements made by the CBCT protocols was comparable to the physical measurements); Friedman test (determine whether a difference was present in the accuracy among the 3 different protocols); ICC (intra- and inter-examiner variability); Bland-Altman plot (inter-examiner variability)	No significant differences were found in the absolute values of the differences between the physical measurements and CBCT measurements (p > 0.05); No differences were found between the medians of the physical measurements and the medians of any of the CBCT measurements made by either examiner (P > .05); Bland-Altman plot indicated very good reproducibility between examiners for smaller measurements. As the measurements became larger, larger discrepancies emerged between the examiners	ICC ranging from 0.978 to 0.985	ICC =0.961	All imaging protocols were reliable for linear measurements. It can be reasonably concluded that a 0.3 mm voxel size is sufficient for the purpose of implant site assessment without exposing the patient to an additional radiation dose

Abbreviations: PCC = Pearson correlation coefficient, ANOVA = analysis of variance, ME = measurement error, LCC = Lin concordance coefficient, ICC = intraclass correlation coefficient.

Risk of bias and meta-analysis

Five studies were classified as presenting a low risk of bias (>70% QUIN Tool score),^{24–26,31,35} and eight studies were classified as presenting a medium risk of bias (50%-70% QUIN Tool score).^{27–30,32–34,36} All studies adequately described their aims/objectives, details of the comparison group, and results.^24–36 Only one study reported sample size calculation.²⁴ Sampling technique and information regarding inclusion and exclusion criteria were reported in detail in five studies.^{24–26,31,33} Two studies^30,32 did not clarify the method of standardization/method of measurement of outcome used to guarantee that the same regions were measured on the different CBCT protocols.

Details regarding examiner calibration were provided in one study.³⁵ Four studies reported sample randomization, but details regarding sequence generation and allocation concealment were not provided.^24,28,30,32 Five studies reported blinding of examiners.^{24–26,31,35} One study²⁵ did not provide details of the software used for statistical analysis. Table 5 contains the QUIN Tool scores for the 13 studies.

Table 5.

Risk of bias assessment using the QUIN Tool.

References	Waltrick et al24	Torres et al25	Luangchana et al26	Moshfeghi et al27	Liljeholm et al28	Al-Ekrish et al29	El Sahili et al30	Sener et al31	El Sahili et al32	de Castro et al33	Neves et al34	Vasconcelos et al35	Ganguly et al36
Clearly stated aims/objectives	2	2	2	2	2	2	2	2	2	2	2	2	2
Detailed explanation of sample size calculation	1	0	0	0	0	0	0	0	0	0	0	0	0
Detailed explanation of sampling technique	2	2	2	1	1	1	1	2	1	2	1	1	1
Details of comparison group	2	2	2	2	2	2	2	2	2	2	2	2	2
Detailed explanation of methodology	2	2	2	2	2	2	1	2	1	2	2	2	2
Examiner details	1	1	1	1	1	1	1	1	1	1	1	2	1
Randomization	1	0	0	0	1	0	1	0	1	0	0	0	0
Method of measurement of outcome	2	2	2	2	2	2	0	2	0	2	2	2	2
Blinding	2	2	2	0	0	0	0	2	0	0	0	2	0
Statistical analysis	2	1	2	2	2	2	2	2	2	2	2	2	2
Presentation of results	2	2	2	2	2	2	2	2	2	2	2	2	2
Score	19	16	17	14	15	14	12	17	12	15	14	17	14
%	86.36	72.72	77.27	63.63	68.18	63.63	54.54	77.27	54.54	68.18	63.63	77.27	63.63
Risk of bias	low	low	low	medium	medium	medium	medium	low	medium	medium	medium	low	medium

Nine studies were included in the meta-analysis.^{24,26,27,29,31,33–36} The variation in the number of incorporated reports is due to the absence of mean data, which was not provided in the remaining four studies.^25,28,30,32Table 6 provides details on the source of maximum error used in the analysis. The overall pooled error in planning measurements obtained from low-dose CBCT protocols was −0.24 mm (95% CI, −0.52 to 0.04) (Figure 2). In addition, the meta-analysis exhibited remarkable levels of heterogeneity, reflected by the I² statistic of 99.9%.

