Unenhanced CT quantitative vascular morphometry for detecting central pulmonary embolism: a multicenter study

Introduction

Acute central pulmonary embolism (ACPE) is defined as a pulmonary embolism (PE) occurring in either unilateral or bilateral pulmonary arteries (1). The mortality rate of ACPE is significantly higher than that of peripheral PE (PE in the lobar arteries and beyond) (1-3). The early detection of ACPE is vital to save patients’ lives through prompt intervention (4).

The clinical manifestations of ACPE, which primarily include dyspnea, chest pain, and syncope, are nonspecific and often overlap with the symptoms of peripheral PE or other cardiopulmonary conditions (1). In primary care settings, general practitioners often misinterpret these symptoms as cardiovascular or pulmonary disorders. This can lead to delays in the diagnosis of PE. When such symptoms occur, unenhanced computed tomography (CT) provides a valuable initial screening tool to exclude alternative cardiopulmonary or mediastinal abnormalities (5). Thus, the accurate identification of ACPE on unenhanced CT could significantly reduce delayed diagnoses. Given these diagnostic complexities, it is imperative that a predictive model for ACPE detection using unenhanced CT imaging be developed.

Previous studies have established unenhanced CT as a reliable alternative for detecting ACPE (6-8). These studies have shown that vascular morphometric parameters, particularly the main pulmonary artery (mPA) diameter, can serve as independent predictive indicators for ACPE (6-8).

This multicenter study expands on previous studies by systematically measuring thoracic vessels related to the pulmonary artery (PA) and introducing several novel parameters. Moreover, this study establishes a diagnostic nomogram. A nomogram is a graphical computational tool that integrates multiple predictive variables to estimate event probabilities. This nomogram underwent rigorous internal and external validation to facilitate its clinical application. We present this article in accordance with the TRIPOD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-702/rc).

Methods

Study design

The study protocol was approved by the Ethics Committee of the Second Affiliated Hospital of Nanchang University (the lead center, approval No. MR-36-24-042480). All the participating hospitals were informed of and agreed to participate in the study. The requirement of individual informed consent was waived due to the retrospective nature of this study, which involved the analysis of existing data without any additional intervention or risk to participants. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

This retrospective multicenter study was conducted across three medical centers (Centers A, B, and C) to develop, validate, and externally test a diagnostic nomogram for ACPE. Center A (the Second Affiliated Hospital of Nanchang University) served as the primary cohort for both model development and internal validation, while a combined cohort from Centers B (Ji’an Central People’s Hospital) and C (Fengcheng People’s Hospital) was used as an independent test set for external validation. This approach ensured the robust evaluation of the nomogram’s generalizability across diverse clinical settings.

Patient data from Center A were retrospectively collected from January 2018 to December 2023, providing a sufficiently large and temporally broad dataset for model construction. While data from Centers B and C were collected from August 2023 to February 2024, forming a distinct external validation cohort.

Participants

This multicenter study enrolled 374 consecutive patients from three institutions: Center A (n=302) for model training, internal validation, and nomogram development; and Centers B (n=44) and C (n=28) for external validation. The cohort included 218 PE cases (105 central, and 113 peripheral) and 156 non-PE cases.

Patients were included in the study if they: (I) underwent unenhanced CT scans within 24 hours preceding computed tomography pulmonary angiography (CTPA) examination (9), irrespective of their final PE diagnosis status; and (II) were aged ≥18 years. Patients were excluded from the study if they: (I) had compromised image quality due to severe artifacts; (II) had advanced pulmonary pathologies, including severe pneumonia, atelectasis, primary vascular endothelioma, or lung cancer where PA tumor could not be definitively excluded; and/or (III) had severe congenital cardiovascular anomalies (e.g., atrial septal defect, ventricular septal defect, tetralogy of Fallot, or patent ductus arteriosus), or aortic dissection resulting in significant morphological alterations of major cardiothoracic vasculature. The inclusion and exclusion criteria are detailed in Figure 1.

Figure 1 Flowchart of patient enrollment in the current study. Center A: the Second Affiliated Hospital of Nanchang University; Center B: Ji’an Central People’s Hospital; Center C: Fengcheng People’s Hospital. ACPE, acute central pulmonary embolism; CT, computed tomography; CTPA, computed tomography pulmonary angiography; PE, pulmonary embolism.

