{"id":"b29bbe8c-40fd-46e4-ac16-79ae10b970db","arxiv_id":"2507.01411","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"A 3D CNN trained on voxel-wise hippocampal functional connectivity predicted chronological age with a 7.30 year MAE on adults aged 36-90, and LayerCAM saliency maps showed age-sensitive connections with precuneus, cuneus, PCC, PHC, left SPL, and right STS.","lead":"This study used a 3D convolutional neural network to predict a person's age from brain scans that map how the hippocampus communicates with the rest of the cortex. The model's saliency maps highlighted hippocampal connections to specific regions that change most with age, offering a finer-grained view of brain aging than standard region-based methods.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim hinges on LayerCAM saliency maps that are never validated as faithful explanations of the CNN; no occlusion, perturbation, or alternative-explainer test is reported.","rationale":"The reader's weakest assumption—that averaged LayerCAM maps truthfully reflect neurobiological age-sensitivity—is also the premise that carries the paper's novel claim. If that premise is false, the list of regions and the anterior/posterior dissociation are not supported, even though the age-prediction numbers themselves may be valid. I agree with the reader's assessment. The paper deserves credit for a well-specified prediction pipeline, fivefold CV, and anatomical findings consistent with prior aging literature, but those positives do not validate the saliency maps. A conditional verdict is appropriate: the interpretability claims should be held until occlusion/perturbation or an equivalent faithfulness check is provided. My proposed occlusion test would settle whether the concern lands; if it passes, the conditional can be upgraded.","tokens_in":20229,"tokens_out":6479,"duration_ms":75805,"concrete_test":"On the HCP-A validation folds, run a voxel-wise occlusion/perturbation test: for each subject, compute the predicted age from the original hippocampal-FC input, then zero out FC values in (a) the six saliency-highlighted ROIs and (b) matched control ROIs of equal size and similar baseline FC but low saliency; recompute predicted age. Use a paired permutation test across subjects to compare the mean absolute change in predicted age between saliency and control ROIs. If saliency ROIs do not produce significantly larger changes, the LayerCAM maps do not reflect the model's actual input dependence and the central interpretation claim fails. As a secondary check, compare LayerCAM maps against ridge-regression weights trained on the same FC input; near-zero spatial correlation would indicate saliency captures artifacts rather than age signal.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's headline result—that hippocampal-cortical connections to precuneus, cuneus, PCC, PHC, left SPL, and right STS are 'most influential' for age prediction—depends on the premise that the LayerCAM saliency maps in Sec. 2.2/Fig. 2a are faithful to the trained 3D CNN's actual decision process. That premise is asserted, not tested. Section 5.5 defines the adapted LayerCAM (Eqs. 2–6), but no experiment shows that these maps track the model's input-output dependence. The comparison with Grad-CAM (Suppl. Sec. 5) only demonstrates higher spatial resolution, not that either method is correct. The comparison with average FC and with linear-regression z-scores (Figs. 2e–f) shows that saliency differs from those maps, not that it is more valid. For regression, the authors keep only positive gradients (Eq. 2) and then apply a ReLU (Eq. 4), so negative-age-direction contributions are discarded; combined with the modified summation fusion, this is an unvalidated variant of a method that itself is only heuristic. Saliency maps can be visually plausible while being uncorrelated with the true features the model relies on (e.g., due to gradient noise or ReLU saturation). Moreover, the anterior/posterior comparison (Sec. 2.4, Fig. 5c) does not control for the different baseline FC distributions of the two seeds, so the differences could reflect known anatomical connectivity rather than age-specific sensitivity. Because all interpretability conclusions are downstream of these maps, this is the most load-bearing weakness.