Improving the Robustness of Deep Learning Models in Predicting Hematoma Expansion from Admission Head CT

References

1. 1.↵
   1. Paschali M,
   2. Conjeti S,
   3. Navarro F, et al

   . Generalizability vs. robustness: adversarial examples for medical imaging. arXiv.org 2018. https://doi.org/10.48550/arXiv.1804.00504. Accessed May 22, 2025
2. 2.↵
   1. Apostolidis KD,
   2. Papakostas GA

   . A survey on adversarial deep learning robustness in medical image analysis. Electronics 2021;10:2132 doi:10.3390/electronics10172132

   CrossRef
3. 3.↵
   1. Xu W,
   2. Evans D,
   3. Qi Y

   . Feature squeezing: detecting adversarial examples in deep neural networks. arXiv.org 2017. https://doi.org/10.48550/arXiv.1704.01155. Accessed May 22, 2025
4. 4.↵
   1. Li X,
   2. Zhu D

   . Robust Detection of Adversarial Attacks on Medical Images. In: 2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI), Iowa City, Iowa. 2020:1154–58 doi:10.1109/ISBI45749.2020.9098628

   CrossRef
5. 5.↵
   1. Joel MZ,
   2. Umrao S,
   3. Chang E, et al

   . Using adversarial images to assess the robustness of deep learning models trained on diagnostic images in oncology. JCO Clin Cancer Inform 2022;6:e2100170 doi:10.1200/CCI.21.00170 pmid:35271304

   CrossRefPubMed
6. 6.↵
   1. Rodriguez D,
   2. Nayak T,
   3. Chen Y, et al

   . On the role of deep learning model complexity in adversarial robustness for medical images. BMC Med Inform Decis Mak 2022;22:160 doi:10.1186/s12911-022-01891-w pmid:35725429

   CrossRefPubMed
7. 7.↵
   1. Abou Karam G,
   2. Chen MC,
   3. Zeevi D, et al

   . Time-dependent changes in hematoma expansion rate after supratentorial intracerebral hemorrhage and its relationship with neurological deterioration and functional outcome. Diagnostics (Basel) 2024;14:308 doi:10.3390/diagnostics14030308 pmid:38337824

   CrossRefPubMed
8. 8.↵
   1. Li G,
   2. Lin Y,
   3. Yang J, et al

   ; INTERACT4 Investigators. Intensive ambulance-delivered blood-pressure reduction in hyperacute stroke. N Engl J Med 2024;390:1862–72 doi:10.1056/NEJMoa2314741 pmid:38752650

   CrossRefPubMed
9. 9.↵
   1. Ma L,
   2. Hu X,
   3. Song L, et al

   ; INTERACT3 Investigators. The Third Intensive Care Bundle with Blood Pressure Reduction in Acute Cerebral Haemorrhage Trial (INTERACT3): an international, stepped wedge cluster randomised controlled trial. Lancet 2023;402:27–40 doi:10.1016/S0140-6736(23)00806-1 pmid:37245517

   CrossRefPubMed
10. 10.↵
    1. Haider SP,
    2. Qureshi AI,
    3. Jain A, et al

    . Radiomic markers of intracerebral hemorrhage expansion on non-contrast CT: independent validation and comparison with visual markers. Front Neurosci 2023;17:1225342 doi:10.3389/fnins.2023.1225342 pmid:37655013

    CrossRefPubMed
11. 11.↵
    1. Tran AT,
    2. Zeevi T,
    3. Haider SP, et al

    . Uncertainty-aware deep-learning model for prediction of supratentorial hematoma expansion from admission non-contrast head computed tomography scan. NPJ Digit Med 2024;7:26 doi:10.1038/s41746-024-01007-w pmid:38321131

    CrossRefPubMed
12. 12.↵
    1. Qureshi AI,
    2. Palesch YY,
    3. Barsan WG, et al

    ; ATACH-2 Trial Investigators and the Neurological Emergency Treatment Trials Network. Intensive blood-pressure lowering in patients with acute cerebral hemorrhage. N Engl J Med 2016;375:1033–43 doi:10.1056/NEJMoa1603460 pmid:27276234

