{"id":"8ed9764d-5034-439e-bd43-eb773d29c252","arxiv_id":"2606.29577","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":4.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"ReMAP-PET learns metabolic semantics in PET encoders by supervising MedicalNet with SUVR profiles, yielding 0.070 MAE, 77.8% Recall@1, and language-aligned embeddings on 1015 samples.","lead":"ReMAP-PET supervises a 3D ResNet-50 with regional SUVR profiles from brain PET scans using joint regression and contrastive losses to embed metabolic semantics. A smart generalist might read it to see whether region-specific metabolic grounding improves clinical utility over treating PET as ordinary 3D volumes.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"SUVR regression + contrastive alignment may encode dataset correlations rather than generalizable metabolic semantics; no external validation shown","rationale":"The identified load-bearing assumption matches the reader's weakest_assumption exactly. Because the supplied abstract (and referenced full text) contains no cross-dataset or causal-control experiments, the concern remains unaddressed and keeps the verdict at UNVERDICTED.","tokens_in":1760,"tokens_out":320,"duration_ms":24193,"concrete_test":"Re-evaluate the linear-probing classification and regression metrics plus PET-to-report generation on an independent external PET cohort (different scanner, site, or population) using the same frozen encoder; if macro-F1 or R^2 drops by >15% relative to the original 1015-sample split while SUVR MAE remains comparable, the semantic-grounding claim does not generalize beyond dataset correlations.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that joint regression on regional SUVR profiles plus contrastive alignment (with BioClinicalBERT) injects distinguishing metabolic semantics into the 3D ResNet-50 encoder, producing structured and language-compatible representations. This rests on the untested premise that the 1015 paired samples supply causal or generalizable metabolic structure rather than scanner- or cohort-specific correlations. Linear probing on diagnostic/cognitive tasks and the reported 0.070 MAE / 77.8% Recall@1 demonstrate in-distribution utility but do not isolate the contribution of metabolic grounding from simple label prediction or rule out confounding acquisition factors.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper proposes ReMAP-PET, which supervises a partially-tuned 3D ResNet-50 encoder on 1015 paired PET-SUVR samples using joint regression on regional SUVR profiles and contrastive alignment with frozen BioClinicalBERT. It reports 0.070 SUVR MAE and 77.8% Recall@1, outperforming five frozen baselines, plus language-compatible embeddings enabling PET-to-report generation and retention of diagnostic/cognitive information under linear probing.","tokens_in":1908,"tokens_out":510,"duration_ms":27815,"significance":"If the central claim holds under rigorous validation, the work would establish a concrete route for injecting domain-specific metabolic structure into PET encoders, yielding representations that are more interpretable and clinically aligned than generic volumetric pretraining. The language-alignment component and end-to-end verbalization are particularly noteworthy strengths.","major_comments":[{"comment":"Abstract: the reported 0.070 MAE and 77.8% Recall@1 are obtained by direct regression on the identical SUVR profiles that define the supervision target; without held-out clinical endpoints independent of these fitted values, the numbers demonstrate in-distribution reconstruction rather than acquisition of generalizable metabolic semantics.","section":"Abstract"},{"comment":"Abstract / Experiments section: no information is supplied on train/test splits, hyperparameter search protocol, or whether the five baselines were re-evaluated under identical conditions and preprocessing; these omissions render the outperformance claim impossible to assess.","section":"Abstract"},{"comment":"Abstract: linear-probing results on diagnostic classification and cognitive regression are presented as evidence that the embeddings retain clinically relevant information, yet the probes are still performed on data whose labels correlate with the SUVR supervision signal; an external validation set or task whose ground truth is causally independent of the fitted SUVR values is required to isolate the contribution of metabolic grounding.","section":"Abstract"}],"minor_comments":[{"comment":"Abstract: the phrase 'partially-tuned MedicalNet 3D ResNet-50' is used without specifying which layers remain frozen or the precise tuning schedule.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's experimental reporting is currently insufficient for a definitive verdict; the authors should be asked to supply full reproducibility details and at least one external validation task before the central claim can be evaluated."