Optimizing CBCT acquisition for detecting non-displaced dentoalveolar fractures: effects of field of view and resolution mode

Abstract

Background

To evaluate the diagnostic accuracy of cone-beam computed tomography for detecting non-displaced dentoalveolar process fractures using an ex vivo sheep head model, and to compare acquisition protocols differing in field of view (FOV) and manufacturer-defined resolution mode.

Methods

Thirty sheep heads underwent baseline cone-beam computed tomography to confirm the absence of pre-existing fractures, followed by controlled mechanical loading. Fracture status (present/absent) was established by direct surgical inspection after imaging. Each specimen was scanned using four predefined protocols combining two FOV settings (8 × 16 cm and 18 × 16 cm) and two resolution modes (standard and high), with voxel size varying across protocols. Four blinded experts assessed fracture presence (binary decision) and confidence (five-point scale); repeat readings after one month were used to estimate intraobserver agreement. Diagnostic discrimination was assessed using the area under the receiver operating characteristic curve derived from confidence scores, while sensitivity and specificity were calculated from binary decisions.

Results

All protocols showed high discrimination (area under the curve, 0.817–0.917), with the best performance for the 18 × 16 cm high-resolution protocol (sensitivity 83.3%; specificity 100.0%). Within the same FOV, high-resolution imaging improved discrimination compared with standard-resolution imaging; however, no meaningful difference was observed between the two high-resolution protocols. Interobserver agreement was substantial (Fleiss kappa 0.67–0.84) and intraobserver agreement was substantial to almost perfect (Cohen kappa 0.605–1.000).

Conclusions

Spatial resolution was the dominant driver of diagnostic performance and reader confidence for non-displaced dentoalveolar fractures, whereas enlarging the FOV provided limited additional benefit once high-resolution imaging was used.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12903-026-09192-8.

Introduction

Dentoalveolar fractures, although a significant component of maxillofacial trauma, are not adequately represented in current classification systems [1, 2]. These injuries, reported to occur in 11–30% of children and 5–20% of adults, can lead to functional and aesthetic complications by affecting the teeth, alveolar bone, and surrounding soft tissues [2, 3]. One of the greatest challenges in clinical practice is recognizing non-displaced dentoalveolar process fractures. Their presentation with minimal clinical signs, together with the limited contrast resolution and superimposition inherent to two-dimensional panoramic and periapical imaging, can result in delayed diagnosis and treatment [4–9].

Studies published over the past decade demonstrate that cone-beam computed tomography (CBCT) offers higher sensitivity and overall accuracy than two-dimensional techniques for multiple components of dentoalveolar trauma. In anterior maxillary trauma, CBCT has shown clear superiority over conventional radiography for detecting total traumatic dental injuries, lateral/extrusive luxations, and alveolar cortical fractures. For example, for cortical plate fractures, sensitivity with CBCT was reported as 58% compared with 1% for two-dimensional radiography [9]. Pediatric series likewise indicate that low-dose CBCT outperforms periapical radiographs in identifying root and bone fractures [10].

Nevertheless, the literature contains a notable paucity of comprehensive studies on detecting non-displaced dentoalveolar fractures with CBCT in maxillofacial trauma. An in vitro study reported that, even for non-displaced cases, CBCT sensitivity can be limited depending on the scenario, with low interobserver agreement—underscoring the difficulty of diagnosing lesions without a distinct fracture line and the dependence on acquisition protocol (voxel size/resolution, FOV) [11]. From a technical perspective, FOV and resolution may influence fracture detection through related but distinct mechanisms. Non-displaced dentoalveolar fractures often appear as thin, low-contrast cortical or trabecular discontinuities without a clearly separated fracture gap. Higher-resolution protocols with smaller voxel sizes may reduce partial-volume averaging and improve depiction of these fine fracture lines, although this may be accompanied by increased image noise. FOV selection, in turn, determines whether the suspected injury region is adequately included within the scan volume and, on many CBCT systems, co-varies with voxel size and exposure settings. Larger volumes may also increase scattered radiation and artefacts or alter the position of the region of interest within the reconstructed volume, thereby affecting image contrast and fracture conspicuity. Therefore, both resolution mode and FOV may alter the visibility of subtle dentoalveolar fracture lines and warrant systematic evaluation. This conflicting and somewhat limited evidence base highlights the need for targeted research, particularly to optimize resolution and FOV so as to reduce the risk of missed non-displaced dentoalveolar process fractures. Recent trauma-focused reviews reinforce CBCT’s role in maxillofacial trauma and emphasize the importance of protocol standardization and indication-specific imaging [12].

