USDL: Inexpensive medical imaging using Deep learning techniques and ultrasound technology

Manish Balamurugan, Kathryn Chung, Venkat Kuppoor, Smruti Mahapatra, Aliaksei Pustavoitau, Amir Manbachi

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this study, we present USDL, a novel model that employs deep learning algorithms in order to reconstruct and enhance corrupted ultrasound images. We utilize an unsupervised neural network called an autoencoder which works by compressing its input into a latent-space representation and then reconstructing the output from this representation. We trained our model on a dataset that compromises of 15,700 in vivo images of the neck, wrist, elbow, and knee vasculature and compared the quality of the images generated using the structural similarity index (SSIM) and peak to noise ratio (PSNR). In closely simulated conditions, the architecture exhibited an average reconstruction accuracy of 90% as indicated by our SSIM. Our study demonstrates that USDL outperforms state of the art image enhancement and reconstruction techniques in both image quality and computational complexity, while maintaining the architecture efficiency.

Original languageEnglish (US)
Title of host publicationFrontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020
PublisherAmerican Society of Mechanical Engineers (ASME)
ISBN (Electronic)9780791883549
DOIs
StatePublished - 2020
Event2020 Design of Medical Devices Conference, DMD 2020 - Minneapolis, United States
Duration: Apr 6 2020Apr 9 2020

Publication series

NameFrontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020

Conference

Conference2020 Design of Medical Devices Conference, DMD 2020
CountryUnited States
CityMinneapolis
Period4/6/204/9/20

Keywords

  • Autoencoders
  • Deep Learning
  • Denoising
  • In Vivo Ultrasounds
  • MSE
  • PSNR
  • SSIM
  • Speckle Noise
  • Ultrasound imaging

ASJC Scopus subject areas

  • Biomedical Engineering

Fingerprint Dive into the research topics of 'USDL: Inexpensive medical imaging using Deep learning techniques and ultrasound technology'. Together they form a unique fingerprint.

Cite this