Image acquisition and analysis

Image acquisition

Unenhanced CT scans at Center A were performed using the following three CT scanners: Philips Brilliance iCT, Philips IQon CT, and General Energy Revolution CT. The CT imaging protocols at Center A were standardized, and the following acquisition parameters were used: tube voltage, 120 kV; tube current modulation, 30–35 mA, and standardized pitch, 1.0 mm.

For CTPA examinations at Center A, the same CT platforms were employed with the following contrast-enhanced protocol: intravenous administration of iodinated contrast medium (350–370 mgI/mL) at an injection rate of 4.0–4.5 mL/s via an 18–20 gauge peripheral venous catheter. Automated bolus tracking was implemented with the region of interest (ROI) positioned in the PA trunk, using a trigger threshold of 100 Hounsfield units (HU) and a 5-second scan delay.

The imaging equipment configurations and acquisition parameters for Centers B and C are detailed in Table S1 and Figures S1,S2.

Image analysis

In accordance with established diagnostic standards (4), CTPA served as the reference standard for PE diagnosis in this study. Based on the CTPA findings, the patients were stratified into two distinct diagnostic categories:

• Category 1: ACPE, characterized by unilateral or bilateral filling defects in the mPA or saddle emboli at the bifurcation.
• Category 2: non-ACPE, defined as either peripheral PE or non-PE, comprising:
  • Peripheral PE, characterized by filling defects in lobar arteries or distal branches;
  • Non-PE, characterized by an absence of pulmonary arterial filling defects.

CTPA interpretation was performed independently by two board-certified radiologists with 4 years of specialized experience in CTPA. Any diagnostic discrepancies were resolved through consensus review with a senior radiologist with 30 years of experience in chest CT. All the interpreting radiologists were blinded to the clinical information and laboratory test results of the patients during the evaluation of the CTPA images.

Unenhanced CT quantitative vascular measurements

Blood vessel diameter

Based on comprehensive consideration of the pathophysiological alterations of acute PE and previous research findings (8,10,11), this study employed the following vascular parameters:

• Main pulmonary artery (mPA) diameter, mm: measured in the transverse plane at the bifurcation level of the mPA, corresponding to the mid-ascending portion of the mPA;
• Ascending aorta (AO) diameter, mm: measured on the transverse plane at the same level as the mPA measurement, representing the anteroposterior diameter of the AO;
• Main pulmonary artery-to-aorta (mPA:AO) diameter ratio: the transverse plane measurements of the mPA and AO diameters;
• Azygos vein (AzV) arch diameter, mm: measured at the level where the AzV arch joins the superior vena cava (SVC), representing the maximum short-axis diameter on the transverse plane;
• SVC diameter, mm: measured on the transverse plane approximately 1 cm below the carina, representing the maximum short-axis diameter of the SVC;
• Inferior vena cava (IVC) diameter, mm: the maximum short-axis diameter measured in the thorax, anterior and to the right of the lower esophagus.

This study innovatively quantified the longitudinal diameter of the main pulmonary artery (L-mPA) using multiplanar reconstruction (MPR). Specifically, at the level of the mPA bifurcation, a reconstruction plane analogous to the coronal plane was obtained using the perpendicular line to the long axis of the mPA as the reference, enabling the measurement of the L-mPA on this reconstructed plane (the measurement methodology is illustrated in detail in Figure 2).

Figure 2 Methodology for L-mPA diameter measurement using MPR. (A) Axial reference image demonstrating the orientation plane (red line) for MPR reconstruction. (B) Reconstructed MPR image orthogonal to the main pulmonary artery axis, with measurement calipers (green line) positioned using the Picture Archiving and Communication System software. L-mPA, longitudinal diameter of the main pulmonary artery; MPR, multiplanar reconstruction.

To maximize reproducibility and minimize interobserver variability, all the measurements were consistently performed using a fixed mediastinal window setting (window width: 40 HU; window level: 350 HU).

In summary, the vascular measurements in this study comprised the mPA, L-mPA, AO, mPA: AO ratio, AzV arch, SVC, and IVC. It should be noted that the AO was exclusively used to calculate the mPA:AO ratio rather than for model development.

CT attenuation at the origin of the pulmonary artery (CT-OPA)

The pathophysiological sequelae of ACPE include thrombotic occlusion of the PA, leading to hemodynamic disturbances (4), which may manifest as density fluctuations at the origin of the pulmonary artery (OPA). Accordingly, this study integrated the CT-OPA to conduct a comparative analysis among the patients.