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an interpretable deep-learning pipeline for studying hippocampal aging. A 3D CNN is trained on voxel-wise seed-based functional connectivity maps from left and right hippocampal seeds, pretrained on HCP-Y and fine-tuned on HCP-A with fivefold cross-validation, achieving a validation MAE of 7.30 years and Spearman's rho of 0.79. The authors adapt LayerCAM to 3D regression to produce saliency maps, which they interpret as identifying hippocampal-cortical connections most influential for age prediction, specifically to precuneus, cuneus, PCC, PHC, left SPL, and right STS. They further compare whole-hippocampus, anterior-hippocampus, and posterior-hippocampus models, reporting similar prediction performance but distinct saliency patterns, which they interpret as reflecting known anterior/posterior functional specializations. Secondary analyses address sex differences, comparison with linear-regression age effects, and regional polynomial regressions.","tokens_in":20562,"tokens_out":3426,"duration_ms":40500,"significance":"If the saliency maps are faithful to the model's decision process, the paper offers a voxel-level, interpretable account of hippocampal-cortical connectivity changes in aging, and the anterior/posterior dissociation would be a useful contribution to the aging literature. The predictive performance is on par with or better than several region-based resting-state fMRI models, and the explicit adaptation of LayerCAM to 3D regression is a useful methodological step. The paper benefits from standard fivefold cross-validation, pretraining/fine-tuning, and comparison against several 3D CNN baselines. However, the central neurobiological claims rest entirely on saliency maps that are never validated as faithful explanations, and the performance claims lack fold-wise variance and may be affected by model selection on the same dataset. The significance is therefore conditional on additional validation experiments.","major_comments":[{"comment":"The central interpretability claim—that the highlighted regions are 'most influential' and 'age-sensitive'—rests on LayerCAM saliency maps (Eqs. 2-6) that are never validated as faithful explanations of the trained 3D CNN. No occlusion, perturbation, label-shuffling, or weight-randomization experiment is reported to show that these maps track the model's actual input-output dependence. The comparison with Grad-CAM (Suppl. Sec. 5) only shows higher spatial resolution, not correctness; the comparison with average FC and linear-regression z-scores (Figs. 2e-f) shows that saliency is different, not that it is more valid. Please add quantitative sanity checks, such as measuring the change in predicted age after occluding highlighted versus non-highlighted regions, or comparing saliency maps from a model trained on shuffled ages to rule out dataset-specific artifacts.","section":"Section 2.2 and Section 5.5"},{"comment":"The adaptation of LayerCAM to regression keeps only positive gradients (Eq. 2) and applies a ReLU to the combined map (Eq. 4), thereby discarding all negative-age-direction contributions. For a regression task, a voxel whose hippocampal FC decreases with age can produce negative gradients, and truncating them biases the saliency map toward positive-direction effects. The claim that the highlighted regions are 'most influential' is therefore incomplete; please justify the sign truncation or compare the reported maps with full-gradient or integrated-gradient alternatives to show that the discarded information is nonessential.","section":"Section 5.5, Eqs. 2 and 4"},{"comment":"The anterior versus posterior saliency comparison (Fig. 5c) does not control for the different baseline FC distributions of the two seed regions. The anterior and posterior hippocampus have known anatomical connectivity differences (ref. [18]), so the observed saliency differences may simply mirror these baseline connectivity differences rather than age-specific sensitivity. A control analysis, such as matching the FC-strength distributions or comparing the saliency differences against a null distribution obtained from label-shuffled models, is needed before interpreting Fig. 5c as evidence of 'distinct functional trajectories in normal aging.'","section":"Section 2.4 and Section 5.8"},{"comment":"The reported validation MAE of 7.30 years may be optimistically biased because the architecture was selected on the same HCP-A dataset via fivefold cross-validation (Suppl. Table 1), and no fold-wise variance or confidence intervals are reported. Please provide per-fold MAE values and either use nested cross-validation or clearly state the model-selection procedure so readers can gauge the selection-induced optimism.","section":"Section 2.1 and Suppl. Sec. 3"}],"minor_comments":[{"comment":"The phrase 'remain s limited' in the abstract contains a typo and should read 'remains limited.'","section":"Abstract"},{"comment":"The text says 'excluding subjects aged 100-year-old,' which is inconsistent with Section 5.1's description of excluding 12 subjects aged above 90 because detailed age information was not provided; please reconcile the age cutoff.","section":"Section 5.6"},{"comment":"The caption states that 'MAE values (in years) were denoted on the right side of each