{"id":"e955dd31-283c-4f6c-8f3d-7d745fbd6157","arxiv_id":"1908.04538","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A variational autoencoder with a regression loss learns a blood-pressure axis in latent space from cardiac MRI biomarkers, revealing that left ventricular diastolic biomarkers change most with systolic blood pressure.","lead":"This paper combines automatic heart MRI measurements with a machine learning model that links cardiac biomarkers to blood pressure. It shows higher blood pressure is mainly associated with left ventricular diastolic changes, and it uses the model to label prehypertensive people as higher or lower risk.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Decoded SBP trajectories may be driven by age: no age adjustment in R-VAE despite strong SBP-age correlation in 40-69 cohort.","rationale":"The paper's method is clearly presented and the segmentation/biomarker pipeline draws on previously validated work, so the engineering contribution is not in question. The concern is about the inference from the decoded latent trajectories to a physiological statement. The central claim that increases in SBP are mainly linked to diastolic LV dysfunction requires that the latent axis trained against SBP represents SBP-specific biology. The model's only covariate besides SBP is gender; age is not included or even discussed. Given the UK Biobank age range (40-69), SBP is strongly age-dependent, and the very biomarkers highlighted (iLVEDV, LVPAFR, LVPER) are known to change with age. A latent regression line that maximally predicts SBP can therefore absorb age-related variance. Experiment 2's decoded 'changes with increasing SBP' are thus not interpretable as SBP effects without age adjustment or age-stratified replication. This is distinct from the reader's longitudinal concern, which is about within-subject progression versus cross-sectional differences; the age confound threatens even the cross-sectional association itself. I would keep the verdict CONDITIONAL because the issue is testable and the methodological framework remains promising, but the physiological conclusion needs an age-controlled analysis before it can be accepted.","tokens_in":6435,"tokens_out":4281,"duration_ms":44381,"concrete_test":"Re-run Experiment 2 with age included as an additional covariate in the latent-space regression, and also compute direct cross-sectional regressions of iLVEDV, LVPAFR, and LVPER on SBP with and without adjustment for age (and, if available, heart rate and antihypertensive medication) in the same 3,600 subjects. If the SBP coefficients or decoded trajectories attenuate substantially after adjustment, or if trajectories at fixed age no longer show the reported pattern, the SBP-specific interpretation fails. A complementary check is to train a model using age as the regression target instead of SBP and compare decoded directions: a high correlation would confirm confounding.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing assumption behind the central physiological conclusion is that the latent SBP axis identified by the R-VAE isolates blood-pressure-specific variation in cardiac biomarkers. The paper does not establish this. Section 3.2 describes a regression model with 'confounders,' but the only covariate besides SBP is a gender dummy. Age is never mentioned, despite the UK Biobank cohort spanning 40-69 years and SBP rising steeply with age. In cross-sectional data, the same diastolic LV changes (lower iLVEDV, higher LVPAFR and LVPER, stiffer LV) also occur with normal aging. Experiment 2 samples the latent regression line from 100 to 170 mmHg and attributes the decoded biomarker changes to SBP, but if the latent direction encodes age-related cardiac remodeling because age is correlated with SBP yet omitted from the model, the decoded trajectories are confounded. The Discussion's statement that 'SBP results in slowly progressive changes in the myocardium' is causal and requires that these trajectories reflect SBP, not age or other omitted factors. Without age adjustment, age stratification, or a negative-control experiment, the central physiological claim is not yet supported. This concern is distinct from, and more basic than, the reader's longitudinal concern: even a purely associational claim about SBP would be threatened by age confounding.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents R-VAE, a variational autoencoder with a latent-space regression loss, applied to 3,600 UK Biobank subjects to relate 13 automatically estimated cardiac MRI biomarkers to systolic blood pressure (SBP). The authors report that R-VAE predicts SBP with R2=0.69, outperforming Lasso and Random Forest, and then use the model to decode how cardiac biomarkers change along the fitted SBP axis in men and women (Experiment 2) and to identify biomarkers distinguishing prehypertensive subjects predicted as normotensive versus hypertensive (Experiment 3). The central physiological claim is that increasing SBP is mainly linked to changes in left ventricular diastolic function, suggesting progressive myocardial stiffening.","tokens_in":6735,"tokens_out":3222,"duration_ms":35501,"significance":"If the results are supported, the interpretable R-VAE framework would be a useful tool for exploring associations between risk factors and cardiac structure/function in large cohorts. The paper has clear strengths: it builds on a fully automated, quality-controlled biomarker extraction pipeline, uses well-known clinical