The primary aim of this study is to evaluate the diagnostic accuracy of CBCT for detecting non-displaced dentoalveolar process fractures induced by standardised mechanical loading in an ex vivo sheep head model. Our secondary objectives are to: (1) compare the effects of high- versus standard-resolution (SR) CBCT protocols on sensitivity, specificity, and overall diagnostic accuracy; (2) compare diagnostic performance between two FOV settings, namely 8 × 16 cm and 18 × 16 cm, referred to here as smaller- and larger-FOV protocols, respectively; and (3) examine interobserver agreement. Our hypothesis is that higher-resolution (HR) CBCT protocols would improve diagnostic accuracy and reader confidence for detecting subtle non-displaced dentoalveolar process fractures, while the incremental benefit of increasing FOV would depend on the balance between anatomical coverage and protocol-related image quality factors.

Materials and methods

Study design and setting

This was an ex vivo diagnostic accuracy study using fresh sheep heads. Thirty fresh sheep heads (bred for human consumption) were obtained from a local butcher (Istanbul, Türkiye) on the day of imaging and sequentially numbered for identification purposes (ID001–ID030). Specimens were kept moist with saline-soaked gauze and stored at 4 °C; all procedures were completed within 24 h of procurement to minimise tissue desiccation. Baseline CBCT scans were acquired before mechanical loading to confirm the absence of pre-existing fractures; baseline datasets were not included in diagnostic accuracy calculations. This report follows STARD 2015 principles where applicable for diagnostic accuracy studies [13]. No live animals were used; specimens were obtained post-mortem from animals bred for human consumption; therefore institutional animal ethics approval was not required.

Specimens and fracture creation protocol

Mechanical loading was applied to the maxillary premolar dentoalveolar region to simulate clinically relevant, non-displaced dentoalveolar trauma. A single experienced oral and maxillofacial surgeon (Z.A.A.A.) performed all loading procedures to ensure procedural consistency. Premolar extraction forceps were used with gauze padding on both beaks to standardise force transmission and minimise local soft-tissue damage. The forceps were positioned so that one beak contacted the buccal/vestibular aspect and the other contacted the palatal aspect of the maxillary premolar dentoalveolar segment. Controlled bucco-palatal loading was then applied gradually until a loss of resistance or subtle dentoalveolar disruption was clinically suspected. A schematic illustration of the fracture creation protocol is provided in Supplementary Figure S1.

Immediately after each loading attempt, the alveolar segment was assessed by visual inspection, palpation, and mobility testing. A fracture was considered non-displaced for the purposes of the experimental model when internal dentoalveolar disruption was suspected. At the same time, the buccal and palatal cortical plates remained anatomically aligned, with no visible separation of the dentoalveolar segment, cortical step-off, gross displacement, or tooth avulsion. Definitive fracture status was not assigned at this stage; fracture presence or absence was established only after CBCT acquisition by direct surgical inspection as the reference standard. No specimen included in the final dataset demonstrated macroscopic displacement, cortical separation, or tooth avulsion.

CBCT acquisition, reconstruction, and reading conditions

CBCT imaging was performed using a Scanora^® 3D-X unit (Soredex, Tuusula, Finland). Sheep heads were positioned upright on a transparent support and stabilised; alignment was verified using scout views before exposure.

