Investigating Radiation Dose and Signal-To-Noise-Ratio in Pediatric and Adult Head Photon-Counting CT: A Phantom-Based Study

Abstract

Purpose

In photon-counting CT (PCCT), image quality is adjusted using image quality levels (IQLs). While the radiation dose increases linearly with IQLs, the signal-to-noise ratio (SNR) shows a non-linear trend. This study aims to investigate the relationship between radiation dose and SNR in adult and pediatric head phantoms using PCCT showing potential IQL ranges for radiation dose optimization.

Materials and Methods

Adult and pediatric anthropomorphic phantoms were scanned across multiple IQLs. The relationship between IQL, radiation dose (mAs, CTDI_vol, DLP, organ doses), and SNR (brain parenchyma, bone) was assessed. Group comparisons were performed at matched IQLs.

Results

Radiation dose increased linearly with IQL (all R²=1.00), whereas SNR demonstrated a more variable course with minor deviations from linearity. Quadratic or exponential fits provided slightly better modeling for some pediatric phantoms (5-year-old and 10-year-old), whereas 1-year-old and adult phantoms followed almost linear trends (R² ≥ 0.90). Pediatric phantoms showed a significantly higher SNR in brain and bone at lower effective mAs and radiation dose levels compared to adults (p<0.05). Among pediatric phantoms, the SNR values for brain parenchyma and bone differed significantly (p=0.002 and p=0.021), with the 1-year-old phantom exhibiting the highest SNR values in both tissues.

Conclusion

In PCCT head imaging, pediatric phantoms reach higher SNR values at lower radiation doses than adult phantoms, suggesting the potential for further protocol optimization. SNR-radiation dose curves indicate diminishing returns at higher IQLs, highlighting the importance of cautious radiation dose management to avoid unnecessary exposure.

Key Points

• In PCCT, image quality levels correlate linearly with radiation dose.

• Higher image quality levels give diminishing SNR gains, indicating optimization thresholds.

• Pediatric protocols achieve equal or higher SNR values at lower doses than adult protocols.

• Phantom-based results support age-specific PCCT protocols to avoid unnecessary exposure.

Citation Format

• Klüner LV, Opitz MK, Peuster H et al. Investigating Radiation Dose and Signal-To-Noise-Ratio in Pediatric and Adult Head Photon-Counting CT: A Phantom-Based Study. Rofo 2025; DOI 10.1055/a-2740-4588

Zusammenfassung

Ziel

In der Photon-Counting-CT (PCCT) wird die Bildqualität über Image Quality Levels (IQLs) eingestellt. Während die Strahlendosis linear mit den IQLs zunimmt, zeigt das Signal-Rausch-Verhältnis (SNR) einen nichtlinearen Verlauf. Ziel dieser Studie war es, den Zusammenhang zwischen Strahlendosis und SNR bei erwachsenen und pädiatrischen Kopfphantomen in der PCCT zu untersuchen und mögliche IQL-Bereiche für eine Dosisoptimierung aufzuzeigen.

Material und Methoden

Erwachsene und pädiatrische anthropomorphe Phantome wurden mit unterschiedlichen IQLs gescannt. Der Zusammenhang zwischen IQL, Strahlendosis (mAs, CTDI_vol, DLP, Organdosen) und SNR (Hirnparenchym, Knochen) wurde analysiert. Gruppenvergleiche erfolgten bei übereinstimmenden IQLs.

Ergebnisse

Die Strahlendosis nahm linear mit den IQLs zu (alle R² = 1,00), während das SNR einen variableren Verlauf mit geringen Abweichungen von der Linearität zeigte. Quadratische oder exponentielle Anpassungen lieferten für einige pädiatrische Phantome (5- und 10-jährig) leicht bessere Modellierungen, während das 1-Jährige und die Erwachsenenphantome fast linearen Trends folgten (R² ≥0,90). Pädiatrische Phantome wiesen bei niedrigerem effektivem mAs und geringerer Strahlendosis signifikant höhere SNR-Werte für Gehirn und Knochen auf als erwachsene Phantome (p<0,05). Innerhalb der pädiatrischen Gruppe unterschieden sich die SNR-Werte für Hirnparenchym und Knochen signifikant (p=0,002 bzw. p=0,021), wobei das Phantom eines 1-jährigen Kindes in beiden Geweben die höchsten SNR-Werte zeigte.

Schlussfolgerung

In der PCCT-Kopfdiagnostik erreichen pädiatrische Phantome bei geringerer Strahlendosis höhere SNR-Werte als erwachsene Phantome, was das Potenzial für eine weitere Protokolloptimierung nahelegt. SNR-Dosis-Kurven weisen auf abnehmende Zugewinne bei höheren IQLs hin, was die Bedeutung einer vorsichtigen Dosissteuerung unterstreicht, um unnötige Exposition zu vermeiden.

