{"id":"bd99b205-66c8-42d3-b970-5caf58de9a42","arxiv_id":"2509.07128","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"HR-XPD combines radiality weighting and Doppler coherence to improve resolution and contrast in contrast-free ultrasound microvascular imaging, validated in simulations and in vivo.","lead":"This paper introduces a new ultrasound processing method, HR-XPD, that sharpens microvascular images without contrast agents by combining radiality weighting with cross-correlation between two angle subgroups. It reports 2 to 3 times sharper vessel images and up to 20 dB better background suppression in simulations and in vivo liver and kidney scans.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed 2–3x resolution gain is not established: radiality/similarity weighting can narrow -3 dB vessel profiles by contrast enhancement alone; an isolated point-target/wire-phantom test is needed.","rationale":"Reading in good faith: the pipeline is well described, the combination of radiality and coherence weighting is plausible for suppressing incoherent noise, and the simulation shows robust background suppression across SNR. The qualitative in vivo images are visibly cleaner. But the abstract and conclusion rest on 'up to 2-3 fold enhancement in spatial resolution,' and that number is load-bearing. The metrics used to support it are FWHM and PVL at fixed dB thresholds; both can improve with contrast enhancement even when the PSF is unchanged. The internal discrepancy between about 1.2-1.4x simulation gains and 2.46-3.65x in vivo gains is consistent with this confound, because in vivo vessel profiles are not point-like and background/sidelobe suppression is much stronger relative to conventional PD. This is an unvalidated inference from image profiles to spatial resolution, not an internal inconsistency. The proposed phantom test would settle it. Since the qualitative contrast benefit is still credible and the needed validation is feasible, I would keep the paper's CONDITIONAL status and make the phantom calibration an explicit acceptance condition. The reader identified the same core weakness, so my assessment agrees.","tokens_in":14980,"tokens_out":6162,"duration_ms":60750,"concrete_test":"Use a calibrated wire/point phantom (20-50 µm nylon wires or equivalent) imaged with the same transducers, steering angles, and local-SVD/radiality/similarity pipeline. Measure PSF FWHM of isolated wires and two-point PVL for wire pairs at separations of roughly 0.5λ, 1λ, and 2λ, using identical fixed dynamic-range settings and no per-profile renormalization. Include a contrast-matched control: apply a simple threshold or sharpening to conventional PD that reproduces HR-XPD's background level. If HR-XPD does not reduce isolated-point FWHM or improve two-point PVL beyond this control, the 2-3x resolution claim should be restated as contrast enhancement, not spatial-resolution improvement.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim, a 2- to 3-fold spatial-resolution improvement over PD, is measured only through -3 dB FWHM and PVL on manually selected vessel cross-sections (Quantitative Evaluation; Figs. 7, 9, 11). Both metrics are contrast-dependent. HR-XPD multiplies the cross-correlated IQ product by a radiality weight derived from local gradient alignment and by a normalized similarity map (Eqs. 1-4). At a bright vessel center the combined weight is near one; in the surrounding sidelobes and background it is lower. Consequently, the -3 dB contour of any bright structure shrinks even if the underlying diffraction-limited PSF is unchanged, and a valley between two nearby vessels can rise above the 3 dB PVL threshold purely because the noise floor is suppressed. This matters because the simulation-based FWHM gain is only about 1.2-1.4-fold (Fig. 2e), whereas the headline 2-3x figures come from in vivo vessel segments whose true cross-sections are unknown; a 1 mm simulated tube is also not a point target, so its FWHM is geometry-dependent. No point-target or wire-phantom calibration is reported, so the resolution claim is not separated from nonlinear contrast enhancement.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes HR-XPD, a contrast-free microvascular ultrasound imaging method that combines radiality weighting (derived from local gradient convergence) with cross-correlation between two interleaved angular Doppler subsets and a similarity map. The claimed contributions are improved spatial resolution, lower background intensity, and enhanced vascular contrast compared with conventional power Doppler (PD), SW-XPD, and HR-PD. Validation is provided through Field II simulations of two crossing 1-mm tubes under varying noise levels and in vivo imaging of a transplanted human kidney, a healthy human liver, and a pig kidney. The central quantitative claim is a 2- to 3-fold spatial-resolution improvement over