Pith. sign in

REVIEW 3 major objections 5 minor 39 references

Edge Computing for Physics-Driven AI in Computational MRI: A Feasibility Study

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read A physics-driven MRI reconstruction network can be cut to 8-bit complex precision and stripped of repeated FFTs without meaningful quality loss, bringing PD-AI within reach of FPGA-based edge devices near the scanner.

desk verdict A credible algorithmic feasibility study on CPU whose headline FPGA-suitability claim outruns the evidence by a wide margin. read the letter →

arxiv 2506.03183 v1 pith:Y5H26VFO submitted 2025-05-30 eess.IV cs.AIcs.ARcs.CVcs.LGphysics.med-ph

classification eess.IVcs.AIcs.ARcs.CVcs.LGphysics.med-ph
keywords physics-drivenAIcomputationalMRIedgecomputingFPGA8-bitquantizationimage-domainreconstructionequispacedundersamplingunrolledoptimization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that physics-driven AI (PD-AI) reconstruction, the current state-of-the-art approach for accelerated MRI, can be made to run on resource-constrained edge devices that could sit next to the scanner. The authors show that for equispaced undersampling with no calibration lines, the data-fidelity step can be rewritten entirely in the image domain using a sparse precomputed matrix, so that the whole pipeline needs only one IFFT per coil at the start. They further show that a pre-trained 32-bit network can be post-hoc quantized to 8-bit complex weights and activations without retraining, losing only about 0.5 dB PSNR. On high-resolution brain T2-weighted data at acceleration R=4, the edge-amenable version reaches 35.21 dB PSNR and 0.924 SSIM, compared with 35.68 dB and 0.932 for conventional PD-AI and 29.05 dB and 0.816 for standard clinical reconstruction, while cutting per-slice CPU inference time from 4.92 s to 2.49 s. The authors take this as evidence that high-volume applications such as fMRI could reconstruct data locally, reducing data transfer and storage.

What carries the argument

The load-bearing object is the image-domain encoding operator $R_\Omega = \sqrt{R} F_M^{-1} P_\Omega F_N$, which expresses how an equispaced, calibration-free undersampling pattern folds the true image R times over in the image domain after a single M-point IFFT. Because its rows contain only zeros and ones at positions $\{k, k+R, \dots, k+(R-1)R\}$, this operator can be built directly as a sparse matrix, eliminating the need for FFTs inside the data-fidelity updates; concatenating the per-coil operators gives $B_\Omega$ and preserves the data-consistency norm exactly. The second mechanism is per-tensor affine 8-bit quantization of both activations and weights, applied post-training to the CNN that acts as the proximal operator, cutting memory roughly fourfold and enabling integer arithmetic. Together these two changes reduce the data-fidelity computation to fixed sparse linear operations and the network to low-precision integer arithmetic, which is what makes an FPGA implementation plausible.

What would settle it

Implement the proposed pipeline on a real FPGA or an RTL-accurate simulator, using fixed-point arithmetic for the image-domain data-fidelity operator and 8-bit integer CNN inference, and measure PSNR/SSIM and latency on the same multi-coil brain slices. If the fixed-point operator introduces artifacts not seen on CPU, or if the FPGA cannot process a 320x320 slice within the per-slice time needed for real-time use at acceptable power, the central feasibility claim would be refuted.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that the two main obstacles to putting PD-AI reconstruction on an FPGA are removable without degrading output. First, the repeated FFT/IFFT pairs in the conjugate-gradient data-fidelity step, roughly 100 per coil, can be replaced by a single per-coil M-point IFFT followed by a fixed sparse matrix multiplication, because equispaced undersampling without a calibration region produces a simple R-fold foldover in the image domain. Second, applying per-tensor affine quantization to both weights and activations of a pre-trained 32-bit network, with no quantization-aware training, brings the CNN to 8-bit complex precision with a small quality cost. Validated on 320x320 multi-coil brain T2-weighted images at R=4, the approach yields PSNR 35.21 dB and SSIM 0.924, close to the 35.68 dB and 0.932 of conventional PD-AI and clearly above the 29.05 dB and 0.816 of clinical reconstruction, with roughly half the CPU inference time. From this the paper concludes that PD-AI is feasible for FPGA-based edge computing for many high-resolution MRI applications.