Each specimen was scanned using four clinically available protocols combining two FOV settings (8 × 16 cm and 18 × 16 cm) and two resolution modes (standard and high), with protocol-specific voxel sizes as follows: 8 × 16-SR (voxel = 0.35 mm), 8 × 16-HR (voxel = 0.15 mm), 18 × 16-SR (voxel = 0.5 mm), and 18 × 16-HR (voxel = 0.3 mm).

The corresponding manufacturer-reported exposure times and dose–area product (DAP) values were as follows: 8 × 16-SR, 2.4 s / 925.61 mGy·cm²; 8 × 16-HR, 4.0 s / 1542.68 mGy·cm²; 18 × 16-SR, 3.2 s / 1523.14 mGy·cm²; and 18 × 16-HR, 3.6 s / 1713.13 mGy·cm². For all protocols, tube voltage was 90 kVp and tube current was 10 mA; rotation angle and exposure mode (pulsed/continuous) followed manufacturer defaults for each preset. Reconstructed volumes were generated using the manufacturer’s software and evaluated in OnDemand3D^® (Cybermed, Seoul, Republic of Korea).

Images were reviewed on diagnostic-grade monitors (Beacon HL2416SH; 1920 × 1200) in a dimly lit room. Multiplanar reconstructions were available, and readers were permitted to adjust window/level, magnification, and slab thickness to reflect routine diagnostic practice (Fig. 1).

Fig. 1.

Representative CBCT multiplanar reconstructions (MPR) of a non-displaced dentoalveolar fracture. A High-resolution (HR) imaging improves visualization of the subtle fracture line compared with (B) standard-resolution (SR) imaging. Images were evaluated in MPR mode in OnDemand3D^® and are shown from left to right in the sagittal, coronal, and axial planes; readers were allowed to adjust window/level and magnification to reflect routine diagnostic conditions

Blinding and randomization

To prevent cross-recognition between repeated acquisitions of the same specimen, each CBCT series received a new random identification code for each protocol. Case order was randomised using a computer-generated sequence. Readers were blinded to fracture status, acquisition protocol, and fracture prevalence.

Reference standard

After completion of all CBCT acquisitions, the maxillary premolar dentoalveolar segment was surgically exposed using the same procedure in all specimens. A full-thickness mucoperiosteal flap was carefully elevated on both the buccal and palatal aspects of the target region to expose the underlying cortical bone. Any adherent soft tissue was gently removed as needed, without altering the bone surface. The exposed buccal and palatal cortical plates were examined under direct illumination. Visual inspection, tactile probing with a fine dental explorer, gentle palpation, and mobility testing of the dentoalveolar segment were performed in each specimen using the same sequence. Fracture presence or absence was determined by direct visual and tactile inspection and recorded as the surgical reference standard. According to this reference standard, 15 specimens were classified as fracture-positive and 15 as fracture-negative. CBCT findings were not used to guide surgical localisation or the extent of exposure, thereby avoiding incorporation bias.

Observers and assessment

Two oral and maxillofacial surgeons and two oral and maxillofacial radiologists participated as observers in this study. All observers were blinded to the fracture status of the specimens and had a minimum of five years of experience in maxillofacial trauma assessment. A pre-calibration session was conducted before the evaluation to ensure familiarity with the anatomical structures of the sheep head model. To replicate standard clinical practice, observers were permitted to adjust the brightness and contrast settings of the CBCT images. This adjustment was intended to reflect the routine diagnostic conditions that clinicians encounter, where they modify image display settings to enhance visualization. Observers were instructed to assess the maxillary premolar region because the experimental fracture model was confined to this predefined anatomical site. Each observer received a printed guide detailing the assessment tasks along with a brief clinical scenario to provide context:

"Each scan represents a patient who has recently sustained trauma to the maxilla. The patient presents with trauma and pain, and although conventional radiographs have been obtained, there remains a concern that undetected root and dentoalveolar fractures may be present. Therefore, a CBCT examination is being evaluated to assess the suspected dentoalveolar injury."