Kernaussagen

• In der PCCT korrelieren Image Quality Levels linear mit der Strahlendosis.

• Höhere Image Quality Levels führen zu abnehmenden SNR-Zugewinnen und deuten auf Optimierungsschwellen hin.

• Pädiatrische Protokolle erreichen bei geringerer Dosis ein gleiches oder höheres SNR als Erwachsene.

• Phantom-basierte Ergebnisse unterstützen altersangepasste PCCT-Protokolle zur Vermeidung unnötiger Exposition.

Keywords

photon-counting CT - radiation dose - phantom - image quality - head - SNR

Introduction

CT examinations are the first-line imaging modality in emergency diagnostic imaging. Overall, CT examinations account for more than 60% of medical radiation exposure and are, therefore, the largest contributor to the average effective dose for both adults and children worldwide [1]. Due to the potentially harmful consequences of radiation exposure, the principle of “as low as reasonably achievable (ALARA)” is applied to optimize the radiation dose required for CT examinations [2]. This approach is based on the linear no-threshold model, which states that any level of radiation exposure can be harmful [3]. In particular for children, it has been shown that there is a risk of hematological malignancies even after a single exposure to radiation [4]. In order to be able to better monitor the radiation exposure for an examination, diagnostic reference levels (DRLs) have been established, and apply to a standard application and a standard patient. DRLs indicate the 75^th percentile of a mean dose for a specific examination area [5].

With the ongoing development of CT scanners, photon-counting CT (PCCT) is currently the latest advancement in which photons can be converted directly into an electronic signal in contrast to energy-integrating detector (EID) CT, which needs to apply a two-step approach to convert X-rays into a digital signal [6] [7]. Several studies have shown that PCCT leads to a reduction in radiation dose with at least the same image quality [8] [9] [10]. For paranasal sinus imaging, dose reductions of over 60% have been reported [11], while a dose reduction of 43–66% in low-dose chest imaging was observed [8] [12] and a mean dose reduction ranging from 19% to 54% in liver imaging compared to EID-CT was measured without compromising diagnostic accuracy [13]. The image quality of CT examinations can be objectively determined based on quantitative parameters. Two commonly used image quality measurements are signal-to-noise ratio (SNR), which is the ratio of the signal and the standard deviation of the noise [14], and contrast-to-noise ratio (CNR), which is defined as the ratio of the difference between signal intensities of two regions of interest and the standard deviation of noise [15]. In head CT imaging, PCCT has been shown to provide a higher SNR than EID-CT in the evaluation of focal brain lesions [16] and in the contrast between gray and white matter [17]. Furthermore, recent studies in pediatric PCCT have demonstrated that, at similar radiation doses, photon-counting CT provides a higher SNR and CNR compared to dual-source CT, resulting in improved cardiovascular imaging quality in children suspected of having congenital heart defects [18]. Complementing these findings, Hellms et al. [19] showed that in pediatric patients with congenital heart disease, PCCT achieved a radiation dose reduction exceeding 40% compared to EID-CT, while maintaining a comparable SNR, CNR, and image quality. Importantly, in the context of pediatric head imaging, Naceanceno et al. [20] demonstrated that PCCT provides similar noise texture with reduced noise magnitude compared to EID-CT at comparable or reduced radiation doses in pediatric head CT exams. Together, these studies highlight the potential of PCCT to reduce radiation exposure in children without compromising diagnostic performance.

In order to be able to determine system-independent image quality before an examination, an image quality control parameter can be set on PCCT scanners. This corresponds to the formerly used “image quality reference mAs”, representing the effective mAs for an average-sized patient. These image quality levels (IQLs) on PCCT and the effective mAs are reported to exhibit a linear relationship [21], whereas, according to physical principles, the SNR shall increase as the square root of the radiation dose [14]. For a risk-benefit analysis, it is, therefore, relevant to decide which radiation dose increase still leads to a relevant image improvement. This study aimed to quantify the relationship between radiation dose and SNR for brain parenchyma and bone in adult and pediatric head phantoms using PCCT, and to assess differences that may guide dose optimization.

Materials and Methods

Given the exclusive use of phantoms in this study, ethical approval by the institutional review board was not required.

Image acquisition and CT protocols

The study used five anthropomorphic phantoms (ATOM Dosimetry Phantoms, CIRS, Norfolk, USA) for head PCCT scans ([Fig. 1]a), with each phantom representing a different adult or pediatric patient group (adult male, adult female, 1 year, 5 years, and 10 years). All scans were acquired on a PCCT system (Siemens NAEOTOM Alpha, Siemens Healthineers, Erlangen, Germany). Pediatric phantoms were scanned using 70 kVp, while adult phantoms were scanned using 120 kVp, in accordance with our clinical routine head CT protocols. The image quality control parameters were called IQLs for the automatic exposure control (AEC) CARE kV (Siemens Healthineers, Erlangen, Germany). The tube voltage is automatically preselected and the tube current is automatically modulated by CARE kV. Depending on age-specific CT protocols, technical parameters of PCCT for phantom head CT scans were automatically set ([Table 1]). Each phantom was scanned with a stepwise increase in IQLs up to the protocol-defined DRLs. These DRLs are defined as 35 mGy for 1–5-year-old children, 40 mGy for 5–10-year-old children, 45 mGy for 10–15-year-old children, and 55 mGy for adults [22].