PD (up to 3.65-fold in the liver) and contrast increases of up to 20 dB, achieved with 0.3-1.2 s of acquisition without contrast agents.","tokens_in":15196,"tokens_out":4338,"duration_ms":39268,"significance":"If the resolution improvement is real, HR-XPD would be a practically valuable addition to microvascular ultrasound because it is contrast-free, uses short acquisition times, and is evaluated across multiple in vivo settings. The method is a transparent nonlinear processing recipe: the radiality annulus radius, sampling-point count, and SVD block/rank settings are explicit, and no parameters are fitted to the reported outcomes, which is a strength. The paper also provides useful comparisons against three baselines (PD, SW-XPD, HR-PD) and quantifies robustness to noise. However, the headline resolution claim rests on metrics that are inherently contrast-dependent, and no point-target or wire-phantom experiment is reported to separate genuine spatial-resolution gain from nonlinear background suppression. The simulation results themselves show only a modest FWHM gain (about 1.2-1.4x), which is inconsistent with the much larger in vivo gains and underscores the need for a direct resolution test.","major_comments":[{"comment":"The central 2-3x spatial-resolution claim is measured exclusively through the -3 dB FWHM of manually selected vessel cross-sections (Eq. 7 definition; Figs. 7, 9, 11). This metric is contrast-dependent: HR-XPD strongly suppresses background and sidelobe intensity, so the -3 dB contour of any bright structure narrows even if the underlying point spread function (PSF) is unchanged. The simulation FWHM gains are only about 1.2-1.4x (Fig. 2e, Fig. 4c), while the in vivo gains are 2.46x, 3.65x, and 1.5-2.8x in Figs. 7b, 9b, and 11e, respectively. This inconsistency suggests that most of the reported in vivo gain is a contrast-enhancement artifact rather than a true spatial-resolution improvement. A point-target or wire-phantom experiment, or an isolated sub-resolution scatterer simulation, is needed to measure the PSF width directly and separate resolution gain from contrast enhancement.","section":"Quantitative Evaluation; Figs. 7, 9, 11"},{"comment":"The peak-to-valley level (PVL) metric is used to claim that HR-XPD resolves adjacent vessels that PD and SW-XPD fail to resolve (PVL < 3 dB). PVL compares the valley intensity to the peak intensity, but a valley can rise above the 3 dB threshold purely because the noise floor between two unresolved PSF peaks has been suppressed by the similarity weighting. PVL therefore does not constitute a two-point resolution test unless the center-to-center separation of a phantom target is known. A resolution phantom with echo-free point scatterers or wires at known separations is required to support the resolvability claim.","section":"Eq. (9); Figs. 7c, 8d, 11c"},{"comment":"The simulation geometry consists of two crossing tubes with a diameter of 1 mm, not point targets. The FWHM of a 1-mm tube profile is determined by the convolution of the PSF with the tube's finite spatial extent and the flow velocity profile, so the measured FWHM and resolution-improvement ratio in Fig. 2(e) do not directly quantify PSF narrowing. A simulation with a point scatterer or a sub-resolution wire, or a deconvolution-based resolution estimator, would provide a cleaner measure of the method's true resolution capability.","section":"Simulation (Materials and Methods)"},{"comment":"The radiality G_q in Eq. (1) is the average of normalized dot products of gradient vectors with radial vectors and can take negative values for outward-pointing gradients. The text states that the radiality maps R1 and R2 are 'log-compressed and normalized' before being applied as weights, but no mapping is specified for negative values. Since radiality weighting is the core mechanism for resolution enhancement, this ambiguity makes the method irreproducible as written and should be clarified with the exact transformation (e.g., absolute value, positive-part, or offset before log compression).","section":"Eq. (1); Materials and Methods, 'Principle of HR-XPD'"}],"minor_comments":[{"comment":"The abstract claims 'up to a 2 to 3-fold enhancement' in spatial resolution, but the liver results in Fig. 9(b) report a 3.65-fold mean improvement over PD; the claim should be updated to reflect the full range or the abstract should be qualified.","section":"Abstract"},{"comment":"There is a mismatch between the caption of Fig. 2 and the main text: the caption labels (d) as 'Estimated resolution improvement' and (e) as 'PVL', while the main text refers to PVL in Fig. 2(d) and resolution improvement in Fig. 2(e). The labeling should be corrected.","section":"Fig. 2 caption and main text"},{"comment":"In the sentence after Eq. (2), 'Details in [5]' is too vague for a described mechanism; a brief derivation or intuitive