Each observer independently evaluated all 120 post-trauma CBCT datasets, corresponding to 30 specimens scanned with four acquisition protocols. Baseline CBCT scans were not included in the observer assessment. In the first session, observers recorded two outcomes for each dataset: first, a binary classification regarding the presence (1) or absence (0) of a dentoalveolar fracture in the maxillary premolar region; and second, their diagnostic confidence using a five-point confidence scale: 1, definitely absent; 2, probably absent; 3, not sure; 4, probably present; and 5, definitely present.

To evaluate intraobserver reliability, the same 120 post-trauma CBCT datasets were re-randomised and reassessed in a second session conducted one month after the initial assessment. In this second session, observers recorded only the binary fracture presence/absence decision. Thus, intraobserver agreement was calculated from binary decisions obtained in Sessions 1 and 2, whereas confidence scores were obtained only in Session 1 and used for ROC/AUC and confidence analyses.

Statistical analyses

The data obtained in this study were analyzed using the MedCalc 11.4 software package (MedCalc Software bv, Ostend, Belgium). Diagnostic performance metrics were calculated using post-trauma CBCT datasets according to the surgical reference standard, which classified 15 specimens as fracture-positive and 15 specimens as fracture-negative. Baseline scans were used only to confirm the absence of pre-existing fractures and were not included in diagnostic accuracy calculations. For diagnostic accuracy, receiver operating characteristic (ROC) curves were generated and the area under the curve (AUC) was estimated for each CBCT protocol. ROC analyses were conducted at the protocol level using pooled Session 1 confidence scores from all four observers, corresponding to 120 reader-by-case observations per protocol (30 specimens × 4 observers). This approach was used to summarise the overall diagnostic discrimination of each acquisition protocol across the observer panel. Because all observers had experience in maxillofacial trauma assessment, underwent the same pre-calibration process, and evaluated the same randomised datasets under identical viewing conditions, pooled protocol-level AUC estimates were considered appropriate for the primary protocol comparison. Pairwise AUC comparisons were performed with the DeLong test (two-sided, α = 0.05), with 95% confidence intervals reported [14]. The AUC results were interpreted as descriptive protocol-level estimates rather than reader-specific performance measures. Sensitivity, specificity, and predictive values were calculated from the binary fracture decisions obtained during Session 1. Interobserver agreement for binary fracture detection was quantified using Fleiss’ kappa across four readers. Intraobserver agreement was assessed using Cohen’s kappa by comparing Session 1 and Session 2 binary decisions within each observer and protocol. According to Landis and Koch, the Kappa values were interpreted as follows: 0.00–0.20 = poor agreement; 0.21–0.40 = fair; 0.41–0.60 = moderate; 0.61–0.80 = substantial; 0.81–1.00 = almost perfect [15]. For ordinal self-confidence scores (1–5), Friedman’s test assessed overall differences across the four protocols (blocking by reader×case), with Kendall’s W as effect size; pairwise Wilcoxon signed-rank tests with Holm adjustment were used for post-hoc comparisons. The confidence interval was set at 95%, and a p-value of < 0.05 was considered statistically significant.

Results

All 30 specimens were included in the final analysis and were successfully scanned using the four CBCT protocols (8 × 16-SR, 8 × 16-HR, 18 × 16-SR, and 18 × 16-HR), yielding 120 post-trauma datasets (30 specimens × 4 protocols).

All protocols demonstrated high discrimination for fracture detection, with AUC values ranging from 0.817 to 0.917 (Table 1). The highest performance was observed for the 18 × 16-HR protocol (AUC 0.917, 95% CI 0.869–0.964), followed by 8 × 16-HR (AUC 0.908, 95% CI 0.857–0.960), 8 × 16-SR (AUC 0.867, 95% CI 0.809–0.924), and 18 × 16-SR (AUC 0.817, 95% CI 0.750–0.883). AUC estimates were derived from pooled reader-by-case confidence ratings (N = 120 per protocol; 30 cases × 4 readers).