Table 1 Technical scan parameters on the photon-counting CT scanner. BMI: body mass index.

Technical parameters on photon-counting CT scanner
Phantom	1-year-old	5-years-old	10-years-old	Adult female	Adult male
kVp	70	70	70	120	120
Scan length [mm]	124.4 (0.4)	122.4 (0.5)	134.4 (0.9)	145.3 (0.4)	140.9 (0.3)
Reconstructed slice thickness [mm]	1.5	1.5	1.5	1.0	1.0
Iterative reconstruction strength	2	2	2	4 (parenchyma), 2 (bone)
Reconstruction kernel	Brain parenchyma: Hr40 Bone: Hr60			Brain parenchyma: Hr44 Bone: Hr64
Pitch	0.35
Rotation time [s]	0.5
Collimation	57.6 mm (144 × 0.4 mm)

Fig. 1 Visualization of the 5 anthropomorphic phantoms and the ROI measurements of the brain parenchyma. a: two adult and three pediatric phantoms (adult male, adult female, 10 years, 5 years, and 1 year). b: ROI measurements of brain parenchyma tissue equivalent for IQL 10 (b), IQL 100 (c) and IQL 200 (d) of the adult male phantom. IQL: image quality level, ROI: region of interest.

Quantitative data and dose assessment

Across varying IQLs, different parameters for image quality and radiation dose were directly or indirectly obtained. This includes the volume computed tomography dose index (CTDI_vol), the dose length product (DLP), and the organ dose. The CTDI_vol is an estimation of the radiation dose calculated by the scanner taking into account the pitch factor. The DLP captures the radiation dose along the length of the scan and was derived by multiplying the CTDI_voland the scan length. The organ dose considers conversion factors of different sensitivities of different organs to radiation in addition to the DLP [1]. Automated dose monitoring software (Radimetrics Enterprise Platform, Bayer Healthcare, Leverkusen, Germany) applying Monte Carlo simulation techniques was used to calculate organ doses, particularly of the brain and the lens for head CT scans. As quantitative image quality parameters, signal and noise were measured in the brain parenchyma and bone tissue equivalents of the phantoms, and based on these values, the SNR was calculated for each IQL ([Fig. 1]). For the measurements, a 5 cm² region of interest (ROI) was placed at a consistent location in the brain parenchyma-equivalent material. For bone, a 1 cm² ROI was positioned in the sphenoid bone equivalent.

Statistical analysis

Statistical analyses were performed using SPSS version 29.0 (IBM Corp., New York, USA) and Python (v3.11). Normality of data distribution was evaluated using the Shapiro-Wilk test. For small sample sizes (n<30), normality was not assumed due to the limited power of the test [23]. Data following a normal distribution were reported as mean ± standard deviation (SD) while non-normally distributed data were presented as median and interquartile range (IQR). Group comparisons were restricted to matching IQLs enabling paired analyses. The Wilcoxon signed-rank test was applied for two-group comparisons, and the Friedman test for comparisons involving more than two groups. To compare pediatric and adult phantoms, values at each matched IQL were averaged within groups resulting in a single pediatric and a single adult value per IQL, which were then compared as paired samples. Percentage differences between male and female phantoms were assessed using an area under the curve (AUC) analysis via trapezoidal integration in Python. For measurements showing a linear association with IQL, linear regression models without an intercept were fitted. Assumptions regarding the normality and homoscedasticity of residuals were checked, and robust standard errors (HC3) were used when these assumptions were not met. Differences in slopes were tested using a t-test for two groups or an F-test on interaction terms for three groups. While Wilcoxon and Friedman tests assess differences at discrete IQLs, they do not capture trends across IQLs. Slope comparisons thus offer additional insight into group differences in the rate of change over the entire IQL range. Due to the non-linear and variable nature of SNR patterns, only non-parametric methods were used for group comparisons in these cases. Statistical significance was defined as p < 0.05 (two-sided).

Results

First, the relationship between IQL, radiation dose, and SNR for both brain parenchyma and bone was analyzed. Second, differences in these parameters between adult and pediatric phantoms were assessed.