bar,' but these values are not visible in the text version; please ensure the figure actually displays the numerical MAE values.","section":"Fig. 3 caption"},{"comment":"The phrase 'one-way paired t-tests' should be 'one-sided paired t-tests' or 'one-tailed paired t-tests' to correctly describe the statistical test used.","section":"Sections 2.4, 5.7, 5.8"},{"comment":"The text mentions 'left superior temporal sulcus (right STS)' and 'right superior parietal lobule (right SPL),' which contradict the main-text naming of 'left SPL and right STS'; please correct these labels.","section":"Suppl. Sec. 6"},{"comment":"Supplementary table numbering is inconsistent (a 'Suppl. Table1' is referenced for model selection in Sec. 3, and the regression results are labeled both 'Suppl. Table1' and 'Suppl. Table2'); please harmonize all supplementary table and figure references.","section":"Suppl. Sec. 6"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the journal's scope and the predictive results are plausible. The main risk is overinterpretation of unvalidated saliency maps; this is addressable with relatively standard sanity checks, so I do not recommend rejection. I would also encourage the editors to ask the authors to release code and per-fold results, as neither is currently stated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper does something genuinely new: it applies a 3D CNN to voxel-wise seed-based hippocampal FC maps and uses LayerCAM to localize age-sensitive hippocampal-cortical connections. That combination, plus the anterior/posterior split, is new, and the authors do it cleanly. They use two public datasets (HCP-Y for pretraining, HCP-A for finetuning), fivefold CV, and they compare their saliency maps against average FC and linear regression, and their architecture against several standard CNNs. That is more than most deep-learning neuroimaging papers do. The writing is clear, and the supplementary material is thorough.\n\nThe main finding is plausible: the model predicts age with MAE 7.30 years on HCP-A, and the saliency maps highlight precuneus, PCC, PHC, cuneus, left SPL, right STS, consistent with prior literature. The anterior/posterior differences also line up with known functional specialization.\n\nThe soft spots are real but not fatal. The saliency maps are never validated as faithful explanations. The stress-test note is correct: no occlusion, perturbation, or alternative explainer test is reported. LayerCAM's positive-gradient ReLU and the modified fusion are heuristic, and the comparison with Grad-CAM only shows higher resolution, not correctness. The paper even acknowledges low R^2 in the regional regressions, which undermines the claim that these regions are nonlinearly informative. Also, they don't report fold-wise variance or error bars, and the architecture was selected on the same HCP-A data (though compared without pretraining, which is a partial safeguard). The anterior/posterior comparison doesn't control for baseline FC differences between the two seeds.\n\nThe good news is that all of these are fixable. A few occlusion experiments, a perturbation test, and reporting fold-by-fold MAE would substantially raise confidence. I would not reject the paper on the current evidence; it deserves a serious referee. The interpretability claims should be treated as hypotheses, not conclusions, and the discussion should be toned down accordingly.\n\nFor whom: readers interested in explainable deep learning for fMRI or hippocampal aging. It's a useful example, not a game-changer. I'd bring it to a reading group to discuss what counts as validation of explanation methods.\n\nMy recommendation: send it to peer review, with a request for validation of the saliency maps before acceptance. It's a solid contribution that needs a modest amount of additional evidence.","headline":"A solid application of 3D CNN plus LayerCAM to hippocampal aging FC, with real novelty in the anterior/posterior split, though the saliency maps need validation before their specific regions are treated as findings.","tokens_in":21114,"tokens_out":2233,"would_cite":false,"duration_ms":25952,"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":"Hippocampal connectivity lets a 3D CNN predict brain age within 7.3 years, and saliency maps reveal which connections matter most.","keywords":["brain age prediction","hippocampal functional connectivity","3D convolutional neural network","LayerCAM","saliency mapping","seed-based functional connectivity","anterior-posterior hippocampus","healthy aging"],"falsifier":"Replace the hippocampal FC values in the six highlighted regions with their mean (an occlusion test) and measure the change in predicted age; if the model's prediction is insensitive to those regions, the saliency maps do not reflect the model's true