Table 1.

Diagnostic accuracy analysis of CBCT protocols using two fields of view and two resolutions for fracture detection

CBCT Protocols	ROC analysis			Descriptive analysis
	AUC	SEa	95% CIb	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
8 × 16-SR	0.867	0.0294	0.809–0.924	75.00	98.33	97.83	79.73
18 × 16-SR	0.817	0.0340	0.750–0.883	70.00	93.33	91.30	75.68
8 × 16-HR	0.908	0.0263	0.857–0.960	86.67	95.00	94.55	87.70
18 × 16-HR	0.917	0.0243	0.869–0.964	83.33	100.00	100.00	85.71

Diagnostic performance metrics were calculated using Session 1 reads. ^aStandard error (SE) of AUC calculated using the nonparametric DeLong method [14]. ^b95% confidence intervals (CI) computed as Wald-type intervals: AUC ± 1.96 × SE. AUC values were derived from the 5-point confidence scores; sensitivity/specificity (and PPV/NPV) were computed from the binary fracture decision (present/absent) in Session 1

Abbreviations: CBCT cone beam computed tomography, ROC receiver operating characteristic, PPV positive predictive value, NPV negative predictive value, AUC area under ROC curve, CI confidence interval, 8 × 16-SR 8 × 16 cm FOV-Standard resolution, 8 × 16-HR 8 × 16 cm FOV-High resolution, 18 × 16-SR 18 × 16 cm FOV-Standard resolution, 18 × 16-HR 18 × 16 cm FOV-High resolution

Pairwise DeLong comparisons (Table 2) demonstrated a significant improvement for 18 × 16-HR compared with 18 × 16-SR (ΔAUC 0.100, p = 0.0003) and for 8 × 16-HR compared with 18 × 16-SR (ΔAUC 0.092, p = 0.003). No significant difference was observed between the two high-resolution protocols (18 × 16-HR vs. 8 × 16-HR: ΔAUC 0.008, p = 0.650). The comparison of standard-resolution protocols (8 × 16-SR vs. 18 × 16-SR) was borderline (ΔAUC 0.050, p = 0.056), and the difference between 8 × 16-HR and 8 × 16-SR did not reach statistical significance (ΔAUC 0.042, p = 0.081). These findings should be interpreted at the protocol level because voxel size differed across the acquisition settings.

Table 2.

Pairwise comparison of ROC curves

Comparison	Difference between AUCs	z	p-value
18 × 16-HR vs. 18 × 16-SR	0.100	3.64	0.0003
18 × 16-HR vs. 8 × 16-HR	0.008	0.453	0.650
18 × 16-HR vs. 8 × 16-SR	0.050	2.52	0.011
8 × 16-HR vs. 18 × 16-SR	0.092	2.92	0.003
8 × 16-SR vs. 18 × 16-SR	0.050	1.91	0.056
8 × 16-HR vs. 8 × 16-SR	0.042	1.74	0.081

Pairwise comparisons of AUCs were performed using the two-tailed DeLong test based on Session 1 reads [14]. The z statistic represents the standardized difference between two correlated AUC estimates, and p < 0.05 was considered statistically significant

Abbreviations: AUC area under ROC curve, CI confidence interval, 8 × 16-SR 8 × 16 cm FOV-Standard resolution, 8 × 16-HR 8 × 16 cm FOV-High resolution, 18 × 16-SR 18 × 16 cm FOV-Standard resolution, 18 × 16-HR 18 × 16 cm FOV-High resolution

Interobserver agreement across four readers was substantial, with Fleiss’ kappa ranging from 0.67 to 0.84. Intraobserver agreement (Session 1 vs. Session 2) was substantial to almost perfect (Table 3): median (IQR) Cohen’s kappa across readers was 0.896 (0.861–0.928) for 8 × 16-SR, 0.932 (0.850–0.932) for 8 × 16-HR, 0.794 (0.787–0.850) for 18 × 16-SR, and 0.932 (0.930–0.950) for 18 × 16-HR; overall reader-by-protocol kappa ranged from 0.605 to 1.000.