Load-bearing premise

The conclusion that this approach is suitable for FPGAs rests on an unverified transfer: the authors run the quantized no-FFT pipeline on a CPU and assume that the same quality and efficiency will carry over to an actual FPGA implementation, which they do not build or measure.

Editorial extensions

If this is right

  • At acceleration R=4 on 320x320 T2-weighted brain data, the 8-bit no-FFT version stays within 0.47 dB PSNR of the full-precision baseline, so the quality cost of edge deployment is small.
  • Removing repeated FFTs from data fidelity and quantizing the CNN roughly halves per-slice CPU inference time, from 4.92 s to 2.49 s.
  • The image-domain reformulation applies directly to echo-planar imaging and other equispaced, no-calibration-line acquisitions, which covers typical fMRI protocols.
  • 8-bit complex quantization reduces CNN memory by roughly 4x, which is what makes the network fit on resource-constrained edge hardware.
  • The proposed method outperforms clinical parallel-imaging reconstruction by about 6.2 dB PSNR, so even with quantization it remains clinically meaningful.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: the authors measured only CPU inference, so the decisive test of the feasibility claim is whether the 8-bit model and the sparse image-domain operator map onto FPGA fabric without precision loss; if they do, the roughly 2x CPU speedup could become a much larger edge gain because the CNN becomes pure integer arithmetic.
  • A natural next experiment, acknowledged in the paper, is 4-bit quantization and quantizing the data-fidelity unit; the sparse-matrix structure of $B_\Omega$ makes fixed-point data fidelity a tractable target that would complete the edge pipeline.
  • Because the image-domain operator is precomputable and sparse, compute time after the initial IFFT is dominated by the CNN; thus further latency reductions would come from network compression or distillation rather than from FFT optimization.
  • The foldover-matrix idea may extend to other regular 2D undersampling patterns with known aliasing structure, which would test whether the FFT-free reformulation generalizes beyond 1D equispaced masks.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The manuscript proposes an edge-computing-oriented modification of a physics-driven deep learning (PD-AI) MRI reconstruction method. The key ideas are (i) an image-domain reformulation of the data fidelity term for equispaced Cartesian subsampling with no calibration region, which removes the repeated FFT/IFFT operations inside the unrolled network, and (ii) 8-bit post-training quantization of the CNN regularizer. Evaluated on fastMRI axial T2-weighted data at R=4, the proposed method yields PSNR 35.21 dB and SSIM 0.924, close to a conventional PD-AI baseline (35.68 dB, 0.932) and above clinical reconstruction (29.05 dB, 0.816), with CPU inference time 2.49 s versus 4.92 s per slice. The authors conclude the method is suitable for FPGA-based edge computing devices.

Significance. If the reported results are reproducible, the paper makes a useful algorithmic contribution: the image-domain operator R_Omega is an exact algebraic reformulation for the considered sampling pattern, and the 8-bit quantization appears to preserve most of the reconstruction quality. The study is transparent about using public fastMRI data and reports both quality and speed. Its main value is as a first step toward deployment of PD-AI reconstruction on resource-constrained hardware. However, the headline conclusion about FPGA suitability currently rests on CPU measurements, so the practical significance is not yet demonstrated.