dependence. A second check would be to train the same pipeline on an independent aging sample and see whether the same cortical regions survive in the averaged saliency maps.","tokens_in":20049,"feed_emoji":"🧠","tokens_out":8490,"duration_ms":83573,"temperature":0.7,"pith_summary":"The paper tests whether the functional connections between the hippocampus and the rest of the brain contain enough information about healthy aging to predict a person's chronological age. Using voxel-wise hippocampal-cortical connectivity maps from 708 adults aged 36–90, the authors train a three-dimensional convolutional neural network (3D CNN) and report a validation mean absolute error of 7.30 years and a Spearman correlation of 0.79. To make the model interpretable, they adapt LayerCAM to produce voxel-level saliency maps, which highlight hippocampal connections with the precuneus, cuneus, posterior cingulate cortex, parahippocampal cortex, left superior parietal lobule, and right superior temporal sulcus as the most influential for age prediction. They further show that splitting the hippocampus into anterior and posterior parts gives comparable prediction accuracy but distinctly different saliency patterns, aligned with the known functional specializations of those subregions. If these results hold, they offer a fine-grained, interpretable view of how hippocampal connectivity reorganizes with age and a possible functional marker for age-related cognitive decline.","feed_headline":"Hippocampal wiring predicts brain age within 7.3 years","feed_subtitle":"Saliency maps show which hippocampal connections carry the aging signal.","key_machinery":"The machinery is a combination of three pieces. First, seed-based functional connectivity: the time series of the left and right hippocampus (or their anterior/posterior halves) are averaged and correlated with every brain voxel, producing a pair of whole-brain 3D correlation volumes per subject. Second, a five-block 3D CNN with alternating 3×3×3 and 5×5×5 convolution kernels, batch normalization, ReLU, and max pooling, trained with mean absolute error as the loss for the regression task of predicting age. Third, LayerCAM, a class-activation-mapping variant that fuses positively weighted activation maps from all convolutional layers to yield high-resolution voxel-level saliency maps; the paper adapts it from 2D to 3D and from classification to regression. The saliency maps are the central interpretative device: they are averaged across subjects and used to identify which hippocampal–cortical connections are age-sensitive.","core_discovery":"The central claim is that seed-based hippocampal functional connectivity, when fed to a 3D CNN as a regression target, can predict chronological age with an error comparable to region-based connectivity models, and that the model's LayerCAM saliency maps identify specific hippocampal–cortical connections whose age sensitivity is not captured by linear regression. The whole-hippocampus model achieves a validation MAE of 7.30 years and Spearman rho 0.79 on the HCP-A cohort after pretraining on young adults and fivefold cross-validation fine-tuning. The averaged saliency map emphasizes the precuneus, cuneus, posterior cingulate cortex, parahippocampal cortex, left superior parietal lobule, and right superior temporal sulcus. The paper also claims that anterior and posterior hippocampal FC produce statistically indistinguishable prediction errors but significantly different saliency patterns, with anterior hippocampus more influential through prefrontal, anterior cingulate, entorhinal, and temporal pole connections, and posterior hippocampus more influential through precuneus, inferior parietal, and superior temporal connections. The authors interpret this as evidence that the model has captured functionally distinct aging trajectories along the hippocampal long axis.","pith_inferences":["If the saliency maps truly reflect age-sensitive connectivity, the same pipeline could be applied to clinical cohorts (e.g., mild cognitive impairment or Alzheimer's disease) to see whether the highlighted hippocampal–cortical links are accelerated or redirected in pathological aging; this is an extension the paper does not test.","Because the reported 7.30-year MAE comes from cross-validation within a single dataset, an independent external cohort test would be a stronger check of generalization; the paper's comparisons to other studies use different datasets, so the numbers are not directly comparable.","The anterior/posterior dissociation suggests a testable prediction: if anterior hippocampus supports emotion and memory encoding and posterior supports spatial navigation, then the relative saliency of the two maps should predict which cognitive domain declines first in longitudinal data.","The nonlinear, spatially