Relationship between IQL, radiation dose, and SNR

The relationship between radiation dose, in particular for mAs, CTDI_vol, DLP, and organ doses, and increasing IQLs was analyzed across all phantoms ([Fig. 2]). A perfect linear relationship between IQL and effective mAs was observed (all R² = 1.00). Furthermore, this linear relationship also applied for other radiation dose metrics. In contrast to the linear radiation dose relationships, the SNR also increased with the IQL but showed minor deviations from linearity ([Fig. 3]). To further characterize these trends, SNR data were fitted using linear, quadratic, and exponential saturation models ([Fig. 4]). In all phantom types, the SNR increased with a rising IQL, while the exact shape of the curves varied slightly. For brain parenchyma, all models demonstrated high coefficients of determination (R² ≥ 0.94), with only subtle differences between them. The 5-year-old and 10-year-old phantoms showed a mild tendency toward saturation behavior that was best approximated by the quadratic or exponential model, whereas the 1-year-old and adult phantoms followed an almost linear trend. For bone SNR, both linear and quadratic fits yielded comparable results across all phantoms (R² ≥ 0.84).

Fig. 2 Linear relationship between IQLs and radiation dose parameters across different phantoms and protocols. All pediatric phantoms were scanned at 70 kVp, all adult phantoms at 120 kVp. a: Relationship between IQL and effective mAs. b: Relationship between IQL and CTDI_vol. CTDI_vol: volume computed tomography dose index, IQL: image quality level. All other dose metrics (DLP and organ doses) follow a similar linear trend with IQL and are not shown for clarity.

Fig. 3 Less linear relationship between IQL and SNR for brain parenchyma. a: Relationship between IQL and SNR for parenchyma. b: Relationship between IQL and SNR for bone. c: Relationship between IQL and signal of parenchyma. d: Relationship between IQL and signal of bone. e: Relationship between IQL and noise of parenchyma. f: Relationship between IQL and noise of bone. DRL: diagnostic reference level, IQL: image quality level, SNR: signal-to-noise ratio.

Fig. 4 Relationship between SNR and IQL across five phantom types, modeled using linear, quadratic, and exponential saturation fits, with goodness-of-fit quantified by the coefficient of determination (R²). a: SNR_{brain parenchyma} for adult male phantom, b: SNR_bone for adult male phantom, c: SNR_{brain parenchyma} for adult female phantom, d: SNR_bone for adult female phantom, e: SNR_{brain parenchyma} for the 10-year-old phantom, f: SNR_bone for the 10-year-old phantom, g: SNR_{brain parenchyma} for the 5-year-old phantom, h: SNR_bone for the 5-year-old phantom, i: SNR_{brain parenchyma} for the 1-year-old phantom, j: SNR_bone for the 1-year-old phantom. IQL: image quality level, SNR: signal-to-noise ratio.

Comparison between IQL, radiation dose, and SNR between pediatric and adult phantoms

First, general differences between pediatric and adult phantoms were analyzed ([Table 2]). Significant differences were observed for effective mAs and all radiation dose parameters (p<0.05). While adult phantoms required a significantly lower effective mAs, they exhibited significantly higher radiation doses. Pediatric phantoms scanned at 70 kVp generally exhibited a higher SNR in brain parenchyma compared to adult phantoms scanned at 120 kVp, reflecting both the lower kVp and smaller phantom size. Significant differences were found in SNR values for both brain parenchyma and bone (p=0.018) with adult phantoms showing overall lower SNR values compared to pediatric phantoms.

Table 2 Comparison of different radiation dose parameters and SNR between pediatric and adult phantoms on PCCT. To ensure comparability, only IQLs at which all phantoms were scanned were included in the analysis (IQL 10–300, n=7 per group). CTDI_vol: volume computed tomography dose index, DLP: dose length product, IQL: image quality level, PCCT: photon-counting CT, SNR: signal-to-noise ratio.

Category across different IQLs	Pediatric phantoms on PCCT	Adult phantoms on PCCT	p-value
mAs	170.18 (225.06)	54.41 (72.50)	0.018
Effective mAs	59.56 (78.77)	19.04 (25.37)	0.018
CTDIvol[mGy]	18.03 (23.75)	26.55 (35.56)	0.018
DLP [mGy·cm]	229.67 (303.63)	380.00 (508.50)	0.018
Effective dose [mSv]	1.07 (1.41)	0.98 (1.32)	0.028
SNRbrain parenchyma	7.51 (4.42)	6.96 (4.85)	0.018
SNRbone	27.51 (9.52)	16.05 (9.50)	0.018
Signalbrain parenchyma	50.00 (0.33)	44.50 (1.00)	0.018
Noisebrain parenchyma	6.67 (4.67)	6.50 (5.50)	0.551
Signalbone	1041.00 (5.00)	904.00 (5.00)	0.018
Noisebone	39.00 (12.33)	56.50 (35.00)	0.018

The comparison between the male and female adult phantoms is summarized in [Table 3]. Significant differences in effective mAs and radiation dose metrics were observed. CTDI_vol and DLP were significantly higher in the female phantom (p=0.028 and p=0.018, respectively), with absolute differences in AUC of 2.94% for CTDI_vol and 6.05% for DLP across the IQL range. Differences in organ doses were more pronounced, particularly for red bone marrow, where the AUC difference reached 38.40%. The slopes of radiation dose increase across IQLs also differed significantly between male and female phantoms for all dose parameters (p<0.001, suppl. Tab. 1). Regarding SNR, significantly higher values for brain parenchyma were observed in the female phantom (p=0.028) corresponding to an AUC difference of 13.11%. No significant differences were observed in SNR for bone (p=0.063).

