Advertisement

Deep Learning-Based Pneumothorax Detection in Ultrasound Videos

  • Courosh MehanianEmail author
  • Sourabh Kulhare
  • Rachel Millin
  • Xinliang Zheng
  • Cynthia Gregory
  • Meihua Zhu
  • Hua Xie
  • James Jones
  • Jack Lazar
  • Amber Halse
  • Todd Graham
  • Mike Stone
  • Kenton Gregory
  • Ben Wilson
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 11798)

Abstract

Pneumothorax (PTX) is a medical and surgical emergency that can lead to hemodynamic instability and life-threatening collapse of the lung. PTX is usually detected using chest X-ray but can be detected using lung ultrasound, which requires interpretation by an expert radiologist. We are developing an AI based algorithm for the automated interpretation of lung ultrasound video to enable fast diagnosis of pneumothorax at the point of care by health care providers without extensive training in ultrasound. In this work, we developed and compared several deep learning methods for identifying pneumothoraces in 3-s ultrasound videos collected with a handheld ultrasound system. The first group of methods were based on convolutional neural networks (CNNs) paired with time-mapping preprocessing algorithms, including reconstructed M-mode and the proposed simplified optical flow transform (SOFT). These preprocessing methods were either used alone or in combination in a single “fusion” CNN. The second class of algorithm used a Deep Learning architecture that combines a CNN for processing spatial information (Inception V3) with a recurrent network (long-short-term-memory, or LSTM) for temporal analysis, enabling raw video to be fed directly into the neural network without preprocessing. We used data from a swine pneumothorax model to train and test the proposed algorithms, comparing their performance. Despite limited data, all algorithms achieved an AUC for pneumothorax detection greater than 0.83.

Keywords

Deep Learning Pneumothorax Lung ultrasound 

Notes

Acknowledgments

The project is supported by Agreement # HR0011-17-3-001 between the Defense Advanced Research Project Agency and Inventive Government Solutions, LLC (IGS). Use, duplication, or disclosure is subject to the restrictions of the agreement. This project does not necessarily reflect the position or policy of the government. No official endorsement should be inferred. This work was also supported by the Global Good Fund I, LLC through IGS.

Conflict of Interest

Drs. Kenton Gregory and Cynthia Gregory have a Significant Financial Interest in Intellectual Ventures Laboratory, a company that may have commercial interest in the results of this research and technology. This potential conflict of interest has been reviewed and managed by OHSU.

