© A. Khallouqi et al., Published by EDP Sciences 2025

1 Introduction

Pulmonary embolism (PE) remains a significant global health concern, with particular relevance in regions like Morocco where its prevalence continues to be high. The recent COVID-19 pandemic has further complicated this landscape, as SARS-CoV-2 infection has been associated with an increased incidence of thromboembolic events, including PE. This relationship between COVID-19 and coagulation disorders puts the attention to the critical importance of accurate and timely PE diagnosis, especially in the post-pandemic era (Ataalla, 2022; Mouzarou et al., 2022)⁠.

In this context, Computed Tomography Pulmonary Angiography (CTPA) has emerged as a focal diagnostic tool, offering unparalleled accuracy in detecting PE and other thoracic pathologies (Karimizarchi and Chaparian, 2017; Semghouli et al., 2024)⁠. CTPA employs intravenous contrast media to visualize the pulmonary arterial system, enabling the delineation of the main pulmonary artery and its branching network⁠. The technique’s diagnostic power lies in its ability to reveal filling defects within the pulmonary vasculature, allowing radiologists to identify potentially life-threatening conditions with high precision (Saeedi-Moghadam et al., 2021)⁠.

The COVID-19 pandemic has expanded CTPA’s role beyond PE diagnosis to include differentiation between COVID-19-related lung changes and other pulmonary complications. However, this increased utilization raises significant concerns regarding cumulative radiation exposure to patients. As CTPA examinations become more frequent, the medical physics community faces a pressing need to optimize radiation dose while keeping the required diagnostic efficacy (Yeung, 2019).

Traditional methods of estimating radiation dose in CT examinations, such as the volume computed tomography dose index (CTDI_vol) and dose length product (DLP), rely on standardized phantoms. These approaches fail to account for the diverse range of patient body habitus encountered in clinical practice, potentially leading to inaccuracies in dose estimation. To address this limitation, the concept of Size-Specific Dose Estimates (SSDE) was introduced by Association of Physicists in Medicine (AAPM) (Hadipour et al., 2022; Hakme et al., 2023; Khallouqi et al., 2024).

The conventional SSDE methodology utilizes the effective diameter (D_eff), derived from anteroposterior (d_AP) and lateral (d_LAT) measurements, to estimate patient-specific radiation dose (El Fahssi et al., 2024; Sekkat et al., 2024b) ⁠. While this approach has proven effective in abdominal and pediatric imaging, it may underestimate the absorbed radiation dose in regions with significant tissue inhomogeneities, such as the chest (Xu et al., 2019)⁠. To overcome these limitations, particularly in thoracic imaging, the water-equivalent diameter (D_W) has been proposed as an alternative metric for SSDE calculation accounting for both patient size and tissue composition (Gabusi et al., 2016a).

The aim of this study is to enhance the accuracy of radiation dose estimation in CTPA by comparing two SSDE methods: one based on effective diameter (SSDE_Deff) and the other on water-equivalent diameter (SSDE_Dw). Additionally, the study aims to compare the effective diameter (D_eff) and water-equivalent diameter (D_w) with lateral (d_LAT) and anteroposterior (d_AP) diameters to assess their correlation and reliability in patient size assessment. The research also evaluates the underestimation of conventional CT dose measurements (CTDIvol) relative to SSDE_Dw.

2 Materials and methods

In this retrospective study, a cohort of 200 CT pulmonary angiography scans was curated from an initial pool of 230 examinations. The final sample comprised 96 male and 104 female patients, all of whom underwent imaging in accordance with the standardized CT CTPA helical scanning protocol delineated in Table 1. Exclusion criteria were assessed by removing from consideration any scans that failed to capture the patient’s entire body within the field of view or those compromised by artifacts resulting from metallic implants. This screening process was essential to maintain the integrity of this study’s data and the validity of subsequent analyses.

All examinations were performed using a specific CT scanner model Optima CT520 Series from G.E. Healthcare. The utilization of a single scanner model across all examinations minimized potential variability in image acquisition parameters. The contrast injection technique involved tracking a bolus by placing a region of interest (ROI) on the main pulmonary artery. Patients received an injection of iodinated contrast media (40–80 ml/s). The scan was automatically triggered when the density in the ROI reached 100 Hounsfield units, typically 3–12 seconds after injection.

For each patient, anonymized CT scans were retrieved from the hospital’s Picture Archive and Communication System (PACS), ensuring patient confidentiality throughout the analysis.

