Lumbar Spine CT synthesis from MR images using CycleGAN: A preliminary study

Mariusz Bajger; Minh Son To; Gobert Lee; Adam Wells; Chee Chong; Marc Agzarian; Santosh Poonnoose

doi:10.1109/DICTA52665.2021.9647237

Lumbar Spine CT synthesis from MR images using CycleGAN: A preliminary study

Mariusz Bajger, Minh Son To, Gobert Lee, Adam Wells, Chee Chong, Marc Agzarian, Santosh Poonnoose

College of Science and Engineering

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

2 Citations (Scopus)

Abstract

In this paper, we investigate the generation of Lumbar Spine synthetic CT (sCT) images based on MR images for MR-only spinal cord injury treatment and surgery planning. CT and MRI provide complementary information and are both important for spine treatment and surgery planning. However, the acquisition of images of two different modalities interrupts the clinical workflow, adds to health care cost and poses challenges in registering the images for analysis. Translating MR images to CT images would result in seamless correlation between images and also save patients from exposure to ionizing radiation due to a CT examination. Using a large clinical dataset of 800 patients, we showed that a cycle consistent generative adversarial network (CycleGAN) can be trained with the unpaired, unaligned MR and CT images of the Lumbar Spine in generating realistic synthetic CT images. The trained model was evaluated using the paired MR and CT images of 8 patients, the average Mean Absolute Error (MAE) was found to be 184 HU with a standard deviation of 24 HU.

Original language	English
Title of host publication	2021 Digital Image Computing
Subtitle of host publication	Techniques and Applications (DICTA)
Editors	Jun Zhou, Olivier Salvado, Ferdous Sohel, Paulo Borges, Shilin Wang
Place of Publication	New York, USA
Publisher	Institute of Electrical and Electronics Engineers
Number of pages	8
ISBN (Electronic)	9781665417099, 9781665417082
DOIs	https://doi.org/10.1109/DICTA52665.2021.9647237
Publication status	Published - 2021
Event	2021 International Conference on Digital Image Computing: Techniques and Applications, DICTA 2021 - Gold Coast, Australia Duration: 29 Nov 2021 → 1 Dec 2021

Conference

Conference	2021 International Conference on Digital Image Computing: Techniques and Applications, DICTA 2021
Country/Territory	Australia
City	Gold Coast
Period	29/11/21 → 1/12/21

Keywords

CT
CycleGAN
Deep Learning
GAN
Generative Adversarial Networks
MR

Access to Document

10.1109/DICTA52665.2021.9647237Licence: In Copyright

Cite this

Bajger, M., To, M. S., Lee, G., Wells, A., Chong, C., Agzarian, M., & Poonnoose, S. (2021). Lumbar Spine CT synthesis from MR images using CycleGAN: A preliminary study. In J. Zhou, O. Salvado, F. Sohel, P. Borges, & S. Wang (Eds.), 2021 Digital Image Computing: Techniques and Applications (DICTA) Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/DICTA52665.2021.9647237

@inproceedings{4e8241e192244b3da2807bee22d34c3b,

title = "Lumbar Spine CT synthesis from MR images using CycleGAN: A preliminary study",

abstract = "In this paper, we investigate the generation of Lumbar Spine synthetic CT (sCT) images based on MR images for MR-only spinal cord injury treatment and surgery planning. CT and MRI provide complementary information and are both important for spine treatment and surgery planning. However, the acquisition of images of two different modalities interrupts the clinical workflow, adds to health care cost and poses challenges in registering the images for analysis. Translating MR images to CT images would result in seamless correlation between images and also save patients from exposure to ionizing radiation due to a CT examination. Using a large clinical dataset of 800 patients, we showed that a cycle consistent generative adversarial network (CycleGAN) can be trained with the unpaired, unaligned MR and CT images of the Lumbar Spine in generating realistic synthetic CT images. The trained model was evaluated using the paired MR and CT images of 8 patients, the average Mean Absolute Error (MAE) was found to be 184 HU with a standard deviation of 24 HU.",

keywords = "CT, CycleGAN, Deep Learning, GAN, Generative Adversarial Networks, MR",

author = "Mariusz Bajger and To, {Minh Son} and Gobert Lee and Adam Wells and Chee Chong and Marc Agzarian and Santosh Poonnoose",

year = "2021",

doi = "10.1109/DICTA52665.2021.9647237",

language = "English",

editor = "Jun Zhou and Olivier Salvado and Ferdous Sohel and Paulo Borges and Shilin Wang",

booktitle = "2021 Digital Image Computing",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

note = "2021 International Conference on Digital Image Computing: Techniques and Applications, DICTA 2021 ; Conference date: 29-11-2021 Through 01-12-2021",