biomarkers for interpretability, integrates regression and generative modeling in a principled way, and stratifies by gender. The decoded trajectories are a falsifiable prediction that could be tested against longitudinal or interventional data. However, the central physiological conclusions currently rest on cross-sectional associations, omit age as a likely confounder, and are extracted from the model's own fitted regression and decoder, so the evidence as presented does not yet support the causal or temporal language used in the Discussion.","major_comments":[{"comment":"The text states that 'all methods performed similarly with regard to regression', but Table 1 reports R2=0.69 for R-VAE versus 0.35 for Lasso and 0.33 for Random Forest, which is not similar. No confidence intervals, cross-validation variance, or significance tests are reported for RMSD, nRMSD, or R2, so the comparative claim is unsupported. The authors should report uncertainty intervals and a test of whether the R2 difference is statistically significant, or temper the comparison.","section":"Section 4, Experiment 1, Table 1"},{"comment":"The latent regression model includes only SBP and a gender dummy variable; age is never mentioned as a covariate. The UK Biobank cohort spans ages 40-69, SBP increases steeply with age, and the diastolic LV changes decoded along the SBP axis (lower iLVEDV, higher LVPAFR and LVPER) also occur with normal aging. Because age is omitted and is correlated with SBP, the decoded trajectories may reflect age-related cardiac remodeling rather than SBP-specific effects. A concrete remedy is to include age in the regression, stratify the decoding by age decade, or run a negative control in which the latent regression is trained on age instead of SBP and the decoded patterns are compared.","section":"Section 3.2 and Section 4, Experiment 2"},{"comment":"Cross-sectional differences between subjects are interpreted as temporal progression: the Discussion states that 'SBP results in slowly progressive changes in the myocardium' and describes 'deterioration of cardiac function with increasing SBP'. No longitudinal follow-up, causal inference method, or intervention data are provided, so this causal and temporal claim is not supported by the evidence. The authors should either reframe the findings as associations between SBP and cardiac biomarkers in a cross-sectional population or add an explicit causal modeling approach.","section":"Section 4, Experiment 2, and Section 5 Discussion"},{"comment":"The abnormal-response analysis is circular with respect to the model: under- and over-prediction are defined by the same latent-space regression that is then used to decode the biomarkers and attribute differences to those biomarkers. Without external validation, such as follow-up outcome data or an independent cohort, the claim that specific LV diastolic biomarkers 'might be effective when trying to stratify cardiac risk' is exploratory, not a validated finding. The authors should clearly label this experiment as internal model analysis and avoid using it as independent evidence for the biomarkers' clinical utility.","section":"Section 4, Experiment 3"}],"minor_comments":[{"comment":"The paragraph beginning 'We applied our R-VAE network to investigate how cardiac function changes with increasing SBP' is duplicated verbatim; one copy should be deleted.","section":"Section 4, Experiment 2"},{"comment":"The reference to the image acquisition protocol appears as '[ ?]'; a proper reference should be inserted.","section":"Section 2"},{"comment":"The text first defines L_regression as mean squared error but then refers to it as the Huber loss in the joint loss equation; the notation should be made consistent.","section":"Section 3.2, Eq. (1)"},{"comment":"The figures show standard deviations but no confidence intervals, and the 'linear tendency curves' plotted with black dotted lines are not defined in the text or captions; adding a description would improve interpretability.","section":"Section 4, Figures 2 and 3"},{"comment":"The hyperparameters alpha=0.3, beta=2, dropout=0.3, and latent dimensionality 2 are introduced without sensitivity analysis; a short robustness check or a sentence justifying these choices would strengthen confidence in the qualitative results.","section":"Section 3.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a compact application paper whose main contribution is the interpretable R-VAE framework and its application to cardiac biomarkers. The core method is reasonable, but the physiological conclusions are stated more strongly than the cross-sectional evidence supports, and the age-confounding issue is a substantive threat. I recommend major revision rather than rejection because the issues are addressable by adding covariates, reframing causal language, and adding uncertainty quantification. I have no concerns about citation practices or novelty disclosure."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this is a legitimate but limited framework paper. The new part is a VAE with a latent-space regression loss (plus a gender dummy) applied to thirteen interpretable cardiac biomarkers, and the first demonstration of that on 3,600 UK Biobank subjects. The idea of decoding along the regression line is a neat way to visualize multivariate associations. The authors cite the relevant autoencoder-regression literature and are clear that the biomarker extraction pipeline comes from their earlier validated work.