Table 6.

Maximum error by lowest dose protocol in each study.

References	Lowest dose protocol	Source of maximum error in implant planning measurements	Sample size (region of maximum error)	ME (mean)	ME (SD)	Variancea	ME (SE) b	Under or overestimation
Waltrick et al24	Voxel size: 0.4 Exposure time: 20 s	Alveolar ridge width on the buccal-lingual aspect in the second molar region, measured by observer 1	12	−0.54	0.15	0.243	0.043	Underestimation
Luangchana et al26	Voxel size: 0.3	Maxilla (absolute error)	72	−1.2	0.14	0.02	0.016	Underestimation
Moshfeghi et al27	Voxel size: 0.3 Exposure time: 18 s	Coronal images	180	−0.28	0.20	0.000	0.014	Underestimation
Al-Ekrish et al29	Voxel size: 0.29 Exposure time: 7 s	Overall sample	83	0.51	0.06	0.313	0.06	Overestimation
Sener et al31	Voxel size: 0.2 Exposure time: 32.4 s mA: 2	Posterior mandible height	60	−0.2	0.02	0.666	0.002	Underestimation
de Castro et al33	Voxel size: 0.2 Exposure time: 15 s mA: 4	Bone thickness in all regions	40	0.52	0.08	0.676	0.012	Overestimation
Neves et al34	Voxel size: 0.2 Exposure time: 14.7 s (half scan) mA: 20.27	Mandibular canal height in the second molar region	32	−0.3	0.05	0.002	0.008	Underestimation
Vasconcelos et al35	Voxel size: 0.2 Exposure time: 10.8 s mA: 2	Canine region	48	0.79	0.11	0.130	0.015	Overestimation
Ganguly et al36	Voxel size: 0.3 Exposure time: 20 s	Width of the bone at the alveolar crest at the right maxillary molar site, measured by observer 1	8	−1.83	0.65	0.418	0.229	Underestimation

Abbreviations: ME = measurement error, SD = standard deviation, SE = standard error, and Variance are all shown in millimeters.

^aVariance was calculated using the package scalar variance (Stata software version 14.3) based on the mean and sample size. In this study, the sample size was calculated by multiplying the number of examiners, the number of sites and the number of measurements.

^bSE was calculated from the standard deviation (SD) and sample size (n) using the formula: SE = SD/$\sqrt{n}$.

Figure 2.

Random effect meta-analysis of the error in planning measurements obtained from low-dose CBCT protocols in comparison with those performed on standard and high-resolution CBCT protocols. Study data, mean, 95%CI, weights, heterogeneity I², and overall effect statistics are shown.

Discussion

Accurate surgical planning before implant placement should include a thorough assessment of alveolar bone conditions, such as a qualitative assessment of bone quality and a quantitative assessment of bone dimensions, including alveolar height and width.^38,39 CBCT has become a standard imaging modality for pre-implant bone assessment due to its ability to provide multi-planar images with submillimeter resolution. A recent systematic review analysed the accuracy of CBCT images for alveolar bone measurements of height and thickness, showing that the difference between real measurements (physical direct measurements) and CBCT measurements was not significant for either height or width, indicating that CBCT offers good accuracy.³⁸ However, the image quality of CBCT scans can be influenced by various factors, including the scanning unit, FOV, examined object, exposure time, tube voltage, amperage, and voxel size.^40,41

Low-dose CBCT protocols have emerged as a recent advancement, leading to reduced radiation exposure to patients.⁴² These protocols are made possible by hardware and software enhancements, including more sensitive detectors and improved reconstruction algorithms, as well as reduced exposure time²⁸ or reduced amperage. However, it is important to acknowledge that a consequence of reducing radiation dose is lower image quality.⁸

The present review systematically assessed the accuracy of linear measurements performed on low-dose CBCT images for implant planning. The selected studies compared these measurements to either direct physical measurements or higher-dose CBCT protocols, which served as the reference method. Since Standard and High-Resolution CBCT protocols have demonstrated accuracy for linear measurements,³⁸ they were considered as reference methods in this review, but the term “gold standard” was not applied in these cases. Ten studies included in the present review compared low-dose protocols to direct measurements,^{24–27,29,31,33–36} which can be considered as the gold standard reference method.

