Cross-modality image synthesis from unpaired data using cyclegan

Effects of gradient consistency loss and training data size

Yuta Hiasa, Yoshito Otake, Masaki Takao, Takumi Matsuoka, Kazuma Takashima, Aaron Carass, Jerry Ladd Prince, Nobuhiko Sugano, Yoshinobu Sato

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

Abstract

CT is commonly used in orthopedic procedures. MRI is used along with CT to identify muscle structures and diagnose osteonecrosis due to its superior soft tissue contrast. However, MRI has poor contrast for bone structures. Clearly, it would be helpful if a corresponding CT were available, as bone boundaries are more clearly seen and CT has a standardized (i.e., Hounsfield) unit. Therefore, we aim at MR-to-CT synthesis. While the CycleGAN was successfully applied to unpaired CT and MR images of the head, these images do not have as much variation of intensity pairs as do images in the pelvic region due to the presence of joints and muscles. In this paper, we extended the CycleGAN approach by adding the gradient consistency loss to improve the accuracy at the boundaries. We conducted two experiments. To evaluate image synthesis, we investigated dependency of image synthesis accuracy on (1) the number of training data and (2) incorporation of the gradient consistency loss. To demonstrate the applicability of our method, we also investigated segmentation accuracy on synthesized images.

Original languageEnglish (US)
Title of host publicationSimulation and Synthesis in Medical Imaging - Third International Workshop, SASHIMI 2018, Held in Conjunction with MICCAI 2018, Proceedings
EditorsOrcun Goksel, Ipek Oguz, Ali Gooya, Ninon Burgos
PublisherSpringer Verlag
Pages31-41
Number of pages11
ISBN (Print)9783030005351
DOIs
StatePublished - Jan 1 2018
Event3rd International Workshop on Simulation and Synthesis in Medical Imaging, SASHIMI 2018 Held in Conjunction with 21st International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2018 - Granada, Spain
Duration: Sep 16 2018Sep 16 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11037 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other3rd International Workshop on Simulation and Synthesis in Medical Imaging, SASHIMI 2018 Held in Conjunction with 21st International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2018
CountrySpain
CityGranada
Period9/16/189/16/18

Fingerprint

Modality
Synthesis
Gradient
Magnetic resonance imaging
Muscle
Bone
Orthopedics
Tissue
Soft Tissue
Training
Segmentation
Experiments
Unit
Evaluate
Demonstrate
Experiment

Keywords

  • CT
  • CycleGAN
  • Image synthesis
  • MR
  • Musculoskeletal image
  • Segmentation

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Hiasa, Y., Otake, Y., Takao, M., Matsuoka, T., Takashima, K., Carass, A., ... Sato, Y. (2018). Cross-modality image synthesis from unpaired data using cyclegan: Effects of gradient consistency loss and training data size. In O. Goksel, I. Oguz, A. Gooya, & N. Burgos (Eds.), Simulation and Synthesis in Medical Imaging - Third International Workshop, SASHIMI 2018, Held in Conjunction with MICCAI 2018, Proceedings (pp. 31-41). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11037 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-00536-8_4

Cross-modality image synthesis from unpaired data using cyclegan : Effects of gradient consistency loss and training data size. / Hiasa, Yuta; Otake, Yoshito; Takao, Masaki; Matsuoka, Takumi; Takashima, Kazuma; Carass, Aaron; Prince, Jerry Ladd; Sugano, Nobuhiko; Sato, Yoshinobu.

Simulation and Synthesis in Medical Imaging - Third International Workshop, SASHIMI 2018, Held in Conjunction with MICCAI 2018, Proceedings. ed. / Orcun Goksel; Ipek Oguz; Ali Gooya; Ninon Burgos. Springer Verlag, 2018. p. 31-41 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11037 LNCS).