References

  1. 1.
    Wilkerson, R.G., Stone, M.B.: Sensitivity of bedside ultrasound and supine anteroposterior chest radiographs for the identification of pneumothorax after blunt trauma. Acad. Emerg. Med. 17(1), 11–17 (2010)CrossRefGoogle Scholar
  2. 2.
    Lichtenstein, D.A.: Lung ultrasound in the critically ill. Ann. Intensive Care 4(1), 1–12 (2014)MathSciNetCrossRefGoogle Scholar
  3. 3.
    Stone, M.B.: Ultrasound diagnosis of traumatic pneumothorax. Pictorial CME 1(1), 19–20 (2008)Google Scholar
  4. 4.
    Brattain, L., Telfer, B., Dhyani, M., Grajo, J., Samir, A.: Machine learning for medical ultrasound: status, methods, and future opportunities. Abdom. Radiol. 43(4), 786–799 (2018)CrossRefGoogle Scholar
  5. 5.
    Liu, S.F., et al.: Deep learning in medical ultrasound analysis: a review. Engineering 5(2), 261–275 (2019)MathSciNetCrossRefGoogle Scholar
  6. 6.
    Lindsey, T., Lee, R., Grisell, R., Vega, S., Veazey, S.: Automated pneumothorax diagnosis using deep neural networks. In: Vera-Rodriguez, R., Fierrez, J., Morales, A. (eds.) CIARP 2018. LNCS, vol. 11401, pp. 723–731. Springer, Cham (2019).  https://doi.org/10.1007/978-3-030-13469-3_84CrossRefGoogle Scholar
  7. 7.
    Volpicelli, G., Elbarbary, M., Blaivas, M., Lichtenstein, D., et al.: International evidence-based recommendations for point-of-care lung ultrasound. Intensive Care Med. 38(4), 577–591 (2012)CrossRefGoogle Scholar
  8. 8.
    Kulhare, S., et al.: Ultrasound-based detection of lung abnormalities using single shot detection convolutional neural networks. In: Stoyanov, D., et al. (eds.) POCUS/BIVPCS/CuRIOUS/CPM 2018. LNCS, vol. 11042, pp. 65–73. Springer, Cham (2018).  https://doi.org/10.1007/978-3-030-01045-4_8CrossRefGoogle Scholar
  9. 9.
    Liu, W., et al.: SSD: Single Shot MultiBox Detector. In: Leibe, B., Matas, J., Sebe, N., Welling, M. (eds.) ECCV 2016. LNCS, vol. 9905, pp. 21–37. Springer, Cham (2016).  https://doi.org/10.1007/978-3-319-46448-0_2CrossRefGoogle Scholar
  10. 10.
    Horn, B., Schunk, B.G.: Determining optical flow. Artif. Intell. 17, 185–203 (1981)CrossRefGoogle Scholar
  11. 11.
    Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9(8), 1735–1780 (1997)CrossRefGoogle Scholar
  12. 12.
    Bengio, Y., Simard, P., Frasconi, P.: Learning long-term dependencies with gradient descent is difficult. IEEE Trans. Neural Networks 5(2), 157–166 (1994)CrossRefGoogle Scholar
  13. 13.
    Szegedy, C., Vanhoucke, V., Loffe, S., Shlens, J., Wojna, Z.: Rethinking the inception architecture for computer vision. In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2016, Las Vegas, pp. 2818–2826. IEEE (2016)Google Scholar
  14. 14.
    Jabbar, S., Day, C., Heinz, N., Chadwick, E.: Using Convolutional Neural Network for edge detection in musculoskeletal ultrasound images. In: International Joint Conference on Neural Networks (IJCNN) 2016, Vancouver, pp. 4619–4626. IEEE (2016)Google Scholar
  15. 15.
    Shin, J.Y., Tajbakhsh, N., Hurst, R.T., Kendall, C.B., Liang, J.M.: Automating carotid intima-media thickness video interpretation with convolutional neural networks. In: Conference on Computer Vision and Pattern Recognition (CVPR) 2016, Las Vegas, pp. 2526–2535. IEEE (2016)Google Scholar
  16. 16.
    Chen, H., et al.: Standard plane localization in fetal ultrasound via domain transferred deep neural networks. IEEE J. Biomed. Health Inform. 19(5), 1627–1636 (2015)CrossRefGoogle Scholar
  17. 17.
    Lichtenstein, D.A., Menu, Y.: A bedside ultrasound sign ruling out pneumothorax in the critically ill. Lung sliding. Chest 108(5), 1345–1348 (1995)CrossRefGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Courosh Mehanian
    • 1
    • 2
    Email author
  • Sourabh Kulhare
    • 1
    • 2
  • Rachel Millin
    • 1
    • 2
  • Xinliang Zheng
    • 1
    • 2
  • Cynthia Gregory
    • 3
  • Meihua Zhu
    • 3
  • Hua Xie
    • 3
  • James Jones
    • 3
  • Jack Lazar
    • 3
  • Amber Halse
    • 3
  • Todd Graham
    • 3
  • Mike Stone
    • 3
    • 4
  • Kenton Gregory
    • 1
  • Ben Wilson
    • 1
    • 2
  1. 1.Inventive Government Solutions, LLCBellevueUSA
  2. 2.Intellectual Ventures LaboratoryBellevueUSA
  3. 3.Oregon Health Sciences UniversityPortlandUSA
  4. 4.Legacy Emanuel HospitalPortlandUSA

Personalised recommendations