Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Sep 2;25(1):366.
doi: 10.1186/s12880-025-01911-x.

Automated quantification of abdominal aortic calcification using 3D nnU-Net: a novel approach to assess AAA rupture risk

Yuan-Lin Luo #  1   2 , Yi-Fan Liu #  1 , Zhi Huang  1 , Chu Wang  1 , Ling-Yue Zhang  3 , Shui-Chuan Huang  4
Affiliations

Automated quantification of abdominal aortic calcification using 3D nnU-Net: a novel approach to assess AAA rupture risk

Yuan-Lin Luo et al. BMC Med Imaging. .

Abstract

Background: Abdominal aortic aneurysms (AAA) pose a serious rupture risk, heightened by aortic calcification. Traditional calcification scoring methods are slow and require expertise. This study aims to construct a convolutional neural network (nnU-Net) model for automatic quantification and segmentation of abdominal aortic calcification from a single CTA scan.

Methods: This retrospective study included 100 patients who underwent abdominal aortic CTA between January 2018 and October 2023, meeting specific inclusion criteria. Vessel and calcification segmentation were manually scored by two physicians, and an nnU-Net deep learning model was developed to automate calcification measurement. Model performance was assessed using Dice scores. Agreement between manual and model-based scoring was assessed using Spearman rank correlation and Bland-Altman analysis.

Results: The nnU-Net model achieved median Dice scores of 93.60% for blood vessels and 81.06% for calcification. Average Dice scores were 92.37 ± 4.87% for blood vessel segmentation and 81.03 ± 5.11% for calcified plaque. The model's Agatston scores correlated closely with manual scores (Spearman's ρ = 0.969), with a mean difference of -229.51 (95% limits of agreement: -6003.92 to 5544.90). The model's evaluation time was also shorter than manual scoring (112 ± 4.4 s vs. 3796 ± 6.6 s, p < 0.001).

Conclusion: The nnU-Net-based model shows potential as an automated tool for accurately segmenting and quantifying abdominal aortic calcification, offering comparable results to manual scoring with significantly reduced evaluation time. This approach may assist in more efficient assessment of AAA rupture risk, supporting clinical decision-making in patient management.

Keywords: Abdominal aortic aneurysm; Calcification; Convolutional neural network (nnU-Net); Rupture.

PubMed Disclaimer

Conflict of interest statement

Declarations. Ethical approval: The studies involving human participants were reviewed and approved by the Medical Ethics Committee of Nanfang Hospital (approval number: NFEC-202504-K3-01). Written informed consent was obtained from all individual participants included in this study. All procedures complied with the Declaration of Helsinki. Consent for publication: The dataset used in this study was fully anonymized and complied with ethical standards for retrospective analysis. No direct patient identifiers or sensitive clinical details were included in the manuscript. Data sharing statement: Due to the nature of this research, participants of this study did not agree for their data to be shared publicly, so supporting data is not available. Competing interests: The authors declare no competing interests.

Figures

Fig. 1
Fig. 1
Comparison of manual and model segmentation: A-C. Manual segmentation, D-F. Model segmentation
Fig. 2
Fig. 2
3D reconstruction of artificial and model segmentation: A. Manual reconstruction, B. Model reconstruction
Fig. 3
Fig. 3
A. First cross-validation loss function curve; B. Second cross-validation loss function curve; C. Third cross-validation loss function curve; D. Fourth cross-validation loss function curve; E. The fifth cross-validation loss function; F. 5-fold loss function curve in training set; G. 5-fold loss function curve in validation set
Fig. 4
Fig. 4
Dice distribution of blood vessels and calcification
Fig. 5
Fig. 5
Correlation between manual and model score (PCScore: Physical calcification score; ACSore: Automated calcification score)
Fig. 6
Fig. 6
Bland-Altman plots of manual and model scores
Fig. 7
Fig. 7
Distribution of manual and model agatston scores (PCScore: Physical calcification score; ACSore: automated calcification score)
Fig. 8
Fig. 8
Confusion matrix
See this image and copyright information in PMC

References

    1. Isselbacher EM, Preventza O, Hamilton Black J 3rd, Augoustides JG, Beck AW, Bolen MA, et al. 2022 ACC/AHA guideline for the diagnosis and management of aortic disease: A report of the American heart association/american college of cardiology joint committee on clinical practice guidelines. Circulation. 2022;146(24):e334–482. - PMC - PubMed
    1. Wanhainen A. Response to re ‘european society for vascular surgery (ESVS) 2019 clinical practice guidelines on the management of abdominal Aorto-Iliac artery aneurysms’. Eur J Vasc Endovasc Surg. 2020;60(6):951. - PubMed
    1. Vorp DA, Vande Geest JP. Biomechanical determinants of abdominal aortic aneurysm rupture. Arterioscler Thromb Vasc Biol. 2005;25(8):1558–66. - PubMed
    1. Buijs RV, Willems TP, Tio RA, Boersma HH, Tielliu IF, Slart RH, et al. Current state of experimental imaging modalities for risk assessment of abdominal aortic aneurysm. J Vasc Surg. 2013;57(3):851–9. - PubMed
    1. Nordon IM, Hinchliffe RJ, Loftus IM, Thompson MM. Pathophysiology and epidemiology of abdominal aortic aneurysms. Nat Rev Cardiol. 2011;8(2):92–102. - PubMed

MeSH terms

LinkOut - more resources