{"id":"d29ccb0a-36d8-4f6e-a3ed-92d9e8f0ecac","arxiv_id":"2501.04457","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Power Doppler ultrasound during normothermic machine perfusion of porcine kidneys correlated with renal function and oxygen consumption, and distinguished functional from non-functional kidneys with AUCs up to 0.943.","lead":"This paper tested whether Power Doppler ultrasound, a standard imaging technique, can judge the quality of donor kidneys while they are kept alive and perfused at body temperature before transplant. In 22 pig kidneys, the Doppler signal matched kidney function and oxygen use well enough to separate healthy from damaged kidneys with about 82% accuracy.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The functional/non-functional labels are defined by the very CrCl/VO2 thresholds PD is compared against, and the PD cutoffs are optimized on the same data; without transplant outcomes or out-of-sample validation, the reported AUCs and 100% specificity are not established.","rationale":"The reader's weakest assumption is exactly the most load-bearing concern: the entire discrimination and correlation analysis depends on functional/non-functional labels that are defined by the same CrCl/VO2 thresholds being studied, with no external anchor such as transplant outcome. Our stress test confirms this and adds two reinforcing details: the cutoffs for PD are optimized on the same data, guaranteeing overoptimistic specificity, and the pooled repeated-measures correlation overstates the strength of evidence relative to the true sample size of 22 kidneys. The concern is not that the authors are dishonest; they explicitly list the lack of transplant as a limitation. But the central claim that 'PD imaging can be a valuable alternative to RBF' goes beyond what the data can support: the study shows that PD metrics track a particular internal classification, not that they diagnose real graft viability. A condition for accepting the paper as a hypothesis-generating exploratory study is to soften the conclusion and call for external validation. Since the reader already assigned CONDITIONAL with these caveats, and the paper's own limitations align with this concern, the verdict remains CONDITIONAL. No change to the reader's decision is warranted, but the rationale is reinforced with more specific statistical and definitional weaknesses.","tokens_in":9323,"tokens_out":6539,"duration_ms":68534,"concrete_test":"Evaluate PD metrics against transplant outcomes. In a follow-up cohort, subject porcine (or human) kidneys to the same NMP protocol with PD imaging, then transplant them and assess graft function (e.g., delayed graft function, 6-month serum creatinine/eGFR). If the 120-min VI, FI, VFI do not distinguish kidneys with good versus poor post-transplant function, then the internal CrCl/VO2 labels used here did not track true viability, and the reported discrimination is not meaningful. This directly tests the load-bearing assumption that the classification reflects actual graft quality rather than agreement with an unvalidated surrogate.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that PD metrics correlate with renal viability and differentiate functional from non-functional kidneys. The ground truth for this claim is a classification based on CrCl>1 ml/min/100g and VO2>2.6 ml/min/100g at 120 min. This label is unvalidated: the authors state in Limitations that they did not transplant the kidneys, so the thresholds may not track true graft viability. If the thresholds are wrong, the entire ROC analysis is measuring agreement with an arbitrary internal surrogate, not real transplant-relevant quality. The classification definition also has a gap: a kidney with CrCl>1 but VO2≤2.6 meets neither stated criterion, yet all 22 kidneys were assigned to a group, implying an unreported rule or unusual concordance that could affect reproducibility. Finally, the PD cutoffs (VI=17%, FI=50, VFI=9) are selected on the same dataset to maximize specificity, so the 100% specificity and AUC 0.943 are in-sample estimates subject to optimism; out-of-sample accuracy would be lower. The reported correlation r=0.839 between VI and VO2 pools 66 repeated measurements from 22 kidneys, inflating statistical significance beyond the true independent sample size. These issues do not disprove an association, but they mean the paper has not established PD as a validated alternative to RBF.