Scatter-to-primary ratio in dentomaxillofacial cone-beam CT: effect of field of view and beam energy

Abstract

Objective:

The aim of this study was to evaluate the effect of field of view (FOV) and beam energy on the scatter-to-primary ratio (SPR) in dental cone-beam CT (CBCT).

Methods:

An anthropomorphic phantom representing an adult male (ATOM Max 711-HN, Norfolk, VA, USA) was scanned using the 3D Accuitomo 170 CBCT (J. Morita, Kyoto, Japan) using 11 FOVs. During each scan, half of the X-ray beam was blocked. Each scan was performed at three exposure settings with varying beam energy and equal radiation dose: 90 kV 5 mA, 77 kV 7.5 mA and 69 kV 10 mA. The SPR was estimated by measuring the grey values in the blocked and non-blocked regions of the RAW data. The effect of FOV on SPR was evaluated using Dunn’s multiple comparison test, and the effect of the exposure settings was compared using a Wilcoxon signed rank test.

Results:

Larger FOVs showed increased scatter. FOVs with a shorter isocenter-detector distance showed a particularly high SPR. Most intercomparisons between FOVs were statistically significant. The largest difference was found between 17 × 12 cm and 6 × 6 cm (lower jaw), with the former showing a 4.9-fold higher SPR. The effect of beam energy was relatively small and varied between FOV sizes and positions.

Conclusion:

While the choice of FOV size and position is determined by the diagnostic region of interest, the image quality deterioration for large FOVs due to scatter provides another incentive to limit the FOV size as much as possible.

Introduction

Cone-beam computed tomography (CBCT) has evolved into an indispensable tool in dentomaxillofacial radiology, being widely used for a variety of clinical applications. Whereas CBCT can yield high-resolution 3D images, there are several limitations inherent to the technology that lead to noise or artefacts.¹ One of the major causes of image quality degradation in CBCT is X-ray scatter. Most of the generated X-ray scatter is absorbed within the patient or deflected at angles that are not oriented towards the detector; however, a fraction of scatter reaches the detector and contributes to the total signal. In CBCT, the scatter-to-primary ratio (SPR) is presumed to be relatively high due to the close proximity of the detector to the patient and the absence of detector-side scatter grids.²

There are two exposure parameters in CBCT that affect SPR: field of view (FOV) size and tube voltage (kV). A larger FOV increases both the magnitude of generated scatter (due to the increased total number of X-ray photons) and the proportion of detected scatter (due to the larger active detector area).³ The tube voltage affects the X-ray energy distribution and thus the probability of scatter occurring.⁴ Furthermore, the angular distribution of Compton and Rayleigh scatter is energy-dependent; at higher X-ray energies, the probability of forward scattering increases.^5,6 To the best of our knowledge, the effect of FOV on scatter in dental CBCT has only been measured indirectly using signal-difference-to-noise ratio measurements on a geometric phantom.⁷ In addition, prior studies have evaluated the effect of kV on the image quality of reconstructed CBCT images,⁸ but the effect on X-ray scatter has not yet been characterized. Thus, the aim of this study was to evaluate the effect of FOV and beam energy on SPR in dental CBCT.

Methods and materials

Phantom – CBCT scanning

An anthropomorphic phantom representing an adult male (ATOM Max 711-HN, CIRS, Norfolk, VA, USA) was used. The phantom consists of tissue-equivalent materials and provides an anatomically accurate representing of the bones, teeth, soft tissues and air cavities. The phantom was scanned using the 3D Accuitomo 170 CBCT (J. Morita, Kyoto, Japan) using the FOV sizes and positions indicated in Table 1, representing the most common clinical indications for CBCT. During each scan, half of the X-ray beam was covered by several sheets of stainless steel, ensuring total attenuation of the beam (Figure 1). Each scan was performed at standard exposure settings for adults: 90 kV, 5 mA and 17.5 s exposure time. In addition, scans were performed at 77 kV 7.5 mA and 69 kV 10 mA, resulting in a total of 33 scans. All three combinations of kV and mA corresponded with the same radiation dose, as indicated by the computed tomography dose index (CTDI). CTDI values for each combination of FOV, kV and mAs were determined by the manufacturer and are displayed on the CBCT unit’s console. For the 14 × 10 cm and 17 × 12 cm FOVs, 70 kV 10 mA was used instead of 69 kV 10 mA, as it provided a closer match with the other beam energies in terms of radiation dose.

Table 1.