explanation of why the cross-terms involving noise are suppressed would make the method more self-contained.","section":"Materials and Methods, 'Principle of HR-XPD'"},{"comment":"The text says that FWHM, CR, and resolution improvement ratios were compared using 'one-way analysis of variance (RM one-way ANOVA)'; the abbreviation RM should be defined as repeated-measures, and the specific post-hoc correction should be stated consistently.","section":"Statistics"},{"comment":"The text and figure captions repeatedly use 'randomly selected' to describe vessel cross-sections (e.g., Fig. 7c, Fig. 8d, Fig. 11c), but the selection is manual and therefore subjective; the term 'randomly' should be replaced with 'manually selected' to avoid implying an objective sampling procedure.","section":"Fig. 6, Fig. 9"}],"recommendation":"major_revision","confidential_remarks":"The core idea of combining radiality and similarity weighting is reasonable and the empirical comparisons are extensive, but the paper's most prominent claim—a 2-3x spatial-resolution improvement—is not adequately supported by the metrics used. The absence of any point-target or wire-phantom measurement is a significant gap for a method whose stated purpose is resolution enhancement. I would ask for either such a phantom experiment or a careful reanalysis that decomposes the resolution gain from the contrast gain, perhaps by measuring PSF width on isolated vessel segments with approximately flat intensity profiles. The work is otherwise technically sound and potentially publishable after this central issue is addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper combines two known ingredients — radiality weighting from CEUS super-resolution and similarity/cross-correlation weighting from the authors' own Doppler work — into a contrast-free pipeline, and the in vivo images are visibly cleaner. That combination is genuinely new as far as I know, and the simulation study across SNR levels is a solid piece of evidence that the method is noise-robust.\n\nThat said, the headline claim of a 2–3x spatial resolution improvement is not established. FWHM is measured at −3 dB on vessel profiles, and HR-XPD strongly suppresses background and sidelobes. A nonlinear weighting that suppresses the surrounding floor will narrow the −3 dB contour even if the underlying PSF is unchanged. The simulations show only ~1.2–1.4x FWHM gain at high SNR, and the larger gains appear in noise-limited and in vivo settings where contrast suppression does most of the work. Without a point-target or wire-phantom calibration, the resolution claim is confounded with contrast enhancement. This is the main soft spot, and it is addressable.\n\nOther issues are minor by comparison: single subjects for the human organs, no code or full parameter set, and a mild self-reinforcement in using the same subsets to compute the similarity map and the weighted product. The paper does acknowledge the lack of sensitivity to the smallest vessels relative to ULM, which is honest.\n\nWhat the paper does well: the method is clearly described, the comparisons against PD, SW-XPD, and HR-PD are fair, and the short acquisition times (0.3–1.2 s) are practically relevant. The qualitative improvement in vessel delineation is consistent across three organ types.\n\nWho should read it: anyone working on contrast-free microvascular Doppler or functional ultrasound. It deserves a serious referee, but the authors should be asked to either provide a point-target phantom experiment or soften the resolution claim to 'apparent resolution/contrast improvement.'\n\nI'd bring it to a reading group to discuss the metric issue.","headline":"Clever combination of radiality and coherence weighting that visibly cleans up contrast-free Doppler, but the 2–3x resolution gain is an artifact-prone FWHM claim until a point-target phantom is run.","tokens_in":15804,"tokens_out":2519,"would_cite":true,"duration_ms":22368,"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":"The paper argues that a new ultrasound processing method, HR-XPD, can sharpen contrast-free microvascular images two- to threefold by combining radiality weighting with cross-subset coherence, while suppressing background noise.","keywords":["Microvascular imaging","Power Doppler","Resolution enhancement","Contrast-free imaging","Cross-correlation","Radiality weighting","Similarity weighting","Ultrafast ultrasound"],"falsifier":"Place a wire or microsphere phantom at several depths and noise levels, reconstruct with both conventional PD and HR-XPD, and measure the full width of the point-spread function at multiple thresholds; if HR-XPD leaves the point-spread width unchanged while lowering the background floor, the reported resolution improvement is a contrast effect, whereas a narrowing of the point-spread itself would confirm true resolution