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript evaluates Power Doppler (PD) ultrasound metrics (VI, FI, VFI) as real-time imaging markers of kidney quality during normothermic machine perfusion (NMP) in 22 slaughterhouse porcine kidneys. Kidneys were classified as functional (n=7) or non-functional (n=15) according to perfusion-based thresholds for creatinine clearance (CrCl > 1 ml/min/100g) and oxygen consumption (VO2 > 2.6 ml/min/100g) at 120 minutes. PD metrics were measured at 30, 60, and 120 minutes and compared with CrCl, VO2, and renal blood flow (RBF). The authors report moderate-to-strong correlations between PD metrics and CrCl/VO2, and present ROC analyses with high AUCs at 120 minutes (up to 0.943), proposing cutoffs with 100% specificity for detecting non-functional kidneys. They conclude that PD imaging can be a valuable alternative to RBF during NMP.","tokens_in":9608,"tokens_out":2766,"duration_ms":28528,"significance":"If the claims were substantiated, PD imaging would provide a non-invasive, real-time, and relatively inexpensive tool for pre-transplant kidney quality assessment during NMP, potentially reducing discard rates of marginal kidneys. The study is exploratory and addresses a clinically relevant gap. Strengths include the side-by-side comparison of PD metrics with RBF at multiple time points, the use of blinded histopathology scoring, and the clear description of the perfusion setup. However, the central diagnostic-accuracy claims are currently compromised by in-sample cutoff optimization, an unvalidated composite reference standard, and statistical analyses that pool repeated measurements; these issues mean the reported AUCs, specificities, and correlation p-values are optimistic and may not reflect real-world performance.","major_comments":[{"comment":"The cutoff values for VI (17%), FI (50 a.u.), and VFI (9 a.u.) were selected on the same 22-kidney dataset used to compute the reported 100% specificity, 73% sensitivity, NPV of 64%, and overall accuracy of 82%. This is in-sample threshold optimization, which systematically overestimates diagnostic performance; the stated values are therefore not reliable estimates of out-of-sample accuracy. The authors should either report optimism-corrected estimates (e.g., via bootstrap or cross-validation) or explicitly label these as exploratory in-sample cutoffs that require external validation before any clinical or discard decisions can be supported.","section":"Statistical analysis / Results (Cutoff values for intergroup differentiation)"},{"comment":"The functional/non-functional classification rests entirely on internally defined CrCl and VO2 thresholds measured at 120 minutes of NMP, with no transplant or post-reperfusion outcome to validate these labels. The authors acknowledge in Limitations that they 'did not transplant the kidneys and hence not able to measure the posttransplant renal function.' Consequently, the ROC analyses evaluate PD agreement with an unvalidated surrogate, not with actual graft viability; if the thresholds do not track transplant-relevant viability, the reported discriminative performance is not meaningful. In addition, the classification criteria as written (functional: CrCl>1 AND VO2>2.6; non-functional: CrCl≤1 AND VO2≤2.6) leave an undefined middle category (e.g., CrCl>1 with VO2≤2.6), yet all 22 kidneys were assigned to a group. The authors should report how mixed profiles were handled and how the classification was applied in practice.","section":"Study design / Materials and Methods (renal function classification) and Limitations"},{"comment":"The correlations reported in Figure 5 pool all 66 measurements (22 kidneys × 3 time points) as independent observations. This violates the independence assumption of Spearman correlation and inflates the effective sample size, making the reported p-values (e.g., r=0.839, p<0.0001) overly precise. Because measurements from the same kidney are correlated, the true number of independent kidneys for the association between VI and VO2 is 22, not 66. The authors should use a mixed-effects model or per-kidney summary measures (e.g., averaged or time-matched values) to account for the repeated-measures structure.","section":"Results / Correlation analysis (Figure 5)"},{"comment":"The conclusion that PD imaging 'can be a valuable alternative to RBF' is not directly supported by a statistical comparison of PD metrics against RBF in the same predictive framework. Although the 120-minute AUCs appear numerically higher for PD (0.943 vs. 0.886), the 95% confidence intervals in Tables S3 and S4 overlap substantially (e.g., VI: 0.848-1.000; RBF: 0.747-1.000), and no formal test of