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

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

doi:10.1115/DMD2020-9109

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

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

School of Medicine

Research output: Chapter in Book/Report/Conference proceeding › Conference 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 language	English (US)
Title of host publication	Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020
Publisher	American Society of Mechanical Engineers (ASME)
ISBN (Electronic)	9780791883549
DOIs	https://doi.org/10.1115/DMD2020-9109
State	Published - 2020
Event	2020 Design of Medical Devices Conference, DMD 2020 - Minneapolis, United States Duration: Apr 6 2020 → Apr 9 2020

Publication series

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

Conference

Conference	2020 Design of Medical Devices Conference, DMD 2020
Country	United States
City	Minneapolis
Period	4/6/20 → 4/9/20

Keywords

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

ASJC Scopus subject areas

Biomedical Engineering

Access to Document

10.1115/DMD2020-9109

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

Balamurugan, M., Chung, K., Kuppoor, V., Mahapatra, S., Pustavoitau, A., & Manbachi, A. (2020). USDL: Inexpensive medical imaging using Deep learning techniques and ultrasound technology. In Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020 (Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020). American Society of Mechanical Engineers (ASME). https://doi.org/10.1115/DMD2020-9109

USDL : Inexpensive medical imaging using Deep learning techniques and ultrasound technology. / Balamurugan, Manish; Chung, Kathryn; Kuppoor, Venkat; Mahapatra, Smruti; Pustavoitau, Aliaksei ; Manbachi, Amir.

Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020. American Society of Mechanical Engineers (ASME), 2020. (Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Balamurugan, M, Chung, K, Kuppoor, V, Mahapatra, S, Pustavoitau, A & Manbachi, A 2020, USDL: Inexpensive medical imaging using Deep learning techniques and ultrasound technology. in Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020. Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020, American Society of Mechanical Engineers (ASME), 2020 Design of Medical Devices Conference, DMD 2020, Minneapolis, United States, 4/6/20. https://doi.org/10.1115/DMD2020-9109

Balamurugan M, Chung K, Kuppoor V, Mahapatra S, Pustavoitau A , Manbachi A. USDL: Inexpensive medical imaging using Deep learning techniques and ultrasound technology. In Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020. American Society of Mechanical Engineers (ASME). 2020. (Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020). https://doi.org/10.1115/DMD2020-9109

Balamurugan, Manish ; Chung, Kathryn ; Kuppoor, Venkat ; Mahapatra, Smruti ; Pustavoitau, Aliaksei ; Manbachi, Amir. / USDL : Inexpensive medical imaging using Deep learning techniques and ultrasound technology. Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020. American Society of Mechanical Engineers (ASME), 2020. (Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020).

@inproceedings{bec8cd1f70074b41b5add883b8a12f77,

title = "USDL: Inexpensive medical imaging using Deep learning techniques and ultrasound technology",

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.",

keywords = "Autoencoders, Deep Learning, Denoising, In Vivo Ultrasounds, MSE, PSNR, SSIM, Speckle Noise, Ultrasound imaging",

author = "Manish Balamurugan and Kathryn Chung and Venkat Kuppoor and Smruti Mahapatra and Aliaksei Pustavoitau and Amir Manbachi",

note = "Publisher Copyright: Copyright {\textcopyright} 2020 ASME Copyright: Copyright 2020 Elsevier B.V., All rights reserved.; 2020 Design of Medical Devices Conference, DMD 2020 ; Conference date: 06-04-2020 Through 09-04-2020",

year = "2020",

doi = "10.1115/DMD2020-9109",

language = "English (US)",

series = "Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020",

publisher = "American Society of Mechanical Engineers (ASME)",

booktitle = "Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020",

}

TY - GEN

T1 - USDL

T2 - 2020 Design of Medical Devices Conference, DMD 2020

AU - Balamurugan, Manish

AU - Chung, Kathryn

AU - Kuppoor, Venkat

AU - Mahapatra, Smruti

AU - Pustavoitau, Aliaksei

AU - Manbachi, Amir

PY - 2020

Y1 - 2020

N2 - 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.

AB - 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.

KW - Autoencoders

KW - Deep Learning

KW - Denoising

KW - In Vivo Ultrasounds

KW - MSE

KW - PSNR

KW - SSIM

KW - Speckle Noise

KW - Ultrasound imaging

UR - http://www.scopus.com/inward/record.url?scp=85090665395&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85090665395&partnerID=8YFLogxK

U2 - 10.1115/DMD2020-9109

DO - 10.1115/DMD2020-9109

M3 - Conference contribution

AN - SCOPUS:85090665395

T3 - Frontiers in Biomedical Devices, BIOMED - 2020 Design of Medical Devices Conference, DMD 2020

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

PB - American Society of Mechanical Engineers (ASME)

Y2 - 6 April 2020 through 9 April 2020

ER -