Table 3.

Summary of intraobserver agreement (Session 1 vs. Session 2) by CBCT protocol

CBCT Protocol	Cohen’s κ, median [IQR]
8 × 16-SR	0.896 [0.861–0.928]
8 × 16-HR	0.932 [0.850–0.932]
18 × 16-SR	0.794 [0.787–0.850]
18 × 16-HR	0.932 [0.930–0.950]

Intraobserver agreement was assessed using Cohen’s κ for fracture detection performance by comparing Session 1 vs. Session 2 for the same 30 cases. κ values were calculated separately for each reader and protocol and are presented as median [IQR] across four readers

Abbreviations: IQR Interquartile Range, 8×16-SR 8 × 16 cm FOV-Standard resolution, 8×16-HR 8 × 16 cm FOV-High resolution, 18×16-SR 18 × 16 cm FOV-Standard resolution, 18×16-HR 18 × 16 cm FOV-High resolution

Self-confidence ratings differed significantly across protocols (Friedman χ²(3, N = 120) = 52.02, p < 0.001; Kendall’s W = 0.144). Median confidence was 4 (IQR 3–5) for both standard-resolution protocols and 4 (IQR 4–5) for 8 × 16-HR, while 18 × 16-HR showed the highest confidence with a median of 5 (IQR 4–5). Post-hoc Wilcoxon tests with Holm correction confirmed higher confidence for high-resolution versus standard-resolution protocols within each FOV (8 × 16-HR vs. 8 × 16-SR, p = 5.53 × 10⁻⁶; 18 × 16-HR vs. 18 × 16-SR, p = 9.23 × 10⁻⁷). No significant difference was observed between 8 × 16-HR and 18 × 16-HR after correction (p = 0.078) (Figs. 2 and 3).

Fig. 2.

Observer self-confidence across CBCT protocols. A Median self-confidence (Likert 1–5) with interquartile ranges (IQR) per protocol. B Stacked proportions of ratings by confidence category (1–5). Both panels are derived from pooled reader-by-case data (N = 120 per protocol; 30 cases × 4 readers) and show overall higher confidence for high-resolution (HR) protocols compared with standard-resolution (SR) protocols

Fig. 3.

ROC curves by CBCT protocol using observer confidence scores (Session 1). ROC analysis was performed using 5-point confidence ratings (N = 120 per protocol), illustrating protocol-level diagnostic discrimination for dentoalveolar fracture detection

Discussion

Diagnosis of non-displaced dentoalveolar fractures remains clinically challenging because these injuries are frequently overlooked during the initial post-trauma assessment. Conventional two-dimensional radiography is limited by anatomical superimposition and restricted spatial resolution; accordingly, periapical radiographs have been reported to detect dentoalveolar fracture lines in only 30–40% of cases, whereas small FOV, high-resolution CBCT protocols may identify the same fractures in up to 90% of cases [12]. Even with CBCT, however, non-displaced fractures can be missed when the fracture line is extremely thin (< 0.5 mm) and confined to a single cortical plate [16]. Against this background, our study assessed how FOV and manufacturer-defined resolution mode influence diagnostic performance for non-displaced dentoalveolar fractures. The key finding is that high-resolution imaging improves diagnostic discrimination and reader confidence relative to standard-resolution imaging, while enlarging the FOV provides limited incremental benefit once high-resolution acquisition is used. Consistent with this interpretation, the large-FOV high-resolution protocol showed significantly higher discrimination than its standard-resolution counterpart (18 × 16 h vs. 18 × 16 SR; p = 0.0003), whereas the two high-resolution protocols did not differ (18 × 16 h vs. 8 × 16 h; p = 0.650), indicating that accuracy gains are predominantly driven by resolution rather than FOV expansion at high resolution.