Overall, the results of the 13 studies included in this review showed no statistical difference between linear measurements performed on low-dose CBCT images and the reference method. However, in four studies^24,30,34,35 some statistically significant differences were observed, mainly related to specific anatomical regions^24,34,35 or among different observers.³⁰ El Sahili et al³⁰ reported that the experience of the observer is a factor that interfered with the results. They observed that junior observers exhibited less agreement when there was an increase in radiation dose, suggesting that aptitude and interest in implant dentistry impacted image evaluation. The studies that found differences regarding the measurement region^24,34,35 attributed it to the partial volume effect^24,41 and to possible bone damage caused by sectioning with an electric saw.³⁴ Vasconcelos et al³⁵ reported that the visibility of the posterior region of the mandible presents more significant degradation than the anterior region due to increased bone density and that lower milliamperage may have resulted in greater image noise.

Moreover, according to Waltrick et al,²⁴ despite the voxel size interfering with image quality, voxels of larger sizes are still reliable for carrying out linear measurements. Neves et al³⁴ reported that the differences found between half and full scanning modes were smaller than 1 mm, proving that both half scan and full scan can be clinically useful in the preoperative evaluation for implant placement, since many authors have previously reported that bone measurements presenting errors smaller than 1 mm can be tolerated in CBCT imaging.^25,43,44 Given these considerations, we emphasize that even though some regions showed statistical differences in measurements, these differences do not seem to be clinically relevant.

Regarding the meta-analyisis, the overall pooled measurement error from low-dose CBCT protocols was −0.24 mm (95% CI, −0.52 to 0.04), showing that low-dose CBCT had a tendendy to underestimate the true values. Two studies presented mean measurement error larger than 1 mm,^26,36 while the other seven studies ranged between −0.2 and 0.79 mm. Luangchana et al²⁶ attributes the larger error found in maxilla to the difficulty in distinguishing anatomic structures on the sectional images of the maxillary bone, which is lined with a thin layer of cortical bone. Ganguly et al³⁶ reports that the embalming fluid on tissues may have been at least partially responsible for the reduced accuracy of the measurements, claiming that better accuracy may be expected when imaging patients.

The high heterogeneity of this meta-analysis can be attributed to the fact that different exposure parameters (ie, voxel size, exposure time, mA), anatomical regions and standardization techniques were compared in order to guarantee that the largest measurement error in the lowest-dose protocol were considered for each study. Even though heterogeneous, these results support the evidence that linear measurements performed on low dose images can offer adequate accuracy.

In the present review, twelve studies were classified as ex vivo studies^{24–32,34–36} since they used dry human mandibles, hemimandibles, dry full skulls, or embalmed cadaver heads as sample materials. Conversely, only one study was classified as in vitro since it used synthetic polyurethane mandibles as a sample, a material that offers similar density to bone tissue.³³ Despite the differences in classification, all studies were conducted in a controlled environment outside of a living organism, enabling the application of the QUIN risk of bias assessment tool. This tool allows for the exclusion of criteria that are not relevant to a specific set of studies, making it a universally applicable tool for dental in vitro studies.¹⁹

There was a considerable variation in the methods used to standardize the regions where the linear measurements were performed. Considering the studies that used direct measurements as the gold standard, different markings that could be detected on CBCT images were added to the sample, such as: gutta-percha markings,^{24,26,27,29,31,34,35} hollow spherical markers,²⁵ carbide drill perforations³³ and stainless steel fiduciary markers.³⁶ However, in the three studies that used higher-dose CBCT protocols as the reference standard, this type of standardization was not performed. Consequently, they had to adopt different methods to ensure that the different observers on various CBCT protocols were measuring the exact same region.

Liljeholm et al²⁸ reported that the 2D cross-sectional sequences that were constructed from each CBCT volume were oriented according to the inclination of the neighbouring roots in the region. Notwithstanding, it was observed that the alignment of one cross-sectional sequence in the standard CBCT images deviated from those of the other three protocols. As a result, the incorrectly aligned measurements from the standard protocol and the corresponding measurements from other protocols had to be excluded from the statistical analysis. This indicates that the standardization method did not guarantee that the same regions were measured on the different CBCT protocols. Regarding the other two studies^30,32 that did not performed standardization on the sample before CBCT acquisitions, there was no mention of how they guaranteed the selection of the exact same region for measurement on both the low-dose CBCT protocol and the reference CBCT protocol. This is an essential methodological detail that was not addressed and may affect the accuracy and comparability of the results.