AUC difference or added predictive value is reported. Without such a comparison, the claim of superiority or equivalence with respect to RBF remains unsubstantiated.","section":"Discussion / Conclusions"}],"minor_comments":[{"comment":"In the text immediately after Figure 3, the sentence 'The functional group showed a progressive increase in signal intensity and vascularization across the time points' is descriptive; the corresponding quantitative values at 60 and 120 minutes are not shown in Figure 4, which only displays the 30-minute comparisons. Please report all time points or clarify the figure legend.","section":"Results / Power Doppler metrics assessment"},{"comment":"The statement that PD acquisition settings were 'chosen manually to optimize the visualization of the renal cortical vasculature' (Table S2) raises reproducibility concerns. Since PD intensity values are compared across kidneys and time points, the authors should specify whether the same settings were used for all acquisitions and whether any gain or depth adjustments were made between measurements.","section":"Materials and Methods / Power Doppler measurement"},{"comment":"The abstract reports 'FI and VFI moderately correlated with CrCl (r=0.537... r=0.536...)' but omits the correlation of VI with CrCl (r=0.528) that appears in the main text. For consistency, either include all three metrics or state that all PD metrics were correlated.","section":"Abstract / Results"},{"comment":"There are minor typographical issues, such as 'scatter' in Figure 5 (should be 'scatter') and the parenthetical '(shown?)' in the Results section, which should be corrected.","section":"General"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague—\n\nWorth a look if you work on machine perfusion or organ assessment. The group at Erasmus is the first to use Power Doppler (VI/FI/VFI) to score cortical perfusion during normothermic machine perfusion, and the idea is sensible: RBF is a global measurement, while PD looks at the cortical microcirculation where nephrons live. In 22 slaughterhouse pig kidneys they show PD metrics correlate with CrCl and VO2, and that PD separates their functional/non-functional groups earlier than RBF does. That is a genuine new application of an established imaging technique, and the imaging setup is described carefully enough to reproduce.\n\nThe soft spots are real and mostly acknowledged. The functional/non-functional labels rest on CrCl and VO2 thresholds measured at 120 minutes of perfusion, with no transplant outcome to validate those thresholds. The authors say this in the limitations, so they are not hiding it, but it means the ROC analysis measures agreement with an internal surrogate, not with graft viability. The PD cutoffs that give 100% specificity are chosen on the same 22 kidneys they are then evaluated on, so those numbers are optimistic; out-of-sample specificity will be lower. The correlation coefficients pool 66 repeated measurements from 22 kidneys, which inflates significance beyond the true sample size. And the classification description has a small gap: functional requires both CrCl>1 and VO2>2.6, non-functional requires both ≤, yet all 22 kidneys were assigned to a group. Either the two criteria never disagreed in this dataset or an unreported rule was used; the paper should say which.\n\nNone of this sinks the basic association. The direction is physiologically coherent, the imaging is clean, and the authors are appropriately cautious in their conclusion. But the phrase \"effectively differentiate\" is stronger than the evidence supports. This is a proof-of-concept, not a validated alternative to RBF.\n\nWho is this for: anyone working on NMP-based organ assessment, or on Doppler imaging as a viability marker. It deserves a serious referee; the novelty is real and the methods are mostly transparent. I would send it to review with a request to address the classification rule, report per-kidney rather than pooled correlations, and reframe the diagnostic accuracy claims as hypothesis-generating. If the authors can add even a small validation cohort, the paper would be much stronger.","headline":"A plausible first use of Power Doppler for kidney quality assessment during NMP, but the headline accuracy numbers are in-sample and the ground-truth labels are unvalidated, so treat the AUCs as suggestive, not established.","tokens_in":10156,"tokens_out":1937,"would_cite":false,"duration_ms":18882,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Power Doppler ultrasound during machine perfusion can identify