Table 3 Sub-analysis within adult phantoms: Comparison of different dose parameters and SNR between the female and male adult phantoms. Comparison of the same IQLs within these groups. AUC: area under the curve, CTDI_vol: volume computed tomography dose index, DLP: dose length product, IQL: image quality level, PCCT: photon-counting CT, SNR: signal-to-noise ratio.

	Median (IQL) and p-value within IQL 10–300 (n=7 per group)
Category across different IQLs	Adult phantom, female	Adult phantom, male	Difference in AUC [%]	p-value
mAs	54.92 (73.05)	53.90 (71.94)	1.60	0.028
Effective mAs	19.22 (25.57)	18.87 (25.18)	1.60	0.028
CTDIvol[mGy]	26.90 (36.18)	26.20 (34.94)	2.94	0.028
DLP [mGy·cm]	391.00 (524.00)	369.00 (493.00)	6.05	0.018
Brain [mSv]	25.38 (34.10)	22.00 (29.37)	15.57	0.018
Lens [mSv]	33.24 (44.70)	31.49 (41.99)	5.84	0.018
Salivary glands [mSv]	25.38 (34.10)	22.00 (29.37)	15.57	0.018
Skin [mSv]	2.50 (3.35)	2.01 (2.69)	24.37	0.018
Muscle [mSv]	1.07 (1.43)	0.86 (1.15)	24.32	0.018
Red bone marrow [mSv]	2.55 (3.42)	1.85 (2.47)	38.40	0.018
Skeleton [mSv]	10.43 (13.98)	8.39 (11.21)	24.37	0.018
Effective dose [mSv]	1.08 (1.45)	0.89 (1.18)	22.07	0.018
SNRbrain parenchyma	7.50 (4.60)	6.43 (4.91)	13.11	0.028
SNRbone	16.87 (11.04)	15.23 (7.87)	8.22	0.063
Signalbrain parenchyma	44.00 (1.00)	45.00 (0.00)	1.75	0.157
Noisebrain parenchyma	6.00 (5.00)	7.00 (6.00)	15.66	0.023
Signalbone	894.00 (5.00)	914.00 (6.00)	2.44	0.017
Noisebone	53.00 (37.00)	60.00 (33.00)	8.35	0.027

Comparisons among the three pediatric phantoms are summarized in [Table 4] and showed consistently significant differences in effective mAs (p<0.001), CTDI_vol, and DLP (both p<0.001) as well as in organ doses (all p<0.001), including the lens (p=0.011). The slopes of the IQL-radiation dose relationships also differed significantly between all pediatric phantoms (p<0.001, suppl. Tab. 2). In general, older age-specific phantoms were associated with higher effective mAs and dose values. However, regarding organ doses, the 1-year-old phantom showed the highest values, particularly for red bone marrow. Significant differences were also observed in SNR for both brain parenchyma (p=0.002) and bone (p=0.021), with the 1-year-old phantom exhibiting the highest SNR values in both tissues. Notably, despite the higher overall SNR, this phantom showed the lowest signal values in bone compared to the 5- and 10-year-old phantoms.

Table 4 Sub-analysis within pediatric phantoms: Comparison of different dose parameters and SNR between the three different pediatric phantoms. Comparison of the same IQLs within these groups. AUC: area under the curve, CTDI_vol: volume computed tomography dose index, DLP: dose length product, IQL: image quality level, PCCT: photon-counting CT, SNR: signal-to-noise ratio.

