Deep Residual Inception Encoder-Decoder Network for Medical Imaging Synthesis

Fei Gao, Teresa Wu, Xianghua Chu, Hyunsoo Yoon, Yanzhe Xu, Bhavika Patel

Research output: Contribution to journalArticle

Abstract

Image synthesis is a novel solution in precision medicine for scenarios where important medical imaging is not otherwise available. Convolutional neural network is an ideal model for this task because of its powerful learning capabilities through the large number of layers and trainable parameters. In this research, we propose a new architecture of Residual Inception Encoder-Decoder Neural Network (RIED-Net) to learn the nonlinear mapping between the input images and targeting output images. To evaluate the validity of the proposed approach, it is compared with 2 models from the literature: sCT-DCNN and Shallow CNN, using both an institutional mammogram dataset from Mayo Clinic Arizona and a public neuroimaging dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Experimental results show that the proposed RIED-Net outperforms the two models on both datasets significantly in terms of structural similarity index (SSIM), mean absolute percent error (MAPE), and peak signal to noise ratio (PSNR).

Original languageEnglish (US)
JournalIEEE Journal of Biomedical and Health Informatics
DOIs
StatePublished - Jan 1 2019

Fingerprint

Medical imaging
Diagnostic Imaging
Neuroimaging
Neural networks
Precision Medicine
Signal-To-Noise Ratio
Medicine
Signal to noise ratio
Learning
Research
Datasets

Keywords

  • deep learning
  • image synthesis
  • inception
  • medical imaging and residual net

ASJC Scopus subject areas

  • Biotechnology
  • Computer Science Applications
  • Electrical and Electronic Engineering
  • Health Information Management

Cite this

Deep Residual Inception Encoder-Decoder Network for Medical Imaging Synthesis. / Gao, Fei; Wu, Teresa; Chu, Xianghua; Yoon, Hyunsoo; Xu, Yanzhe; Patel, Bhavika.

In: IEEE Journal of Biomedical and Health Informatics, 01.01.2019.

Research output: Contribution to journalArticle

@article{fa372a0dbb5842649a5c5efb34b6e732,
title = "Deep Residual Inception Encoder-Decoder Network for Medical Imaging Synthesis",
abstract = "Image synthesis is a novel solution in precision medicine for scenarios where important medical imaging is not otherwise available. Convolutional neural network is an ideal model for this task because of its powerful learning capabilities through the large number of layers and trainable parameters. In this research, we propose a new architecture of Residual Inception Encoder-Decoder Neural Network (RIED-Net) to learn the nonlinear mapping between the input images and targeting output images. To evaluate the validity of the proposed approach, it is compared with 2 models from the literature: sCT-DCNN and Shallow CNN, using both an institutional mammogram dataset from Mayo Clinic Arizona and a public neuroimaging dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Experimental results show that the proposed RIED-Net outperforms the two models on both datasets significantly in terms of structural similarity index (SSIM), mean absolute percent error (MAPE), and peak signal to noise ratio (PSNR).",
keywords = "deep learning, image synthesis, inception, medical imaging and residual net",
author = "Fei Gao and Teresa Wu and Xianghua Chu and Hyunsoo Yoon and Yanzhe Xu and Bhavika Patel",
year = "2019",
month = "1",
day = "1",
doi = "10.1109/JBHI.2019.2912659",
language = "English (US)",
journal = "IEEE Journal of Biomedical and Health Informatics",
issn = "2168-2194",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - JOUR

T1 - Deep Residual Inception Encoder-Decoder Network for Medical Imaging Synthesis

AU - Gao, Fei

AU - Wu, Teresa

AU - Chu, Xianghua

AU - Yoon, Hyunsoo

AU - Xu, Yanzhe

AU - Patel, Bhavika

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Image synthesis is a novel solution in precision medicine for scenarios where important medical imaging is not otherwise available. Convolutional neural network is an ideal model for this task because of its powerful learning capabilities through the large number of layers and trainable parameters. In this research, we propose a new architecture of Residual Inception Encoder-Decoder Neural Network (RIED-Net) to learn the nonlinear mapping between the input images and targeting output images. To evaluate the validity of the proposed approach, it is compared with 2 models from the literature: sCT-DCNN and Shallow CNN, using both an institutional mammogram dataset from Mayo Clinic Arizona and a public neuroimaging dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Experimental results show that the proposed RIED-Net outperforms the two models on both datasets significantly in terms of structural similarity index (SSIM), mean absolute percent error (MAPE), and peak signal to noise ratio (PSNR).

AB - Image synthesis is a novel solution in precision medicine for scenarios where important medical imaging is not otherwise available. Convolutional neural network is an ideal model for this task because of its powerful learning capabilities through the large number of layers and trainable parameters. In this research, we propose a new architecture of Residual Inception Encoder-Decoder Neural Network (RIED-Net) to learn the nonlinear mapping between the input images and targeting output images. To evaluate the validity of the proposed approach, it is compared with 2 models from the literature: sCT-DCNN and Shallow CNN, using both an institutional mammogram dataset from Mayo Clinic Arizona and a public neuroimaging dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Experimental results show that the proposed RIED-Net outperforms the two models on both datasets significantly in terms of structural similarity index (SSIM), mean absolute percent error (MAPE), and peak signal to noise ratio (PSNR).

KW - deep learning

KW - image synthesis

KW - inception

KW - medical imaging and residual net

UR - http://www.scopus.com/inward/record.url?scp=85065147734&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85065147734&partnerID=8YFLogxK

U2 - 10.1109/JBHI.2019.2912659

DO - 10.1109/JBHI.2019.2912659

M3 - Article

JO - IEEE Journal of Biomedical and Health Informatics

JF - IEEE Journal of Biomedical and Health Informatics

SN - 2168-2194

ER -