enhancement.","tokens_in":14757,"feed_emoji":"🩸","tokens_out":6951,"duration_ms":60223,"temperature":0.7,"pith_summary":"Ultrasound power Doppler can show blood flow but is limited by diffraction and background noise. This paper proposes HR-XPD, a processing method that splits ultrafast angled transmissions into two interleaved subsets, clutter-filters each, weights each by a local radiality map measuring the convergence of intensity gradients, cross-correlates the weighted subsets, and then applies a similarity map that keeps coherent blood signal and rejects uncorrelated noise. The paper argues that this joint use of spatial symmetry and signal coherence resolves small vessels more sharply than conventional Doppler without injected contrast agents. In simulations and in vivo images of human liver, transplanted human kidney, and pig kidney, HR-XPD is reported to reduce vessel profile width by roughly two to threefold and raise contrast by up to 20 dB, using only 0.3 to 1.2 seconds of acquisition.","feed_headline":"Contrast-free microvessel Doppler sharpened up to 3x","feed_subtitle":"HR-XPD pairs radial symmetry with signal coherence to reveal small liver and kidney vessels in under 1.2 seconds.","key_machinery":"The load-bearing object is the radiality map $G_q$, defined at each pixel as the average over eight points on a surrounding annulus of the normalized dot product between the local image gradient and the outward radial vector. A value near 1 means gradients converge symmetrically, marking the center of a vessel's point-spread function. The companion machinery is the similarity map $M_0(x,z)$, the normalized cross-correlation between two clutter-filtered in-phase and quadrature (IQ) subsets over the frame ensemble. Radiality sharpens the spatial peak, while cross-correlation and similarity weighting suppress the uncorrelated noise that radiality alone amplifies; multiplying the radiality-weighted cross-correlation (HR-XPD*) by $M_0$ produces the final HR-XPD image.","core_discovery":"The central claim is that HR-XPD (high-resolution cross-correlation power Doppler) makes contrast-free microvascular ultrasound images both sharper and cleaner than conventional power Doppler. The method works by splitting ultrafast plane-wave transmissions into two subsets, clutter-filtering each subset separately, and combining two complementary weightings: a radiality map, which rewards pixels whose surrounding intensity gradients point inward like a point-spread-function center, and a similarity map, the normalized cross-correlation between the two subsets over time, which keeps coherent blood flow and rejects random noise. The product is a power Doppler image in which vessel centers are emphasized while sidelobes and background are suppressed. The paper supports this with simulations across signal-to-noise levels and in vivo demonstrations, reporting roughly 2.5- to 3.7-fold full-width-at-half-maximum resolution improvements over conventional PD, higher peak-to-valley separation of adjacent vessels, and background intensity reductions of about 20 dB.","pith_inferences":["A direct experimental separation of true resolution from contrast would be to run HR-XPD on a wire or point-scatterer phantom: if the point-spread function width is unchanged while background drops, the two- to threefold FWHM figure is mainly contrast enhancement rather than actual resolution gain.","The 2-3x resolution claim likely depends on where the -3 dB contour sits on a vessel profile; measuring widths at a lower threshold, such as -10 dB, would test whether the apparent sharpening persists across the full point-spread function.","Splitting transmissions into two angle subsets reduces the signal available to each subset, so the method's advantage may depend on having enough frames for the similarity map to stabilize; this can be tested by sweeping frame count.","If confirmed, HR-XPD could be combined with spectral decomposition or deconvolution approaches, since radiality and coherence suppression are complementary to model-based resolution restoration."],"forward_implications":["If the reported gains hold, microvascular anatomy can be imaged without contrast agents in under a second, making high-resolution vascular imaging practical in breath-held liver and kidney scans.","Adjacent vessels that conventional PD leaves unresolved (valley intensity below the 3 dB threshold) become separable, enabling more reliable vessel counting and microvascular density quantification.","The short acquisition time and high frame rate open a path to functional ultrasound activation mapping, where spatial sharpness and temporal resolution both matter.","The same radiality-plus-coherence weighting could be applied to microbubble contrast-enhanced data, potentially stabilizing radiality-based super-resolution in lower-signal-to-noise clinical settings, a possibility the paper notes.","Because