To ensure the precision of the measurements, a standardized protocol was instituted. A radiologist, with three years of specialized experience in thoracic CT interpretation, conducted all the initial measurements on unenhanced chest CT scans. Employing a standardized circular ROI of approximately 2 cm², measurements were extracted 1–2 cm distal to the pulmonary valve plane, meticulously avoiding the vascular walls, perivascular tissues, and calcific deposits. For each patient, three consecutive measurements were obtained, and the mean value was adopted as the final measurement.

Variable repeatability

To ensure measurement reliability, two board-certified thoracic radiologists independently performed all the measurements on unenhanced CT scans under a strict blinding procedure (i.e., without access to the CTPA results or clinical data). Interobserver agreement was subsequently calculated by comparing the measurements of the radiologists.

Statistical analysis

The statistical analyses were performed using R version 4.1.0 (https://www.r-project.org/), SPSS Statistics version 26.0 and MedCalc version 20.0. The normality of the continuous variables was assessed using the Shapiro-Wilk test. The normally distributed variables were expressed as the mean ± standard deviation, while the non-normally distributed variables were reported as the median with the interquartile range (Q1, Q3). The categorical and ordinal variables were presented as the frequency with the percentage.

Comparative analyses were conducted using appropriate statistical tests based on data distribution characteristics. The continuous variables were analyzed using the independent samples t-test for the normally distributed data or the Mann-Whitney U test for the non-normally distributed data. The categorical variables were compared using Pearson’s Chi-squared test or Fisher’s exact test as appropriate. A two-tailed P value <0.05 was considered statistically significant.

L-mPA, mPA, and the mPA:AO ratio are PA diameter measurements. These measurements were organized into the following three distinct feature pools for predictive modeling:

• Feature Pool 1: L-mPA, AzV arch, SVC, IVC, and CT-OPA;
• Feature Pool 2: mPA, AzV arch, SVC, IVC, and CT-OPA;
• Feature Pool 3: mPA:AO ratio, AzV arch, SVC, IVC, and CT-OPA.

The dataset from Center A was divided into training and validation subsets using a 7:3 ratio. Data from Centers B and C were pooled and randomly shuffled to create an independent test set. Feature selection and model construction were performed using a bidirectional stepwise logistic regression analysis on the training set across multiple feature pools. The optimal performing model from the validation set was selected for nomogram development and subsequent external validation in the test set.

The model first calculates a weighted combination of variables, then converts it to a probability [0–1] using the sigmoid function, indicating the likelihood of belonging to the target category. A 0.5 threshold was applied for classification: probabilities ≥0.5 were classified as positive [1], and probabilities <0.5 were classified as negative [0].

Model performance was evaluated using multiple complementary approaches. Diagnostic accuracy was assessed by calculating the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Discriminative ability was determined using the area under the curve (AUC) of the receiver operating characteristic (ROC) curve. Calibration curves were generated to evaluate the agreement between the predicted probabilities and observed outcomes. Finally, a decision curve analysis (DCA) was performed to quantify the clinical net benefit across various probability thresholds.

The sample size for the logistic regression analysis in this study was determined using a widely guideline. This rule stipulates that a minimum of 10 to 20 events per predictor variable (EPV) is required to ensure stable and reliable parameter estimates (12). The following formula was applied:

N=(EPV×p)/r[1]

where N is the total minimum sample size required, EPV is the number of events per variable (set at 10 for a conservative estimate), p is the number of predictor variables included in the model, and r is the estimated probability of the outcome event.

Results

Patients

Comparison of baseline characteristics based on PE distribution

The age, IVC, CT-OPA, and L-mPA values of the PE patients were all significantly higher than those of the patients without PE (P<0.05). The IVC, CT-OPA, L-mPA values of the patients with ACPE were significantly higher than those of the patients with peripheral PE (P<0.05) (Table 1).

Comparative analysis of baseline characteristics across patient sets

The baseline demographic and clinical characteristics of the patients are set out in Table 2. There were no statistically significant differences between all the variables in the training and validation sets, suggesting that the division was free from subjective bias.

Variable repeatability and correlation

In this study, the overall intraclass correlation coefficients of all the variables were greater than 0.80, indicating good repeatability. Prior to model building, a correlation analysis was conducted on the training set specifically. Among the variables, the highest correlation was observed between the mPA and mPA:AO ratio (r=0.65, P<0.001). The correlations among the remaining features were not significant, suggesting no significant multicollinearity.

