Segmentation of Acute Ischemic Stroke Infarct using Multi-scale U-Net Architecture on Non-contrast CT Scan Images
Abstract
Acute ischemic stroke must be treated within a few hours, and the first scan is usually a non-contrast CT (NCCT), on which an early infarct is often hard to see. Automatic segmentation could help readers, and multi-scale convolution is a common proposal for it. This study measured how much such a block improves U-Net on this task, and what it costs. Every convolutional block of U-Net, in the encoder and the decoder, was replaced by a block with three parallel branches using 1×1, 3×3 and 5×5 kernels. Both networks were trained on the public Acute Ischemic Stroke Dataset, on 375 studies with supratentorial infarcts. The data were divided by study, the scans of the dataset’s own test split were kept for testing, and augmentation was applied to the training subset alone. Four configurations were compared: each architecture with and without augmentation. On the 51 test patients the Multi-scale U-Net did not improve overlap. Its Dice score was 0.6372 against 0.6632 for U-Net with augmentation, and the paired difference per patient was 0.0233 in favour of U-Net (95 % CI −0.060 to 0.013, p = 0.19). What the block changed was the balance between the two kinds of error: sensitivity rose by 0.060 (p = 0.002) and precision fell by 0.100 (p < 0.001). Augmentation helped both networks (+0.106 and +0.048 Dice per patient, p < 0.001), mainly by cutting false-positive lesions from 15.5 to 5.6 per patient. The multi-scale block costs 2.31 times the parameters and 3.37 times the inference time. An ablation over the target definition favoured the dataset authors' labels: adding the label for infarct invisible on NCCT cost 0.040 Dice and tripled false-positive lesions. Both networks segment small infarcts poorly, with Dice below 0.16 under 100 pixels, which remains the main obstacle.
Downloads
References
N. A. Hilkens, B. Casolla, T. W. Leung, and F.-E. de Leeuw, “Stroke,” Lancet, vol. 403, no. 10446, pp. 2820–2836, Jun. 2024, doi: 10.1016/S0140-6736(24)00642-1.
World Health Organization, “Stroke,” WHO Fact Sheets, Geneva, Switzerland, 2024. [Online]. Available: https://www.who.int/news-room/fact-sheets/detail/stroke (accessed Oct. 28, 2026).
GBD 2021 Stroke Risk Factor Collaborators, “Global, regional, and national burden of stroke and its risk factors, 1990–2021: a systematic analysis for the Global Burden of Disease Study 2021,” Lancet Neurol., vol. 23, no. 10, pp. 973–1003, Oct. 2024, doi: 10.1016/S1474-4422(24)00369-7.
V. L. Feigin et al., “World Stroke Organization: Global Stroke Fact Sheet 2025,” Int. J. Stroke, vol. 20, no. 2, pp. 132–144, 2025, doi: 10.1177/17474930241308142.
Kementerian Kesehatan Republik Indonesia, “Tekan risiko stroke, masyarakat dianjurkan teratur cek kesehatan,” Jakarta, Indonesia, 2022. [Online]. Available: https://kemkes.go.id/id/tekan-risiko-stroke-masyarakat-dianjurkan-teratur-cek-kesehatan (accessed Oct. 28, 2026).
Kementerian Kesehatan Republik Indonesia, “Cegah stroke dengan aktivitas fisik,” Jakarta, Indonesia, 2022. [Online]. Available: https://kemkes.go.id/eng/cegah-stroke-dengan-aktivitas-fisik (accessed Oct. 28, 2026).
Kementerian Kesehatan Republik Indonesia, “Profil Kesehatan Indonesia 2021,” Jakarta, Indonesia, 2023. [Online]. Available: https://www.kemkes.go.id/id/profil-kesehatan-indonesia-2021 (accessed Oct. 28, 2026).
R. Rigual, B. Fuentes, and E. Díez-Tejedor, “Management of acute ischemic stroke,” Med. Clínica (English Ed., vol. 161, no. 11, pp. 485–492, 2023, doi: https://doi.org/10.1016/j.medcle.2023.06.032.
N. Yang and Y. B. Hartanto, “Characteristics and reasons for delayed presentation in acute ischemic stroke: single-centered study in Indonesia,” Egypt. J. Neurol. Psychiatry Neurosurg., vol. 60, no. 1, p. 67, 2024, doi: 10.1186/s41983-024-00838-4.