major comments (3)
  1. [Section II-C and Table I] The central claim that the method is 'suitable for FPGA-based applications' is not supported by the evidence. All inference is performed on a CPU (AMD EPYC 7352) with software 8-bit quantization; the paper reports no FPGA synthesis, resource utilization (LUT/DSP/BRAM), power consumption, or on-chip latency. Since the feasibility conclusion depends on hardware-specific behavior (fixed-point arithmetic, memory bandwidth, data movement), the CPU proxy is load-bearing but unverified. Either an FPGA implementation or a cycle-accurate hardware estimate is required, or the claims should be explicitly limited to algorithmic amenability.
  2. [Section III-B and Table I] The experimental design does not allow the contributions to be separated. The proposed method differs from the conventional baseline in two ways: 8-bit quantization and removal of FFTs/IFFTs via the image-domain reformulation. The reported quality drop (0.47 dB PSNR) and speed-up (2.49 s vs 4.92 s) could be caused by either factor or their interaction. A 2x2 ablation (quantization on/off, image-domain vs k-space data fidelity) is needed to support the individual claims made in the abstract and Section II-C.
  3. [Section III-A and Table I] The evaluation reports no test-set size, number of slices, or statistical significance. Given the standard deviations (e.g., PSNR 35.68±2.65 vs 35.21±2.44), it is unclear whether the difference between methods is meaningful. The paper should state how many volumes/slices were used, whether the metrics are averaged per slice or per volume, and provide paired significance testing or confidence intervals.
minor comments (5)
  1. [Table I] The header 'Edge Computing Amendable PD-AI' should read 'Amenable' to match the terminology used in the rest of the manuscript.
  2. [Section II-C] The norm-preserving identity ||y_Omega - E_Omega x||^2 = ||s_Omega - B_Omega x||^2 is presented as a claim. Please provide the explicit derivation or a citation, including the definition of the scaling factor in R_Omega.
  3. [Section II-C] Clarify how complex values are quantized to 8 bits (whether real and imaginary parts are each 8-bit, or the complex number is represented differently) and whether the data fidelity unit is also quantized or remains 32-bit.
  4. [Section III-B] Specify the CPU inference conditions (number of threads, batch size, whether the reported time is per slice or for a volume) to allow reproducibility.
  5. [Introduction] The phrase 'suitable for using FPGAs on edge computing devices such as the sensors' overstates the evidence; consider rewording to 'potentially amenable' until FPGA measurements are provided.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the FPGA-suitability conclusion is an extrapolation from a CPU proxy, but no derivation step feeds its own inputs back as outputs.

full rationale

The paper's derivation chain is self-contained. The image-domain reformulation in Section II-C (Eq. 7 and R_Omega = sqrt(R) F_M^{-1} P_Omega F_N) is an exact algebraic identity for equispaced sampling, and the norm equality ||y - E x||^2 = ||s - B x||^2 is stated and used as a mathematical equivalence, not fitted. The 8-bit quantization is applied post-training to a pre-trained 32-bit network with a stated per-tensor affine scheme; no quantization parameter is tuned against the reported PSNR/SSIM or timing numbers. Quality and speed figures in Table I are empirical measurements on an external benchmark (fastMRI) against clinical and conventional PD-AI baselines. The only significant weakness is the extrapolation from CPU inference to 'suitable for FPGA-based applications' (Sections II-C and IV), which is a hardware-transferability gap, not a circular step: no equation or fitted parameter is defined in terms of the conclusion. The paper itself flags this limitation ('as running inference with quantized networks on GPUs is still an area of ongoing development, we focused on performing inference on the CPU for this study') and the Discussion's future-work caveats, so score 0. Self-citations (e.g., [5, 30, 31]) support network architectures and losses but are not load-bearing for the claimed equivalence; independent fastMRI data and quantitative comparison carry the central evaluation.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The method introduces no new physical entities and fits no parameters to the target result. The central claim rests on standard mathematics and several domain assumptions about the acquisition pattern, sensitivity maps, CPU-to-FPGA transferability, dataset representativeness, and quantization adequacy. The absence of FPGA measurements makes the transferability assumption the most fragile.