Development and internal validation of the nomogram

Feature Pool 1

In the training set, Model 1 was developed through bidirectional stepwise logistic regression analysis of Feature Pool 1. The stepwise selection process identified two significant predictors (P<0.05); that is, the IVC and L-mPA diameters. These variables maintained statistical significance (P<0.05) following adjustment for age and gender, demonstrating their robustness as independent predictors of ACPE (Figure 3A). The AzV arch, SVC diameter, and CT-OPA values were not ultimately incorporated into Model 1 (all P>0.10).

Figure 3 Multivariable logistic regression analysis. (A) Forest plot of Model 1. (B) Forest plot of Model 2. Each horizontal line in the plot represents the results of an individual study, with a black square at the midpoint indicating the estimated OR. The length of the horizontal line represents the confidence interval of the OR. aOR, adjusted odds ratio; ap, adjusted P value; AzV, azygos vein arch; CI, confidence interval; CT-OPA, computed tomography attenuation at the pulmonary artery; IVC, inferior vena cava; L-mPA, longitudinal diameter of the main pulmonary artery; OR, odds ratio.

Feature Pools 2 and 3

Consistent results from Feature Pools 2 and 3 led to the retention of three significant parameters (IVC diameter, AzV arch diameter, CT-OPA values) in Model 2 (Figure 3B), while excluding the non-significant mPA (or mPA:AO) and SVC diameters (all P>0.10). These features were used to construct Model 2. Following adjustment for age and gender, the AzV arch demonstrated non-significant association with the ACPE (P>0.05). Consequently, we developed an adjusted Model 2 (a-Model 2), excluding the AzV arch, which was subsequently validated in the validation set.

ROC curve analysis

The performance evaluation revealed significant differences in the predictive accuracy of the three models. On the training dataset, Model 1 demonstrated superior discriminative ability with an AUC of 0.869 [95% confidence interval (CI): 0.816–0.911] (Figure 4A). This performance advantage was maintained in the validation set, where Model 1 achieved an AUC of 0.821 (95% CI: 0.726–0.894) (Figure 4B).

Figure 4 ROC and calibration curves. (A) Training set ROC curves for Models 1 and 2. (B) Validation set ROC curves for Models 1 and 2. (C) Training set calibration curves for Models 1 and 2. (D) Validation set calibration curves for Models 1 and 2. Model 1: IVC and L-mPA; Model 2: IVC, CT-OPA and AzV; a-Model 2: IVC and CT-OPA. AzV, azygos vein arch; CT-OPA, computed tomograohy attenuation at the pulmonary artery; IVC, inferior vena cava; L-mPA, longitudinal diameter of the main pulmonary artery; ROC, receiver operating characteristic.

Conversely, Model 2 performed less favorably on the training dataset, achieving an AUC of 0.728 (95% CI: 0.663–0.787) (Figure 4A). In the validation set, the performance of Model 2 declined further, resulting in an AUC of 0.565 (95% CI: 0.456–0.669) (Figure 4B).

The performance of a-Model 2 on the training set was slightly inferior to that of Model 2, with an AUC of 0.703 (95% CI: 0.637–0.764). On the validation set, its performance was slightly superior to that of Model 2, with an AUC of 0.613 (95% CI: 0.504–0.713).

Calibration curve analysis

Although Model 2 and a-Model 2 demonstrated better alignment with the ideal curve on the training set (Figure 4C), their performance on the validation set revealed significant limitations, characterized by substantial curve fluctuations and a notably narrow range (Figure 4D). This contrast between training and validation performance suggests potential overfitting in Model 2 and a-Model 2. Considering the comprehensive evaluation metrics, Model 1 exhibited superior predictive performance and robustness across both datasets. Based on these findings, Model 1 was selected as the final model for the construction of the nomogram, as it demonstrated more consistent and reliable performance in both the training and validation scenarios.

External validation

This study employed Model 1 to construct a nomogram (Figure 5), whose predictive performance was subsequently assessed using both the training set and an external test set. The ROC curve analysis revealed that the nomogram achieved an AUC of 0.827 (95% CI: 0.720–0.906) (Figure 6A). The calibration curve of the nomogram closely aligned with the ideal 45° line, indicating a high level of predictive accuracy (Figure 6B). Further, the DCA demonstrated that the nomogram exhibited substantial clinical net benefit at specific threshold probabilities (Figure 6C). These findings underscore the potential utility of the nomogram as a predictive tool in clinical settings.