	Median (IQL) and p-value within IQL 1–300 (n=8 per group)
Category across different IQLs	1-year-old phantom	5-year-old phantom	10-year-old phantom	p-value
mAs	130.49 (225.30)	142.05 (245.34)	153.49 (264.92)	<0.001
Effective mAs	45.67 (78.85)	49.72 (85.87)	53.72 (92.72)	<0.001
CTDIvol[mGy]	13.75 (23.81)	15.10 (25.76)	16.30 (28.28)	<0.001
DLP [mGy·cm]	171.00 (296.15)	184.50 (317.33)	219.50 (379.85)	<0.001
Brain [mSv]	16.72 (28.95)	14.35 (24.53)	13.67 (23.68)	<0.001
Lens [mSv]	21.12 (36.57)	21.02 (35.87)	21.37 (37.07)	0.011
Salivary glands [mSv]	16.72 (28.95)	14.35 (24.53)	13.67 (23.68)	<0.001
Skin [mSv]	4.11 (7.11)	2.61 (4.50)	2.11 (3.66)	<0.001
Muscle [mSv]	1.93 (3.35)	1.01 (1.75)	0.98 (1.70)	<0.001
Red bone marrow [mSv]	4.72 (8.18)	2.45 (4.20)	1.78 (3.07)	<0.001
Skeleton [mSv]	13.60 (23.55)	11.88 (20.40)	9.11 (15.77)	<0.001
Effective dose [mSv]	1.18 (2.05)	0.79 (1.35)	0.70 (1.21)	<0.001
SNRbrain parenchyma	7.71 (6.39)	6.71 (5.80)	6.70 (4.61)	0.002
SNRbone	29.06 (15.21)	25.73 (10.78)	22.75 (4.64)	0.021
Signalbrain parenchyma	54.00 (2.50)	46.00 (3.00)	50.00 (3.25)	<0.001
Noisebrain parenchyma	7.00 (7.00)	7.00 (9.50)	7.50 (7.00)	0.024
Signalbone	928.50 (10.25)	1077.50 (10.00)	1114.50 (12.00)	<0.001
Noisebone	32.50 (18.25)	42.00 (15.75)	49.00 (8.25)	<0.001

Discussion

This study examined the relationship between IQL, radiation dose, and SNR in adult and pediatric head phantoms using PCCT. While the radiation dose increased linearly with the IQL, the SNR demonstrated a less linear pattern with characteristic plateaus.

Our analysis confirmed a strong, strictly linear relationship between IQLs and radiation dose, aligning with previous studies [21]. This consistent scaling supports the technical reliability and internal validity of IQL as a measure of dose modulation in PCCT imaging. However, the relationship between IQL and SNR was more complex and showed minor deviations from linearity rather than pronounced intermediate plateaus. While a quadratic relationship has been suggested in the literature [14] [24], our findings indicate only mild quadratic or saturating trends for some pediatric phantoms, particularly the 5-year-old and 10-year-old phantoms, whereas the 1-year-old and adult phantoms followed an almost linear trend. For the SNR in brain parenchyma, both quadratic and exponential saturation models provided slightly better fits for the 5-year-old and 10-year-old phantoms. For the SNR in bone, the increase was generally less pronounced than in the parenchyma, resulting in flatter curves and comparable model fits between linear and quadratic models. The observed mildly nonlinear relationship between IQL and SNR may partly stem from the use of iterative reconstruction algorithms, which are known to reduce image noise more effectively and can exhibit nonlinear characteristics, leading to plateauing of SNR improvements at higher dose levels [25] [26] [27]. These trends may also be influenced by the size and age of the phantoms in combination with the lower kVp used for pediatric scans. Overall, these findings suggest that while the SNR generally increases with the IQL, the deviation from linearity is minor, and the gain in SNR diminishes only slightly at higher IQLs, indicating a potential, but modest, limit beyond which increased radiation exposure yields minimal additional image quality improvement.

When comparing adult and pediatric phantoms, statistically significant differences were observed in effective mAs and radiation dose metrics. Pediatric protocols applied a significantly higher tube current, which can be explained by the use of a lower tube voltage (70 kVp in pediatric protocols vs. 120 kVp in adult protocols). A lower kVp is sufficient to achieve adequate image quality in pediatric imaging due to the smaller body size and less tissue attenuation. However, since radiation exposure increases approximately with the square of the tube voltage [28] [29], adults received higher radiation doses despite a lower mAs. Regarding SNR, pediatric phantoms demonstrated a higher SNR of osseous structures which contain more calcium. This can be attributed to the enhanced photoelectric effect at lower kVp levels, where the photoelectric absorption probability is approximately proportional to Z³/E³ (with Z being the atomic number and E the photon energy) [28]. In the comparison between male and female adult phantoms, a significant difference in the SNR of the brain parenchyma was observed, but not in the SNR of bone. This highlights the fact that image quality evaluation should be tissue-specific since the SNR may not serve as a universal quality metric across all anatomical regions. Among the pediatric phantoms, as expected, larger children required a higher mAs to achieve the same IQL at a constant tube voltage due to increased attenuation with body size. Interestingly, the 1-year-old phantom showed lower signal intensity in bone compared to the older pediatric models. This observation likely reflects the underlying anatomical modeling of pediatric phantoms, where bone structures in younger children are less mineralized and are, therefore, less radiodense leading to lower signal values in CT [30].

