How Can AI Help for MRS?
Esin Ozturk-Isik1
1Institute of Biomedical Engineering, Bogaziçi University, Istanbul, Turkey

Synopsis

Magnetic resonance spectroscopic imaging (MRS/I) has benefited from the advances in the machine learning field. This talk will summarize recent MRSI studies that have used machine learning for data acquisition and reconstruction, preprocessing, super-resolution, quality control, denoising, and spectral quantification, in addition to molecular subtyping of the brain tumors.

Objective
To provide an overview of different applications of artificial intelligence in magnetic resonance spectroscopic imaging (MRS/I)
Course Summary
Magnetic resonance spectroscopic imaging provides biochemical information of the tissue. There have been some limitations of MRSI that have prevented its wider use in clinical settings. One of the major issues of MRSI is its inherently lower signal-to-noise ratio, which has partially been solved with the availability of high field MR scanners. Additionally, MRSI has a lower spatial resolution and requires a longer scan time and specialized software for accurate quantification. Recent advances in artificial intelligence (AI) and its integration with big data have helped MRSI in data acquisition and reconstruction [1-5], preprocessing [6], super-resolution [7-10], quality control [11-13], denoising [14]and spectral quantification [15, 16]. Additional machine learning studies have proposed identifying malignancy [17] or genetic factors that influence prognosis and survival in brain tumors including isocitrate dehydrogenase (IDH) and TERT promoter (TERTp) mutations in gliomas and NF2 copy number loss in meningiomas using machine learning on MRS/I [18-21]. This talk will cover some of the recent studies that have proposed the use of AI in MRS/I.

Acknowledgements

No acknowledgement found.

References

1. Iqbal, Z., et al., Deep learning can accelerate and quantify simulated localized correlated spectroscopy. Sci Rep, 2021. 11(1): p. 8727.

2. Motyka, S., et al., k-Space-based coil combination via geometric deep learning for reconstruction of non-Cartesian MRSI data. Magn Reson Med, 2021. 86(5): p. 2353-2367.

3. Bolan, P.J., et al., Automated Acquisition Planning for Magnetic Resonance Spectroscopy in Brain Cancer. Med Image Comput Comput Assist Interv, 2020. 12267: p. 730-739.

4. Lee, H., H.H. Lee, and H. Kim, Reconstruction of spectra from truncated free induction decays by deep learning in proton magnetic resonance spectroscopy. Magn Reson Med, 2020. 84(2): p. 559-568.

5. Lam, F., Y. Li, and X. Peng, Constrained Magnetic Resonance Spectroscopic Imaging by Learning Nonlinear Low-Dimensional Models. IEEE Trans Med Imaging, 2020. 39(3): p. 545-555.

6. Tapper, S., et al., Frequency and phase correction of J-difference edited MR spectra using deep learning. Magn Reson Med, 2021. 85(4): p. 1755-1765.

7. Cengiz, S., M. Valdes-Hernandez, and E. Ozturk-Isik, Super Resolution Convolutional Neural Networks for Increasing Spatial Resolution of 1H Magnetic Resonance Spectroscopic Imaging, in Medical Image Understanding and Analysis 2017, Communications in Computer and Information Science, M. Valdes Hernandez and V. Gonzalez-Castro, Editors. 2017, Springer: Edinburgh, UK. p. 641–650.

8. Iqbal, Z., et al., Super-Resolution (1)H Magnetic Resonance Spectroscopic Imaging Utilizing Deep Learning. Front Oncol, 2019. 9: p. 1010.

9. Li, X., et al., Super-Resolution Whole-Brain 3D MR Spectroscopic Imaging for Mapping D-2-Hydroxyglutarate and Tumor Metabolism in Isocitrate Dehydrogenase 1-mutated Human Gliomas. Radiology, 2020. 294(3): p. 589-597.

10. Hangel, G., et al., High-resolution metabolic mapping of gliomas via patch-based super-resolution magnetic resonance spectroscopic imaging at 7T. Neuroimage, 2019. 191: p. 587-595.

11. Liu, H., et al. Application of Deep Leaning Model for Quality Control of Short-echo 7T MRSI with Various Disease Types. in Conference Proceedings of 2021 ISMRM & SMRT Annual Meeting. 2021. virtual Vancouver, Canada: ISMRM. p. 2016.

12. Kyathanahally, S.P., et al., Quality of clinical brain tumor MR spectra judged by humans and machine learning tools. Magn Reson Med, 2018. 79(5): p. 2500-2510.

13. Kyathanahally, S.P., A. Doring, and R. Kreis, Deep learning approaches for detection and removal of ghosting artifacts in MR spectroscopy. Magn Reson Med, 2018. 80(3): p. 851-863.

14. Li, Y., et al., Machine Learning-Enabled High-Resolution Dynamic Deuterium MR Spectroscopic Imaging. IEEE Trans Med Imaging, 2021. 40(12): p. 3879-3890.

15. Lee, H.H. and H. Kim, Intact metabolite spectrum mining by deep learning in proton magnetic resonance spectroscopy of the brain. Magn Reson Med, 2019. 82(1): p. 33-48.

16. Lee, H.H. and H. Kim, Deep learning-based target metabolite isolation and big data-driven measurement uncertainty estimation in proton magnetic resonance spectroscopy of the brain. Magn Reson Med, 2020. 84(4): p. 1689-1706.

17. Ranjith, G., et al., Machine learning methods for the classification of gliomas: Initial results using features extracted from MR spectroscopy. Neuroradiol J, 2015. 28(2): p. 106-11.

18. Ozturk-Isik, E., et al., Identification of IDH and TERTp mutation status using (1) H-MRS in 112 hemispheric diffuse gliomas. J Magn Reson Imaging, 2020. 51(6): p. 1799-1809.

19. Bumes, E., et al., Non-Invasive Prediction of IDH Mutation in Patients with Glioma WHO II/III/IV Based on F-18-FET PET-Guided In Vivo (1)H-Magnetic Resonance Spectroscopy and Machine Learning. Cancers (Basel), 2020. 12(11).

20. Bas, A., et al. 1D-CNN for the Detection of IDH and TERTp Mutations in Diffuse Gliomas using Proton Magnetic Resonance Spectroscopy. in Conference Proceedings of ISMRM 2021. p. 0957.

21.Sacli Bilmez, B., et al. Identification of NF2 loss in meningiomas using 1H-MRS at 3T. in Conference Proceedings of ISMRM 2022. London, UK.

Proc. Intl. Soc. Mag. Reson. Med. 30 (2022)