Improvement of artificial intelligence-based computed tomography pulmonary angiography in identifying acute pulmonary embolism
Original Article

Improvement of artificial intelligence-based computed tomography pulmonary angiography in identifying acute pulmonary embolism

Jing Song1, Anqi Chen2, Hanhua Yu1, Lei Song3,4

1Department of Radiology, Wuhan Fourth Hospital, Wuhan, China; 2School of Medicine, Wuhan University of Science and Technology, Wuhan, China; 3Department of Radiology, Huangshi Central Hospital, Affiliated Hospital of Hubei Polytechnic University, Huangshi, China; 4Huangshi Key Laboratory of Cerebrovascular Disease Imaging and Artificial Intelligence, Huangshi, China

Contributions: (I) Conception and design: J Song, L Song; (II) Administrative support: H Yu; (III) Provision of study materials or patients: H Yu; (IV) Collection and assembly of data: J Song; (V) Data analysis and interpretation: J Song, A Chen, L Song; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors.

Correspondence to: Hanhua Yu, MD. Department of Radiology, Wuhan Fourth Hospital, No. 473 Hanzheng Street, Qiaokou District, Wuhan 430000, China. Email: yuhanhua777@163.com; Lei Song, MD, PhD. Department of Radiology, Huangshi Central Hospital, Affiliated Hospital of Hubei Polytechnic University, Huangshi, China; Huangshi Key Laboratory of Cerebrovascular Disease Imaging and Artificial Intelligence, No. 141 Tianjin Road, Huangshigang District, Huangshi 435000, China. Email: song580lei@163.com.

Background: Acute pulmonary embolism (APE) is a potentially fatal condition. Although artificial intelligence (AI) algorithms show promise in diagnosis, overreliance on them may lead to critical errors. This study aimed to improve diagnostic accuracy by integrating AI algorithms with clinical indicators.

Methods: A retrospective analysis was conducted on 329 patients with suspected APE who underwent computed tomography pulmonary angiography (CTPA), lower-extremity venous ultrasound, and laboratory tests. Independent risk factors were identified through logistic regression. The diagnostic performance of individual and combined predictors was assessed via receiver operating characteristic (ROC) curves. Interrater agreement between the AI algorithms and radiologists was evaluated via Kappa analysis. Additionally, subgroup analysis was performed on patients with discordant AI predictions and clinical indicators. Differences in confirmed APE rates among these subgroups were calculated to assess the diagnostic value of AI algorithms in this patient population.

Results: The AI algorithms demonstrated excellent standalone diagnostic performance, with an area under the curve (AUC) of 0.933 [95% confidence intervals (CI): 0.894–0.973; P<0.001] and an odds ratio of 803.28 (95% CI: 163.05–3,957.36; P<0.001). They showed strong agreement with radiologists (κ=0.87; P<0.001). Subgroup analysis revealed that among patients positive for deep vein thrombosis (DVT) or elevated D-dimer levels (defined as >1 mg/L) but with negative AI predictions, the confirmation rates of APE were low (5.1% and 4.6%, respectively). In contrast, among those negative for DVT or normal D-dimer levels (≤1 mg/L) but with positive AI predictions, the APE confirmation rates were significantly higher (75.0% and 57.1%, respectively). These findings suggest that AI algorithms can help identify cases missed by clinical indicators and may reduce unnecessary imaging in high-risk patients without actual embolism. Logistic regression analysis identified DVT, elevated plasma fibrinogen levels (>4 g/L), and male gender as independent risk factors for APE. When AI predictions were combined with these clinical indicators, the diagnostic performance improved markedly, with an AUC of 0.981 (95% CI: 0.967–0.994; P<0.001), outperforming both AI algorithms or clinical indicators alone.

Conclusions: Integrating AI algorithms with clinical indicators significantly enhances the accuracy of diagnosing APE, reduces the risk of misdiagnosis, and improves screening efficiency.