This study is not without limitations. First, this is a solely phantom-based study without patient measurements. Interindividual differences within an age group cannot be taken into account. However, multiple CT measurements would not be ethically justifiable, especially in children, so that phantom measurements are inevitable. Secondly, the phantoms lack internal contrast variations, limiting the use of CNR metrics, which are more directly linked to diagnostic performance. Future studies should incorporate task-based assessments, including reader studies and the evaluation of CNR or detectability indices, to validate the clinical relevance of radiation dose/image quality trade-offs observed here. Third, due to the high standardization of the phantoms, no repeated scans were performed, which is unlikely to affect the observed trends. Nevertheless, future studies including repeated measurements and testing different reconstruction settings could further validate and refine these findings. Fourth, for the comparison between adult and pediatric groups, results were based on averaged phantom data per IQL. While this approach reflects the study’s main focus, individual phantom comparisons with stricter multiple-testing correction (e.g., Bonferroni) yield fewer statistically significant results. Lastly, it should be noted that reconstruction parameters, particularly the use of different kernels (Hr44 vs. Hr40 for brain parenchyma and Hr64 vs. Hr60 for bone) and iterative reconstruction strengths (level 4 in adults vs. level 2 in pediatric protocols for brain parenchyma), may influence the measured SNR. Sharper kernels generally improve spatial resolution but increase image noise, whereas softer kernels yield smoother images with a higher SNR [31] [32]. A higher iterative reconstruction level generally provides stronger noise suppression, which may partially compensate for the noise introduced by the sharper adult kernels [33]. Additionally, the tube voltage and slice thickness differ between groups (120 kVp for adults, 70 kVp for children) to reflect standard clinical practice, as scanning adults at 70 kVp or children at 120 kVp would be clinically inappropriate. This variability may slightly limit direct comparability between adult and pediatric protocols. Nevertheless, pediatric and adult phantoms were each scanned under consistent kernel and reconstruction settings within their respective groups, ensuring valid intra-group comparisons. Furthermore, our approach reflects clinically established and age-appropriate head CT protocols, which were intentionally preserved to maintain clinical relevance rather than testing artificially standardized but non-clinical acquisition settings. Additionally, future work could directly compare PCCT and EID-CT using these phantom protocols and complement objective SNR analyses with radiologist-based image quality assessments.

Conclusion

In conclusion, our results demonstrate clear differences in SNRs and radiation dose performance between pediatric and adult head PCCT protocols, providing a foundation for radiation dose optimization.

Conflict of Interest

The authors declare that they have no conflict of interest.

Acknowledgement

We would like to thank Ann-Christin Jacoby and Melanie Ebenau for their help with this work. We also thank the Federal Office for Radiation Protection (BfS) for providing the phantoms used in this study. In particular, we are grateful to Helmut Schlattl and Patrizia Kunert for their support and valuable contributions.