S. Patil, R. Rossi, D. Jabrah, and K. Doyle, “Detection, diagnosis and treatment of acute ischemic stroke: current and future perspectives,” Front. Med. Technol., vol. 4, art. no. 748949, Jun. 2022, doi: 10.3389/fmedt.2022.748949.
T. F. Hasan, H. Hasan, and R. E. Kelley, “Overview of Acute Ischemic Stroke Evaluation and Management,” Biomedicines, vol. 9, no. 10, 2021, doi: 10.3390/biomedicines9101486.
A. Alsalahat et al., “Current state and advancements of imaging in acute ischemic stroke: a practical review,” Neurol. Sci., vol. 47, 2025, doi: 10.1007/s10072-025-08685-8.
K. Muir et al., “Randomised, controlled Trial of CT perfusion and angiography compared to CT alone in thrombolysis-eligible acute ischaemic stroke patients: The penumbra and recanalisation acute computed tomography in ischaemic stroke evaluation (PRACTISE) trial,” Eur. Stroke J., vol. 11, 2026, doi: 10.1093/esj/23969873251372348.
H. Luecking, P. Hoelter, S. Lang, M. Schmidt, F. Eisenhut, and A. Doerfler, “Change your Angle of View,” Clin. Neuroradiol., vol. 32, pp. 1109–1115, 2022, doi: 10.1007/s00062-022-01172-z.
A. Sarraj, D. Pujara, and B. Campbell, “Current State of Evidence for Neuroimaging Paradigms in Management of Acute Ischemic Stroke,” Ann. Neurol., vol. 95, 2024, doi: 10.1002/ana.26925.
L. M. van Poppel, C. B. L. M. Majoie, H. A. Marquering, and B. J. Emmer, “Associations between early ischemic signs on non-contrast CT and time since acute ischemic stroke onset: A scoping review.,” Eur. J. Radiol., vol. 155, p. 110455, Oct. 2022, doi: 10.1016/j.ejrad.2022.110455.
N. Venketasubramanian, F. L. Yudiarto, and D. Tugasworo, “Stroke Burden and Stroke Services in Indonesia,” Cerebrovasc. Dis. Extra, vol. 12, no. 1, pp. 53–57, 2022, doi: 10.1159/000524161.
C. Capirossi, A. Laiso, L. Renieri, F. Capasso, and N. Limbucci, “Epidemiology, organization, diagnosis and treatment of acute ischemic stroke,” Eur. J. Radiol. Open, vol. 11, p. 100527, 2023, doi: https://doi.org/10.1016/j.ejro.2023.100527.
R. Gauriau et al., “Head CT deep learning model is highly accurate for early infarct estimation,” Sci. Rep., vol. 13, no. 1, p. 189, 2023, doi: 10.1038/s41598-023-27496-5.
N. Kaothanthong, K. Atsavasirilert, S. Sarampakhul, P. Chantangphol, D. Songsaeng, and S. Makhanov, “Artificial intelligence for localization of the acute ischemic stroke by non-contrast computed tomography,” PLoS One, vol. 17, no. 12 December, pp. 1–20, 2022, doi: 10.1371/journal.pone.0277573.
J. Lu et al., “Identification of early invisible acute ischemic stroke in non-contrast computed tomography using two-stage deep-learning model,” Theranostics, vol. 12, no. 13, pp. 5564–5573, 2022, doi: 10.7150/thno.74125.
L. Alzubaidi et al., “Review of deep learning: concepts, CNN architectures, challenges, applications, future directions,” J. Big Data, vol. 8, no. 1, p. 53, 2021, doi: 10.1186/s40537-021-00444-8.
Q. Xia et al., “A comprehensive review of deep learning for medical image segmentation,” Neurocomputing, vol. 613, p. 128740, 2024, doi: 10.1016/j.neucom.2024.128740.
R. Azad et al., “Medical image segmentation review: the success of U-Net,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 46, no. 12, pp. 10076–10095, Dec. 2024, doi: 10.1109/TPAMI.2024.3435571.