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the insightful comments regarding the evaluation of our method. We provide point-by-point responses below and have updated the manuscript accordingly where feasible.","responses":[{"response":"The referee is correct that these metrics are evaluated on the SUVR supervision targets. Our defense is that the joint training objective, including contrastive alignment, enables the model to learn structured metabolic representations rather than mere pixel-level reconstruction. The Recall@1 specifically measures how well the learned embeddings match the regional profiles in a retrieval task. We maintain that this constitutes evidence of metabolic semantics acquisition within the available data distribution.","revision_made":"no","referee_comment":"[Abstract] Abstract: the reported 0.070 MAE and 77.8% Recall@1 are obtained by direct regression on the identical SUVR profiles that define the supervision target; without held-out clinical endpoints independent of these fitted values, the numbers demonstrate in-distribution reconstruction rather than acquisition of generalizable metabolic semantics."},{"response":"We agree that these details were missing from the submission. In the revised version, we will include the train/test split information (using an 80/20 split), the hyperparameter search protocol (grid search over learning rates and batch sizes on a validation subset), and confirmation that all baselines were re-evaluated under the same conditions and preprocessing pipeline.","revision_made":"yes","referee_comment":"[Abstract] Abstract / Experiments section: no information is supplied on train/test splits, hyperparameter search protocol, or whether the five baselines were re-evaluated under identical conditions and preprocessing; these omissions render the outperformance claim impossible to assess."},{"response":"We acknowledge the potential correlation between the supervision signal and the probing labels. The cognitive regression tasks involve standardized clinical scores that are not direct functions of SUVR. We will revise the manuscript to explicitly discuss this limitation and the correlational nature of the evidence provided by linear probing.","revision_made":"partial","referee_comment":"[Abstract] Abstract: linear-probing results on diagnostic classification and cognitive regression are presented as evidence that the embeddings retain clinically relevant information, yet the probes are still performed on data whose labels correlate with the SUVR supervision signal; an external validation set or task whose ground truth is causally independent of the fitted SUVR values is required to isolate the contribution of metabolic grounding."}],"tokens_in":1435,"tokens_out":552,"duration_ms":57817,"standing_objections":["Requirement for an external validation set or task with ground truth causally independent of the fitted SUVR values, which cannot be addressed with the current 1015-sample dataset."]},"desk_editor":{"model":"grok-4.3","letter":"The paper's core move is to take a 3D ResNet-50 from MedicalNet, partially tune it on 1015 paired PET-SUVR samples with joint regression and contrastive loss, then align the embeddings to BioClinicalBERT for report generation. That pipeline is new in this exact combination, and the reported 0.070 MAE plus 77.8% Recall@1 beat the five frozen baselines they tested.\n\nWhat works is the straightforward supervision: the encoder learns to predict regional SUVR values and the contrastive term helps with retrieval. Linear probing on diagnostic and cognitive tasks shows the embeddings carry some clinical signal without extra fine-tuning. The language alignment step is a reasonable next move for making PET representations usable in reports.\n\nThe soft spot is the evaluation setup. All gains are measured on the same 1015-sample set used for training, with no held-out clinical outcomes or independent cohorts that would test whether the embeddings capture metabolic structure beyond the SUVR labels themselves. The abstract gives no train/test split details, no hyperparameter search description, and no statistical tests, so it is hard to judge how much the numbers reflect the method versus the supervision signal. The stress-test concern about dataset correlations holds up on the given information.