Fields of view and phantom positioning

Field of view (cm)	Position	Code
Dental scans
17 × 12a	Dentomaxillofacial	17 × 12_D
14 × 10	Dentomaxillofacial	14 × 10_D
10 × 10	Both jaws	10 × 10_B
10 × 5	Upper jaw	10 × 5_U
	Lower jaw	10 × 5_L
8 × 8	Both jaws	8 × 8_B
6 × 6	Upper jaw, anterior region	6 × 6_U
	Lower jaw, molar region	6 × 6_L
Other scans
14 × 10	Sinuses	14 × 10_S
17 × 5a	Temporomandibular joint (dual)	17 × 5_T
6 × 6	Temporomandibular joint (single)	6 × 6_T

^aShorter isocenter-detector distance (i.e. 204 mm vs 302 mm)

Figure 1.

Measurement set-up. Half of the X-ray beam was blocked at the tube side. As a result, half of the detector area represented primary radiation and scatter, whereas the other half represented scatter only.

Image analysis of RAW data

For this particular CBCT unit, the RAW data are in the form of a single 16-bit unsigned, unencrypted TIF image. The RAW data were analysed using ImageJ (NIH, Bethesda, MD, USA). Two regions of interest (ROIs) were defined as shown in Figure 2. Both ROIs had a width equal to 10% that of the projection, and a height equal to 50% that of the projection. One of the ROIs was placed slightly to the left of the halfway point of the projection, and the other slightly to the right, with the distance between the ROIs equal to 20% of the projection width. A custom macro was written in ImageJ to automatically determine the ROI dimensions (i.e. width, height) and position relative to the dimensions of the projection, ensuring perfect measurement reproducibility. The mean grey value within the ROI was measured for 40 dark projections (representing the background signal transmitted by the detector in the absence of X-rays) and 512 radiographic projections covering a full rotation of the X-ray tube. The former was subtracted from the latter to yield a corrected signal for each radiographic projection.

Figure 2.

Region of interest (ROI) for measurement of total signal (primary plus scatter) (left) and scatter signal (right) on the 10 × 10 cm field of view (FOV). The relative dimension and position of the ROIs were identical for all FOVs.

SPR calculation

The SPR was estimated per projection, using the mean grey values of the unblocked (MGV_u) and blocked (MGV_b) ROIs. MGV_b is equal to the scatter in the blocked region (S_b), whereas MGV_u comprises the primary signal (P_u) and the scatter in the unblocked region (S_u). Due to the energy-dependent angular distribution of Compton and Rayleigh scatter, S_b and S_u are not identical. In addition, their sum is not equal to the total scatter that would reach the ROIs in the absence of a beam blocker (S_t). Therefore, Monte-Carlo simulation was used to estimate S_u and S_t based on S_b. A full description of the simulation set-up and results is provided as Supplementary Material 1. In summary, using the geometry of the 3D Accuitomo 170 CBCT, projections of a cylindrical water phantom were simulated. Simulations were performed for each beam energy and FOV, for conditions with and without a beam blocker, and with and without scatter. From these simulations, S_u/S_b and S_t/S_b could be calculated and applied to the experimentally determined S_b values to calculate S_u and S_b for each scan.

Next, the primary signal P_u for a given projection was determined as the difference between MGV_u of that projection and the estimated scatter in the unblocked region:

$${P_u=MGV}_u-S_u$$ (1)

The average SPR for a condition without beam blocker could then be calculated for each projection as:

$$SPR=\frac{S_t}{P_u}$$ (2)

Statistical analysis

All statistical analyses were performed using Prism 9 (GraphPad Software, San Diego, CA, USA). It was found that SPR values were not normally distributed according to the Kolmogorov-Smirnov test; therefore, non-parametric tests were used. The SPR values for the 11 FOVs were compared after averaging the data for the three exposure settings for each FOV, using Dunn’s multiple comparison test. The SPR values for the three exposure settings were compared by calculating the ratio between the SPRs for each projection and for each FOV, pooling these ratios per eight consecutive projections, and using a one-sample Wilcoxon signed-rank test with a hypothetical value of 1.0 (i.e. ‘no difference’). The overall significance level for each test was 0.05.

Results

Table 2 shows SPR values for each scan, shown as median values and interquartile range. For clarity: higher SPR values indicate more (i.e. worse) scatter.

Table 2.

Scatter-to-primary ratio per field of view (FOV) and exposure setting. Results for each exposure setting are shown as ‘median (Q1-Q3)’.