M. Li, Y. Jiang, Y. Zhang, and H. Zhu, “Medical image analysis using deep learning algorithms,” Front. Public Heal., vol. Volume 11, 2023, doi: 10.3389/fpubh.2023.1273253.
L.-C. Chen, G. Papandreou, I. Kokkinos, K. Murphy, and A. L. Yuille, “DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 40, no. 4, pp. 834–848, 2018, doi: 10.1109/TPAMI.2017.2699184.
J. Schlemper, O. Oktay, M. Schaap, M. Heinrich, B. Kainz, B. Glocker, and D. Rueckert, “Attention gated networks: learning to leverage salient regions in medical images,” Med. Image Anal., vol. 53, pp. 197–207, Apr. 2019, doi: 10.1016/j.media.2019.01.012.
Z. Zhou, M. M. R. Siddiquee, N. Tajbakhsh, and J. Liang, “UNet++: A Nested U-Net Architecture for Medical Image Segmentation,” in Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support, LNCS vol. 11045, 2018, pp. 3–11, doi: 10.1007/978-3-030-00889-5_1.
K. Liang et al., “Symmetry-enhanced attention network for acute ischemic infarct segmentation with non-contrast CT images,” in Medical Image Computing and Computer Assisted Intervention – MICCAI 2021, Lecture Notes in Computer Science, vol. 12907, Cham, Switzerland: Springer, 2021, pp. 432–441, doi: 10.1007/978-3-030-87234-2_41.
J. Chen et al., “TransUNet: rethinking the U-Net architecture design for medical image segmentation through the lens of transformers,” Med. Image Anal., vol. 97, art. no. 103280, Oct. 2024, doi: 10.1016/j.media.2024.103280.
H. Cao et al., “Swin-Unet: Unet-Like Pure Transformer for Medical Image Segmentation,” in Computer Vision – ECCV 2022 Workshops, LNCS vol. 13803, 2023, pp. 205–218, doi: 10.1007/978-3-031-25066-8_9.
C. Szegedy et al., “Going deeper with convolutions,” in 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015, pp. 1–9, doi: 10.1109/CVPR.2015.7298594.
J. Li, K. V. Sarma, K. C. Ho, A. Gertych, B. S. Knudsen, and C. W. Arnold, “A multi-scale U-Net for semantic segmentation of histological images from radical prostatectomies,” AMIA Annu. Symp. Proc., vol. 2017, pp. 1140–1148, Apr. 2018, PMID: 29854182, PMCID: PMC5977596. [Online]. Available: https://pmc.ncbi.nlm.nih.gov/articles/PMC5977596 (accessed Oct. 28, 2026).
R. Su, D. Zhang, J. Liu, and C. Cheng, “MSU-Net: Multi-Scale U-Net for 2D Medical Image Segmentation,” Front. Genet., vol. Volume 12, 2021, doi: 10.3389/fgene.2021.639930.
W. Zhang, H. Cheng, and J. Gan, “MUNet: a multi-scale U-Net framework for medical image segmentation,” in Proc. 2020 Int. Joint Conf. Neural Networks (IJCNN), Glasgow, U.K., Jul. 19–24, 2020, pp. 1–7, doi: 10.1109/IJCNN48605.2020.9206703.
S. Li, J. Zheng, and D. Li, “Precise segmentation of non-enhanced computed tomography in patients with ischemic stroke based on multi-scale U-Net deep network model.,” Comput. Methods Programs Biomed., vol. 208, p. 106278, Sep. 2021, doi: 10.1016/j.cmpb.2021.106278.
A. Sandru, M.-I. Georgescu, and R. T. Ionescu, “Feature-level augmentation to improve robustness of deep neural networks to affine transformations,” in Computer Vision – ECCV 2022 Workshops, Lecture Notes in Computer Science, vol. 13801, Cham, Switzerland: Springer, 2023, pp. 332–341, doi: 10.1007/978-3-031-25056-9_22.
R. Zhang, B. Zhou, C. Lu, and M. Ma, “The Performance Research of the Data Augmentation Method for Image Classification,” Math. Probl. Eng., vol. 2022, no. 1, p. 2964829, Jan. 2022, doi: https://doi.org/10.1155/2022/2964829.
