{"id":"907c04a6-7039-4454-935a-8c0da896f8d5","arxiv_id":"2608.00075","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"KGR fits shared Gabor-Gaussian primitives directly in multi-coil k-space and applies low-rank projection, consistently outperforming representative parallel-MRI baselines on brain datasets.","lead":"An MRI reconstruction method places Gaussian blobs directly in raw frequency (k-space) data, instead of in the final image, to fill in missing measurements for accelerated scans. It combines these frequency-domain primitives with a low-rank constraint and beats several established baselines on brain MRI benchmarks.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 1 assumes K* ∈ C_H, but Eq. 1's noise makes C_H's data-consistency condition incompatible with K*, so the error bound is unsupported.","rationale":"The reader's weakest assumption focused on the low-rank part of C_H and the lack of validation of that model. My concern targets the data-consistency part of C_H, showing that even if the low-rank assumption were valid for the noiseless signal, the noisy measurements in Eq. 1 prevent K* from belonging to C_H. Thus the two concerns are complementary: both center on the unrealistic idealization of C_H in Proposition 1, but the reader's version is about the rank condition, while mine is about the exact data-consistency condition. This is the most load-bearing issue because the paper explicitly lists the 'error inheritance' analysis as a contribution, and a false theorem undermines the theoretical grounding of the method. The empirical results may still be valid, but the claimed guarantee is not. I checked the paper for internal inconsistencies and found this one; I did not find evidence of fabrication or misconduct. The conditional verdict remains appropriate: the authors should either reframe Proposition 1 for noiseless measurements or derive a noise-aware bound that includes ε_Ω. Because the empirical comparison is plausible and the method is essentially a new representation, a rejection would be too harsh; a revision addressing Proposition 1 and adding noise analysis is sufficient. Hence the verdict is unchanged from the reader's CONDITIONAL.","tokens_in":15517,"tokens_out":8625,"duration_ms":99103,"concrete_test":"Re-derive Proposition 1 under Eq. 1's noise model. Set K* = F{sρ} and \\hat{K} = M_Ω⊙K* + ε with ε_Ω ≠ 0. Show that any K ∈ C_H satisfies M_Ω⊙(K - K*) = ε_Ω, so part (i) is false and the bound in part (ii) must include an additional ||ε_Ω|| term. Numerically, on one fastMRI slice, generate noisy undersampled data per Eq. 1 with known ε, run Algorithm 1, and compute ||M_Ω⊙(K_final - K*)||. If this is nonzero and approximately ||ε_Ω||, the proposition's premise is violated, confirming the theoretical analysis is unsupported in the noisy setting.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's theoretical analysis rests on Proposition 1, which assumes the target k-space K* belongs to C_H = {K : M_Ω⊙K = \\hat{K}, rank(H(K)) ≤ q} (Eq. 13). However, the measurement model in Eq. 1 is \\hat{K}_c = M_Ω ⊙ F{s_c ρ} + ε_c. Thus the observed data at acquired locations are the true k-space samples plus noise ε_Ω. A matrix K in C_H must satisfy M_Ω⊙K = \\hat{K}, meaning it reproduces the noisy samples exactly. The true k-space K* = F{s_c ρ} satisfies M_Ω⊙K* = \\hat{K} - ε_Ω, not M_Ω⊙K* = \\hat{K}. Therefore K* ∉ C_H whenever ε_Ω ≠ 0. This invalidates both conclusions of Proposition 1: part (i) claims M_Ω⊙(\\bar{K}-K*) = 0, but in reality M_Ω⊙(\\bar{K}-K*) = ε_Ω, and part (ii)'s bound ||\\bar{K}-K*|| ≤ 2ε_G fails because the distance term dist_F(Kaug, C_H) is no longer bounded by ε_G. The final projection in Eq. (11) explicitly enforces M_Ω⊙K_final = \\hat{K}, so the reconstruction's error on acquired samples is exactly the measurement noise. The paper presents this analysis as a key contribution ('We provide an analysis of the compatibility...'), but the proof silently treats measurements as noiseless. This is an internal inconsistency, not a disagreement with consensus: the theorem's premise is false under the paper's own forward model.