Abstract
Pulmonary embolism (PE) is a life-threatening condition for which computed tomography pulmonary angiography (CTPA) is the standard diagnostic modality. However, conventional CTPA protocols require relatively high iodine contrast and radiation doses, raising concerns about renal injury and radiation exposure. In this study, we propose a deep learning-based framework for PE diagnosis under low-iodine and low-radiation CTPA conditions. The proposed two-stage framework integrates image enhancement and classification by jointly leveraging original low-exposure images and their super-resolved counterparts. We further construct and publicly release a low-iodine, low-radiation CTPA dataset developed in collaboration with a clinical institution to support reproducible research in safe imaging. Experimental results demonstrate that the proposed method substantially improves diagnostic performance compared with single-branch baselines, achieving an area under the ROC curve (AUC) of 0.928 while maintaining balanced sensitivity and specificity. These findings suggest that the proposed framework enables accurate and safer PE diagnosis under reduced contrast and radiation exposure, offering a practical solution for improving diagnostic safety in clinical CTPA imaging.
Data availibility
The present study used CTPA images from publicly available datasets RSNA as well as from Beijing Hospital. Public datasets can be accessed through the respective websites. In addition, the low-iodine, low-radiation CTPA dataset was collected from real clinical acquisitions at Beijing Hospital, with all procedures approved by the institutional ethics committee (Approval No. 2024BJYYEC-KY089-02). Patient data were fully anonymized to protect privacy. The clinical low-iodine, low-radiation CTPA dataset has been made available via a controlled-access repository and can be accessed at: https://pan.baidu.com/s/1PvfT_13mLlcHLLE-C5_g_g. Due to ethical and privacy considerations, access is granted for research and reproducibility purposes only. The access code can be obtained by contacting the corresponding author upon reasonable request.
Code availability
The implementation code for the proposed two-stage deep learning framework, including the Frequency-Aware Super-Resolution Network (FASRN) and Dual-Branch Classification Network (DBCN), will be released upon manuscript acceptance on a public repository. This will allow full reproducibility of the reported results and facilitate future research on low-exposure CTPA imaging.
References
Chernysh, I. N. et al. The distinctive structure and composition of arterial and venous thrombi and pulmonary emboli. Sci. reports 10, 5112 (2020).
Huang, S.-C. et al. Penet-a scalable deep-learning model for automated diagnosis of pulmonary embolism using volumetric ct imaging. NPJ digital medicine 3, 61 (2020).
Hu, Z. et al. High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosis. npj Digital Medicine 8, 254 (2025).
Thomas, S. E., Weinberg, I., Schainfeld, R. M., Rosenfield, K. & Parmar, G. M. Diagnosis of pulmonary embolism: A review of evidence-based approaches. J. Clin. Medicine 13, 3722 (2024).
Shapiro, J., Reichard, A. & Muck, P. E. New diagnostic tools for pulmonary embolism detection. Methodist DeBakey Cardiovasc. J. 20, 5 (2024).
Turetz, M., Sideris, A. T., Friedman, O. A., Triphathi, N. & Horowitz, J. M. Epidemiology, pathophysiology, and natural history of pulmonary embolism. In Seminars in interventional radiology, vol. 35, 92–98 (Thieme Medical Publishers, 2018).
Chen, Q. et al. Incidence, risk factors, and mortality of pulmonary embolism in the netherlands (2015–22): sex differences and shifts during the coronavirus disease 2019 pandemic. Eur. Hear. J. ehaf211 (2025).
Eng, J. et al. Accuracy of ct in the diagnosis of pulmonary embolism: a systematic literature review. Am. J. Roentgenol. 183, 1819–1827 (2004).
Klings, E. S. et al. An official american thoracic society clinical practice guideline: diagnosis, risk stratification, and management of pulmonary hypertension of sickle cell disease. American journal of respiratory and critical care medicine 189, 727–740 (2014).
Shen, L. et al. Deep learning reconstruction combined with contrast-enhancement boost in dual-low dose ct pulmonary angiography: a two-center prospective trial. Eur. Radiol. 1–10 (2025).
Triggiani, S. et al. Comprehensive review of pulmonary embolism imaging: past, present and future innovations in computed tomography (ct) and other diagnostic techniques. Jpn. J. Radiol. 1–15 (2025).
Haubold, J. et al. Contrast agent dose reduction in computed tomography with deep learning using a conditional generative adversarial network. Eur. Radiol. 31, 6087–6095 (2021).
Shin, D.-J. et al. Low-iodine-dose computed tomography coupled with an artificial intelligence-based contrast-boosting technique in children: a retrospective study on comparison with conventional-iodine-dose computed tomography. Pediatr. Radiol. 54, 1315–1324 (2024).
Rawat, W. & Wang, Z. Deep convolutional neural networks for image classification: A comprehensive review. Neural computation 29, 2352–2449 (2017).