L. Huang, J. Qin, Y. Zhou, F. Zhu, L. Liu, and L. Shao, “Normalization Techniques in Training DNNs: Methodology, Analysis and Application,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 45, no. 8, pp. 10173–10196, 2023, doi: 10.1109/TPAMI.2023.3250241.
G. Varoquaux and V. Cheplygina, “Machine learning for medical imaging: methodological failures and recommendations for the future,” npj Digit. Med., vol. 5, no. 1, p. 48, 2022, doi: 10.1038/s41746-022-00592-y.
B. Murugesakumar, A. S., A. Arunmanikandan, M. Chithra, and M. Dhasaranjan, “U-Net Convolutional Networks for Real-Time Biomedical Image Segmentation and Anomaly Detection,” in 2025 Tenth International Conference on Science Technology Engineering and Mathematics (ICONSTEM), 2025, pp. 1–8. doi: 10.1109/iconstem65670.2025.11374860.
X.-C. Tai, H. Liu, R. H. Chan, and L. Li, “A mathematical explanation of UNet,” Math. Found. Comput., vol. 8, no. 5, pp. 874–889, 2025, doi: 10.3934/mfc.2024040.
P. Pan, C. Zhang, J. Sun, and L. Guo, “Multi-scale conv-attention U-Net for medical image segmentation,” Sci. Rep., vol. 15, 2025, doi: 10.1038/s41598-025-96101-8.
M. Yeung, E. Sala, C.-B. Schönlieb, and L. Rundo, “Unified Focal loss: Generalising Dice and cross entropy-based losses to handle class imbalanced medical image segmentation,” Comput. Med. Imaging Graph., vol. 95, p. 102026, 2022, doi: https://doi.org/10.1016/j.compmedimag.2021.102026.
S. S. M. Salehi, D. Erdogmus, and A. Gholipour, “Tversky Loss Function for Image Segmentation Using 3D Fully Convolutional Deep Networks,” in Machine Learning in Medical Imaging, LNCS vol. 10541, 2017, pp. 379–387, doi: 10.1007/978-3-319-67389-9_44.
N. Abraham and N. M. Khan, “A Novel Focal Tversky Loss Function With Improved Attention U-Net for Lesion Segmentation,” in 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019), 2019, pp. 683–687, doi: 10.1109/ISBI.2019.8759329.
S. Yang and G. Berdine, “Confusion matrix,” Southwest Respir. Crit. Care Chronicles, vol. 12, no. 53, pp. 75–79, 2024, doi: 10.12746/swrccc.v12i53.1391.
G. V. Vlăsceanu, N. Tarbă, M.-L. Voncilă, and C. A. Boiangiu, “Selecting the right metric: a detailed study on image segmentation evaluation,” BRAIN. Broad Res. Artif. Intell. Neurosci., vol. 15, no. 4, pp. 295–318, 2024, doi: 10.70594/brain/15.4/20.
L. Maier-Hein et al., “Metrics reloaded: recommendations for image analysis validation,” Nat. Methods, vol. 21, no. 2, pp. 195–212, 2024, doi: 10.1038/s41592-023-02151-z.
J. Sun et al., “Pathological Asymmetry-Guided Progressive Learning for Acute Ischemic Stroke Infarct Segmentation.,” IEEE Trans. Med. Imaging, vol. 43, no. 12, pp. 4146–4160, Dec. 2024, doi: 10.1109/TMI.2024.3414842.
J. Sun et al., “Deep learning for segmenting ischemic stroke infarction in non-contrast CT scans by utilizing asymmetry,” Clin. Neuroradiol., vol. 36, no. 1, pp. 129–142, Mar. 2026, doi: 10.1007/s00062-025-01559-8.
L. Cherikbayeva, Z. Melis, V. Berikov, A. Yeleussinov, and A. Ataniyazova, “Deep Ensemble for Ischemic Stroke Segmentation on Non-Contrast CT Images,” Mater. Proc., vol. 33, no. 1, p. 4, 2026, doi: 10.3390/materproc2026033004.
Copyright (c) 2026 Osmalina Rahma, Aisyah Widayani, Nimas Maharani Putri, Rifai Chai
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution-ShareAlikel 4.0 International (CC BY-SA 4.0) that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
.png)
.png)
.png)
.png)
.png)