assumptions (6)
  • domain assumption Equispaced sub-sampling with no calibration lines is the target acquisition pattern for the proposed method.
    The image-domain folding operator R_Omega relies on equispaced Cartesian sampling. The paper tests only R=4 on T2-weighted data and asserts relevance to fMRI/EPI, which typically uses this pattern, but this is not validated on fMRI data.
  • domain assumption Coil sensitivity maps C_k are known or can be estimated reliably.
    The forward operators E_Omega and B_Omega require coil sensitivities. The paper assumes these are available, as in standard SENSE and PD-AI methods, but does not discuss how they are obtained or the effect of estimation errors.
  • standard math The image-domain reformulation with R_Omega exactly preserves the data fidelity term.
    The identity ||y_Omega - E_Omega x||^2 = ||s_Omega - B_Omega x||^2 follows from the definition of R_Omega and Parseval's theorem, given correct scaling. This is an algebraic fact, not an approximation.
  • domain assumption CPU inference performance and accuracy transfer to FPGA implementation.
    The paper uses CPU execution as a proxy for FPGA feasibility. The claim that the method is suitable for FPGAs depends on this transfer, which is plausible but unmeasured. No FPGA resource, power, or latency data are provided.
  • domain assumption fastMRI T2-weighted data is representative of the target fMRI application.
    The method is motivated by fMRI, but evaluated on T2-weighted brain images. The authors justify this because fMRI lacks ground truth and T2 has higher resolution, but the contrast, noise, and temporal dynamics of fMRI differ, so generalization is not directly established.
  • domain assumption Post-training quantization without quantization-aware training is sufficient for acceptable quality.
    The authors apply PTQ to a pre-trained 32-bit network and show on one dataset that quality is maintained. This assumes the chosen quantization scheme and calibration are adequate, with no comparison to QAT or other quantization settings.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Edge Computing for Physics-Driven AI in Computational MRI: A Feasibility Study." pith.science (2026). https://pith.science/paper/Y5H26VFO

@misc{pith2026250603183,
  author       = {Pith},
  title        = {Pith review of: Edge Computing for Physics-Driven AI in Computational MRI: A Feasibility Study},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y5H26VFO}},
  note         = {Machine review of arXiv:2506.03183}
}
read the original abstract

Physics-driven artificial intelligence (PD-AI) reconstruction methods have emerged as the state-of-the-art for accelerating MRI scans, enabling higher spatial and temporal resolutions. However, the high resolution of these scans generates massive data volumes, leading to challenges in transmission, storage, and real-time processing. This is particularly pronounced in functional MRI, where hundreds of volumetric acquisitions further exacerbate these demands. Edge computing with FPGAs presents a promising solution for enabling PD-AI reconstruction near the MRI sensors, reducing data transfer and storage bottlenecks. However, this requires optimization of PD-AI models for hardware efficiency through quantization and bypassing traditional FFT-based approaches, which can be a limitation due to their computational demands. In this work, we propose a novel PD-AI computational MRI approach optimized for FPGA-based edge computing devices, leveraging 8-bit complex data quantization and eliminating redundant FFT/IFFT operations. Our results show that this strategy improves computational efficiency while maintaining reconstruction quality comparable to conventional PD-AI methods, and outperforms standard clinical methods. Our approach presents an opportunity for high-resolution MRI reconstruction on resource-constrained devices, highlighting its potential for real-world deployment.

Figures

Figures reproduced from arXiv: 2506.03183 by the authors.

Figure 1
Figure 1. Proposed edge computing amenable PD-AI. (a) Initially, the acquired [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Comparison of clinical reconstruction, conventional PD-AI, and the proposed edge computing-amenable PD-AI strategies. For a representative slice, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 38 canonical work pages

  1. [1]

    Learning a variational network for reconstruction of accelerated MRI data,

    K. Hammernik, et al., “Learning a variational network for reconstruction of accelerated MRI data,”Magn. Reson. Med., vol. 79, no. 6, pp. 3055– 3071, 2018

  2. [2]

    A deep cascade of convolutional neural networks for dynamic MR image reconstruction,

    J. Schlemper, J. Caballero, J. V . Hajnal, A. N. Price, and D. Rueckert, “A deep cascade of convolutional neural networks for dynamic MR image reconstruction,”IEEE Trans. Med. Imag., vol. 37, no. 2, pp. 491–503, 2018

  3. [3]

    MoDL: Model-based deep learning architecture for inverse problems,

    H. K. Aggarwal, M. P. Mani, and M. Jacob, “MoDL: Model-based deep learning architecture for inverse problems,”IEEE Trans. Med. Imag., vol. 38, no. 2, pp. 394–405, 2019

  4. [4]

    Deep-learning methods for parallel magnetic resonance imaging reconstruction: A survey of the current approaches, trends, and issues,

    F. Knoll, et al., “Deep-learning methods for parallel magnetic resonance imaging reconstruction: A survey of the current approaches, trends, and issues,”IEEE Signal Process. Mag., vol. 37, no. 1, pp. 128–140, 2020

  5. [5]

    Dense recurrent neural networks for accelerated MRI: History- cognizant unrolling of optimization algorithms,