Keywords: Acute pulmonary embolism (APE); deep learning; artificial intelligence algorithms (AI algorithms); computed tomography pulmonary angiography (CTPA)

Submitted Feb 22, 2025. Accepted for publication Jul 01, 2025. Published online Sep 03, 2025.

doi: 10.21037/qims-2025-459

Introduction

Acute pulmonary embolism (APE) is responsible for roughly 100,000 deaths annually (1), ranking as the third most prevalent acute cardiovascular ailment (2). As a major health concern, APE is also the foremost preventable reason for hospital deaths (3), making its timely identification imperative. The presence of a filling defect in the pulmonary arteries on computed tomography pulmonary angiography (CTPA), which serves as a diagnostic marker for APE, is widely used clinically due to its high sensitivity and specificity, cost-effectiveness, noninvasive nature, and round-the-clock availability (4-6). In particular, the use of artificial intelligence (AI)-based CTPA in detecting APE has surged in recent years due its advantages of reducing waiting times, facilitating rapid risk stratification, and quantifying the fraction of lung tissue at risk distal to pulmonary embolism (PE) (7-11). Nevertheless, a small proportion of APE remains clinically undetected by AI-based CTPA (12,13). Failure to identify an APE may lead to delayed therapy, which could be life-threatening (14). Therefore, this study was conducted to support solutions that could improve the ability of AI-based CTPA in detecting APE. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-459/rc).

Methods

Study population

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Institutional Review Board (IRB) of Wuhan Fourth Hospital (IRB No. KY2025-002-01). The requirement for informed consent in this retrospective analysis was waived due to the use of deidentified data. This retrospective, single-center study included patients who underwent CTPA scans and were diagnosed with AI-assisted CTPA software for suspected acute PE from May to August 2024. The inclusion criteria were as follows: (I) CTPA images that met the technical and quality requirements for AI software processing, including image completeness, recognizable pulmonary arteries, and compliance with the standard Digital Imaging and Communications in Medicine (DICOM) format (see Appendix 1 for details); (II) adequate contrast agent filling, defined as an attenuation value of ≥250 Hounsfield units (HU) in the main pulmonary artery (15); and (III) complete data on D-dimer levels, coagulation function tests, and lower limb venous ultrasound examinations. Meanwhile, the exclusion criteria were as follows: (I) a history of diagnosed PE; (II) poor image quality or artifacts; (III) diffuse pulmonary artery lesions; and (IV) uncertain diagnoses, defined as cases in which CTPA images were nondiagnostic due to motion artifacts or insufficient contrast enhancement iwth no follow-up pulmonary angiography or repeat CTPA being performed. A flowchart of participant selection is shown in Figure 1.

Figure 1 The flowchart of participant selection. AI, artificial intelligence; APE, acute pulmonary embolism; CTPA, computed tomography pulmonary angiography.

Data collection and processing

Wuhan Fourth Hospital is an orthopedic specialty center that admits a large number of patients with fractures and joint replacements, most of whom are at high risk for venous thromboembolism (VTE). Upon admission, symptomatic high-risk patients undergo lower limb venous ultrasound and D-dimer screening. If the D-dimer level exceeds 1 mg/L or ultrasound suggests the presence of deep vein thrombosis (DVT), CTPA is performed within 3 days to evaluate for PE. Collected clinical data included patient history, age, sex, D-dimer level, fibrinogen level (considered positive if >4 g/L), and ultrasound findings.

CT scans were performed with a 640-slice CT device (Aquilion ONE, Toshiba Medical Systems, Tokyo, Japan) under the following parameters: 100–120 kV, adaptive tube current, pitch 1.0–1.2, and slice thickness 1 mm. Iomeron 350 contrast agent (Bracco, Milan, Italy) was injected at 4 mL/s with a volume of 60–70 mL based on patient morphology. All images were processed with AI software developed by Shanghai United Imaging Healthcare Co., Ltd. (Shanghai, China), which utilizes a convolutional neural network (CNN) based on the VB-Net architecture for automated analysis. This architecture has been validated in previous studies conducted by United Imaging Healthcare Co., Ltd. (16,17). The CNN model was trained on 2,424 CTPA scans obtained from both public datasets (e.g., Kaggle) and multiple clinical centers, covering a variety of CT vendors and slice thicknesses (primarily 1 mm, with some scans at 0.625 and 1.25 mm). To ensure robust performance, the dataset was randomly divided into training (n=1,942), validation (n=241), and testing (n=241) sets in an 8:1:1 ratio. In this study, two radiologists with over 10 years of clinical experience retrospectively and independently reviewed the CTPA images for APE diagnosis, without access to the AI analysis results or original diagnostic reports. In case of disagreement, a third senior radiologist with over 20 years of experience was consulted for the final decision.

