XSIM: A structural similarity index measure optimized for MRI QSM

Carlos Milovic; Cristian Tejos; Javier Silva; Karin Shmueli; Pablo Irarrazaval

doi:10.1002/mrm.30271

journal article Aug 23, 2024

XSIM: A structural similarity index measure optimized for MRI QSM

Carlos Milovic

Cristian Tejos Javier Silva

Karin Shmueli

Pablo Irarrazaval

Magnetic Resonance in Medicine Vol. 93 No. 1 pp. 411-421 · Wiley

View at Publisher Save 10.1002/mrm.30271

Abstract

AbstractPurposeThe structural similarity index measure (SSIM) has become a popular quality metric to evaluate QSM in a way that is closer to human perception than RMS error (RMSE). However, SSIM may overpenalize errors in diamagnetic tissues and underpenalize them in paramagnetic tissues, resulting in biasing. In addition, extreme artifacts may compress the dynamic range, resulting in unrealistically high SSIM scores (hacking). To overcome biasing and hacking, we propose XSIM: SSIM implemented in the native QSM range, and with internal parameters optimized for QSM.MethodsWe used forward simulations from a COSMOS ground‐truth brain susceptibility map included in the 2016 QSM Reconstruction Challenge to investigate the effect of QSM reconstruction errors on the SSIM, XSIM, and RMSE metrics. We also used these metrics to optimize QSM reconstructions of the in vivo challenge data set. We repeated this experiment with the QSM abdominal phantom. To validate the use of XSIM instead of SSIM for QSM quality assessment across a range of different reconstruction techniques/algorithms, we analyzed the reconstructions submitted to the 2019 QSM Reconstruction Challenge 2.0.ResultsOur experiments confirmed the biasing and hacking effects on the SSIM metric applied to QSM. The XSIM metric was robust to those effects, penalizing the presence of streaking artifacts and reconstruction errors. Using XSIM to optimize QSM reconstruction regularization weights returned less overregularization than SSIM and RMSE.ConclusionXSIM is recommended over traditional SSIM to evaluate QSM reconstructions against a known ground truth, as it avoids biasing and hacking effects and provides a larger dynamic range of scores.

Topics

No keywords indexed for this article. Browse by subject →

References

40

[1]

10.1002/mrm.20198

[2]

10.1002/mrm.25358

[3]

10.1016/j.mri.2014.09.004

[4]

10.1016/b978-0-12-817057-1.00033-0

[5]

10.1002/nbm.3569

[6]

10.1002/mrm.22135

[7]

10.1016/j.neuroimage.2010.06.070

[8]

10.1002/mrm.22187

[9]

10.1002/cmr.b.10083

[10]

10.1002/cmr.b.20034

[11]

10.1016/j.neuroimage.2012.05.067

[12]

10.1109/tbme.2017.2749298

[13]

10.1002/mrm.26830

[14]

10.1002/mrm.28754

[15]

10.1002/mrm.22816

[16]

10.1109/tmi.2016.2544958

[17]

10.1002/mrm.21828

[18]

10.1002/mrm.22482

[19]

MR Image Reconstruction From Highly Undersampled k-Space Data by Dictionary Learning

Saiprasad Ravishankar, YORAM BRESLER

IEEE Transactions on Medical Imaging 10.1109/tmi.2010.2090538

[20]

Image quality assessment: from error visibility to structural similarity

Zhou Wang, A.C. Bovik, H.R. Sheikh et al.

IEEE Transactions on Image Processing 10.1109/tip.2003.819861

[21]

10.1109/tip.2006.881959

[22]

10.1109/icpr.2010.579

[23]

10.1109/tip.2008.921328

[24]

10.1109/msp.2008.930649

[25]

Wang Z

[26]

10.1002/mrm.28185

[27]

10.1002/mrm.27483

[28]

10.1137/18m1191452

[29]

10.1016/j.neuroimage.2018.06.030

[30]

10.1016/j.mri.2016.05.014

[31]

Bilgic B (2015)

[32]

10.1002/mrm.27073

[33]

10.1002/mrm.29597

[34]

10.1016/j.neuroimage.2007.12.025

[35]

10.1002/mrm.28716

[36]

10.1002/mrm.24272

[37]

10.1002/mrm.26331

[38]

10.1016/j.neuroimage.2019.03.060

[39]

10.1002/nbm.4292

[40]

10.1002/mrm.30006

Metrics

14

Citations

40

References

Details

Published: Aug 23, 2024
Vol/Issue: 93(1)
Pages: 411-421
License: View

Authors

C

Carlos Milovic

School of Electrical Engineering Pontificia Universidad Catolica de Valparaiso Valparaiso Chile

C

Cristian Tejos

Department of Electrical Engineering Pontificia Universidad Catolica de Chile Santiago Chile; Biomedical Imaging Center Pontificia Universidad Catolica de Chile Santiago Chile; Millennium Institute for Intelligent Healthcare Engineering (iHEALTH) Santiago Chile

J

Javier Silva

Department of Electrical Engineering Pontificia Universidad Catolica de Chile Santiago Chile; Biomedical Imaging Center Pontificia Universidad Catolica de Chile Santiago Chile

K

Karin Shmueli

Department of Medical Physics and Biomedical Engineering University College London London UK

P

Pablo Irarrazaval

Department of Electrical Engineering Pontificia Universidad Catolica de Chile Santiago Chile; Biomedical Imaging Center Pontificia Universidad Catolica de Chile Santiago Chile; Millennium Institute for Intelligent Healthcare Engineering (iHEALTH) Santiago Chile; Institute for Biological and Medical Engineering Pontificia Universidad Catolica de Chile Santiago Chile

Funding

European Research Council Award: SFN 770939

Cancer Research UK Award: C53545/A24348

Fondo Nacional de Desarrollo Cientìfico y Tecnológico Award: 1210747

Vicerrectoría de Investigación, Creación e Innovación Award: VINCI‐DI Iniciacion 2023

Cite This Article

Carlos Milovic, Cristian Tejos, Javier Silva, et al. (2024). XSIM: A structural similarity index measure optimized for MRI QSM. Magnetic Resonance in Medicine, 93(1), 411-421. https://doi.org/10.1002/mrm.30271

XSIM: A structural similarity index measure optimized for MRI QSM

You May Also Like