FOVa	90 kV, 5 mA	77 kV, 7.5 mA	69 kV, 10 mA	Relative SPRb	Sig. (FOV)c	Sig. (kV)d
Dental scans
17 × 12_D	0.80 (0.59–1.00)	0.78 (0.61–0.99)	0.79 (0.62–1.00)	4.90	A	69>77>90
14 × 10_D	0.43 (0.29–0.54)	0.43 (0.30–0.54)	0.44 (0.32–0.55)	2.68	B	69>77=90
10 × 10_B	0.35 (0.28–0.43)	0.34 (0.26–0.42)	0.35 (0.26–0.43)	2.16	C	90>77=69
8 × 8_B	0.26 (0.20–0.29)	0.26 (0.20–0.29)	0.26 (0.19–0.29)	1.60	D	90>77>69
10 × 5_U	0.26 (0.18–0.32)	0.26 (0.18–0.32)	0.25 (0.17–0.32)	1.59	D	None
10 × 5_L	0.20 (0.16–0.26)	0.20 (0.17–0.26)	0.21 (0.17–0.26)	1.27	E	None
6 × 6_U	0.17 (0.14–0.24)	0.17 (0.14–0.22)	0.17 (0.14–0.22)	1.07	F	90>77=69
6 × 6_L	0.16 (0.14–0.19)	0.16 (0.14–0.19)	0.16 (0.13–0.19)	1.00	F	90>77>69
Other scans
14 × 10_S	0.58 (0.19–0.83)	0.59 (0.18–0.85)	0.59 (0.18–0.87)	3.64	B	69>77
17 × 5_T	0.72 (0.54–0.95)	0.71 (0.52–0.96)	0.71 (0.51–0.98)	4.43	A	90=69>77
6 × 6_T	0.25 (0.19–0.29)	0.25 (0.19–0.28)	0.25 (0.19–0.28)	1.55	DE	90>77>69

FOV, field of view; SPR, scatter-to-primary ratio.

^aSee Table 1 for definitions of FOVs

^bNormalized to SPR value of the 6 × 6_L protocol, averaged for the three beam energies

^cDunn’s Multiple Comparison test. If any of the letters for two given FOVs are the same, the pairwise intercomparison between them was not significant e.g. ‘A’ vs ‘B’: significantly different, ‘D’ vs ‘DE’: not significantly different

^dWilcoxon-Signed Rank test. Example 1: ‘69>77=90’: 69 kV showed a significantly higher SPR than 77 kV and 90 kV. Example 2: ‘69>77’: 69 kV showed a significantly higher SPR than 77 kV, but neither showed a significant difference with 90 kV

Effect of FOV

A violin plot of the SPR distribution for each FOV is shown in Figure 3. There was a distinct effect of FOV size, FOV position and isocenter-detector distance (IDD) on SPR. Out of 55 pairwise intercomparisons between FOVs, 48 tests showed a significant difference (Table 2). The largest difference was found between 17 × 12_D and 6 × 6_L, with the former showing a 4.90-fold higher median SPR. For the 10 × 5 cm FOVs, significantly less scatter was found for lower jaw scans. For the 14 × 10 cm and 6 × 6 cm FOV, the non-dental scans showed a significantly higher SPR than the dental scans. Furthermore, the two protocols with a shorter IDD (i.e. 17 × 12_D and 17 × 5_T) showed a significantly higher SPR than all other protocols. For dental scans with the same IDD, a linear relation could be seen between FOV size (i.e. “diameter x height”) and median SPR (Figure 4), with R² = 0.96.

Figure 3.

Violin plots of SPR for all fields of view.

Figure 4.

Correlation between field of view (FOV) size and median scatter-to-primary ratio. The trendline refers to dental FOVs with identical isocenter-detector distance.

Effect of beam energy

Figure 5 shows the overall distribution of SPR for each beam energy for all FOVs combined. While the overall distribution is highly similar, it can be seen that the maximum SPR (SPR_max) was inversely proportional to the beam energy (i.e. SPR_max = 2.88 for 90 kV, SPR_max = 3.12 for 77 kV, and SPR_max = 3.49 for 69 kV). The majority (i.e. 21/33) of intercomparisons between the exposure settings showed a significant difference, although differences between median SPR values were small. Furthermore, significant differences were often not dictated by the median, and there was no consistency in terms of the relation between beam energy and SPR; whereas SPR values for 90 kV were found to be significantly higher for smaller FOVs (i.e. 6 × 6 cm to 10 × 10 cm, except 10 × 5 cm), significantly higher SPR values for 69 kV were found for the 17 × 12 cm and 14 × 10 cm FOVs.

Figure 5.

Violin plots of SPR for each beam energy. Values were pooled for all FOVs.

Discussion

To our knowledge, this study was the first to quantify the effect of FOV on scatter in dental CBCT. We evaluated scatter in clinically representative conditions using a phantom that represents an average adult patient in terms of anatomy and attenuation. A wide range in SPR was found; it can be noted that other CBCT models on the market offer even larger or smaller FOV options than those covered in this study, implying that the range in SPR between different CBCT models is likely to be well beyond fivefold. However, a comparison between different CBCTs is hampered by the fact that varying kV and filtration is used.