Statistical analysis

Data analysis was performed with SPSS v. 22.0 (IBM Corp., Armonk, NY, USA), GraphPad Prism v. 10.0 (Dotmatics, Boston, MA, USA), and MedCalc v. 20.03 (MedCalc Software, Ostend, Belgium), and R v.4.0.1 (The R Foundation for Statistical Computing, Vienna, Austria). Continuous variables are expressed as the mean ± standard deviation (SD) or as the median and interquartile range (IQR), while categorical variables are presented as percentages. Comparisons between patient groups with and without APE were conducted with the Student t-test, Mann-Whitney test, Chi-squared test, or Fisher exact test, as appropriate. Subgroup analysis was performed on patients whose AI predictions were inconsistent with clinical indicators. Differences in confirmed APE rates among these subgroups were calculated. Logistic regression identified independent risk factors for APE, with results reported as odds ratio (OR) and 95% confidence interval (CI). Diagnostic performance was evaluated via receiver operating characteristic (ROC) curves and the area under the curve (AUC), with comparisons performed with the DeLong test. Agreement between AI algorithms and expert assessments was measured with Cohen kappa coefficient. A two-sided P value <0.05 was considered statistically significant.

Results

Patient characteristics

Baseline characteristics of the 329 patients included in the study are summarized in Table 1. Among the patients, 86 were diagnosed with APE. Male sex, positive DVT findings, and elevated levels of D-dimer and fibrinogen were significantly more prevalent in the APE group (P<0.05). No significant difference in age was observed between the two groups. Additionally, 176 patients had a history of trauma, and 74 had degenerative osteoarthropathy; however, these factors showed no significant difference in distribution between the groups (P>0.05).

Table 1

Comparison of clinical characteristics between patients with and without APE

Variable Without APE (n=243) With APE (n=86) P value Statistical test
Age (years) 65.01±14.04 66.84±12.84 0.291 Student t-test
Gender (male) 83 (34.2%) 54 (62.8%) <0.001 Chi-squared test
Trauma history 135 (55.6%) 41 (47.7%) 0.208 Chi-squared test
Degenerative osteoarthritis 55 (22.6%) 19 (22.1%) 0.918 Chi-squared test
DVT (positive) 170 (70.0%) 77 (89.5%) <0.001 Chi-squared test
Elevated D-dimer (positive) 192 (79.0%) 82 (95.3%) <0.001 Fisher’s exact test
Elevated fibrinogen (positive) 67 (27.6%) 51 (59.3%) <0.001 Chi-squared test
D-dimer value (mg/L) 2.40 (1.11–6.47) 3.90 (2.56–8.36) <0.001 Mann-Whitney test
Fibrinogen value (g/L) 3.32 (2.65–4.32) 4.27 (2.98–5.47) <0.001 Mann-Whitney test

Continuous variables with normal distribution are presented as the mean ± standard deviation, while nonnormally distributed variables are presented as the median with interquartile range, unless otherwise stated. APE, acute pulmonary embolism; DVT, deep vein thrombosis.

Consistency analysis

The expert-confirmed final diagnostic results from CTPA images were used as the gold standard and compared with the AI algorithms results. The kappa coefficient was 0.873 (P<0.001), indicating almost perfect agreement between the two diagnostic methods.

CTPA-confirmed APE in patients with discordance between the AI and clinical indicators

In patients with discordant results between the AI prediction and clinical indicators, clear differences in CTPA-confirmed APE rates were observed (Table 2). Among those positive for DVT but with a negative AI prediction, only 9 of 175 (5.1%) were confirmed to have APE. In contrast, 9 of 12 (75.0%) patients negative for DVT but with a positive AI prediction were confirmed cases. A similar pattern was found for D-dimer: 9 of 197 (4.6%) patients with higher D-dimer levels but with a negative AI prediction had APE, while 4 of 7 (57.1%) with negative for high D-dimer levels but with a positive AI prediction were confirmed.

Table 2

CTPA-confirmed APE in patients with discrepant AI and clinical indicators

Indicator Status AI prediction Patients, n CTPA-confirmed APE, n Confirmation rate, %
DVT Positive Negative 175 9 5.1
Negative Positive 12 9 75.0
Elevated D-dimer Positive Negative 197 9 4.6
Negative Positive 7 4 57.1

AI, artificial intelligence; APE, acute pulmonary embolism; CTPA, computed tomography pulmonary angiography; DVT, deep vein thrombosis.