non-functional kidneys from cortical blood flow images.","keywords":["power doppler","normothermic machine perfusion","kidney viability","kidney quality assessment","vascularization index","flow index","renal blood flow","porcine model"],"falsifier":"Perfuse a set of kidneys with identical metabolic status but artificially varied pump flow and compare PD indices; if VI, FI, and VFI track pump flow more tightly than they track CrCl and VO2, then PD is a perfusion meter rather than a viability test. The stronger falsifier is a transplant follow-up: if PD metrics taken during NMP fail to predict post-transplant urine output or graft survival, the central claim that PD can replace RBF for kidney-quality assessment is refuted.","tokens_in":9139,"feed_emoji":"🩸","tokens_out":4716,"duration_ms":47031,"temperature":0.7,"pith_summary":"Power Doppler ultrasound, a real-time imaging technique that shows blood moving through tissue, can grade kidney quality during normothermic machine perfusion. The paper reports that in 22 porcine kidneys, the Doppler metrics vascularization index, flow index, and vascularization flow index correlate with creatinine clearance and oxygen consumption, the standard perfusion-based markers of kidney function and viability. At 120 minutes of perfusion, these metrics separate kidneys deemed functional from non-functional with an area under the ROC curve near 0.94 and, at chosen cutoffs, 100% specificity. The authors conclude that Power Doppler can serve as a practical alternative to renal blood flow measurement for pre-transplant kidney assessment, with the advantage of imaging the cortical microcirculation directly. If true, this would give transplant teams a non-invasive, repeatable way to watch organ recovery without repeated blood sampling.","feed_headline":"Power Doppler can grade kidneys during machine perfusion","feed_subtitle":"Three ultrasound flow indices matched kidney viability markers and beat renal blood flow in a 22-kidney study.","key_machinery":"The central object is Power Doppler ultrasound, an imaging mode in which the ultrasound system assigns pixel intensity proportional to the amplitude of the Doppler signal from moving red blood cells. Three summary metrics are computed from the renal cortex, defined as a 5 mm layer below the kidney surface: the vascularization index (VI), the percentage of colored pixels in the region of interest; the flow index (FI), the mean signal intensity of those colored pixels; and the vascularization flow index (VFI), the product of VI and FI. These indices convert a qualitative perfusion image into numbers that can be tracked over time and compared across kidneys, and it is this quantification that lets the authors treat PD as a functional assay rather than just a picture.","core_discovery":"The paper's claim is that Power Doppler imaging during normothermic machine perfusion captures the same biological signal that clinicians currently read from renal blood flow, and does so earlier and more precisely. Using 22 slaughterhouse pig kidneys perfused for two hours, the authors show that the three quantitative PD metrics differ significantly between kidneys labelled functional (n=7) and non-functional (n=15) by creatinine clearance above 1 ml/min/100g and oxygen consumption above 2.6 ml/min/100g. Pooled over all time points, VI, FI and VFI correlate moderately with CrCl (r≈0.54) and strongly with VO2, with VI reaching r=0.839. At 120 minutes, VI and VFI reach AUC 0.943 and FI 0.924 for intergroup separation, and cutoffs of VI above 17%, FI above 50 a.u., and VFI above 9 a.u. identify non-functional kidneys with 100% specificity and 73% sensitivity, an overall accuracy of 82% versus 77% for renal blood flow. The authors state that PD metrics reflect renal cortical perfusion specifically, which explains why they outperform whole-organ flow.","pith_inferences":["A testable extension would be to acquire PD images during hypothermic machine perfusion as well, since the same vascular index might predict post-reperfusion function without the cost of warming the organ.","If PD correlates with VO2 through microvascular density, then PD might also serve as a surrogate for viability in other perfused organs, such as livers or hearts, during normothermic preservation.","The 100% specificity cutoffs are tuned to avoid discarding viable kidneys, but a clinical protocol would still need to weigh the 18% false-negative rate against the consequences of transplanting a marginal organ, a trade-off the paper mentions but does not resolve.","Because the