Our results indicate that scatter comprises a considerable contribution to the total detector signal, even for FOVs as small as 6 × 6 cm. For the largest FOVs, it was found that the contribution of the scatter to the signal could be equal or more to that of primary X-rays in regions with high attenuation. A prior simulation study involving a dental CBCT unit found relatively high degrees of scatter in a bone-equivalent (i.e. polytetrafluoroethylene) portion of a contrast phantom.⁹ Similar levels of scatter were found in studies involving CBCT units for image-guided radiation therapy (IGRT), which typically use large FOVs. SPR values close to, or exceeding, 1.0 has been shown for large-FOV CBCT of the head in IGRT.^10–12 A recent study showed that, in the absence of scatter correction, SPR values over 2.0 can occur.¹³ In multi-detector CT (MDCT), lower scatter levels are found owing to the use of detector rows rather than flat panel detectors as well as the relative long object-detector distance; a simulation study involving two MDCT scanners found SPR values ranging from 0.08 to 0.14 for a head phantom,¹⁴ which is less than the median SPR for the smallest FOVs in our study.

The primary finding of this study is the use of smaller FOVs can lead to improved image quality due to reduced SPR. A previous study showed that signal difference to noise ratio (SDNR) at the detector could be improved considerably by reducing FOV size⁷; however, the effects of scatter on image quality are not limited to noise. Although our study did not involve an assessment of image quality of reconstructed images, previous studies involving CBCT for IGRT have shown that scatter affects contrast-to-noise ratio, Hounsfield Unit accuracy, and cupping artefacts.^{10–13,15,16} Although these effects are expected to be the same for dental CBCT scans, further study is warranted to quantify the influence of scatter on image quality after reconstruction.

Whereas the findings from this study could lead to a recommendation towards using small FOVs, it is important to note that the main rationale behind FOV collimation is the reduction of patient radiation dose in accordance with the ‘as low as reasonably achievable’ (ALARA) principle.¹⁷ A previous study showed a 2.7-fold difference in effective dose between the largest (17 × 12 cm; 303 µSv) and smallest (6 × 6 cm, upper jaw; 113 µSv) FOV used in this study at 90 kV and 87.5 mAs.¹⁸ Regardless of the obvious effect of FOV size on patient dose, literature and anecdotal evidence shows the consistent use of disproportionately large FOVs by a fraction of CBCT users. This can be attributed to several factors. First, larger FOVs can increase the frequency of incidental findings. However, while several studies have shown that such findings are common,^19–21 the majority are of relatively minor concern and do not affect the treatment plan.²² Furthermore, findings outside the dentoalveolar region may be missed when an oral or head and neck radiologist is not involved,²³ or when the radiological report covers only the referred indication rather than the entire volume. A second, more concerning reason for the routine use of large FOVs is a lack of adherence to the principles of radiation protection. This issue is compounded by the high degree of self-referral in dentistry (Farman 2009),²⁴ essentially allowing the clinical practitioner to bypass the justification and optimization principles with little or no accountability. Although figures regarding the radiation protection culture in dentistry are lacking, it is realistic to assume that certain end-users will be more eager to use smaller FOVs if there is a benefit in terms of image quality. While previous studies have shown that smaller FOVs can lead to increased spatial resolution owing to the use of smaller voxel sizes,^25,26 this is only applicable for CBCT models in which the voxel size is affected by the FOV, and if the voxel size is the limiting factor for the spatial resolution rather than other factors for example, focal spot size, effective detector pixel size and motion blurring.²⁷ The current study provides evidence of a more fundamental effect of FOV size on image quality and may alter the perception of end-users regarding the use of optimized vs. ‘one-size-fits-all’ FOVs.

Interestingly, no consistent effect of beam energy on SPR was found, with median values being highly similar. When the beam energy is changed by altering the kV and/or filtration, an interplay of different factors occurs. For Compton as well as Rayleigh scatter, the probability decreases at higher X-ray energies. However, because the probability of photoelectric absorption is inversely proportional to the cube of the X-ray energy, the contribution of Compton scatter to the total attenuation increases from ~ 60% to~80% between 40 and 60 keV for a head-size object.²⁸ Although the contribution of Rayleigh scatter slightly decreases at higher X-ray energies, the total scatter at the detector for diagnostic X-ray energies is mainly determined by Compton scatter.³ On the other hand, the increased beam penetration at higher energies leads to a higher primary signal for a given dose, especially in regions with high total attenuation; as a result, SPR peaked higher for lower beam energies in our study. Although the effect of this change in SPR distribution on reconstructed image quality requires further study, prior research using the same CBCT unit and beam energies as the current study showed that 90 kV results in the highest subjective image quality in terms of sharpness, noise and artefacts at a given radiation dose.²⁹

In conclusion, whereas the choice of FOV size and position is determined by the diagnostic region of interest, the image quality deterioration for large FOVs due to scatter provides another incentive to limit the FOV size as much as possible. In terms of beam energy, further evidence regarding the effect of beam energy on reconstructed image quality is needed.