These findings suggest that the AI model may help avoid unnecessary imaging in some clinically positive patients while also identifying APE in cases negative for DVT or elevated D-dimer levels.

Logistic regression analysis

Logistic regression analysis (Figure 2) revealed that a positive AI algorithms result was strongly associated with an increased APE risk (OR =803.28; 95% CI: 163.05–3,957.36; P<0.001). The presence of lower-extremity DVT (OR =18.02; 95% CI: 2.94–110.41; P=0.002), elevated plasma fibrinogen levels (>4 g/L) (OR =10.18; 95% CI: 2.61–39.70; P=0.001), and male gender (OR =5.82; 95% CI: 1.68–20.21; P=0.006) were also significant independent predictors of APE. These findings highlight the importance of AI, DVT, plasma fibrinogen, and gender as key factors in predicting APE.

Figure 2 Logistic regression analysis of AI models, DVT, fibrinogen, and gender for predicting APE. AI, artificial intelligence; APE, acute pulmonary embolism; CI, confidence interval; DVT, deep vein thrombosis; OR, odds ratio.

Differences between the patients with and without APE

ROC analysis (Figure 3 and Table 3) indicated that the AUCs for the AI algorithms, DVT, fibrinogen, gender, and their combination were 0.933, 0.598, 0.659, 0.643, and 0.981, respectively. The AUC of the combined model was significantly higher than that of any individual indicator (P<0.05), indicating superior diagnostic performance. For the AI algorithms alone, the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were 89.5%, 97.1%, 91.7%, and 96.3%, respectively. The corresponding false-negative rate (FNR) and false-positive rate (FPR) were 10.5% and 2.9%, respectively. For the combined model, the sensitivity, specificity, PPV, and NPV were 89.5%, 98.4%, 95.1%, and 96.4%, respectively. The FNR was 10.5%, and the FPR was 1.6%, respectively. These results demonstrate that the combined approach achieves the highest overall accuracy with excellent discrimination and a notably low rate of misclassification.

Figure 3 Receiver operating characteristic curve analysis of the ability of AI models, DVT, elevated fibrinogen level, male gender, and the combination of all indicators (combination) to detect APE. AI, artificial intelligence; APE, acute pulmonary embolism; AUC, area under the curve; DVT, deep vein thrombosis.

Table 3

Diagnostic performance of individual indicators

Indicator AUC (95% CI) Sensitivity Specificity PPV NPV Delong test
AI 0.933 (0.894–0.973) 0.895 0.971 0.917 0.963 Reference
DVT level 0.598 (0.532–0.663) 0.895 0.300 0.312 0.890 <0.001
Fibrinogen level 0.659 (0.590–0.728) 0.593 0.724 0.432 0.834 <0.001
Gender 0.643 (0.575–0.712) 0.628 0.658 0.394 0.833 <0.001
Combination 0.981 (0.967–0.994) 0.895 0.984 0.951 0.964 <0.001

AI, artificial intelligence; AUC, area under the curve; CI, confidence interval; combination, combination of all indicators; DVT, deep vein thrombosis; NPV, negative predictive value; PPV, positive predictive value.