To manually measure the effective diameter (D_eff), each CT pulmonary angiography (CTPA) scan was processed to extract the lateral (d_LAT) and anteroposterior (d_AP) diameters of the patient. This involved selecting the central axial slice that best represented the midsection of the patient’s chest knowing that there is no established protocol for measuring these dimensions until nowdays (Sekkat et al., 2024). On this slice, the d_LAT and d_AP dimensions were measured using a digital caliper tool within the imaging software, ensuring alignment with the outer contours of the patient’s body (AAPM, 2011) (Fig.1)⁠:

$$D_{\text{eff}}=\sqrt{d_{Ap}\times d_{Lat}}.$$(1)

The patient D_W was determined for each scan location following the methodology outlined in AAPM (2014)⁠. This calculation involved the formula (Eq. (2)):

$$D_W=2\sqrt{\left(\frac{1}{1000}{\displaystyle \overline{H{U}_{ROI}}}+1\right)\frac{A_{ROI}}{\pi }}.$$(2)

A_ROI is the patient’s area and ***${\displaystyle \overline{H{U}_{\text{ROI}}}}$ is the mean Hounsfield Unit value of the patient’s image.

SSDE was derived both from D_eff and D_w (respectively SSDE_Deff and SSDE_Dw. In the first case, SSDE_Deff is computed following equation (3):

$$SSD{E}_{deff}={\displaystyle \overline{CTD{I}_{vol}}}\times f\left(D_{eff}\right),$$(3)

where CTDI_vol was the mean over the whole scan range, while f(D_eff) was the size-dependent conversion factor given in AAPM (2011)⁠, depending on D_eff, calculated in the middle of the scan interval.

In addition, at each image location, SSDE was calculated as a function of D_w where f(D_w) is the conversion factor based on the measured D_w (AAPM, 2014) (Eq. (4)):

$$SSD{E}_{dw}=CTD{I}_{vol}\times f\left(D_w\right).$$(4)

Due to numerous low-density tissues and elements within the chest, which can significantly reduce radiation attenuation, D_eff demonstrates marked divergence from D_w, rendering it unsuitable for precise dosimetric calculations. To address this, the relative contribution of low-attenuating regions, represented by F_LA, was calculated for each patient based on the image at the center of the scan range. Specifically, F_LA was determined by counting the number of pixels within the body contour that had values lower than −700 HU (indicating low-attenuating tissues such as fat), and normalizing this count to the total number of pixels in the region (Fig. 2). Additionally, the relationship between ***$F_D=\frac{d_{EFF}}{d_W}$ ratio and F_LA was analyzed for both male and female patients.

The data was analyzed using IBM SPSS version 23.0 (IBM Corporation, Armonk, NY, USA). Descriptive statistics, including the mean, standard deviation, minimum, and maximum values, were computed.

Fig. 1 Effective diameter measurement at the mid-slice of the CT lung images.

Fig. 2 Hounsfield unit histogram for the central slice, highlighting low-attenuating regions (HU < −700).

3 Results

The gender distribution within the study population (48% male, 52% female) closely approximates the general demographic balance, potentially offering insights into gender-specific variations in CTPA findings. The mean age was 60.5 years and between 28 and 94 years old. Analysis of dosimetric data reveals a correlation between dose parameters and patient morphology. The CTDIvol shows a notable increase as a function of the patient’s lateral diameter (Fig. 3).

Figure 3 illustrates the relationship between D_LAT and dose measurements in CTPA, demonstrating a positive correlation for both CTDIvol and SSDE_W. A dispersion of data was observed during the analysis, with CTDIvol values ranging from 5.1 to 11.5 mGy and SSDE values extending from 5.0 to 15.8 mGy, across a LAT (lateral) range of 22 to 38 cm. Trend lines indicate a more pronounced increase in SSDE compared to CTDIvol as d_LAT increases, evidenced by the slopes of their respective regression lines. This divergence becomes more accentuated at higher LAT values due to the importance of taking into consideration attenuation in dose estimation, particularly for patients with larger body habitus.

Comparison between SSDE_Deff and SSDE_Dw reveals an average overestimation of 10% for SSDE_Deff. However, in-depth analysis of the distribution highlights considerable variability, ranging from −12% (obese patients) to +41% (thin patients). This dispersion shows the importance of accounting for body composition in dose estimation.

In Figure 4, a scatterplot of SSDE using a linear regression revealed a squared correlation coefficient of R² = 0.8108 between SSDE computed from D_eff against SSDE computed using D_w.

In Figure 5, the distributions (density) of the D_W and D_eff measurements are presented. The average value of D_W was calculated to be 23.75 cm, with a standard deviation of 3.24 cm. In contrast, the average value of D_eff reached 25.91 cm, accompanied by a standard deviation of 2.70 cm. It is noteworthy that there were no D_W values exceeding 32 cm, while 3% of the D_eff measurements were observed beyond this threshold, indicating a limited occurrence of high values.

Figure 6 illustrates the relationship between F_D and F_LA, derived from 96 male and 104 female examinations. Distinct marker shapes were employed to differentiate between genders. The corresponding fit coefficients are presented in Table 2.

