A deeper convolutional neural network for denoising low-dose CT images

Byeongjoon Kim, Hyunjung Shim, Jongduk Baek

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

2 Citations (Scopus)

Abstract

In recent years, CNN has been gaining attention as a powerful denoising tool after the pioneering work [7], developing 3-layer convolutional neural network (CNN). However, the 3-layer CNN may lose details or contrast after denoising due to its shallow depth. In this study, we propose a deeper, 7-layer CNN for denoising low-dose CT images. We introduced dimension shrinkage and expansion steps to control explosion of the number of parameters, and also applied the batch normalization to alleviate difficulty in optimization. The network was trained and tested with Shepp-Logan phantom images reconstructed by FBP algorithm from projection data generated in a fan-beam geometry. For a training set and a test set, the independently generated uniform noise with different noise levels was added to the projection data. The image quality improvement was evaluated both qualitatively and quantitatively, and the results show that the proposed CNN effectively reduces the noise without resolution loss compared to BM3D and the 3-layer CNN.

Original languageEnglish
Title of host publicationMedical Imaging 2018
Subtitle of host publicationPhysics of Medical Imaging
EditorsTaly Gilat Schmidt, Guang-Hong Chen, Joseph Y. Lo
PublisherSPIE
Volume10573
ISBN (Electronic)9781510616356
DOIs
Publication statusPublished - 2018 Jan 1
EventMedical Imaging 2018: Physics of Medical Imaging - Houston, United States
Duration: 2018 Feb 122018 Feb 15

Other

OtherMedical Imaging 2018: Physics of Medical Imaging
CountryUnited States
CityHouston
Period18/2/1218/2/15

Fingerprint

Noise
Neural networks
dosage
Explosions
Quality Improvement
projection
Image quality
Fans
fans
shrinkage
explosions
education
Geometry
optimization
expansion
geometry

All Science Journal Classification (ASJC) codes

  • Electronic, Optical and Magnetic Materials
  • Biomaterials
  • Atomic and Molecular Physics, and Optics
  • Radiology Nuclear Medicine and imaging

Cite this

Kim, B., Shim, H., & Baek, J. (2018). A deeper convolutional neural network for denoising low-dose CT images. In T. G. Schmidt, G-H. Chen, & J. Y. Lo (Eds.), Medical Imaging 2018: Physics of Medical Imaging (Vol. 10573). [105733P] SPIE. https://doi.org/10.1117/12.2286720
Kim, Byeongjoon ; Shim, Hyunjung ; Baek, Jongduk. / A deeper convolutional neural network for denoising low-dose CT images. Medical Imaging 2018: Physics of Medical Imaging. editor / Taly Gilat Schmidt ; Guang-Hong Chen ; Joseph Y. Lo. Vol. 10573 SPIE, 2018.
@inproceedings{1589b2a845564f55ad8d23298c9cdb52,
title = "A deeper convolutional neural network for denoising low-dose CT images",
abstract = "In recent years, CNN has been gaining attention as a powerful denoising tool after the pioneering work [7], developing 3-layer convolutional neural network (CNN). However, the 3-layer CNN may lose details or contrast after denoising due to its shallow depth. In this study, we propose a deeper, 7-layer CNN for denoising low-dose CT images. We introduced dimension shrinkage and expansion steps to control explosion of the number of parameters, and also applied the batch normalization to alleviate difficulty in optimization. The network was trained and tested with Shepp-Logan phantom images reconstructed by FBP algorithm from projection data generated in a fan-beam geometry. For a training set and a test set, the independently generated uniform noise with different noise levels was added to the projection data. The image quality improvement was evaluated both qualitatively and quantitatively, and the results show that the proposed CNN effectively reduces the noise without resolution loss compared to BM3D and the 3-layer CNN.",
author = "Byeongjoon Kim and Hyunjung Shim and Jongduk Baek",
year = "2018",
month = "1",
day = "1",
doi = "10.1117/12.2286720",
language = "English",
volume = "10573",
editor = "Schmidt, {Taly Gilat} and Guang-Hong Chen and Lo, {Joseph Y.}",
booktitle = "Medical Imaging 2018",
publisher = "SPIE",
address = "United States",

}

Kim, B, Shim, H & Baek, J 2018, A deeper convolutional neural network for denoising low-dose CT images. in TG Schmidt, G-H Chen & JY Lo (eds), Medical Imaging 2018: Physics of Medical Imaging. vol. 10573, 105733P, SPIE, Medical Imaging 2018: Physics of Medical Imaging, Houston, United States, 18/2/12. https://doi.org/10.1117/12.2286720

A deeper convolutional neural network for denoising low-dose CT images. / Kim, Byeongjoon; Shim, Hyunjung; Baek, Jongduk.

Medical Imaging 2018: Physics of Medical Imaging. ed. / Taly Gilat Schmidt; Guang-Hong Chen; Joseph Y. Lo. Vol. 10573 SPIE, 2018. 105733P.

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

TY - GEN

T1 - A deeper convolutional neural network for denoising low-dose CT images

AU - Kim, Byeongjoon

AU - Shim, Hyunjung

AU - Baek, Jongduk

PY - 2018/1/1

Y1 - 2018/1/1

N2 - In recent years, CNN has been gaining attention as a powerful denoising tool after the pioneering work [7], developing 3-layer convolutional neural network (CNN). However, the 3-layer CNN may lose details or contrast after denoising due to its shallow depth. In this study, we propose a deeper, 7-layer CNN for denoising low-dose CT images. We introduced dimension shrinkage and expansion steps to control explosion of the number of parameters, and also applied the batch normalization to alleviate difficulty in optimization. The network was trained and tested with Shepp-Logan phantom images reconstructed by FBP algorithm from projection data generated in a fan-beam geometry. For a training set and a test set, the independently generated uniform noise with different noise levels was added to the projection data. The image quality improvement was evaluated both qualitatively and quantitatively, and the results show that the proposed CNN effectively reduces the noise without resolution loss compared to BM3D and the 3-layer CNN.

AB - In recent years, CNN has been gaining attention as a powerful denoising tool after the pioneering work [7], developing 3-layer convolutional neural network (CNN). However, the 3-layer CNN may lose details or contrast after denoising due to its shallow depth. In this study, we propose a deeper, 7-layer CNN for denoising low-dose CT images. We introduced dimension shrinkage and expansion steps to control explosion of the number of parameters, and also applied the batch normalization to alleviate difficulty in optimization. The network was trained and tested with Shepp-Logan phantom images reconstructed by FBP algorithm from projection data generated in a fan-beam geometry. For a training set and a test set, the independently generated uniform noise with different noise levels was added to the projection data. The image quality improvement was evaluated both qualitatively and quantitatively, and the results show that the proposed CNN effectively reduces the noise without resolution loss compared to BM3D and the 3-layer CNN.

UR - http://www.scopus.com/inward/record.url?scp=85049260438&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85049260438&partnerID=8YFLogxK

U2 - 10.1117/12.2286720

DO - 10.1117/12.2286720

M3 - Conference contribution

VL - 10573

BT - Medical Imaging 2018

A2 - Schmidt, Taly Gilat

A2 - Chen, Guang-Hong

A2 - Lo, Joseph Y.

PB - SPIE

ER -

Kim B, Shim H, Baek J. A deeper convolutional neural network for denoising low-dose CT images. In Schmidt TG, Chen G-H, Lo JY, editors, Medical Imaging 2018: Physics of Medical Imaging. Vol. 10573. SPIE. 2018. 105733P https://doi.org/10.1117/12.2286720