    S. A. H. Hosseini, B. Yaman, S. Moeller, M. Hong, and M. Akc ¸akaya, “Dense recurrent neural networks for accelerated MRI: History- cognizant unrolling of optimization algorithms,”IEEE J. Sel. Topics Signal Process., vol. 14, no. 6, pp. 1280–1291, Oct. 2020

  6. [6]

    Physics-driven deep learning for computational magnetic resonance imaging: Combining physics and machine learning for improved medical imaging,

    K. Hammernik, et al., “Physics-driven deep learning for computational magnetic resonance imaging: Combining physics and machine learning for improved medical imaging,”IEEE Signal Process. Mag., vol. 40, no. 1, pp. 98–114, 2023

  7. [7]

    Advancing machine learning for MR image recon- struction with an open competition: Overview of the 2019 fastMRI challenge,

    F. Knoll, et al., “Advancing machine learning for MR image recon- struction with an open competition: Overview of the 2019 fastMRI challenge,”Magn. Reson. Med., vol. 84, no. 6, pp. 3054–3070, 2020

  8. [8]

    Results of the 2020 fastMRI challenge for machine learning MR image reconstruction,

    M. J. Muckley, et al., “Results of the 2020 fastMRI challenge for machine learning MR image reconstruction,”IEEE Trans. Med. Imag., vol. 40, no. 9, pp. 2306–2317, 2021

Show all 39 references
  1. [9]

    Akc ¸akaya, M

    M. Akc ¸akaya, M. Doneva, and C. Prieto (eds.),Magnetic resonance image reconstruction: theory, methods, and applications, vol. 7 ofAd- vances in Magnetic Resonance Technology and Applications, Academic Press, 2022

  2. [10]

    Self-supervised physics-guided deep learning recon- struction for high-resolution 3D LGE CMR,

    B. Yaman, et al., “Self-supervised physics-guided deep learning recon- struction for high-resolution 3D LGE CMR,” inProc. IEEE Int. Symp. Biomed. Imag., 2021, pp. 100–104

  3. [11]

    High-quality 0.5 mm isotropic fMRI: Random matrix theory meets physics-driven deep learning,

    ¨O. B. Demirel, et al., “High-quality 0.5 mm isotropic fMRI: Random matrix theory meets physics-driven deep learning,” inProc. 11th Int. IEEE/EMBS Conf. Neural Eng. (NER), 2023, pp. 1–6

  4. [12]

    Highly-accelerated high-resolution multi-echo fMRI using self-supervised physics-driven deep learning reconstruction,

    M. G ¨ulle, et al., “Highly-accelerated high-resolution multi-echo fMRI using self-supervised physics-driven deep learning reconstruction,” in Proc. IEEE 9th Int. Workshop CAMSAP, 2023, pp. 196–200

  5. [13]

    Non-cartesian self-supervised physics-driven deep learn- ing reconstruction for highly-accelerated multi-echo spiral fMRI,

    H. Gu, C. Zhang, Z. Yu, C. Rettenmeier, V . A. Stenger, and M. Akc ¸akaya, “Non-cartesian self-supervised physics-driven deep learn- ing reconstruction for highly-accelerated multi-echo spiral fMRI,” in Proc. IEEE Int. Symp. Biomed. Imag., 2024, pp. 1–5

  6. [14]

    20-fold accelerated 7T fMRI using referenceless self-supervised deep learning reconstruction,

    ¨O. B. Demirel, et al., “20-fold accelerated 7T fMRI using referenceless self-supervised deep learning reconstruction,” inProc. 43rd Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. (EMBC), 2021, pp. 3765–3769

  7. [15]

    Lowering the thermal noise barrier in functional brain mapping with magnetic resonance imaging,

    L. Vizioli, et al., “Lowering the thermal noise barrier in functional brain mapping with magnetic resonance imaging,”Nature Communications, vol. 12, no. 1, pp. 1–15, 2021

  8. [16]

    SENSE: Sensitivity encoding for fast MRI,

    K. P. Pruessmann, M. Weiger, M. B. Scheidegger, and P. Boesiger, “SENSE: Sensitivity encoding for fast MRI,”Magn. Reson. Med., vol. 42, no. 5, pp. 952–962, 1999