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes K-space Gaussian Representation (KGR), a scan-specific, network-free method that models multi-coil k-space as a continuous signal using Gabor-Gaussian primitives with geometry shared across coils. The pipeline comprises a Hankel low-rank initialization, frequency-adaptive fitting of band-specific primitives, a reliability-gated augmentation, and a final Hankel low-rank compatibility projection. Experiments on fastMRI and CC359 under Poisson and radial undersampling at R=6 and R=10 report that KGR achieves the best PSNR, SSIM, and NMSE among all compared methods, with gains up to about 0.9 dB over SAKE and about 2.4 dB over an image-domain Gaussian baseline. The authors also provide a theoretical Proposition 1 intended to characterize how Gaussian approximation error and low-rank inconsistency affect the final reconstruction error.","tokens_in":15938,"tokens_out":6773,"duration_ms":74305,"significance":"If the empirical results hold, KGR is a useful contribution to calibrationless, network-free parallel MRI reconstruction: it connects explicit continuous k-space modeling with established structured low-rank priors, and it ships with code, paired t-tests, and an ablation study. The frequency-adaptive, cross-coil shared-geometry formulation is a reasonable and somewhat novel design. However, the theoretical analysis is a stated contribution and is currently not sound under the paper's own noisy forward model; additionally, the empirical comparison would be stronger with sensitivity analysis for per-dataset hyperparameters and with measures of variability for the reported means. The core reconstruction idea is plausible, but the paper needs revision before the claims can be accepted at face value.","major_comments":[{"comment":"The proposition assumes K* ∈ C_H, where C_H = {K : M_Ω⊙K = \\hat K, rank(H(K)) ≤ q}. Under the paper's own forward model Eq. (1), \\hat K = M_Ω⊙K* + ε_Ω, so M_Ω⊙K* = \\hat K − ε_Ω; hence K* ∉ C_H whenever ε_Ω ≠ 0. This invalidates both conclusions: part (i) is false on Ω, and part (ii)'s bound dist_F(Kaug,C_H) ≤ ε_G relies on K*∈C_H and is unsupported. The final projection in Eq. (11) enforces M_Ω⊙K_final = \\hat K, so the reconstructed error on acquired samples is exactly the measurement noise. Please state Proposition 1 under a noiseless model and analyze the noisy case explicitly, or modify the compatibility set to a noise-aware constraint and re-derive the bound.","section":"Section III-D, Eq. (13), Proposition 1"},{"comment":"Hyperparameters are tuned separately for each dataset: fastMRI uses band counts 50/600/2000/10000, reliability gate δ_max=0.40, 10×10 final-projection window and threshold 2.0, while CC359 uses 10/80/300/1100, δ_max=0.70, 8×8, and threshold 1.8. No sensitivity analysis is provided for δ_max, band counts, λ_Ω, low-rank thresholds, or window sizes. Since the reliability gate in Eq. (10) and the final projection in Eq. (12) are central to the claimed gains, the 'consistent improvements' claim is confounded with per-dataset tuning. A sensitivity study, or at least a single fixed setting across both datasets, is needed to support the general claim.","section":"Section IV-A, implementation details"},{"comment":"Proposition 1 does not analyze the Gaussian fitting error ε_G; it merely assumes ∥Kaug−K*∥ ≤ ε_G and then applies the triangle inequality. No bound on ε_G is derived from the KGR fitting procedure, so the theorem amounts to a conditional error-inheritance statement rather than a characterization of when the Gaussian stage benefits reconstruction. The contributions section says 'the derived error analysis characterizes how representation accuracy and low-rank consistency affect reconstruction fidelity,' which overstates what Proposition 1 establishes. Please either derive a nontrivial bound on ε_G or rephrase the theoretical contribution to match the conditional nature of the result.","section":"Section III-D, Proposition 1 (proof and interpretation)"}],"minor_comments":[{"comment":"The sentence 'The factor of two comes from the nonconvex rank constraint: Best approximation onto C_H is not firmly nonexpansive' is inaccurate: the factor 2 in Eq. (17) follows directly from the triangle inequality together with dist_F(Kaug,C_H) ≤ ε_G, and does not depend on nonconvexity. Please correct or remove this explanation.","section":"Section III-D, explanation after Proposition 1"},{"comment":"The tables report only mean PSNR/SSIM/NMSE. Paired t-test p-values show that the means differ, but no standard deviations or confidence intervals are given. Since the central claim is that KGR is best 'across all eight settings,' reporting effect-size uncertainty would strengthen the comparison.","section":"Section IV-B, Tables I and II"},{"comment":"The p-value column is described as a 'two-sided paired t-test against KGR on PSNR averaged over the selected slices and the four sampling settings.' Please clarify whether the test is performed on per-case PSNR values or on