Discussion

Our study demonstrates that combining AI algorithms with clinical indicators significantly enhances the accuracy of diagnosing APE. Although the AI algorithms alone achieved high diagnostic accuracy (AUC =0.933, sensitivity =89.5%, specificity =97.1%), the integration of clinical and laboratory data further improved model performance, yielding an AUC of 0.981 (sensitivity =89.5%, specificity =98.4%). This finding highlights several important points: first, AI algorithms are a reliable tool for diagnosing APE, offering high sensitivity, specificity, OR value, and excellent consistency. Our models are comparable to previously reported high-performing AI models, which achieved sensitivities ranging from 92.6% to 96.8% and specificities from 95.5% to 99.9% (18-20). Moreover, our model outperformed the pooled diagnostic performance reported in the meta-analysis by Soffer et al. (sensitivity =0.88, specificity =0.86), which may be partly explained by the substantial heterogeneity among the studies included in their analysis (12). Furthermore, the AI algorithms identified some true APE cases in patients negative for DVT or D-dimer (75.0% and 57.1%, respectively), helping to reduce unnecessary missed diagnoses. However, relying solely on the AI algorithms is insufficient for detecting all APE cases. As shown in Table 2 and Figure S1, a small proportion of patients positive for DVT or high D-dimer level but with a negative AI prediction were confirmed to have APE, indicating that without clinical information, some cases may still be missed. Specifically, AI algorithms mainly use CNNs to analyze CTPA images, identifying and extracting embolus features (12,21). However, when artifacts, poor image quality, lung masses, and small branch lesions are present, the FPR and FNR of AI detection are higher (22,23). Additionally, CNNs, as “black-box” models, lack interpretability, making it difficult for clinicians to fully trust AI’s results, especially in emergencies (24,25). The most concerning problem in the use of CNNs is that they typically analyze imaging data in isolation, without integrating other valuable clinical information such as patient history, symptoms, or laboratory and US results (26). Previous studies, such as the work by Huang et al., have attempted to address this issue by leveraging multimodal fusion of CT imaging and electronic health records (EHRs) using deep neural networks in the context of PE detection (27). Building on this concept, our study further validates the incremental value of combining clinical indicators with AI models, demonstrating that the integration of clinical data significantly enhances the diagnostic performance of AI in the detection of PE. Clinical data remain irreplaceable in the diagnosis and treatment of PE. Studies suggest that when the D-dimer values are below the cutoff of 500 µg/L, CTPA scans are unnecessary (28). Furthermore, males over 60 years of age, individuals with a history of DVT, and those with high D-dimer or fibrinogen levels are at a greater risk for PE (29-32). Therefore, when assessing whether a patient has APE, radiologists can interpret AI-generated results in conjunction with relevant clinical data when available, rather than relying solely on the algorithm. In clinical practice, especially under busy and time-constrained conditions, it is often difficult for radiologists to thoroughly review every image. The integration of clinical indicators—when accessible—can help identify high-risk patients who require closer attention, while radiologists can feel more confident in accepting AI findings for low-risk patients with both negative clinical indicators and negative AI results. This integrated diagnostic strategy can enhance diagnostic safety and improve workflow efficiency. Although we acknowledge that AI models do not always have access to clinical data, incorporating such variables when available may further improve their performance and clinical utility in the diagnosis of PE.

Limitations

Certain limitations of this study should be acknowledged. First, we employed a retrospective, single-center design with a small sample size. Second, patients diagnosed with APE were not confirmed by pulmonary arteriography due to its invasiveness. Third, some participants were excluded due to artifacts, severe lung infection or atelectasis, and insufficient pulmonary artery opacification, preventing an evaluation of the AI algorithm’s performance in cases of poor image quality. To address this limitation, future studies should focus on strategies to improve the robustness of AI models under suboptimal imaging conditions.

Conclusions

AI algorithms combined with clinical predictors can further improve the accuracy in diagnosing APE.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-459/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-459/dss

Funding: This work was supported by the Hubei Provincial Natural Science Foundation of China (grant numbers: 2024AFB431 and 2024AFD004).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-459/coif). L.S. reports that this work was supported by the Hubei Provincial Natural Science Foundation of China (grant numbers: 2024AFB431 and 2024AFD004). 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 was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of Wuhan Fourth Hospital (IRB No. KY2025-002-01), and individual informed consent was waived due to the retrospective nature of the study and the use of de-identified data.