    CrossRefPubMed
13. 13.↵
    1. Haider SP,
    2. Qureshi AI,
    3. Jain A, et al

    . The coronal plane maximum diameter of deep intracerebral hemorrhage predicts functional outcome more accurately than hematoma volume. Int J Stroke 2022;17:777–84 doi:10.1177/17474930211050749 pmid:34569877

    CrossRefPubMed
14. 14.↵
    1. Dowlatshahi D,
    2. Demchuk AM,
    3. Flaherty ML, et al

    ; VISTA Collaboration. Defining hematoma expansion in intracerebral hemorrhage: relationship with patient outcomes. Neurology 2011;76:1238–44 doi:10.1212/WNL.0b013e3182143317 pmid:21346218

    CrossRefPubMed
15. 15.↵
    1. Rorden C,
    2. Bonilha L,
    3. Fridriksson J, et al

    . Age-specific CT and MRI templates for spatial normalization. Neuroimage 2012;61:957–65 doi:10.1016/j.neuroimage.2012.03.020 pmid:22440645

    CrossRefPubMed
16. 16.↵
    1. Isensee F,
    2. Jaeger PF,
    3. Kohl SAA, et al

    . nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat Methods 2021;18:203–11 doi:10.1038/s41592-020-01008-z pmid:33288961

    CrossRefPubMed
17. 17.↵
    1. Goodfellow IJ,
    2. Shlens J,
    3. Szegedy C

    . explaining and harnessing adversarial examples. arXiv.org 2014. https://arxiv.org/abs/1412.6572. Accessed May 22, 2025
18. 18.↵
    1. Madry M AA,
    2. Schmidt L,
    3. Tsipras D, et al

    . Towards deep learning models resistant to adversarial attacks. arXiv.org 2017. https://arxiv.org/abs/1706.06083. Accessed May 22, 2025
19. 19.↵
    1. Otsu N

    . A threshold selection method from gray-level histograms. IEEE Trans Syst Man Cybern 1979;9:62–66 doi:10.1109/TSMC.1979.4310076

    CrossRefPubMedWeb of Science
20. 20.↵
    1. Jing Z,
    2. Tang B

    . Improved image segmentation method based on Otsu thresholding and level set techniques. J Phys: Conf Ser 2024;2813:012017 doi:10.1088/1742-6596/2813/1/012017

    CrossRef
21. 21.↵
    1. Bortsova G,
    2. Gonzalez-Gonzalo C,
    3. Wetstein SC, et al

    . Adversarial attack vulnerability of medical image analysis systems: unexplored factors. Med Image Anal 2021;73:102141 doi:10.1016/j.media.2021.102141 pmid:34246850

    CrossRefPubMed
22. 22.↵
    1. Muoka GW,
    2. Yi D,
    3. Ukwuoma CC, et al

    . A comprehensive review and analysis of deep learning-based medical image adversarial attack and defense. Mathematics 2023;11:4272 doi:10.3390/math11204272

    CrossRef
23. 23.↵
    1. Gladstone DJ,
    2. Aviv RI,
    3. Demchuk AM, et al

    ; SPOTLIGHT and STOP-IT Investigators and Coordinators. Effect of recombinant activated coagulation factor vii on hemorrhage expansion among patients with spot sign-positive acute intracerebral hemorrhage: the SPOTLIGHT and STOP-IT Randomized Clinical Trials. JAMA Neurol 2019;76:1493–501 doi:10.1001/jamaneurol.2019.2636 pmid:31424491

    CrossRefPubMed
24. 24.↵
    1. Meretoja A,
    2. Yassi N,
    3. Wu TY, et al

    . Tranexamic acid in patients with intracerebral haemorrhage (STOP-AUST): a multicentre, randomised, placebo-controlled, phase 2 trial. Lancet Neurol 2020;19:980–87 doi:10.1016/S1474-4422(20)30369-0 pmid:33128912