\n\nThe soft spots are in the analysis, and the stress-test lands. The cohort is 40–69 years old, SBP rises steeply with age, and the model controls for gender only. Age is never included as a confounder. So the decoded trajectories—lower iLVEDV, higher LVPAFR, higher LVPER—could encode age-related remodeling just as plausibly as SBP-specific effects. The Discussion goes further and says SBP 'results in slowly progressive changes in the myocardium.' That is causal language on cross-sectional data, and it is not supported. The reader's circularity concern is secondary: decoding from the fitted model is a legitimate way to visualize what the model has captured, provided the model actually controls for confounders. Here it does not, so the visualization is not trustworthy as a statement about SBP.\n\nThere are also no confidence intervals or significance tests on the regression comparison in Table 1 or on the decoded trends. R2=0.69 for R-VAE versus 0.35 for Lasso might be real, but you cannot tell from the paper. Experiment 3 is an interesting proof-of-concept, but the same age-confounding caveat applies.\n\nVerdict: this deserves serious peer review because the method is relevant and the application is important, but a referee should require an age-adjusted model, an age-stratified analysis, or a negative-control experiment before accepting the physiological claim. As it stands, the paper is a promising methods demonstration, not an established clinical finding.","headline":"A useful framework paper whose central physiological claim is undercut by missing age adjustment and by decoding the model's own fitted regression line.","tokens_in":7243,"tokens_out":1785,"would_cite":false,"duration_ms":19099,"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":"Rising blood pressure mainly alters the heart's filling phase","keywords":["variational autoencoder","latent space regression","cardiac function","systolic blood pressure","interpretable biomarkers","diastolic function","cardiac MRI","risk stratification"],"falsifier":"Repeatedly measure systolic blood pressure and the same cardiac biomarkers in the same individuals over several years; if a person's within-person rise in blood pressure does not produce the decoded changes (falling indexed end-diastolic volume, rising peak atrial filling and emptying rates), the physiological interpretation fails. A blood-pressure-lowering intervention that shifts systolic pressure but leaves these biomarkers unchanged would likewise refute the claim.","tokens_in":6279,"feed_emoji":"🫀","tokens_out":9566,"duration_ms":81188,"temperature":0.7,"pith_summary":"Using a variational autoencoder whose latent space is forced to regress systolic blood pressure, the paper maps how cardiac function changes as blood pressure climbs. In 3,600 healthy subjects, the model decodes a consistent pattern: increasing systolic blood pressure is associated with decreased left-ventricular end-diastolic volume, increased peak atrial filling rate, and increased peak emptying rate, while the right ventricle stays largely unchanged. The authors interpret this as evidence that rising blood pressure chiefly compromises left-ventricular diastolic function, pointing to myocardial stiffening as a key disease process. The same model flags prehypertensive individuals whose predicted blood pressure is higher or lower than measured, offering a biomarker-based way to separate subgroups at differential risk.","feed_headline":"Rising blood pressure mainly alters the heart's filling phase","feed_subtitle":"Decodes how rising systolic pressure alters heart biomarkers, pointing to left-ventricular stiffening as the key process","key_machinery":"The central object is the R-VAE: a variational autoencoder with a two-dimensional latent space and a linear regression term added to the loss so that one direction in latent space encodes systolic blood pressure. The regression uses a dummy variable for gender, and the total loss is $L_{\\text{R-VAE}} = L_{\\text{recon}} + \\alpha L_{\\text{KL}} + \\beta L_{\\text{regression}}$. Because the decoder maps latent coordinates back to the 13 interpretable biomarkers, sampling along the regression line lets the model decode how each biomarker responds to blood pressure while holding other latent variation fixed. This combination of a generative decoder with a supervised regression axis is what turns the model into a tool for physiological interpretation rather than a black-box predictor.","core_discovery":"The central claim is that a regression-constrained variational autoencoder (R-VAE) learns a latent axis that tracks systolic blood pressure, and that decoding clinical cardiac biomarkers along this axis reveals a coherent physiological trajectory: as systolic blood pressure rises, indexed left-ventricular end-diastolic volume falls while left-ventricular peak atrial filling rate and peak emptying rate rise, with the right ventricle unaffected. The paper further claims that these decoded changes are larger in males for two of the biomarkers, and that the same latent regression can separate prehypertensive subjects whose cardiac biomarkers resemble normotensive or hypertensive patterns, with diastolic biomarkers again dominating the separation. The paper states this pattern suggests that stiffening of the left-ventricular myocardium could be an important disease process in deterioration of cardiac function in the light of increased systolic blood pressure.","pith_inferences":["A testable extension: apply the same