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. Freund Y, Cohen-Aubart F, Bloom B. Acute Pulmonary Embolism: A Review. JAMA 2022;328:1336-45. [Crossref] [PubMed]
  2. Wendelboe A, Weitz JI. Global Health Burden of Venous Thromboembolism. Arterioscler Thromb Vasc Biol 2024;44:1007-11. [Crossref] [PubMed]
  3. Lapner ST, Kearon C. Diagnosis and management of pulmonary embolism. BMJ 2013;346:f757. [Crossref] [PubMed]
  4. Moore AJE, Wachsmann J, Chamarthy MR, Panjikaran L, Tanabe Y, Rajiah P. Imaging of acute pulmonary embolism: an update. Cardiovasc Diagn Ther 2018;8:225-43. [Crossref] [PubMed]
  5. Henzler T, Barraza JM Jr, Nance JW Jr, Costello P, Krissak R, Fink C, Schoepf UJ. CT imaging of acute pulmonary embolism. J Cardiovasc Comput Tomogr 2011;5:3-11. [Crossref] [PubMed]
  6. Huisman MV, Klok FA. How I diagnose acute pulmonary embolism. Blood 2013;121:4443-8. [Crossref] [PubMed]
  7. Lareyre F, Chaudhuri A, Behrendt CA, Pouhin A, Teraa M, Boyle JR, Tulamo R, Raffort J. Artificial intelligence-based predictive models in vascular diseases. Semin Vasc Surg 2023;36:440-7. [Crossref] [PubMed]
  8. de Jong CMM, Kroft LJM, van Mens TE, Huisman MV, Stöger JL, Klok FA. Modern imaging of acute pulmonary embolism. Thromb Res 2024;238:105-16. [Crossref] [PubMed]
  9. Wang Q, Yuan L, Ding X, Zhou Z. Prediction and Diagnosis of Venous Thromboembolism Using Artificial Intelligence Approaches: A Systematic Review and Meta-Analysis. Clin Appl Thromb Hemost 2021;27:10760296211021162. [Crossref] [PubMed]
  10. Sadegh-Zadeh SA, Sakha H, Movahedi S, Fasihi Harandi A, Ghaffari S, Javanshir E, Ali SA, Hooshanginezhad Z, Hajizadeh R. Advancing prognostic precision in pulmonary embolism: A clinical and laboratory-based artificial intelligence approach for enhanced early mortality risk stratification. Comput Biol Med 2023;167:107696. [Crossref] [PubMed]
  11. Zhao Y, Malkasian S, Hubbard L, Molloi S. Validation of an automated technique for quantification of pulmonary perfusion territories using computed tomography angiography. Quant Imaging Med Surg 2023;13:3115-26. [Crossref] [PubMed]
  12. Soffer S, Klang E, Shimon O, Barash Y, Cahan N, Greenspana H, Konen E. Deep learning for pulmonary embolism detection on computed tomography pulmonary angiogram: a systematic review and meta-analysis. Sci Rep 2021;11:15814. [Crossref] [PubMed]
  13. Mohanarajan M, Salunke PP, Arif A, Iglesias Gonzalez PM, Ospina D, Benavides DS, Amudha C, Raman KK, Siddiqui HF. Advancements in Machine Learning and Artificial Intelligence in the Radiological Detection of Pulmonary Embolism. Cureus 2025;17:e78217. [Crossref] [PubMed]
  14. Mansella G, Keil C, Nickel CH, Eken C, Wirth C, Tzankov A, Peterson CJ, Aujesky D, Bingisser R. Delayed Diagnosis in Pulmonary Embolism: Frequency, Patient Characteristics, and Outcome. Respiration 2020;99:589-97. [Crossref] [PubMed]
  15. Schönfeld T, Seitz P, Krieghoff C, Ponorac S, Wötzel A, Olthoff S, Schaudt S, Steglich J, Gutberlet M, Gohmann RF. High-pitch CT pulmonary angiography (CTPA) with ultra-low contrast medium volume for the detection of pulmonary embolism: a comparison with standard CTPA. Eur Radiol 2024;34:1921-31. [Crossref] [PubMed]
  16. Kline TL, Edwards ME, Fetzer J, Gregory AV, Anaam D, Metzger AJ, Erickson BJ. Automatic semantic segmentation of kidney cysts in MR images of patients affected by autosomal-dominant polycystic kidney disease. Abdom Radiol (NY) 2021;46:1053-61. [Crossref] [PubMed]
  17. Han M, Yao G, Zhang W, Mu G, Zhan Y, Zhou X, Gao Y. Segmentation of CT thoracic organs by multi-resolution VB-nets. Stud Health Technol Inform 2019;2019:1-4.
  18. Weikert T, Winkel DJ, Bremerich J, Stieltjes B, Parmar V, Sauter AW, Sommer G. Automated detection of pulmonary embolism in CT pulmonary angiograms using an AI-powered algorithm. Eur Radiol 2020;30:6545-53. [Crossref] [PubMed]