Figure 5 A nomogram based on Model 1 was developed. To use it, a line is drawn from the “Regimen” axis directly upwards to the “Points” axis to ascertain the score corresponding to the regimen. This process is repeated for the two variables: L-mPA and IVC. The scores are then aggregated, and the total score is identified on the “Total points” axis. A straight line is drawn downwards to the “Risk of central PE” axis to predict the probability of ACPE. Model 1: IVC and L-mPA. ACPE, acute central pulmonary embolism; IVC, inferior vena cava; L-mPA, longitudinal diameter of the main pulmonary artery; PE, pulmonary embolism.

Figure 6 The performance of Model 1 on the external test set: (A) ROC curve, (B) calibration curve, (C) decision curve analysis. Model 1: IVC and L-mPA. AUC, area under the curve; IVC, inferior vena cava; L-mPA, longitudinal diameter of the main pulmonary artery; ROC, receiver operating characteristic.

Diagnostic accuracy metrics of the three models across validation datasets demonstrated significant variations, as detailed in Table 3. Model 1 demonstrated moderate sensitivity (54.6%) but achieved high specificity (96.0%) in the validation cohort.

Sample size calculation

For the training set, which incorporated five predictor variables, the calculation yielded a requirement of 167 samples. For the validation and test sets, which were designed for evaluation with a streamlined set of two key predictors, the required size for each was calculated as 67 samples.

Discussion

This study identified the L-mPA and IVC diameters on unenhanced chest CT as independent predictors of ACPE. The nomogram based on Model 1 exhibited high specificity and a strong ability to accurately identify ACPE (AUC >0.80), indicating its clinical utility in confirming diagnosis and potentially reducing unnecessary CTPA scans. This nomogram provides a straightforward and intuitive tool for rapid ACPE risk stratification using quantitative vascular parameters (13), facilitating timely therapeutic decisions and optimized patient management.

Previous ACPE diagnostic studies using unenhanced CT were limited by small sample sizes (7,8,11). Its diagnostic potential of ACPE may be under-recognized, as these scans are not routinely performed before CTPA. Our study addressed this issue through an expanded cohort, demonstrating robust nomogram performance across internal and external validation sets, ensuring its clinical applicability across diverse institutions.

Another departure from previous studies was our exclusion of subjective CT signs, such as mosaic perfusion, pulmonary infarction, or hyperattenuation (8,11,14-16). These signs based on attenuation measurements have been shown to be influenced by considerable interobserver variability and scanner parameters (17). Moreover, all measurements were consistently performed using a fixed mediastinal window (width, 350 HU; level, 40 HU), minimizing interobserver variability. Consequently, Model 1, which exclusively incorporated vascular morphological parameters, achieved robust diagnostic performance (AUC 0.821 validation set, AUC 0.827 test set). These quantitative approaches not only increased the objectivity and reliability of our results but also facilitated more consistent and reproducible assessments across different clinical settings.

Currently, few clinical scoring systems have been specifically designed for the diagnosis of ACPE. While tools like the Wells Score are widely used for risk stratification in acute PE, they lack specificity for ACPE. Chien et al. confirmed this limitation, demonstrating that the Wells score achieved only a modest AUC of 0.688 in the diagnosis of ACPE (11). This diagnostic gap underscores the clinical value of our study, which investigated the utility of unenhanced CT. In contrast to the limited specificity of clinical scores, our findings suggest that non-contrast CT offers a more direct and accessible imaging method for identifying ACPE.

In recent years, artificial intelligence–driven algorithms have been developed to automatically compute quantitative morphological parameters of pulmonary arteries, with preliminary investigations exploring their use in diagnosing chronic pulmonary hypertension (18,19). Such quantitative vascular parameters also hold potential value in the diagnosis of ACPE. In contrast, this study, employed simple and easily obtainable vessel diameter measurements, facilitating rapid clinical application. Notably, the nomogram developed in this study has a number of practical advantages, as it requires minimal computational resource deployment and can even be printed for use at the point of care. In the future, we plan to integrate artificial intelligence algorithms to incorporate more complex metrics, such as arterial tortuosity and bronchovascular bundle tortuosity difference (18), for the diagnosis of ACPE.

Notably, the newly proposed L-mPA index demonstrates significant discriminative power for ACPE identification. Compared with the transverse diameter of the mPA, the difference in the L-mPA between the central and peripheral PE was more significant. The underlying mechanism for this phenomenon may be attributed to the heterogeneous stress distribution across different orientations in the PA. An in vitro study using porcine PA revealed heterogeneous stress distribution across different orientations in the mPA (20). However, further research needs to be conducted to determine whether human PA exhibit similar biomechanical characteristics.