  9. [17]

    Automatic compilation of diverse CNNs onto high-performance FPGA accelerators,

    Y . Ma, Y . Cao, S. Vrudhula, and J.-s. Seo, “Automatic compilation of diverse CNNs onto high-performance FPGA accelerators,”IEEE Trans. Comput.-Aided Design Integr. Circuits Syst., vol. 39, no. 2, pp. 424–437, Feb. 2018

  10. [18]

    ALAMO: FPGA acceleration of deep learning algorithms with a modularized RTL compiler,

    Y . Ma, N. Suda, Y . Cao, S. Vrudhula, and J.-s. Seo, “ALAMO: FPGA acceleration of deep learning algorithms with a modularized RTL compiler,”Integration, vol. 62, pp. 14–23, 2018

  11. [19]

    Caffeine: Toward uniformed representation and acceleration for deep convolutional neural networks,

    C. Zhang, G. Sun, Z. Fang, P. Zhou, P. Pan, and J. Cong, “Caffeine: Toward uniformed representation and acceleration for deep convolutional neural networks,”IEEE Trans. Comput.-Aided Design Integr. Circuits Syst., vol. 38, no. 11, pp. 2072–2085, Nov. 2019

  12. [20]

    Algorithm-hardware co-optimization for energy-efficient drone detec- tion on resource-constrained FPGA,

    H.-s. Suh, J. Meng, T. Nguyen, V . Kumar, Y . Cao, and J.-S. Seo, “Algorithm-hardware co-optimization for energy-efficient drone detec- tion on resource-constrained FPGA,”ACM Trans. Reconfig. Technol. Syst., vol. 16, no. 2, pp. 1–25, May. 2023

  13. [21]

    FPGA acceleration of GCN in light of the symmetry of graph adjacency matrix,

    G. R. Nair, H.-S. Suh, M. Halappanavar, F. Liu, J.-s. Seo, and Y . Cao, “FPGA acceleration of GCN in light of the symmetry of graph adjacency matrix,” inIn Proc. DATE, 2023, pp. 1–6

  14. [22]

    Array compression for mri with large coil arrays,

    M. Buehrer, K. P. Pruessmann, P. Boesiger, and S. Kozerke, “Array compression for mri with large coil arrays,”Magn. Reson. Med., vol. 57, no. 6, pp. 1131–1139, 2007

  15. [23]

    The WU-Minn Human Connectome Project: An overview,

    D. C. Van Essen, et al., “The WU-Minn Human Connectome Project: An overview,”Neuroimage, vol. 80, pp. 62–79, 2013

  16. [24]

    Optimization methods for magnetic resonance image reconstruction,

    J. A. Fessler, “Optimization methods for magnetic resonance image reconstruction,”IEEE Signal Process. Mag., vol. 37, no. 1, pp. 33–40, 2020

  17. [25]

    Self-supervised learning of physics-guided recon- struction neural networks without fully sampled reference data,

    B. Yaman, S. A. H. Hosseini, S. Moeller, J. Ellermann, K. U ˘gurbil, and M. Akc ¸akaya, “Self-supervised learning of physics-guided recon- struction neural networks without fully sampled reference data,”Magn. Reson. Med., vol. 84, no. 6, pp. 3172–3191, Dec. 2020

  18. [26]

    Multi-mask self-supervised learning for physics- guided neural networks in highly accelerated magnetic resonance imag- ing,

    B. Yaman, et al., “Multi-mask self-supervised learning for physics- guided neural networks in highly accelerated magnetic resonance imag- ing,”NMR Biomed, vol. 35, no. 12, 2022, Art. no. e4798

  19. [27]

    Zero-shot self- supervised learning for MRI reconstruction,

    B. Yaman, S. A. H. Hosseini, and M. Akcakaya, “Zero-shot self- supervised learning for MRI reconstruction,” inProc. 10th Int. Conf. Learn. Represent., 2022

  20. [28]

    Unsupervised deep learning methods for biological image reconstruction and enhancement: An overview from a signal processing perspective,