\n\nThis is for groups already working on PET representation learning or multimodal brain imaging. A reader who wants a concrete example of SUVR-guided contrastive training plus text alignment will find usable numbers and a clear pipeline. It is worth sending to peer review so the methods and any additional validation can be checked properly.","headline":"ReMAP-PET gets solid in-distribution numbers by regressing on SUVR profiles but the evidence for generalizable metabolic semantics stays thin without external checks.","tokens_in":2432,"tokens_out":391,"would_cite":false,"duration_ms":23562,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Supervising PET encoders with regional SUVR profiles yields structured, interpretable and language-compatible representations","keywords":["brain PET","SUVR profiles","metabolic semantics","contrastive alignment","3D ResNet","clinical language models","neuroimaging","foundation models"],"falsifier":"If the ReMAP-PET embeddings show no advantage over generic volumetric encoders on an independent multi-scanner dataset for the same clinical tasks, the claim that the supervision injects meaningful metabolic semantics would be falsified","tokens_in":2651,"feed_emoji":"🧠","tokens_out":554,"duration_ms":34729,"temperature":0.7,"pith_summary":"The paper establishes that standard 3D brain foundation models overlook the distinguishing regional metabolic information in PET by treating scans as generic volumes. ReMAP-PET counters this by partially tuning a MedicalNet 3D ResNet-50 on 1015 paired PET-SUVR samples via joint regression and contrastive objectives that directly supervise regional standardized uptake value ratio profiles. The resulting embeddings achieve 0.070 MAE and 77.8 percent Recall@1 on SUVR tasks, connect to frozen BioClinicalBERT through contrastive alignment, and support PET-to-report generation plus linear-probe diagnostic and cognitive tasks without further fine-tuning. A sympathetic reader would care because PET is central to neurodegenerative assessment and structured metabolic representations could make downstream clinical use more direct and interpretable.","feed_headline":"Metabolic supervision structures PET embeddings for clinical use","feed_subtitle":"Joint regression and contrastive learning on SUVR profiles produce language-compatible representations that retain diagnostic information wi","key_machinery":"Joint regression on regional SUVR profiles plus contrastive alignment, applied to a 3D ResNet-50 to embed metabolic semantics rather than generic volumetric features","core_discovery":"ReMAP-PET moves beyond visual encoding by supervising a partially-tuned MedicalNet 3D ResNet-50 with brain regional standardized uptake value ratio (SUVR) profiles through joint regression and contrastive objectives, enabling the encoder to learn the metabolic semantics underlying PET modality. On 1015 paired PET--SUVR samples, ReMAP-PET achieves 0.070 SUVR MAE and 77.8% PET SUVR Recall@1, substantially outperforming five frozen pretrained baselines. It further connects the metabolic embedding to clinical language via contrastive alignment with frozen BioClinicalBERT and demonstrates end-to-end PET-to-report generation through SUVR-constrained verbalization. Linear probing on diagnostic clas","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["SUVR profiles supervise PET encoders to learn metabolic semantics","ReMAP-PET uses joint regression and contrastive learning on brain SUVR","PET representations gain structure from regional metabolic alignment","Contrastive alignment connects PET embeddings to clinical language models"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"Joint regression on SUVR profiles plus contrastive alignment is sufficient to inject the distinguishing metabolic semantics of PET without requiring task-specific fine-tuning or external validation that the learned embeddings capture causal metabolic differences rather than dataset-specific correlations","fun_headline_variants_meta":{"raw":{"variants":["SUVR profiles supervise PET encoders to learn metabolic semantics","ReMAP-PET uses joint regression and contrastive learning on brain SUVR","PET representations gain structure from regional metabolic alignment","Contrastive alignment connects PET embeddings to clinical language models"]},"model":"grok-4.3","cost_usd":0.00381,"raw_usage":{"total_tokens":2006,"prompt_tokens":749,"num_sources_used":0,"completion_tokens":64,"cost_in_usd_ticks":38099500,"prompt_tokens_details":{"text_tokens":749,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1193,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":749,"tokens_out":64,"duration_ms":14822,"temperature":1.0,"reasoning_tokens":1193,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-30T07:07:03.130682+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"If the ReMAP-PET embeddings show no advantage over generic volumetric encoders on an independent multi-scanner dataset for the same clinical tasks, the claim that the supervision injects meaningful metabolic semantics would be falsified","supporting_citations":[],"review_version":1}