setting-level means, and how the four masks are pooled.","section":"Section IV-A, p-value description"},{"comment":"In Eq. (10), δ_r is defined as a band-level scalar in Eq. (9), but the condition is written as δ_r(k) ≤ δ_max with r(k) the band containing k. Clarify that the gate applies to all unacquired points in a band based on that band's scalar δ_r.","section":"Section III-C, Eq. (10)"},{"comment":"The claim of being 'the first explicit continuous representation formulated directly in the native k-space domain' should be carefully qualified relative to neural implicit k-space methods cited as [22], [23]. The authors distinguish explicit from implicit representations, but a reader may see the abstract claim as overly broad; a sentence clarifying the distinction would help.","section":"Abstract and Section I"}],"recommendation":"major_revision","confidential_remarks":"The empirical contribution appears potentially solid, but the theoretical Proposition 1 is currently inconsistent with the paper's noisy forward model and is largely conditional rather than a substantive guarantee. The per-dataset hyperparameter tuning and absence of sensitivity analysis further weaken the generalizability claim. I would support reconsideration after the authors revise the theory to address the noise issue and provide sensitivity or fixed-hyperparameter evidence. The novelty claim regarding 'first explicit continuous k-space representation' should be checked carefully against the cited neural implicit k-space works during editorial review."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take. The core idea is fresh: instead of modeling k-space as discrete pixels or implicit network weights, KGR fits explicit Gabor-Gaussian primitives directly in native multi-coil k-space, with shared geometry across coils and band-wise frequency normalization. That is a real representation gap relative to image-space Gaussians and neural implicit methods. The reliability gate and low-rank projection are sensible. Empirically, the gains over SAKE (up to ~0.9 dB) and Image-GR (~2.4 dB) are consistent across eight dataset/mask settings, and the ablation shows that the frequency-adaptive decomposition, low-rank projection, and directional Gabor all contribute. Public code is a plus.\n\nThe main soft spot is Proposition 1. The theorem assumes K* belongs to C_H, where C_H enforces M_Ω ⊙ K = \\hat{K} exactly. But Eq. (1) defines \\hat{K} as true k-space plus noise ε. So the true K* does not satisfy M_Ω ⊙ K* = \\hat{K} whenever ε ≠ 0; the premise is false under the paper's own forward model. Consequently part (i)'s claim that error vanishes on acquired samples is wrong (it equals the noise), and part (ii)'s bound ||\\bar{K} - K*|| ≤ 2ε_G unravels because the distance from K_aug to C_H is not bounded by ε_G. The proof is fine under a noiseless assumption, but that assumption is never stated and contradicts Eq. (1). This is a real flaw, though not fatal to the empirical method—the reconstruction still works because data consistency is enforced against the measurements, which is standard practice. But the theory as written is unsupported.\n\nOther soft spots are reporting quality: no error bars or confidence intervals, hyperparameters (δ_max, band counts, low-rank thresholds) tuned per dataset without sensitivity analysis, and the novelty claim vs ref [16] (Gabor primitives in cardiac MRI) needs a crisper separation. These are fixable.\n\nBottom line: the paper is a solid empirical contribution with a novel representation, and it deserves a serious referee. But it needs major revision on the theory and the reporting before I'd trust it. Send it to review, and ask the authors to fix Proposition 1, add variance reporting and sensitivity analysis, and sharpen the comparison with [16].","headline":"KGR is a genuinely new k-space continuous representation with consistent empirical gains, but Proposition 1's noiseless premise contradicts Eq. (1) and needs fixing before the theory is credible.","tokens_in":16416,"tokens_out":4079,"would_cite":true,"duration_ms":42556,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"KGR reconstructs accelerated parallel MRI by modeling k-space as an explicit continuous Gabor-Gaussian signal rather than discrete samples, and pairs it with Hankel low-rank projection.","keywords":["parallel MRI","k-space reconstruction","Gaussian representation","structured low-rank","Gabor primitives","calibrationless reconstruction","frequency-adaptive fitting","continuous signal representation"],"falsifier":"Compute the block-Hankel