Soffer, S. et al. Deep learning for pulmonary embolism detection on computed tomography pulmonary angiogram: a systematic review and meta-analysis. Sci. reports 11, 15814 (2021).
Tachibana, Y., Takaji, R., Shiroo, T. & Asayama, Y. Deep-learning reconstruction with low-contrast media and low-kilovoltage peak for ct of the liver. Clin. Radiol. 79, e546–e553 (2024).
Bae, K., Kim, T. H. & Jeon, K. N. Deep learning-based iodine contrast augmentation for suboptimally enhanced ct pulmonary angiography: Implications for pulmonary embolism diagnosis. Diagnostics 15, 2325 (2025).
Singh, G. et al. Various image enhancement techniques-a critical review. Int. J. Innov. Sci. Res. 10, 267–274 (2014).
Qi, Y. et al. A comprehensive overview of image enhancement techniques. Arch. Comput. Methods Eng 29, 583–607 (2022).
Yan, Q. et al. Hvi: A new color space for low-light image enhancement. In Proceedings of the Computer Vision and Pattern Recognition Conference, 5678–5687 (2025).
Park, M. et al. Application of a deep learning–based contrast-boosting algorithm to low-dose computed tomography pulmonary angiography with reduced iodine load. J. Comput. Assist. Tomogr. 10–1097 (2022).
Sun, J. et al. Improving the image quality of pediatric chest ct angiography with low radiation dose and contrast volume using deep learning image reconstruction. Quant. Imaging Medicine Surg. 11, 3051 (2021).
Colak, E. et al. The rsna pulmonary embolism ct dataset. Radiol. Artif. Intell. 3, e200254 (2021).
Ahmed, N., Natarajan, T. & Rao, K. R. Discrete cosine transform. IEEE transactions on Comput. 100, 90–93 (2006).
Wang, Q. et al. Eca-net: Efficient channel attention for deep convolutional neural networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 11534–11542 (2020).
Keys, R. Cubic convolution interpolation for digital image processing. IEEE transactions on acoustics, speech, and signal processing 29, 1153–1160 (2003).
Duchon, C. E. Lanczos filtering in one and two dimensions. J. Appl. Meteorol. 1962–1982, 1016–1022 (1979).
Lai, W.-S., Huang, J.-B., Ahuja, N. & Yang, M.-H. Deep laplacian pyramid networks for fast and accurate super-resolution. In Proceedings of the IEEE conference on computer vision and pattern recognition, 624–632 (2017).
Shi, W. et al. Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network. In Proceedings of the IEEE conference on computer vision and pattern recognition, 1874–1883 (2016).
Lim, B., Son, S., Kim, H., Nah, S. & Mu Lee, K. Enhanced deep residual networks for single image super-resolution. In Proceedings of the IEEE conference on computer vision and pattern recognition workshops, 136–144 (2017).
Dong, C., Loy, C. C. & Tang, X. Accelerating the super-resolution convolutional neural network. In European conference on computer vision, 391–407 (Springer, 2016).
Abdelhamid, A., El-Ghamry, A., Abdelhay, E. H., Abo-Zahhad, M. M. & Moustafa, H.E.-D. Improved pulmonary embolism detection in ct pulmonary angiogram scans with hybrid vision transformers and deep learning techniques. Sci.Reports 15, 31443 (2025).
Zhong, Z. et al. Vision-language model for report generation and outcome prediction in ct pulmonary angiogram. NPJ Digit. Medicine 8, 432 (2025).
Acknowledgements
The authors thank the clinicians and research staff who contributed to data collection and analysis, and colleagues who provided valuable feedback during manuscript preparation.
Funding
This work was supported by the National Natural Science Foundation of China (Grant No. 62521007, 62431011), the Fundamental Research Funds for the Central Universities (E2ET1104) and National High Level Clinical Research Funding (BJ-2025-213).
Author information
Authors and Affiliations
Contributions
X.Z. organized the entire project and supervised the experimental design and methodological framework. M.H. is responsible for the main experiments, including image classification model development, training, evaluation, and manuscript writing. T.G. is responsible for clinical data acquisition, annotation, and coordination, ensuring high-quality low-iodine, low-radiation CTPA dataset. H.A. and X.F. are responsible for the design of the super-resolution model and related experiments. All authors reviewed and approved the final version of the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Consent to participate
The requirement for informed consent was waived by the Beijing Hospital Ethics Committee (2024BJYYEC-KY089-02) because this study was retrospective and used fully anonymized clinical data.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Hong, M., Gu, T., An, H. et al. Enhancing diagnostic safety with low iodine, low radiation CTPA classification using deep learning. Sci Rep (2026). https://doi.org/10.1038/s41598-026-38223-1
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-026-38223-1