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

Hiasa, Y, Otake, Y, Takao, M, Matsuoka, T, Takashima, K, Carass, A, Prince, JL, Sugano, N & Sato, Y 2018, Cross-modality image synthesis from unpaired data using cyclegan: Effects of gradient consistency loss and training data size. in O Goksel, I Oguz, A Gooya & N Burgos (eds), Simulation and Synthesis in Medical Imaging - Third International Workshop, SASHIMI 2018, Held in Conjunction with MICCAI 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11037 LNCS, Springer Verlag, pp. 31-41, 3rd International Workshop on Simulation and Synthesis in Medical Imaging, SASHIMI 2018 Held in Conjunction with 21st International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2018, Granada, Spain, 9/16/18. https://doi.org/10.1007/978-3-030-00536-8_4
Hiasa Y, Otake Y, Takao M, Matsuoka T, Takashima K, Carass A et al. Cross-modality image synthesis from unpaired data using cyclegan: Effects of gradient consistency loss and training data size. In Goksel O, Oguz I, Gooya A, Burgos N, editors, Simulation and Synthesis in Medical Imaging - Third International Workshop, SASHIMI 2018, Held in Conjunction with MICCAI 2018, Proceedings. Springer Verlag. 2018. p. 31-41. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-00536-8_4
Hiasa, Yuta ; Otake, Yoshito ; Takao, Masaki ; Matsuoka, Takumi ; Takashima, Kazuma ; Carass, Aaron ; Prince, Jerry Ladd ; Sugano, Nobuhiko ; Sato, Yoshinobu. / Cross-modality image synthesis from unpaired data using cyclegan : Effects of gradient consistency loss and training data size. Simulation and Synthesis in Medical Imaging - Third International Workshop, SASHIMI 2018, Held in Conjunction with MICCAI 2018, Proceedings. editor / Orcun Goksel ; Ipek Oguz ; Ali Gooya ; Ninon Burgos. Springer Verlag, 2018. pp. 31-41 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{301c889c2fb74be89d5afc1e3efbd4fb,
title = "Cross-modality image synthesis from unpaired data using cyclegan: Effects of gradient consistency loss and training data size",
abstract = "CT is commonly used in orthopedic procedures. MRI is used along with CT to identify muscle structures and diagnose osteonecrosis due to its superior soft tissue contrast. However, MRI has poor contrast for bone structures. Clearly, it would be helpful if a corresponding CT were available, as bone boundaries are more clearly seen and CT has a standardized (i.e., Hounsfield) unit. Therefore, we aim at MR-to-CT synthesis. While the CycleGAN was successfully applied to unpaired CT and MR images of the head, these images do not have as much variation of intensity pairs as do images in the pelvic region due to the presence of joints and muscles. In this paper, we extended the CycleGAN approach by adding the gradient consistency loss to improve the accuracy at the boundaries. We conducted two experiments. To evaluate image synthesis, we investigated dependency of image synthesis accuracy on (1) the number of training data and (2) incorporation of the gradient consistency loss. To demonstrate the applicability of our method, we also investigated segmentation accuracy on synthesized images.",
keywords = "CT, CycleGAN, Image synthesis, MR, Musculoskeletal image, Segmentation",
author = "Yuta Hiasa and Yoshito Otake and Masaki Takao and Takumi Matsuoka and Kazuma Takashima and Aaron Carass and Prince, {Jerry Ladd} and Nobuhiko Sugano and Yoshinobu Sato",
year = "2018",
month = "1",
day = "1",
doi = "10.1007/978-3-030-00536-8_4",
language = "English (US)",
isbn = "9783030005351",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "31--41",
editor = "Orcun Goksel and Ipek Oguz and Ali Gooya and Ninon Burgos",
booktitle = "Simulation and Synthesis in Medical Imaging - Third International Workshop, SASHIMI 2018, Held in Conjunction with MICCAI 2018, Proceedings",

}

TY - GEN

T1 - Cross-modality image synthesis from unpaired data using cyclegan

T2 - Effects of gradient consistency loss and training data size

AU - Hiasa, Yuta

AU - Otake, Yoshito

AU - Takao, Masaki

AU - Matsuoka, Takumi

AU - Takashima, Kazuma

AU - Carass, Aaron

AU - Prince, Jerry Ladd

AU - Sugano, Nobuhiko

AU - Sato, Yoshinobu

PY - 2018/1/1

Y1 - 2018/1/1

N2 - CT is commonly used in orthopedic procedures. MRI is used along with CT to identify muscle structures and diagnose osteonecrosis due to its superior soft tissue contrast. However, MRI has poor contrast for bone structures. Clearly, it would be helpful if a corresponding CT were available, as bone boundaries are more clearly seen and CT has a standardized (i.e., Hounsfield) unit. Therefore, we aim at MR-to-CT synthesis. While the CycleGAN was successfully applied to unpaired CT and MR images of the head, these images do not have as much variation of intensity pairs as do images in the pelvic region due to the presence of joints and muscles. In this paper, we extended the CycleGAN approach by adding the gradient consistency loss to improve the accuracy at the boundaries. We conducted two experiments. To evaluate image synthesis, we investigated dependency of image synthesis accuracy on (1) the number of training data and (2) incorporation of the gradient consistency loss. To demonstrate the applicability of our method, we also investigated segmentation accuracy on synthesized images.

AB - CT is commonly used in orthopedic procedures. MRI is used along with CT to identify muscle structures and diagnose osteonecrosis due to its superior soft tissue contrast. However, MRI has poor contrast for bone structures. Clearly, it would be helpful if a corresponding CT were available, as bone boundaries are more clearly seen and CT has a standardized (i.e., Hounsfield) unit. Therefore, we aim at MR-to-CT synthesis. While the CycleGAN was successfully applied to unpaired CT and MR images of the head, these images do not have as much variation of intensity pairs as do images in the pelvic region due to the presence of joints and muscles. In this paper, we extended the CycleGAN approach by adding the gradient consistency loss to improve the accuracy at the boundaries. We conducted two experiments. To evaluate image synthesis, we investigated dependency of image synthesis accuracy on (1) the number of training data and (2) incorporation of the gradient consistency loss. To demonstrate the applicability of our method, we also investigated segmentation accuracy on synthesized images.

KW - CT

KW - CycleGAN

KW - Image synthesis

KW - MR

KW - Musculoskeletal image

KW - Segmentation

UR - http://www.scopus.com/inward/record.url?scp=85053912661&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053912661&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-00536-8_4

DO - 10.1007/978-3-030-00536-8_4

M3 - Conference contribution

SN - 9783030005351

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 31

EP - 41

BT - Simulation and Synthesis in Medical Imaging - Third International Workshop, SASHIMI 2018, Held in Conjunction with MICCAI 2018, Proceedings

A2 - Goksel, Orcun

A2 - Oguz, Ipek

A2 - Gooya, Ali

A2 - Burgos, Ninon

PB - Springer Verlag

ER -