singular value spectrum of fully sampled k-space slices from the same datasets under 6x6 and 10x10 windows. If the energy captured by the top-q singular values, at the truncation thresholds used, is much less than the reported PSNR gains imply, then the true k-space is not in the low-rank Hankel set and Proposition 1's bound is vacuous. A second check: run KGR under uniform Cartesian R=8; the paper concedes robustness there is limited, so a strong performance collapse would mark the method's coverage boundary.","tokens_in":15421,"feed_emoji":"🧠","tokens_out":5975,"duration_ms":64278,"temperature":0.7,"pith_summary":"This paper proposes K-space Gaussian Representation (KGR), a way to reconstruct accelerated parallel MRI that treats the measured multi-coil k-space as a continuous signal built from Gabor-Gaussian primitives, not as discrete Fourier samples to be filled in. It claims that placing an explicit continuous representation directly in native k-space is new, and that pairing it with Hankel structured low-rank projection makes a scan-specific, calibration-free, network-free reconstruction pipeline. On eight dataset-by-mask combinations from two brain MRI benchmarks, it reports the best PSNR, SSIM, and NMSE among compared methods, with gains up to about 0.9 dB over SAKE and about 2.4 dB over an image-domain Gaussian baseline. The paper also proves that if the true k-space lies in a low-rank Hankel set and the Gaussian fit is within epsilon, then the final projection error is at most 2 epsilon, with all error confined to unacquired samples. A sympathetic reader would care because it opens a representation axis—explicit continuous k-space modeling—that existing interpolation, low-rank, and learned approaches do not occupy.","feed_headline":"Continuous k-space Gaussians beat discrete-grid MRI reconstruction","feed_subtitle":"Network-free and scan-specific, the method leads all eight accelerated MRI settings tested, up to 2.4 dB over an image-domain rival.","key_machinery":"The central object is the k-space Gabor-Gaussian primitive phi_j(k) = exp(-1/2 d_j^T Sigma_j^{-1} d_j) exp(i xi_j^T d_j), with d_j = k - nu_j, whose center nu_j, covariance Sigma_j, and carrier xi_j are shared across coils while complex amplitudes remain coil-specific. It does the work of representing local frequency variation continuously, including an anisotropic radial-tangential form in the high-frequency band. The second load-bearing mechanism is the block-Hankel low-rank projection Pi_H, which enforces the algebraic redundancy of multi-coil k-space by lifting local neighborhoods into matrices of low rank; KGR alternates between fitting the Gaussians with a frequency-adaptive, band-RMS-","core_discovery":"On its own terms, the central claim is that k-space itself is a continuous complex signal and can be parameterized directly by shared-geometry Gabor-Gaussian primitives: for each coil c, K_theta,c(k) is a sum over radial frequency bands of coil-specific complex amplitudes times a common Gaussian envelope modulated by a sinusoidal carrier. All coils share centers, covariances, and carriers, while each coil keeps its own complex amplitudes, so inter-coil correlations are preserved without estimating sensitivity maps. The fitted continuous prediction is not trusted everywhere: a per-band relative-MSE gate accepts it only where it agrees with acquired data, and the accepted k-space is returned t","pith_inferences":["A testable extension is to feed KGR's continuous predictions into calibration-based interpolation under uniform Cartesian undersampling, since the paper concedes that setting remains a weakness and its reliability gate would likely reject predictions in large uniform gaps.","The reliability threshold and per-band primitive counts are hand-set; a scan-specific search over them could shift the reported margins, and the ablation suggests the final result is sensitive to at least the band decomposition.","The shared-geometry assumption is untested for coils with strongly different sensitivity profiles; a direct check is to compare shared-geometry against per-coil geometry on a dataset with a heterogeneous coil array.","Because the continuous representation can be evaluated off-grid, it could extend to non-Cartesian trajectories without separate gridding, but only if the low-rank operator is replaced by a trajectory-aware version—the authors flag this as future work."],"forward_implications":["KGR can be used as a drop-in continuous representation module: it augments structured low-rank reconstruction rather than replacing it, so