Three studies did not report performing soft-tissue simulation on the sample,^24,27,33 and even though it is known that there could be difference on measurements performed on dry skulls and on live patients, due to the soft tissues affecting the attenuation coefficient and X-ray beam transmission through the skull,⁴⁵ the accuracy results of these three studies^24,27,33 were similar to those performed with soft-tissue simulation^{25,26,28,29,31,34,35} and to those that performed CBCT acquisitions with soft tissue preserved on the sample.^30,32,36

Two studies^20,21 were excluded during the screening phase because they lacked a reference method for comparison. Although the authors did not directly compare specific low-dose protocols to a gold standard, they conducted comparisons of various milliamperage and kilovoltage settings amongst themselves. In one of these studies, Panmekiate et al²¹ found no significant differences in linear measurements among different radiographic parameter combinations. They concluded that low peak kilovoltage and milliampere values could be employed for linear measurements in the posterior mandible. On the other hand, Alawaji et al²⁰ concluded that since changes in the exposure parameters affect radiation dose but do not necessarily produce better images, the image quality could be optimized for implant treatment planning at lower exposure settings and dose.

A limitation of the present systematic review is that only quantitative data regarding linear measurements was systematically assessed, not taking into consideration the diagnostic image quality associated with low-dose protocols. Four studies^28,30,32,35 included in this review also evaluated subjective bone quality. Liljeholm et al²⁸ compared three low-dose protocols on CBCT unit ProMax 3D Classic regarding overall image quality and precision of anatomical landmarks, concluding that the ultra-low low definition protocol was not diagnostically acceptable for pre-implant assessment, since the bone quality was significantly inferior compared to the standard protocol. Two studies performed by El Sahili et al^30,32 assessed image quality based on evaluators with varying experience levels. They found that the image quality assessment for almost all observers was independent of each other, but increased dosage improved concordance and kappa values, resulting in higher consistency. Vasconcelos et al³⁵ performed a subjective image quality evaluation of anatomical structures using a 4-point scale. They concluded that there was a negative influence of the milliamperage setting only when the lowest values (2 and 4 mA) were used, which resulted in poor-quality images.

These four studies^28,30,32,35 demonstrated that although there were no statistically significant differences in linear measurements, there was a point where reducing radiation dose was detrimental to image quality, resulting in images that would impair diagnostic accuracy. Thus, a balance must be reached in regards to diagnostic needs and radiation dose, depending on each patient. In cases where conventional radiographs can sufficiently provide information regarding bone quality,⁴⁶ lower-dose CBCT protocols could be suitable for linear measurements. However, for patients with significant periodontal diseases, precise implant planning requirements³⁸ or possible bone lesions needing thorough assessment, lower-dose CBCT protocols may not offer sufficient diagnostic accuracy.

Across the 13 studies, the main conclusions underline the adequate precision and accuracy of linear measurements performed using various low-dose CBCT protocols. The findings suggest that voxel sizes ranging from 0.2 to 0.4 mm are sufficient for accurate linear measurements, even in regions such as the posterior mandible. Importantly, several studies emphasize that employing lower-dose protocols or lower milliamperage settings does not compromise the reliability or accuracy of measurements, making them preferable for reducing patient radiation exposure. This includes the use of larger voxel sizes, reduced exposure times, and lower tube currents.

Conclusion

Current evidence shows that linear measurements performed on low-dose CBCT protocols are not statistically different from those performed on higher-dose CBCT protocols or directly on the sample (gold standard), suggesting that low-dose CBCT protocols can provide accurate linear measurements for implant planning. Notwithstanding, image quality and diagnostic needs must be taken into consideration when selecting low-dose CBCT protocols.

Funding

This study was financed by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—Brasil (CAPES).

Conflicts of interest

The authors deny any conflicts of interest related to this study.