R-VAE to a dataset with repeated measurements to check whether the decoded cross-sectional trajectory matches within-person progression; the paper's physiological story depends on this match.","If the latent axis truly reflects causal SBP effects, a randomized blood-pressure-lowering trial should show the decoded biomarkers reverting toward normotensive values, a prediction not tested in the paper.","The biomarker bottleneck discards raw image information, so a combined image-and-biomarker VAE might reveal whether the SBP axis changes when higher-dimensional features are available.","Because the model encodes SBP in one latent direction, the same framework could regress multiple risk factors simultaneously with orthogonal regression directions, exposing interactions between hypertension and other factors."],"forward_implications":["If correct, routine cardiac MRI could track blood-pressure-related cardiac damage through a small set of interpretable diastolic biomarkers, without extra scans.","The gender differences in the decoded trajectories imply that monitoring and intervention thresholds for blood-pressure-related cardiac remodeling may need to be sex-specific.","Prehypertensive individuals whose biomarker profile predicts hypertension could be prioritized for early treatment, while those whose profile predicts normotension could be reassured.","The R-VAE approach is not limited to blood pressure: the same latent regression with a dummy variable can be repurposed to other risk factors once the regression target is swapped."],"supporting_citations":[{"why":"Supplies the fully automated, quality-controlled biomarker extraction pipeline that produces the 13 cardiac biomarkers from CMR.","marker":"[11]"},{"why":"Provides the fully convolutional network used to segment left and right ventricles from the cine CMR images.","marker":"[12]"},{"why":"Provides the registration algorithm used to align segmentations and correct breath-hold motion artifacts before biomarker calculation.","marker":"[13]"},{"why":"Supplies the blood pressure measurements used as the regression target.","marker":"[10]"},{"why":"Provides the multivariable regression modelling convention the latent-space regression with confounding variables is based on.","marker":"[15]"}],"fun_headline_variants":["BP mainly shrinks heart's filling phase, AI study shows","Rising systolic pressure impacts heart filling most","VAE decodes blood pressure's key cardiac effect: filling","Heart stiffening is key effect of blood pressure, AI finds","Blood pressure's main heart effect: altered filling phase"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The model treats differences in blood pressure between different people at one point in time as if they were changes that happen within a single person over time.","fun_headline_variants_meta":{"raw":{"variants":["BP mainly shrinks heart's filling phase, AI study shows","Rising systolic pressure impacts heart filling most","VAE decodes blood pressure's key cardiac effect: filling","Heart stiffening is key effect of blood pressure, AI finds","Blood pressure's main heart effect: altered filling phase"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000498,"raw_usage":{"total_tokens":2408,"prompt_tokens":885,"completion_tokens":1523,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":501,"completion_tokens_details":{"reasoning_tokens":1443}},"tokens_in":501,"tokens_out":1523,"duration_ms":16117,"temperature":1.0,"reasoning_tokens":1443,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:39:41.717643+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeatedly measure systolic blood pressure and the same cardiac biomarkers in the same individuals over several years; if a person's within-person rise in blood pressure does not produce the decoded changes (falling indexed end-diastolic volume, rising peak atrial filling and emptying rates), the physiological interpretation fails. A blood-pressure-lowering intervention that shifts systolic pressure but leaves these biomarkers unchanged would likewise refute the claim.","supporting_citations":[{"cited_title":"Fully automated, quality-controlled cardiac analysis from cmr: Validation and large-scale application to characterize cardiac function,","cited_arxiv_id":null,"evidence_quote":"Supplies the fully automated, quality-controlled biomarker extraction pipeline that produces the 13 cardiac biomarkers from CMR."},{"cited_title":"Automated cardiovascular magnetic reso- nance image analysis with fully convolutional networks,","cited_arxiv_id":null,"evidence_quote":"Provides the fully convolutional network used to segment left and right ventricles from the cine CMR images."},{"cited_title":"Fully automated segmentation-based respiratory motion correction of multiplanar cardiac magnetic resonance images for large-scale datasets,","cited_arxiv_id":null,"evidence_quote":"Provides the registration algorithm used to align segmentations and correct breath-hold motion artifacts before biomarker calculation."},{"cited_title":"Associations with intraocular pressure in a large cohort: results from the uk biobank,","cited_arxiv_id":null,"evidence_quote":"Supplies the blood pressure measurements used as the regression target."},{"cited_title":"How should meta-regression analyses be undertaken and interpreted?,","cited_arxiv_id":null,"evidence_quote":"Provides the multivariable regression modelling convention the latent-space regression with confounding variables is based on."}],"review_version":1}