  19. Cheikh AB, Gorincour G, Nivet H, May J, Seux M, Calame P, Thomson V, Delabrousse E, Crombé A. How artificial intelligence improves radiological interpretation in suspected pulmonary embolism. Eur Radiol 2022;32:5831-42. [Crossref] [PubMed]
  20. Langius-Wiffen E, de Jong PA, Hoesein FAM, Dekker L, van den Hoven AF, Nijholt IM, Boomsma MF, Veldhuis WB. Retrospective batch analysis to evaluate the diagnostic accuracy of a clinically deployed AI algorithm for the detection of acute pulmonary embolism on CTPA. Insights Imaging 2023;14:102. [Crossref] [PubMed]
  21. Anwar SM, Majid M, Qayyum A, Awais M, Alnowami M, Khan MK. Medical Image Analysis using Convolutional Neural Networks: A Review. J Med Syst 2018;42:226. [Crossref] [PubMed]
  22. Ayobi A, Chang PD, Chow DS, Weinberg BD, Tassy M, Franciosini A, Scudeler M, Quenet S, Avare C, Chaibi Y. Performance and clinical utility of an artificial intelligence-enabled tool for pulmonary embolism detection. Clin Imaging 2024;113:110245. [Crossref] [PubMed]
  23. Robb CL, Bhalla S, Raptis CA. Pitfalls in the diagnosis of acute pulmonary embolism on computed tomography: common pathologic and imaging mimics. Curr Radiol Rep 2018;6:14. [Crossref]
  24. Hassija V, Chamola V, Mahapatra A, Singal A, Goel D, Huang K, Scardapane S, Spinelli I, Mahmud M, Hussain AJ. Interpreting black-box models: a review on explainable artificial intelligence. Comput Commun 2024;16:45-74.
  25. Ghaffar Nia N, Kaplanoglu E, Nasab A. Evaluation of artificial intelligence techniques in disease diagnosis and prediction. Discov Artif Intell 2023;3:5. [Crossref] [PubMed]
  26. Yu H, Yang LT, Zhang Q, Armstrong D, Deen MJ. Convolutional neural networks for medical image analysis: state-of-the-art, comparisons, improvement and perspectives. Neurocomputing 2021;444:92-110. [Crossref]
  27. Huang SC, Pareek A, Zamanian R, Banerjee I, Lungren MP. Multimodal fusion with deep neural networks for leveraging CT imaging and electronic health record: a case-study in pulmonary embolism detection. Sci Rep 2020;10:22147. [Crossref] [PubMed]
  28. Kearon C, de Wit K, Parpia S, Schulman S, Afilalo M, Hirsch A, Spencer FA, Sharma S, D'Aragon F, Deshaies JF, Le Gal G, Lazo-Langner A, Wu C, Rudd-Scott L, Bates SM, Julian JA. Diagnosis of Pulmonary Embolism with d-Dimer Adjusted to Clinical Probability. N Engl J Med 2019;381:2125-34. [Crossref] [PubMed]
  29. Jarman AF, Mumma BE, Singh KS, Nowadly CD, Maughan BC. Crucial considerations: Sex differences in the epidemiology, diagnosis, treatment, and outcomes of acute pulmonary embolism in non-pregnant adult patients. J Am Coll Emerg Physicians Open 2021;2:e12378. [Crossref] [PubMed]
  30. Klovaite J, Nordestgaard BG, Tybjærg-Hansen A, Benn M. Elevated fibrinogen levels are associated with risk of pulmonary embolism, but not with deep venous thrombosis. Am J Respir Crit Care Med 2013;187:286-93. [Crossref] [PubMed]
  31. Jiménez D, Aujesky D, Díaz G, Monreal M, Otero R, Martí D, Marín E, Aracil E, Sueiro A, Yusen RD. RIETE Investigators. Prognostic significance of deep vein thrombosis in patients presenting with acute symptomatic pulmonary embolism. Am J Respir Crit Care Med 2010;181:983-91. [Crossref] [PubMed]
  32. Brink A, Elf J, Svensson PJ, Engström G, Melander O, Zöller B. Sex-Specific Risk Factors for Deep Venous Thrombosis and Pulmonary Embolism in a Population-Based Historical Cohort Study of Middle-Aged and Older Individuals. J Am Heart Assoc 2023;12:e027502. [Crossref] [PubMed]
Cite this article as: Song J, Chen A, Yu H, Song L. Improvement of artificial intelligence-based computed tomography pulmonary angiography in identifying acute pulmonary embolism. Quant Imaging Med Surg 2025;15(10):9729-9737. doi: 10.21037/qims-2025-459

Article Options

Download Citation

Share