Previous studies provide important context for the diagnostic challenge addressed in the present study. Conventional two-dimensional imaging may be limited in the detection of subtle or linear fracture patterns because of anatomical superimposition and restricted spatial information [17]. Cross-sectional imaging with CBCT has been reported to improve diagnostic confidence in traumatic dentoalveolar settings [9, 11]. However, this study did not include a two-dimensional imaging comparator; therefore, this literature should be interpreted only as background context. Instead, our findings provide protocol-level evidence within CBCT, suggesting that optimisation of spatial resolution is particularly important for improving the detectability of subtle, non-displaced dentoalveolar fractures.

From a technical and clinical optimisation perspective, our results reinforce that diagnostic yield in dentoalveolar trauma should be considered at the protocol level, where FOV and spatial resolution are interdependent through voxel size and image quality. Smaller voxel sizes are generally associated with improved depiction of fine structures but may also increase image noise and require appropriate exposure settings; therefore, image quality should be interpreted as the combined effect of acquisition and reconstruction parameters rather than FOV alone [18]. Importantly, although the clinical target differs, both vertical root fractures and non-displaced dentoalveolar fractures rely on detecting thin, low-contrast fracture lines; thus, evidence that technical acquisition parameters materially influence CBCT diagnostic accuracy in vertical root fracture diagnosis is conceptually transferable to the present context [19]. In line with this principle, our results support prioritising spatial resolution/voxel size when detailed depiction of a local dentoalveolar segment is needed, rather than assuming that increasing FOV will inherently improve detectability. Nominal voxel size alone may not fully explain protocol-level differences in diagnostic performance. Although the 8 × 16-SR and 18 × 16-HR protocols had comparable voxel sizes, they differed in acquisition mode, exposure time, DAP, FOV geometry, and potentially reconstruction-related image characteristics. These factors may influence image noise, contrast-to-noise ratio, scatter, and artefact behavior, thereby affecting the conspicuity of fine fracture lines. Therefore, the observed differences between these protocols should be interpreted as protocol-level effects rather than as consequences of voxel size alone.

In traumatic CBCT imaging, selection of the FOV remains pivotal because it determines whether the suspected fracture region is included within the imaging volume and may reduce the need for repeat acquisitions. In a large patient series, incomplete FOV coverage was the most common reason for CBCT retakes (57.5%), implying that inadequate initial coverage can inadvertently increase cumulative radiation exposure by necessitating additional scans [20]. Accordingly, when the extent of injury is uncertain, a wider imaging volume may be justified to assess segmental fracture propagation and adjacent structures. Supporting this concept, a prospective study reported that CBCT revealed additional mandibular fracture lines not detected by conventional assessment in 17.75% of cases and altered management in 9.52%, underscoring the potential for “occult fractures” in the trauma setting [21].

However, the influence of FOV is not limited to coverage; diagnostic visibility may also be shaped by positioning within the FOV and by artefact/exomass dynamics. In a root-fracture model, using a small FOV and positioning the target close to the FOV centre (or an appropriate peripheral region) improved fracture detection, whereas off-centre positioning reduced performance and increased the perception of severe artefacts [22]. Conversely, metallic structures within the exomass can degrade image quality, and a small FOV may exacerbate exomass-related effects in some settings, potentially offsetting the expected benefit of FOV reduction through adverse grey-value and artefact behaviour [23]. Together, these observations support interpreting FOV effects as context-dependent and reinforce the rationale for protocol-level optimisation rather than universal “smallest FOV is best” rules.