    M. Akc ¸akaya, B. Yaman, H. Chung, and J. C. Ye, “Unsupervised deep learning methods for biological image reconstruction and enhancement: An overview from a signal processing perspective,”IEEE Signal Process. Mag., vol. 39, no. 2, pp. 28–44, 2022

  21. [29]

    Cycle-consistent self- supervised learning for improved highly-accelerated MRI reconstruc- tion,

    C. Zhang, O. B. Demirel, and M. Akc ¸akaya, “Cycle-consistent self- supervised learning for improved highly-accelerated MRI reconstruc- tion,” inIEEE Int. Symp. Biomed. Imag., 2024, pp. 1–5

  22. [30]

    A convex compressibility- inspired unsupervised loss function for physics-driven deep learning reconstruction,

    Y . U. Alc ¸alar, M. G¨ulle, and M. Akc ¸akaya, “A convex compressibility- inspired unsupervised loss function for physics-driven deep learning reconstruction,” inProc. IEEE Int. Symp. Biomed. Imag., 2024, pp. 1–5

  23. [31]

    Sparsity-driven parallel imaging consistency for improved self-supervised MRI reconstruction,

    Y . U. Alc ¸alar and M. Akc ¸akaya, “Sparsity-driven parallel imaging consistency for improved self-supervised MRI reconstruction,” inIEEE Int. Conf. Image Process., 2025

  24. [32]

    Revisitingℓ 1-wavelet compressed-sensing MRI in the era of deep learning,

    H. Gu, B. Yaman, S. Moeller, J. Ellermann, K. U ˘gurbil, and M. Akc ¸akaya, “Revisitingℓ 1-wavelet compressed-sensing MRI in the era of deep learning,”Proc. Natl. Acad. Sci., vol. 119, no. 33, 2022, Art. no. e2201062119

  25. [33]

    Signal intensity informed multi-coil encoding operator for physics-guided deep learning reconstruction of highly accelerated myocardial perfusion CMR,

    O. B. Demirel, B. Yaman, C. Shenoy, S. Moeller, S. Weing ¨artner, and M. Akc ¸akaya, “Signal intensity informed multi-coil encoding operator for physics-guided deep learning reconstruction of highly accelerated myocardial perfusion CMR,”Magn. Reson. Med., vol. 89, no. 1, pp. 3...

  26. [34]

    NTIRE 2017 challenge on single image super-resolution: Methods and results,

    R. Timofte, E. Agustsson, L. V . Gool, M.-H. Yang, and L. Zhang, “NTIRE 2017 challenge on single image super-resolution: Methods and results,” inProc. IEEE Conf. Comput. Vis. Pattern Recog. Worksh., 2017, pp. 114–125

  27. [35]

    Smoothquant: Accurate and efficient post-training quantization for large language models,

    G. Xiao, J. Lin, M. Seznec, H. Wu, J. Demouth, and S. Han, “Smoothquant: Accurate and efficient post-training quantization for large language models,” inProc. Int. Conf. Mach. Learn., 2023, pp. 38087– 38099

  28. [36]

    Quantization and training of neural networks for efficient integer-arithmetic-only inference,

    B. Jacob, et al., “Quantization and training of neural networks for efficient integer-arithmetic-only inference,” inProc. IEEE Conf. Comput. Vis. Pattern Recog., 2018, pp. 2704–2713

  29. [37]

    fastMRI: An open dataset and benchmarks for accelerated MRI,

    J. Zbontar, et al., “fastMRI: An open dataset and benchmarks for accelerated MRI,” 2019, arXiv:1811.08839

  30. [38]

    fastMRI: a publicly available raw k-space and DICOM dataset of knee images for accelerated MR image reconstruction using machine learning,

    F. Knoll, et al., “fastMRI: a publicly available raw k-space and DICOM dataset of knee images for accelerated MR image reconstruction using machine learning,”Radiol., Artif. Intell, vol. 2, no. 1, Jan. 2020, Art. no. e190007

  31. [39]

    Advances in sensitivity encoding with arbitrary k-space trajectories,

    K. P. Pruessmann, M. Weiger, P. B ¨ornert, and P. Boesiger, “Advances in sensitivity encoding with arbitrary k-space trajectories,”Magn. Reson. Med., vol. 46, no. 4, pp. 638–651, 2001

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.