    CrossRefPubMed
25. 25.↵
    1. Naidech AM,
    2. Grotta J,
    3. Elm J, et al

    . Recombinant factor VIIa for hemorrhagic stroke treatment at earliest possible time (FASTEST): protocol for a phase III, double-blind, randomized, placebo-controlled trial. Int J Stroke 2022;17:806–09 doi:10.1177/17474930211042700 pmid:34427473

    CrossRefPubMed
26. 26.↵
    1. Leasure AC,
    2. Qureshi AI,
    3. Murthy SB, et al

    . Intensive blood pressure reduction and perihematomal edema expansion in deep intracerebral hemorrhage. Stroke 2019;50:2016–22 doi:10.1161/STROKEAHA.119.024838 pmid:31272326

    CrossRefPubMed
27. 27.↵
    1. Leasure AC,
    2. Qureshi AI,
    3. Murthy SB, et al

    . Association of intensive blood pressure reduction with risk of hematoma expansion in patients with deep intracerebral hemorrhage. JAMA Neurol 2019;76:949–55 doi:10.1001/jamaneurol.2019.1141 pmid:31081862

    CrossRefPubMed
28. 28.↵
    1. Li Q,
    2. Warren AD,
    3. Qureshi AI, et al

    . Ultra-early blood pressure reduction attenuates hematoma growth and improves outcome in intracerebral hemorrhage. Ann Neurol 2020;88:388–95 doi:10.1002/ana.25793 pmid:32453453

    CrossRefPubMed
29. 29.↵
    1. Teng L,
    2. Ren Q,
    3. Zhang P, et al

    . Artificial intelligence can effectively predict early hematoma expansion of intracerebral hemorrhage analyzing noncontrast computed tomography image. Front Aging Neurosci 2021;13:632138 doi:10.3389/fnagi.2021.632138 pmid:34122038

    CrossRefPubMed
30. 30.↵
    1. Chen Q,
    2. Zhu D,
    3. Liu J, et al

    . Clinical-radiomics nomogram for risk estimation of early hematoma expansion after acute intracerebral hemorrhage. Acad Radiology 2021;28:307–17 doi:10.1016/j.acra.2020.02.021 pmid:32238303

    CrossRefPubMed
31. 31.↵
    1. Ma C,
    2. Zhang Y,
    3. Niyazi T, et al

    . Radiomics for predicting hematoma expansion in patients with hypertensive intraparenchymal hematomas. Eur J Radiology 2019;115:10–15 doi:10.1016/j.ejrad.2019.04.001 pmid:31084753

    CrossRefPubMed
32. 32.↵
    1. Liu S,
    2. Setio AAA,
    3. Ghesu FC, et al

    . No surprises: training robust lung nodule detection for low-dose CT scans by augmenting with adversarial attacks. IEEE Trans Med Imaging 2021;40:335–45 doi:10.1109/TMI.2020.3026261 pmid:32966215

    CrossRefPubMed
33. 33.↵
    1. Vatian A,
    2. Gusarova N,
    3. Dobrenko N, et al

    . Impact of adversarial examples on the efficiency of interpretation and use of information from high-tech medical images. In: 24th Conference of Open Innovations Association (FRUCT), Moscow, Russian Federation. 2019:472–478 doi:10.23919/FRUCT.2019.8711974

    CrossRef
34. 34.↵
    1. Hu L,
    2. Zhou DW,
    3. Guo XY, et al

    . Adversarial training for prostate cancer classification using magnetic resonance imaging. Quant Imaging Med Surg 2022;12:3276–87 doi:10.21037/qims-21-1089 pmid:35655831

    CrossRefPubMed
35. 35.↵
    1. Han T,
    2. Nebelung S,
    3. Pedersoli F, et al

    . Advancing diagnostic performance and clinical usability of neural networks via adversarial training and dual batch normalization. Nat Commun 2021;12:4315. doi:10.1038/s41467-021-24464-3 pmid:34262044

    CrossRefPubMed
36. 36.↵
    1. Anastasiou T,
    2. Karagiorgou S,
    3. Petrou P, et al

    . Towards robustifying image classifiers against the perils of adversarial attacks on artificial intelligence systems. Sensors (Basel) 2022;22:6905 doi:10.3390/s22186905 pmid:36146258

    CrossRefPubMed