Figure 7 shows the relationship between D_eff, D_W, and d_LAT. In this graph, both D_eff and D_w are plotted as a function of d_LAT, revealing high positive correlations. Coefficients of determination were calculated as 0.9175 for D_eff and 0.7578 for D_W, indicating a strong relationship between d_LAT and D_eff. Linear trend lines fitted to the data resulted in equations of y = 0.7003x + 3.8434 for D_eff and y = 0.7044x + 2.0385 for D_W. Generally, D_eff values were found to be greater than D_W, primarily due to a higher y-intercept, while the slopes remained very similar. This consistent difference suggests a systematic overestimation of D_eff compared to the mid-height diameter, which could be attributed to differences in measurement methods or intrinsic characteristics of the measured objects. The high correlations between d_LAT and both diameter measurements indicate good linear relationships, with D_eff potentially being a more precise indicator of overall object size due to its stronger correlation with d_LAT.

Figure 8 illustrates the relationship between D_W and D_eff as a function of d_AP. The scatter plot presents a high positive correlation between both diameter measurements and d_AP, with D_eff consistently exceeding D_W by an average of 7.73%. Nearly parallel trend lines for D_W and D_eff suggest a consistent relationship across the AP range.

Fig. 3 dLAT correlation with CTDIvol and SSDEDw in CTPA.

Fig. 4 Correlation between SSDEDw and SSDEDeff: scatter plot with linear regression analysis.

Fig. 5 The distribution of Dw (indicated by the dashed line) and Deff (shown by the dotted line).

Fig. 6 Relationship between FD and FLA based on patient gender, with distinct markers for each gender.

Fig. 7 Comparison of Dw and Deff diameters as function of dLAT.

Fig. 8 Comparison of Dw and Deff diameters as function of dAP.

4 Discussion

This study presents findings that aim to contribute to the understanding of dose estimation complexities in CT, particularly within the context of computed tomography pulmonary angiography examinations. The results highlight the limitations of conventional dose estimation methodologies, especially the reliance on the standard 32-cm-diameter water-filled phantom. This traditional approach has been shown to substantially underestimate the radiation dose received by patients, particularly when considering the diverse range of patient morphologies encountered in clinical practice. The analysis revealed that the CTDIvol consistently underestimates patient dose when compared to more tailored metrics, such as the size-specific dose estimate based on water-equivalent diameter (Ponnusamy et al., 2019; Sekkat et al., 2024a). This discrepancy, which varies from 8% to 26% depending on patient size, with an average difference of approximately 18%, underscores the urgent need for more accurate dose estimation techniques that reflect individual patient characteristics. This is a direct consequence of using Automatic Tube Current Modulation (ATCM). This technology adjusts tube current based on tissue attenuation, resulting in a CTDIvol increase of approximately 60% for patients of larger build compared to those of smaller stature as presented in the previous study (Sekkat et al., 2024c).

The SSDE calculation based on effective diameter (D_eff) marked a significant advance in CT dosimetry. This methodology enables a more patient-specific assessment of radiation exposure while retaining clinical practicality. SSDE_Deff is distinguished by its simplicity and efficiency, requiring only linear measurements from a single CT image to calculate the effective diameter. This approach strikes an optimal balance between improved accuracy and ease of implementation, enhancing the precision of dose estimates without compromising workflow efficiency. However, it is essential to recognize that this method is not without its limitations. The study indicates that in anatomical regions characterized by substantial tissue inhomogeneities, the D_EFF-based approach may lead to either under- or overestimations of patient dose. This variability is contingent upon the average tissue attenuation within the scanned volume, emphasizing the need for ongoing refinement in dose estimation methodologies.

The findings reveal a strong positive correlation between d_LAT and both D_eff and D_w, with determination coefficients of R² = 0.9175 and R² = 0.7578, respectively. These correlations suggest that both D_eff and D_w provide comparable estimates of patient size for SSDE calculations, aligning with recent studies in the field.

However, the analysis also uncovered systematic differences between these metrics. SSDE values derived from D_eff were consistently higher than those based on D_w for the same patient, with an average difference of 7.73%. This systematic overestimation of D_eff relative to D_w is particularly notable in the thoracic region.

The observed discrepancies can be attributed to the distinct calculation methods employed for each metric. Both D_eff and D_w are derived from lateral and anteroposterior diameters, but they utilize different formulae (Pace et al., 2022)⁠. D_eff is calculated as the square root of the product of lateral and anteroposterior diameters, while D_w incorporates tissue attenuation information, effectively replacing heterogeneous tissues (such as lungs) with an equivalent thickness of water.

