Cape Town - 2026 ISMRM-ISMRT Annual Meeting and Exhibition
9 May 2026 – 14 May 2026 · Cape Town, South Africa
568-06-011 ISMRM Abstract

A mixed precision FFT with applications in MRI

Primary: Acquisition & Reconstruction - Non-AI methods
Secondary: Acquisition & Reconstruction - Image Reconstruction: Non-AI
568-06-011 · Image Reconstruction II · Wednesday, 13 May, 4:55 PM–5:50 PM · Digital Posters Row I
Keywords: MRI Microscaling Mixed precision 8-bit floating point FFT
Accepted
Nikhil Deveshwar1, Abhejit Rajagopal2,3, Peder Larson4
1Department of Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, United States of America
2Allen Institute, Seattle, United States of America
3Department of Urology, University of California San Francisco, San Francisco, United States of America
4Department of Radiology and Biomedical Imaging, University Of California, San Francisco (UCSF), United States of America
Presenting Author: Bochun Mei

Synopsis

Motivation:
Goals:
Approach:
Results:
Full abstract & presentation

The full text, figures, and any recorded presentation for this abstract are not shown here. Log in if you are a member or registered attendee with access.

Full abstracts, figures, and presentations for Cape Town - 2026 ISMRM-ISMRT Annual Meeting and Exhibition are available to registered attendees. This content becomes freely available to the public roughly two years after the meeting.

To request or purchase access, contact the ISMRM Central Office at info@ismrm.org.

Log in

References

1. Lawrence L. Wald, Patrick C. McDaniel, Thomas Witzel, Jason P. Stockmann, and Clarissa Zimmerman Cooley, “Low-cost and portable mri,” Journal of Mag- netic Resonance Imaging, vol. 52, no. 3, pp. 686–696, Oct. 2019.
2. Yilong Liu, Alex T. L. Leong, Yujiao Zhao, Linfang Xiao, Henry K. F. Mak, Anderson Chun On Tsang, Gary K. K. Lau, Gilberto K. K. Leung, and Ed X. Wu, “A low- cost and shielding-free ultra-low-field brain mri scan- ner,” Nature Communications, vol. 12, no. 1, Dec. 2021.
3. Paulius Micikevicius, Dusan Stosic, Neil Burgess, Mar- ius Cornea, Pradeep Dubey, Richard Grisenthwaite, Sangwon Ha, Alexander Heinecke, Patrick Judd, John Kamalu, Naveen Mellempudi, Stuart Oberman, Mo- hammad Shoeybi, Michael Siu, and Hao Wu, “Fp8 for- mats for deep learning,” arXiv, 2022, Accessed 2025- 09-15.
4. NVIDIA Corporation, “Transformer engine: User guide and fp8 primer,” 2025, Includes FP8 usage with amax history and scaling; Accessed 2025-09-15.
5. “Ocp 8-bit floating point specification (ofp8), revision 1.0,” Tech. Rep., Open Compute Project, Dec. 2023, Defines FP8 E4M3 and E5M2 encodings; Accessed 2025-09-15.
6. “Ocp microscaling formats (mx) specification, ver- sion 1.0,” Tech. Rep., Open Compute Project, Sept. 2023, Standardizes block-shared-exponent formats MXFP8/MXFP6/MXFP4; Accessed 2025-09-15.
7. Bita Rouhani et al., “Microscaling data formats for deep learning,” arXiv, 2023, Defines MX data-format family; Accessed 2025-09-15.
8. J. Elam and C. Iovescu, “A block floating point imple- mentation for an N-point FFT on TMS320C55x DSPs,” Tech. Rep. SPRA948, Texas Instruments, 2002, Classic BFP-FFT application report; Accessed 2025-09-15.
9. NVIDIA Corporation, cuFFT 13.0 Documentation: Half-precision and Bfloat16 Transforms, 2025, Sec- tion 2.3.1 Half-precision cuFFT Transforms; Accessed 2025-09-15.
10. Binrui Li, Shenggan Cheng, and James Lin, “tcfft: Accelerating half-precision fft through tensor cores,” arXiv, 2021, Mixed/low-precision FFT on GPUs; Ac- cessed 2025-09-15.
11. SeyedAhmadMirsalari,SabaYousefzadeh,AhmedHe- mani, and Giuseppe Tagliavini, “Unleashing 8-bit float- ing point formats out of the deep-learning domain,” in 2024 31st IEEE International Conference on Electron- ics, Circuits and Systems (ICECS). Nov. 2024, p. 1–4, IEEE.
12. Graphcore Research, gfloat v0.5 Documentation, 2025, Version 0.5.
13. Graphcore Research, “gfloat: Floating-point and mi- croscaling formats for research,” 2025, GitHub reposi- tory. Version 0.5 APIs used (encode/decode, block mi- croscaling).
14. Jure Zbontar, Florian Knoll, Anuroop Sriram, Tullie Murrell, Zhengnan Huang, Matthew J. Muckley, Aaron Defazio, Ruben Stern, Patricia Johnson, Mary Bruno, Marc Parente, Krzysztof J. Geras, Joe Katsnelson, Hersh Chandarana, Zizhao Zhang, Michal Drozdzal, Adri- ana Romero, Michael Rabbat, Pascal Vincent, Nafissa Yakubova, James Pinkerton, Duo Wang, Erich Owens, C. Lawrence Zitnick, Michael P. Recht, Daniel K. Sod- ickson, and Yvonne W. Lui, “fastmri: An open dataset and benchmarks for accelerated mri,” 2018.
15. Arjun D Desai, Andrew M Schmidt, Elka B Rubin, Christopher M Sandino, Marianne S Black, Valentina Mazzoli, Kathryn J Stevens, Robert Boutin, Christopher Re ́, Garry E Gold, Brian A Hargreaves, and Akshay S Chaudhari, “Skm-tea: A dataset for accelerated mri reconstruction with dense image labels for quantitative clinical evaluation,” 2022.
16. Michael Lustig, David Donoho, and John M. Pauly, “Sparse mri: The application of compressed sensing for rapid mr imaging,” Magnetic Resonance in Medicine, vol. 58, no. 6, pp. 1182–1195, Oct. 2007.
17. Hemant K. Aggarwal, Merry P. Mani, and Mathews Ja- cob, “Modl: Model-based deep learning architecture for inverse problems,” IEEE Transactions on Medical Imaging, vol. 38, no. 2, pp. 394–405, Feb. 2019.
18. Anuroop Sriram, Jure Zbontar, Tullie Murrell, Aaron Defazio, C. Lawrence Zitnick, Nafissa Yakubova, Flo- rian Knoll, and Patricia Johnson, “End-to-end vari- ational networks for accelerated mri reconstruction,” 2020.

Cite this abstract

http://echo.ismrm.org/p/ISMRM2026/568-06-011