}

Bajger, M, To, MS, Lee, G, Wells, A, Chong, C, Agzarian, M & Poonnoose, S 2021, Lumbar Spine CT synthesis from MR images using CycleGAN: A preliminary study. in J Zhou, O Salvado, F Sohel, P Borges & S Wang (eds), 2021 Digital Image Computing: Techniques and Applications (DICTA). Institute of Electrical and Electronics Engineers, New York, USA, 2021 International Conference on Digital Image Computing: Techniques and Applications, DICTA 2021, Gold Coast, Australia, 29/11/21. https://doi.org/10.1109/DICTA52665.2021.9647237

Lumbar Spine CT synthesis from MR images using CycleGAN: A preliminary study. / Bajger, Mariusz; To, Minh Son; Lee, Gobert et al.
2021 Digital Image Computing: Techniques and Applications (DICTA). ed. / Jun Zhou; Olivier Salvado; Ferdous Sohel; Paulo Borges; Shilin Wang. New York, USA: Institute of Electrical and Electronics Engineers, 2021.

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

TY - GEN

T1 - Lumbar Spine CT synthesis from MR images using CycleGAN

T2 - 2021 International Conference on Digital Image Computing: Techniques and Applications, DICTA 2021

AU - Bajger, Mariusz

AU - To, Minh Son

AU - Lee, Gobert

AU - Wells, Adam

AU - Chong, Chee

AU - Agzarian, Marc

AU - Poonnoose, Santosh

PY - 2021

Y1 - 2021

N2 - In this paper, we investigate the generation of Lumbar Spine synthetic CT (sCT) images based on MR images for MR-only spinal cord injury treatment and surgery planning. CT and MRI provide complementary information and are both important for spine treatment and surgery planning. However, the acquisition of images of two different modalities interrupts the clinical workflow, adds to health care cost and poses challenges in registering the images for analysis. Translating MR images to CT images would result in seamless correlation between images and also save patients from exposure to ionizing radiation due to a CT examination. Using a large clinical dataset of 800 patients, we showed that a cycle consistent generative adversarial network (CycleGAN) can be trained with the unpaired, unaligned MR and CT images of the Lumbar Spine in generating realistic synthetic CT images. The trained model was evaluated using the paired MR and CT images of 8 patients, the average Mean Absolute Error (MAE) was found to be 184 HU with a standard deviation of 24 HU.

AB - In this paper, we investigate the generation of Lumbar Spine synthetic CT (sCT) images based on MR images for MR-only spinal cord injury treatment and surgery planning. CT and MRI provide complementary information and are both important for spine treatment and surgery planning. However, the acquisition of images of two different modalities interrupts the clinical workflow, adds to health care cost and poses challenges in registering the images for analysis. Translating MR images to CT images would result in seamless correlation between images and also save patients from exposure to ionizing radiation due to a CT examination. Using a large clinical dataset of 800 patients, we showed that a cycle consistent generative adversarial network (CycleGAN) can be trained with the unpaired, unaligned MR and CT images of the Lumbar Spine in generating realistic synthetic CT images. The trained model was evaluated using the paired MR and CT images of 8 patients, the average Mean Absolute Error (MAE) was found to be 184 HU with a standard deviation of 24 HU.

KW - CT

KW - CycleGAN

KW - Deep Learning

KW - GAN

KW - Generative Adversarial Networks

KW - MR

UR - http://www.scopus.com/inward/record.url?scp=85124308726&partnerID=8YFLogxK

U2 - 10.1109/DICTA52665.2021.9647237

DO - 10.1109/DICTA52665.2021.9647237

M3 - Conference contribution

AN - SCOPUS:85124308726

BT - 2021 Digital Image Computing

A2 - Zhou, Jun

A2 - Salvado, Olivier

A2 - Sohel, Ferdous

A2 - Borges, Paulo

A2 - Wang, Shilin

PB - Institute of Electrical and Electronics Engineers

CY - New York, USA

Y2 - 29 November 2021 through 1 December 2021

ER -

Bajger M, To MS, Lee G, Wells A, Chong C, Agzarian M et al. Lumbar Spine CT synthesis from MR images using CycleGAN: A preliminary study. In Zhou J, Salvado O, Sohel F, Borges P, Wang S, editors, 2021 Digital Image Computing: Techniques and Applications (DICTA). New York, USA: Institute of Electrical and Electronics Engineers. 2021 doi: 10.1109/DICTA52665.2021.9647237

Lumbar Spine CT synthesis from MR images using CycleGAN: A preliminary study

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this