functional labels come from CrCl and VO2 thresholds, the strongest test of PD would come from a series where kidneys are imaged during NMP and then transplanted; if PD failed to predict delayed graft function, the present AUCs would overstate its accuracy."],"forward_implications":["Power Doppler could replace continuous renal blood flow probes during normothermic machine perfusion, giving a similar viability signal with less hardware and no in-line flow sensor.","Separation between functional and non-functional kidneys was already seen at 30 minutes, suggesting PD could shorten the assessment window before a discard decision.","Because PD images the renal cortex, it may detect regional perfusion defects that whole-organ flow misses, providing information about heterogeneity rather than only average perfusion.","The proposed cutoffs (VI above 17%, FI above 50 a.u., VFI above 9 a.u. at two hours) could be tested prospectively in clinical NMP as a pre-specified rule for whether to accept a kidney.","If PD is integrated into NMP circuits, quality assessment becomes repeatable and non-destructive, allowing the same kidney to be monitored throughout the perfusion period."],"supporting_citations":[{"why":"Supplies the CrCl and VO2 thresholds used to assign kidneys to functional and non-functional groups.","marker":"[24]"},{"why":"Provides a comparable NMP-based classification scheme for kidney function, supporting the chosen cutoffs.","marker":"[25]"},{"why":"The authors' previous laser speckle contrast imaging study of cortical microcirculation during NMP, which PD imaging extends.","marker":"[13]"},{"why":"MRI study documenting heterogeneous corticomedullary flow during NMP, cited to justify why cortical-specific PD can outperform whole-organ RBF.","marker":"[11]"},{"why":"Defines the VI, FI, and VFI equations and their interpretation as vascularity and blood volume indices.","marker":"[27]"},{"why":"Establishes normothermic machine perfusion as a quality-assessment platform with RBF-based scoring that PD aims to improve.","marker":"[8]"}],"fun_headline_variants":["PD imaging rivals renal blood flow for kidney function screening","Power Doppler indices predict kidney viability better than blood flow","Ultrasound metrics spot dead kidneys during machine perfusion","Three PD indices match kidney viability, beat RBF in 22-kidney test","Power Doppler grades kidney quality earlier than renal blood flow"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's classification of kidneys as functional or non-functional is based entirely on creatinine clearance and oxygen consumption thresholds measured during perfusion, not on transplant or post-reperfusion outcomes, so if those thresholds do not reflect true graft viability then the reported Power Doppler accuracy is measuring agreement with an unvalidated label.","fun_headline_variants_meta":{"raw":{"variants":["PD imaging rivals renal blood flow for kidney function screening","Power Doppler indices predict kidney viability better than blood flow","Ultrasound metrics spot dead kidneys during machine perfusion","Three PD indices match kidney viability, beat RBF in 22-kidney test","Power Doppler grades kidney quality earlier than renal blood flow"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000772,"raw_usage":{"total_tokens":3521,"prompt_tokens":1149,"completion_tokens":2372,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":765,"completion_tokens_details":{"reasoning_tokens":2289}},"tokens_in":765,"tokens_out":2372,"duration_ms":14712,"temperature":1.0,"reasoning_tokens":2289,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:31:55.603829+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Perfuse a set of kidneys with identical metabolic status but artificially varied pump flow and compare PD indices; if VI, FI, and VFI track pump flow more tightly than they track CrCl and VO2, then PD is a perfusion meter rather than a viability test. The stronger falsifier is a transplant follow-up: if PD metrics taken during NMP fail to predict post-transplant urine output or graft survival, the central claim that PD can replace RBF for kidney-quality assessment is refuted.","supporting_citations":[{"cited_title":"Magnetic resonance imaging assessment of renal flow distribution patterns during ex vivo normothermic machine perfusion in porcine and human kidneys","cited_arxiv_id":null,"evidence_quote":"MRI study documenting heterogeneous corticomedullary flow during NMP, cited to justify why cortical-specific PD can outperform whole-organ RBF."}],"review_version":1}