Integrating these considerations with our data, the effect of FOV should be interpreted in relation to spatial resolution and voxel size, rather than as an independent advantage. At standard resolution, the smaller FOV showed a tendency toward better discrimination than the larger FOV (8 × 16-SR vs. 18 × 16-SR; p = 0.056). At high resolution, however, expanding the FOV provided little additional benefit in terms of AUC or specificity. Therefore, FOV selection should first be guided by the anatomical coverage required for the suspected injury pattern. A larger FOV may be necessary when the extent of trauma is uncertain or when additional fractures are suspected [21]. For subtle, linear, non-displaced dentoalveolar fractures, resolution mode becomes critical. After the appropriate FOV has been selected, a high-resolution protocol should be preferred when clinically justified. This is consistent with our finding that both high-resolution protocols showed better diagnostic discrimination and higher reader confidence than their standard-resolution counterparts. Because larger FOVs and smaller voxel sizes may increase radiation dose, protocol selection should remain indication-oriented and patient-specific [24, 25]. The ALADAIP framework therefore supports selecting the smallest FOV that answers the clinical question while using a high-resolution protocol when subtle fracture detection is the clinical priority [26].

This study has several limitations. First, because the experiments were performed in an ex vivo model, the findings may not fully reflect clinical trauma conditions, including patient motion, soft-tissue attenuation, bleeding/edema, and heterogeneous metal restoration profiles. Therefore, generalisation to real-world settings should be made with caution. Second, although fractures were generated under standardised conditions, the model cannot capture the full spectrum of dentoalveolar trauma seen in practice, such as combined injuries, comminution, concomitant luxation, or inter-individual anatomical variability. Third, observers were instructed to assess the maxillary premolar region. This approach was necessary to standardise the diagnostic task and to match the predefined experimental fracture model. However, it may have reduced the search component of image interpretation and introduced localisation bias. Consequently, diagnostic accuracy and reader confidence may have been higher than would be expected in routine clinical trauma assessment, in which the fracture location and extent may be less clearly defined. Fourth, evaluations were confined to a single CBCT unit and four predefined, clinically available acquisition protocols, and ROC analyses were based on pooled reader-by-case confidence ratings rather than a formal multi-reader multi-case model. As voxel size and related noise/exposure characteristics varied across FOV settings within each nominal resolution mode, the independent effects of FOV and voxel size could not be fully disentangled. Consequently, results should be interpreted as protocol-level estimates and may not be directly reproducible across CBCT systems with different exposure settings and reconstruction algorithms. Nevertheless, a key strength of this work is the systematic within-design comparison of four clinically relevant protocols using a surgical reference standard under standardised conditions.

Conclusion

In conclusion, our ex vivo findings suggest that protocol optimisation for detecting non-displaced dentoalveolar fractures on CBCT should prioritise spatial resolution, as high-resolution acquisition improves diagnostic discrimination and reader confidence, whereas enlarging the FOV offers limited incremental benefit once high resolution is used. Clinically, FOV selection should be driven by the anatomic coverage required by the suspected injury pattern, while a targeted high-resolution protocol may be preferred when the primary concern is subtle, linear fractures—supporting indication-oriented protocol selection to maximise diagnostic yield while avoiding unnecessary repeat imaging.

Abbreviations

ALADAIP

As Low As Diagnostically Acceptable, Indication-oriented and Patient-specific

ALARA

As Low As Reasonably Achievable

AUC

Area Under the Curve

CBCT

Cone Beam Computed Tomography

FOV

Field of view

HR

High Resolution

IQR

Interquartile Range

NPV

Negative Predictive Value

PPV

Positive Predictive Value

ROC

Receiver Operating Characteristic Curves

SR

Standard Resolution

Authors’ contributions

S.A.T. and Z.A.A.A. contributed to study conception, literature search, data extraction; Z.A.A.A.drafted the manuscript and contributed to the critical review. Writing - original draft: Z.A.A.A. All authors provided substantial review and feedback on the final version of the article. Z.A.A.A. takes responsibility for the paper.

Funding

The author(s) received no financial support for this work, authorship, and/or publication.

Data availability

The authors confirm that the data supporting the findings of this study are available within the article.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.