distributed features the CNN exploits (e.g., in precuneus and PCC, where linear regression finds almost nothing) could be extracted as a feature vector and correlated with known biomarkers of aging, such as hippocampal volume or amyloid burden, as a way to validate the saliency interpretation."],"forward_implications":["The age signal in hippocampal–cortical FC is strong enough to predict chronological age within about 7.3 years, a performance comparable to region-based FC methods while preserving voxel-level spatial detail.","The six highlighted regions (precuneus, cuneus, PCC, PHC, left SPL, right STS) are candidate functional markers of healthy aging; their hippocampal FC declines with age and fits second- or third-order polynomial regressions better than linear ones.","Anterior and posterior hippocampal FC contribute to age prediction through different cortical targets, so analyzing the hippocampus as a single seed conceals subregion-specific aging effects.","The model is most accurate for 55–60-year-olds (MAE 4.69 years), suggesting the hippocampal FC age signal is strongest in middle age and less detectable in younger and older extremes.","Saliency maps are largely stable across sex and age groups, but subtle regional differences (e.g., precuneus and IPL more salient in females) hint that the model picks up sex-specific aging nuances without being trained on sex."],"supporting_citations":[{"why":"Supplies the LayerCAM method that the paper adapts to 3D and to regression to generate voxel-wise saliency maps.","marker":"[17]"},{"why":"Provides the HCP-Y young-adult dataset used for pretraining the model.","marker":"[23]"},{"why":"Provides the HCP-A aging dataset used for fine-tuning, validation, and all main age-prediction and saliency results.","marker":"[24]"},{"why":"Establishes that anterior and posterior hippocampal FC age differently, the prior finding the paper extends with subregion-specific saliency analysis.","marker":"[8]"},{"why":"Provides the anatomical anterior–posterior hippocampal connectivity basis for splitting the hippocampus into two seeds.","marker":"[18]"},{"why":"Documents differential functional connectivity of anterior vs posterior hippocampus with cortex, used to interpret the saliency differences.","marker":"[59]"},{"why":"Provides a lightweight CNN brain-age architecture used as a comparison baseline in model selection.","marker":"[13]"},{"why":"Defines the minimal preprocessing pipeline whose output is used to compute the seed-based FC inputs.","marker":"[74]"}],"fun_headline_variants":["Hippocampal functional connectivity predicts brain age to 7.3 years","Saliency mapping highlights age-sensitive hippocampal-cortical links","Anterior and posterior hippocampus age via distinct connectivity patterns","Explainable 3D CNN uncovers hippocampal aging signatures","Hippocampal wiring age signal focused in precuneus and cingulate"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's biological conclusions depend on the assumption that the subject-averaged LayerCAM saliency maps faithfully reflect which hippocampal–cortical connections actually drive aging, rather than dataset-specific confounds or model artifacts.","fun_headline_variants_meta":{"raw":{"variants":["Hippocampal functional connectivity predicts brain age to 7.3 years","Saliency mapping highlights age-sensitive hippocampal-cortical links","Anterior and posterior hippocampus age via distinct connectivity patterns","Explainable 3D CNN uncovers hippocampal aging signatures","Hippocampal wiring age signal focused in precuneus and cingulate"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000772,"raw_usage":{"total_tokens":3421,"prompt_tokens":951,"completion_tokens":2470,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":567,"completion_tokens_details":{"reasoning_tokens":2382}},"tokens_in":567,"tokens_out":2470,"duration_ms":19523,"temperature":1.0,"reasoning_tokens":2382,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:51:35.170547+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Replace the hippocampal FC values in the six highlighted regions with their mean (an occlusion test) and measure the change in predicted age; if the model's prediction is insensitive to those regions, the saliency maps do not reflect the model's true dependence. A second check would be to train the same pipeline on an independent aging sample and see whether the same cortical regions survive in the averaged saliency maps.","supporting_citations":[{"cited_title":"Differential effect of age on posterior and anterior hippocampal functional connectivity,","cited_arxiv_id":null,"evidence_quote":"Establishes that anterior and posterior hippocampal FC age differently, the prior finding the paper extends with subregion-specific saliency analysis."}],"review_version":1}