Moreover, our findings indicated that the IVC diameter may serve as an independent predictor of ACPE. This observation is conceptually understandable, as the dilation of the IVC may be attributed to congestion resulting from acute right ventricular dysfunction (RVD), a mechanism similar to the IVC contrast reflux observed in high and intermediate-risk PE patients on CTPA (21,22), both reflecting increased right heart load (23,24). Notably, the present study found no significant dilation of the SVC and AzV arch in the patients with ACPE. However, an earlier study found that patients with severe PE exhibited significantly larger diameters in the SVC and AzV arch compared to controls (10). The differential venous dilation patterns likely stem from the substantially greater baseline blood volume of the IVC compared to the SVC (25). This physiological disparity explains why the IVC demonstrates both more pronounced and earlier-onset dilation than the SVC and AzV arch during RVD.

This study introduced CT-OPA as a predictive variable with its potential mechanism attributed to altered blood flow at the PA origin due to ACPE (4). Despite differences in CT-OPA between central and peripheral PE in the total cohort during the development of Model 1 using multivariable stepwise logistic regression, CT-OPA was not retained in the final model after adjustment for various vascular measurements. When constructing Model 2, multivariate logistic regression demonstrated a potential positive trend for CT-OPA in diagnosing ACPE, the 95% CI of its OR value crossed 1 before age and sex adjustment, indicating instability. Further, both Model 2 and a-Model 2 incorporating CT-OPA showed unsatisfactory diagnostic performance in the validation set. The limited performance of CT-OPA in the external validation set can be attributed to several technical and clinical factors. First, CT-OPA measurements are particularly susceptible to respiratory motion artifacts and cardiac pulsation effects. Unenhanced CT scans are typically acquired without electrocardiogram gating, and PE patients often present with dyspnea leading to reduced compliance during scanning, collectively compromising the accuracy and reproducibility of CT-OPA measurements. Further, the observed limitations likely stem from variations in scanning parameters (e.g., tube voltage and current) across different scanners, which may influence CT attenuation measurements (HU values) (26), potentially compromising the generalizability of diagnostic models incorporating CT-OPA in multicenter applications.

Limitations

This study had several limitations. First, the applicability of the nomogram may be limited in patients with chronic pulmonary hypertension, as these patients may also develop PA dilation or systemic venous congestion (e.g., IVC/SVC distension). The diagnostic performance of the nomogram in this subgroup remains limited (27). Second, due to its limited sensitivity, the nomogram cannot reliably exclude ACPE, nor can it provide thrombus localization, potentially necessitating additional CTPA in select cases. Third, despite its diagnostic value, our nomogram does not predict prognosis or provide risk stratification, so additional clinical scores are needed to guide ACPE patient management. Moreover, although this study used CTPA as the only reference standard for diagnosing PE, it is not entirely reliable (28), and further validation (e.g., via pulmonary angiography) was lacking. Finally, our model did not fully capture the effect of heterogeneity in the scanner technology and acquisition protocols at each center. Therefore, future research should standardize imaging protocols and include a wider range of participants to confirm the applicability of the model.

Conclusions

The measurements of IVC and L-mPA diameters on unenhanced chest CT showed significant diagnostic value for ACPE. Further, the developed nomogram, which incorporated these two robust predictors, showed significant utility in the diagnosis of ACPE.

Acknowledgments

This abstract was presented as an e-poster at the 36th Great Wall International Congress of Cardiology (GW-ICC)/Asian Heart Society Congress 2025.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-702/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-702/dss