Radiation attenuation at a given anatomical location, such as the thoracic region, is influenced by tissue composition and density, which can vary significantly between patients. While radiation attenuation is accounted for in the calculation of ***$SSD{E}_{D_W}$, it does not impact the ***$SSD{E}_{D_{\text{eff}}}$ value. Even when D_eff values are the same across patients, the attenuation of radiation can differ due to tissue heterogeneities. This variability in tissue composition results in differences in radiation attenuation, which may affect dose calculations.

The relationship between the ***$\frac{D_{\text{eff}}}{d_W}$ratio (denoted as F_D) and the fraction of low-attenuating regions in the body (represented by F_LA) was examined. A stronger correlation between F_D and F_LA was found in male patients compared to female patients. This difference can likely be attributed to the greater variability in anatomical proportions among women, including differences in fat distribution and the presence of breast tissue, which impact the fraction of low-attenuating tissues in the thoracic region.

More specifically, these gender-related anatomical differences could help explain the biphasic curve observed for D_w in Figure 5 which is not observed for D_eff. The presence of low-attenuating tissues, such as fat or breast tissue in women, may result in the observed bimodal distribution of D_w, whereas D_eff remains more consistent due to its relative insensitivity to tissue heterogeneity.

As the fraction of low-attenuating tissues decreased (F_LA approached 0), both D_eff and D_w values tended to converge, with F_D approaching 1. This convergence is a result of increasing tissue homogeneity, where radiation attenuation becomes more uniform across the body, leading to more consistent dose estimates.

Clinically, these results underscore the importance of clearly specifying the size metric used (D_eff or D_w) when reporting SSDE, as the choice of one or the other can influence the reported numerical value by 5–10%. This is particularly critical when comparing doses across different hospitals or against diagnostic reference levels based on SSDE (Satharasinghe et al., 2022)⁠. Radiologists and medical physicists must be aware of these differences when interpreting and communicating doses to clinicians and patients.

This study’s strength lies in its comprehensive inclusion of diverse patient morphologies, encompassing diameters from 22 to 38 cm. This broad range facilitated an in-depth examination of the size-SSDE relationship across a spectrum of clinically relevant dimensions.

A notable limitation, however, is the absence of analysis regarding the influence of acquisition parameters such as tube voltage, current-time product, and pitch on the size-dose relationship. Further investigation is warranted to elucidate these intricate interactions and their implications for dose estimation accuracy.

5 Conclusion

The findings of this comprehensive study underscore the limitations of traditional methods for estimating radiation dose in CTPA examinations. The observed underestimation of patient-specific exposure using the standard CTDI highlights the need for more personalized dosimetry approaches that account for individual anatomical variations. The use of SSDE based on water-equivalent diameter demonstrated improved accuracy compared to effective diameter-based SSDE, yet notable variability persists, emphasizing the complexities of thoracic tissue composition. The strong correlations between patient lateral diameter and both SSDE metrics reinforce the importance of considering morphological factors in dose calculations. As CTPA utilization rises in the post-COVID-19 era, these insights contribute to ongoing efforts to enhance radiation safety through the refinement of dose estimation methodologies. Implementing personalized, sophisticated techniques will optimize radiation exposure while preserving the diagnostic efficacy of this critical imaging modality, ultimately improving patient care and outcomes.

Acknowledgments

Special thanks to Mr. khallouqi Abdellah for supervising data collection at the university hospital. We acknowledge the financial support from Hassan First University (Grant #FP202006).

Funding

This work was supported by Hassan First University (Grant #FP202006).

Conflicts of interest

The authors declare that they have no conflict of interest.

Data availability statement

The findings presented in this study are based on a retrospective dataset that is not publicly available due to privacy and confidentiality restrictions. Unfortunately, the data cannot be shared or made openly accessible to protect the privacy of the individuals included in the dataset. However, the authors are committed to providing any necessary information or clarification upon reasonable request.

Author contribution statement

A. Khallouqi: Writing original draft, Methodoloy, Investigation, H. sekkat : Visualization, Methodology, A. Halimi and O. El rhazouani : Supervision.

References

All Tables

All Figures

	Fig. 1 Effective diameter measurement at the mid-slice of the CT lung images.
In the text

	Fig. 2 Hounsfield unit histogram for the central slice, highlighting low-attenuating regions (HU < −700).
In the text

	Fig. 3 dLAT correlation with CTDIvol and SSDEDw in CTPA.
In the text

	Fig. 4 Correlation between SSDEDw and SSDEDeff: scatter plot with linear regression analysis.
In the text

	Fig. 5 The distribution of Dw (indicated by the dashed line) and Deff (shown by the dotted line).
In the text

	Fig. 6 Relationship between FD and FLA based on patient gender, with distinct markers for each gender.
In the text

	Fig. 7 Comparison of Dw and Deff diameters as function of dLAT.
In the text

	Fig. 8 Comparison of Dw and Deff diameters as function of dAP.
In the text