• References

• 1 Bos D, Guberina N, Zensen S. et al. Radiation Exposure in Computed Tomography. Dtsch Arztebl Int 2023; 120: 135-141
• 2 Seibert JA. Tradeoffs between image quality and dose. Pediatr Radiol 2004; 34
• 3 Willis CE, Slovis TL. The ALARA concept in pediatric CR and DR: Dose reduction in pediatric radiographic exams – A white paper conference Executive Summary. Pediatr Radiol 2004; 34
• 4 Bosch de Basea Gomez M, Thierry-Chef I, Harbron R. et al. Risk of hematological malignancies from CT radiation exposure in children, adolescents and young adults. Nat Med 2023; 29: 3111-3119
• 5 Schegerer A, Loose R, Heuser LJ. et al. Diagnostic Reference Levels for Diagnostic and Interventional X-Ray Procedures in Germany: Update and Handling. RoFo Fortschritte auf dem Gebiet der Rontgenstrahlen und der Bildgeb Verfahren 2019; 191: 739-751
• 6 Flohr T, Petersilka M, Henning A. et al. Photon-counting CT review. Phys Medica 2020; 79: 126-136
• 7 Siegel MJ, Ramirez-Giraldo JC. Photon counting detector computed tomography in pediatric cardiothoracic CT imaging. Radiol Adv 2024; 1: umae012
• 8 Frings M, Welsner M, Mousa C. et al. Low-dose high-resolution chest CT in adults with cystic fibrosis: intraindividual comparison between photon-counting and energy-integrating detector CT. Eur Radiol Exp 2024; 8: 105
• 9 Pourmorteza A, Symons R, Sandfort V. et al. Abdominal imaging with contrast-enhanced photon-counting CT: First human experience. Radiology 2016; 279: 239-245
• 10 Woeltjen MM, Niehoff JH, Michael AE. et al. Low-Dose High-Resolution Photon-Counting CT of the Lung: Radiation Dose and Image Quality in the Clinical Routine. Diagnostics 2022; 12
• 11 Rajendran K, Voss BA, Zhou W. et al. Dose Reduction for Sinus and Temporal Bone Imaging Using Photon-Counting Detector CT with an Additional Tin Filter. Invest Radiol 2020; 55: 91-100
• 12 Jungblut L, Euler A, Von Spiczak J. et al. Potential of Photon-Counting Detector CT for Radiation Dose Reduction for the Assessment of Interstitial Lung Disease in Patients with Systemic Sclerosis. Invest Radiol 2022; 57: 773-779
• 13 Racine D, Mergen V, Viry A. et al. Photon-Counting Detector CT with Quantum Iterative Reconstruction: Impact on Liver Lesion Detection and Radiation Dose Reduction. Invest Radiol 2023; 58: 245-252
• 14 Webb A. Introduction to biomedical imaging. 2002
• 15 Desai N, Singh A, Valentino DJ. Practical evaluation of image quality in computed radiographic (CR) imaging systems. Med Imaging 2010 Phys Med Imaging 2010; 7622: 76224Q
• 16 Aliukonyte I, Caudri D, Booij R. et al. Unlocking the potential of photon counting detector CT for paediatric imaging: a pictorial essay. BJR|Open 2023; 6
• 17 Pourmorteza A, Symons R, Reich DS. et al. Photon-counting CT of the brain: In Vivo human results and image-quality assessment. Am J Neuroradiol 2017; 38: 2257-2263
• 18 Dirrichs T, Tietz E, Röffer A. et al. Photon-counting versus Dual-Source CT of Congenital Heart Defects in Neonates and Infants: Initial Experience. Radiology 2023; 307
• 19 Hellms S, Werncke T, Böttcher J. et al. Reduction of radiation exposure and preserved image quality using photon-counting detector cardiac computed tomography without electrocardiographic gating in children with congenital heart disease. Eur Radiol 2025;
• 20 Naceanceno K, Thomas A, Zhang Z. et al. Quantifying performance enhancements in photon-counting pediatric head CT: a direct comparison with energy-integrating CT. Pediatr Radiol 2025; 55: 1690-1701
• 21 Tsiflikas I, Thater G, Ayx I. et al. Low dose pediatric chest computed tomography on a photon counting detector system – initial clinical experience. Pediatr Radiol 2023; 53: 1057-1062
• 22 Bundesamt für Strahlenschutz. Bekanntmachung der aktualisierten diagnostischen Referenzwerte für diagnostische und interventionelle Röntgenanwendungen. 2022 Accessed May 30, 2025 at: https://www.bfs.de/SharedDocs/Downloads/BfS/DE/fachinfo/ion/drw-roentgen.pdf?__blob=publicationFile&v=1
• 23 Hanusz Z, Tarasińska J. Normalization of the Kolmogorov–Smirnov and Shapiro–Wilk tests of normality. Biometrical Lett 2015; 52: 85-93
• 24 Infante JC, Liu Y, Rigsby CK. CT image quality in sinogram affirmed iterative reconstruction phantom study – is there a point of diminishing returns?. Pediatr Radiol 2017; 47: 333-341
• 25 Beister M, Kolditz D, Kalender WA. Iterative reconstruction methods in X-ray CT. Phys Medica 2012; 28: 94-108
• 26 Kawashima H, Ichikawa K, Matsubara K. et al. Quality evaluation of image-based iterative reconstruction for CT: Comparison with hybrid iterative reconstruction. J Appl Clin Med Phys 2019; 20: 199-205
• 27 Takata T, Ichikawa K, Mitsui W. et al. Object shape dependency of in-plane resolution for iterative reconstruction of computed tomography. Phys Medica 2017; 33: 146-151
• 28 Bushberg JT, Seibert JA, Leidholdt EM. et al. The essential physics of medical imaging. 3. Aufl. Lippincott Williams and Wilkins; 2012
• 29 Elojeimy S, Tipnis S, Huda W. Relationship between radiographic techniques (kilovolt and milliampere-second) and CTDI_vol . Radiat Prot Dosimetry 2010; 141: 43-49
• 30 CIRS. ATOM Dosimetry Phantoms. 2013 Accessed May 30, 2025 at: https://www.cirsinc.com/wp-content/uploads/2019/05/701–706-ATOM-PB-120418.pdf
• 31 Yang Y, Fink N, Emrich T. et al. Optimization of Kernel Type and Sharpness Level Improves Objective and Subjective Image Quality for High-Pitch Photon Counting Coronary CT Angiography. Diagnostics 2023; 13
• 32 Eldevik K, Nordhøy W, Skretting A. Relationship between sharpness and noise in CT images reconstructed with different kernels. Radiat Prot Dosimetry 2010; 139: 430-433
• 33 Krompaß K, Goldbrunner FA, Hartung V. et al. Combined influence of quantum iterative reconstruction level and kernel sharpness on image quality in photon counting CT angiography of the upper leg. Sci Rep 2024; 14: 27774

Publication History

Received: 15 August 2025

Accepted after revision: 04 November 2025

Article published online:
21 November 2025