Funding: The study was supported by Science and Technology Plan of Jiangxi Provencal Health Commission (No. 202310508) and Internal Funding Project of the Second Affiliated Hospital, Jiangxi Medical College, Nanchang University (No. 2024efyB04).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-702/coif). W.Z. and X.L. report that they are employees of Neusoft Medical Systems Co., Ltd., but they had no control of data or information submitted for publication. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study protocol was approved by the Ethics Committee of the Second Affiliated Hospital of Nanchang University (the lead center, approval No. MR-36-24-042480). All participating hospitals were informed and agreed the study. The need for obtaining individual informed consent was waived due to the retrospective nature of this research, which involved the analysis of existing data without any additional intervention or risk to participants. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Hofstetter RV, Stalder O, Tritschler T, Méan M, Rodondi N, Righini M, Aujesky D. Clinical characteristics and prognosis of patients with central pulmonary embolism. J Thromb Haemost 2025;23:1585-95. [Crossref] [PubMed]
2. Bangalore S, Horowitz JM, Beam D, Jaber WA, Khandhar S, Toma C, Weinberg MD, Mina B. Prevalence and Predictors of Cardiogenic Shock in Intermediate-Risk Pulmonary Embolism: Insights From the FLASH Registry. JACC Cardiovasc Interv 2023;16:958-72. [Crossref] [PubMed]
3. Jia D, Ji C, Zhao M. Saddle pulmonary embolism is not a sign of high-risk deterioration in non-high-risk patients: A propensity score-matched study. World J Emerg Med 2021;12:261-7. [Crossref] [PubMed]
4. Konstantinides SV, Meyer G, Becattini C, Bueno H, Geersing GJ, Harjola VP, et al. 2019 ESC Guidelines for the diagnosis and management of acute pulmonary embolism developed in collaboration with the European Respiratory Society (ERS): The Task Force for the diagnosis and management of acute pulmonary embolism of the European Society of Cardiology (ESC). Eur Respir J 2019;54:1901647. [Crossref] [PubMed]
5. Ducray V, Vlachomitrou AS, Bouscambert-Duchamp M, Si-Mohamed S, Gouttard S, Mansuy A, Wickert F, Sigal A, Gaymard A, Talbot F, Michel C, Perpoint T, Pialat JB, Rouviere O, Milot L, Cotton F, Douek P, Rabilloud M, Boussel L. COVID-Outcomes-HCL Consortium. Chest CT for rapid triage of patients in multiple emergency departments during COVID-19 epidemic: experience report from a large French university hospital. Eur Radiol 2021;31:795-803. [Crossref] [PubMed]
6. Dermesropian F, Ghaye B. Spontaneously hyperattenuating thrombi revealing acute central pulmonary embolism on unenhanced CT. Diagn Interv Imaging 2019;100:729-30. [Crossref] [PubMed]
7. Ehsanbakhsh A, Hatami F, Valizadeh N, Khorashadizadeh N, Norouzirad F. Evaluating the Performance of Unenhanced Computed Tomography in the Diagnosis of Pulmonary Embolism. J Tehran Heart Cent 2021;16:156-61. [Crossref] [PubMed]
8. Guo R, Deng M, Xi L, Zhang S, Xu W, Liu M. Chest non‑contrasted computed tomography in detecting acute pulmonary thromboembolism: A single‑center retrospective study. Exp Ther Med 2024;28:304. [Crossref] [PubMed]
9. Deng M, Liu A, Kang H, Xi L, Yu P, Xu W, Yang H, Xie W, Liu M, Zhang R. Development and validation of a lung graph-based machine learning model to predict acute pulmonary thromboembolism on chest noncontrast computed tomography. Quant Imaging Med Surg 2023;13:6710-23. [Crossref] [PubMed]
10. Ghaye B, Ghuysen A, Willems V, Lambermont B, Gerard P, D'Orio V, Gevenois PA, Dondelinger RF. Severe pulmonary embolism:pulmonary artery clot load scores and cardiovascular parameters as predictors of mortality. Radiology 2006;239:884-91. [Crossref] [PubMed]
11. Chien CH, Shih FC, Chen CY, Chen CH, Wu WL, Mak CW. Unenhanced multidetector computed tomography findings in acute central pulmonary embolism. BMC Med Imaging 2019;19:65. [Crossref] [PubMed]
12. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996;49:1373-9. [Crossref] [PubMed]