existing Hankel-based solvers can adopt it without retraining.","Because the pipeline is scan-specific and needs no coil sensitivity maps or pretrained networks, it is applicable where training data or calibration signal are scarce.","The ablation indicates that frequency-adaptive band decomposition is the largest single contributor, so handling k-space's dynamic range is a key mechanism, not just the primitive form.","If the reported ranking holds, an explicit continuous parameterization is a viable alternative to learned k-space interpolation and self-supervised networks for accelerated MRI reconstruction.","The error analysis provides a design criterion: improve either the Gaussian approximation error or the low-rank compatibility, and the final reconstruction bound improves."],"supporting_citations":[{"why":"Defines the calibrationless structured low-rank matrix completion baseline (SAKE) and the block-Hankel prior KGR reuses for initialization and final projection.","marker":"[3]"},{"why":"Supplies LORAKS, the local k-space neighborhood low-rank model that the generic Hankel projection Pi_H instantiates.","marker":"[9]"},{"why":"Introduces explicit Gaussian primitives as a continuous signal representation; KGR adapts this primitive style to k-space.","marker":"[12]"},{"why":"Provides the Gabor-primitive formulation for high-frequency MRI that motivates the Gabor carrier in Eq. (5).","marker":"[16]"},{"why":"Supplies the compressed-sensing TV baseline (CS-TV) that KGR compares against.","marker":"[17]"},{"why":"Supplies P-LORAKS, the parallel-imaging low-rank baseline that enforces coil constraints in the same local-neighborhood family.","marker":"[10]"},{"why":"Supplies ZS-SSL, the scan-specific self-supervised deep unrolled baseline; KGR claims to beat it without sensitivity maps or pretraining.","marker":"[5]"},{"why":"Provides the fastMRI brain k-space benchmark data used for the main quantitative comparison.","marker":"[34]"},{"why":"Provides the CC359 brain dataset used for the second benchmark and the ablation study.","marker":"[37]"}],"fun_headline_variants":["K-space Gaussians model continuous MRI signal directly","Shared-geometry Gaussian primitives improve parallel MRI","No networks needed: k-space Gaussians beat grid-based MRI","Continuous k-space parameterization surpasses discrete MRI","Gabor-Gaussian k-space model enhances accelerated MRI"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"Everything in the pipeline leans on the true multi-coil k-space being well approximated by a low-rank block-Hankel matrix; if that rank model is wrong for the data, the initialization and final projection both push the reconstruction toward the wrong structure and the paper's 2-epsilon bound has no premise.","fun_headline_variants_meta":{"raw":{"variants":["K-space Gaussians model continuous MRI signal directly","Shared-geometry Gaussian primitives improve parallel MRI","No networks needed: k-space Gaussians beat grid-based MRI","Continuous k-space parameterization surpasses discrete MRI","Gabor-Gaussian k-space model enhances accelerated MRI"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000285,"raw_usage":{"total_tokens":1524,"prompt_tokens":760,"completion_tokens":764,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":689}},"tokens_in":504,"tokens_out":764,"duration_ms":8319,"temperature":1.0,"reasoning_tokens":689,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T04:21:02.537875+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the block-Hankel singular value spectrum of fully sampled k-space slices from the same datasets under 6x6 and 10x10 windows. If the energy captured by the top-q singular values, at the truncation thresholds used, is much less than the reported PSNR gains imply, then the true k-space is not in the low-rank Hankel set and Proposition 1's bound is vacuous. A second check: run KGR under uniform Cartesian R=8; the paper concedes robustness there is limited, so a strong performance collapse would mark the method's coverage boundary.","supporting_citations":[{"cited_title":"Low-rank modeling of local k-space neighborhoods (LO- RAKS) for constrained MRI,","cited_arxiv_id":null,"evidence_quote":"Supplies LORAKS, the local k-space neighborhood low-rank model that the generic Hankel projection Pi_H instantiates."},{"cited_title":"3D Gaussian splatting for real-time radiance field rendering,","cited_arxiv_id":null,"evidence_quote":"Introduces explicit Gaussian primitives as a continuous signal representation; KGR adapts this primitive style to k-space."}],"review_version":2}