the method relies on flow-signal coherence, it may be extended from power Doppler to color Doppler to add velocity information without losing the resolution gain."],"supporting_citations":[{"why":"Establishes the two-interleaved-subset cross-correlation framework and block-wise adaptive SVD clutter filtering that HR-XPD builds on.","marker":"[5]"},{"why":"Supplies the local block-wise SVD clutter filter and the in vivo kidney and liver imaging context.","marker":"[9]"},{"why":"Defines the similarity map used to weight coherent flow and suppress uncorrelated noise.","marker":"[42]"},{"why":"Introduces radiality weighting as a way to shrink the effective point-spread function in contrast-enhanced ultrasound imaging.","marker":"[47]"},{"why":"Provides the entropy-based radiality super-resolution formulation and documents radiality's sensitivity to noise.","marker":"[48]"},{"why":"Originates the radiality measure of gradient convergence that the radiality map uses.","marker":"[50]"},{"why":"Supplies the FWHM, background-intensity, and peak-to-valley metrics used to compare resolution and contrast.","marker":"[34]"}],"fun_headline_variants":["Radiality and similarity weighting sharpen ultrasound microvessels 3x","Contrast-free Doppler: 3x sharper microvasculature in seconds","HR-XPD: contrast-free microvascular imaging with 20 dB less noise","Sharper microvessels without contrast: radiality plus coherence","Microvascular ultrasound: 3x resolution, no contrast agent"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the -3 dB width of manually selected vessel profiles measures true spatial resolution rather than the edge of a suppressed background, because the paper does not calibrate against a point target or resolution phantom.","fun_headline_variants_meta":{"raw":{"variants":["Radiality and similarity weighting sharpen ultrasound microvessels 3x","Contrast-free Doppler: 3x sharper microvasculature in seconds","HR-XPD: contrast-free microvascular imaging with 20 dB less noise","Sharper microvessels without contrast: radiality plus coherence","Microvascular ultrasound: 3x resolution, no contrast agent"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000202,"raw_usage":{"total_tokens":1378,"prompt_tokens":936,"completion_tokens":442,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":552,"completion_tokens_details":{"reasoning_tokens":348}},"tokens_in":552,"tokens_out":442,"duration_ms":4244,"temperature":1.0,"reasoning_tokens":348,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:12:38.949863+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Place a wire or microsphere phantom at several depths and noise levels, reconstruct with both conventional PD and HR-XPD, and measure the full width of the point-spread function at multiple thresholds; if HR-XPD leaves the point-spread width unchanged while lowering the background floor, the reported resolution improvement is a contrast effect, whereas a narrowing of the point-spread itself would confirm true resolution enhancement.","supporting_citations":[{"cited_title":"Enhancement of Ultrasound Microbubble and Blood Flow Imaging using Similarity Measurement,","cited_arxiv_id":null,"evidence_quote":"Defines the similarity map used to weight coherent flow and suppress uncorrelated noise."},{"cited_title":"Simultaneous noise suppression and incoherent artifact reduction in ultrafast ultrasound vascular imaging,","cited_arxiv_id":null,"evidence_quote":"Establishes the two-interleaved-subset cross-correlation framework and block-wise adaptive SVD clutter filtering that HR-XPD builds on."},{"cited_title":"Ultrasound small vessel imaging with block-wise adaptive local clutter filtering,","cited_arxiv_id":null,"evidence_quote":"Supplies the local block-wise SVD clutter filter and the in vivo kidney and liver imaging context."},{"cited_title":"Ultrasound microvascular imaging based on super -resolution radial fluctuations,","cited_arxiv_id":null,"evidence_quote":"Introduces radiality weighting as a way to shrink the effective point-spread function in contrast-enhanced ultrasound imaging."},{"cited_title":"Ultrasound microvasculature imaging with entropy-based radiality super- resolution (ERSR),","cited_arxiv_id":null,"evidence_quote":"Provides the entropy-based radiality super-resolution formulation and documents radiality's sensitivity to noise."},{"cited_title":"Fast live-cell conventional fluorophore nanoscopy with ImageJ through super-resolution radial fluctuations,","cited_arxiv_id":null,"evidence_quote":"Originates the radiality measure of gradient convergence that the radiality map uses."},{"cited_title":"High-quality ultrafast power Doppler imaging based on spatial angular coherence factor,","cited_arxiv_id":null,"evidence_quote":"Supplies the FWHM, background-intensity, and peak-to-valley metrics used to compare resolution and contrast."}],"review_version":2}