13. Xie W, Li Y, Meng X, Zhao M. Machine learning prediction models and nomogram to predict the risk of in-hospital death for severe DKA: A clinical study based on MIMIC-IV, eICU databases, and a college hospital ICU. Int J Med Inform 2023;174:105049. [Crossref] [PubMed]
14. Torres PPTES, Rabahi MF, Mançano AD, Santos SHRD, Marchiori E. Frequency of spontaneous detection of pulmonary arterial thrombi in unenhanced chest computed tomography in patients diagnosed with pulmonary embolism. J Bras Pneumol 2022;48:e20210128. [Crossref] [PubMed]
15. Torres PPTES, Mançano AD, Marchiori E. Hyperattenuating Sign: An Important Finding to Diagnose Pulmonary Embolism at Unenhanced Chest CT. AJR Am J Roentgenol 2017;209:W201. [Crossref] [PubMed]
16. Deng ZX, Wang H, Zhao Z, Zhang HT, Liu AE, Chen L, Peng Y, Gong LG. The diagnostic efficacy of unenhanced CT for detecting acute central pulmonary embolism. BMC Cardiovasc Disord 2025;25:729. [Crossref] [PubMed]
17. Pan Y, Gao Y, Wang Z, Dou Y, Sun X, Yang Z, Pan S, Jia C. Effects of low-tube voltage coronary CT angiography on plaque and pericoronary fat assessment: intraindividual comparison. Eur Radiol 2024;34:5713-23. [Crossref] [PubMed]
18. Xu W, Xi L, Ni Y, Wang J, Yang H, Liu A, Gao Q, Tao X, Huang Q, Liu X, Zhen Y, Xie W, Liu M. Artificial intelligence-driven quantitative analysis of CT morphological differences between chronic thromboembolic pulmonary hypertension and chronic thromboembolic disease. Quant Imaging Med Surg 2025;15:1101-13. [Crossref] [PubMed]
19. Synn AJ, Nardelli P, Renapurkar R, Quesada L, Sanchez-Ferrero GV, Ross JC, San José Estépar R, Hunsaker AR, Waxman AB, Leopold JA, Washko GR, San José Estépar R, Heresi GA, Rahaghi FN. Spatial heterogeneity and distribution of CT-Based pulmonary vascular volumes in chronic thromboembolic pulmonary hypertension. Eur J Radiol 2025;192:112405. [Crossref] [PubMed]
20. Pillalamarri NR, Patnaik SS, Piskin S, Gueldner P, Finol EA. Ex Vivo Regional Mechanical Characterization of Porcine Pulmonary Arteries. Exp Mech 2021;61:285-303. [Crossref] [PubMed]
21. Yuriditsky E, Zhang RS, Zhang P, Horowitz JM, Bernard S, Greco AA, Postelnicu R, Mukherjee V, Hena K, Elbaum L, Alviar CL, Keller NM, Bangalore S. Inferior vena cava contrast reflux grade is associated with a reduced cardiac index in acute pulmonary embolism. Thromb Res 2024;244:109177. [Crossref] [PubMed]
22. Bailis N, Lerche M, Meyer HJ, Wienke A, Surov A. Contrast reflux into the inferior vena cava on computer tomographic pulmonary angiography is a predictor of 24-hour and 30-day mortality in patients with acute pulmonary embolism. Acta Radiol 2021;62:34-41. [Crossref] [PubMed]
23. Marcus JT, Westerhof BE, Groeneveldt JA, Bogaard HJ, de Man FS, Vonk Noordegraaf A. Vena cava backflow and right ventricular stiffness in pulmonary arterial hypertension. Eur Respir J 2019;54:1900625. [Crossref] [PubMed]
24. Giannakoulas G, Farmakis IT, Hobohm L, Verbrugge FH, Tedford RJ, Sanz J. Acute right ventricular failure: pathophysiology, aetiology, assessment, and management. Eur Heart J 2025;46:2520-35. [Crossref] [PubMed]
25. Wehrum T, Lodemann T, Hagenlocher P, Stuplich J, Ngo BTT, Grundmann S, Hennemuth A, Hennig J, Harloff A. Age-related changes of right atrial morphology and inflow pattern assessed using 4D flow cardiovascular magnetic resonance: results of a population-based study. J Cardiovasc Magn Reson 2018;20:38. [Crossref] [PubMed]
26. Phelps ME, Hoffman EJ, Ter-Pogossian MM. Attenuation coefficients of various body tissues, fluids, and lesions at photon energies of 18 to 136 keV. Radiology 1975;117:573-83. [Crossref] [PubMed]
27. Condliffe R, Durrington C, Hameed A, Lewis RA, Venkateswaran R, Gopalan D, Dorfmüller P. Clinical-radiological-pathological correlation in pulmonary arterial hypertension. Eur Respir Rev 2023;32:230138. [Crossref] [PubMed]
28. Rothenberg SA, Savage CH, Abou Elkassem A, Singh S, Abozeed M, Hamki O, Junck K, Tridandapani S, Li M, Li Y, Smith AD. Prospective Evaluation of AI Triage of Pulmonary Emboli on CT Pulmonary Angiograms. Radiology 2023;309:e230702. [Crossref] [PubMed]