{"id":"784b14ff-502a-4f8b-876b-3b4fbdc696f9","arxiv_id":"2605.02230","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":1,"one_line_summary":"InfiltrNet fuses CNN and Swin Transformer encoders via cross-attention to predict infiltration risk zones from BraTS MRI data using distance-transform labels and outperforms five baselines.","lead":"The paper introduces InfiltrNet, a dual-branch CNN and Swin Transformer model that generates three-zone infiltration risk maps from multimodal MRI for gliomas. Clinicians and researchers working on brain tumor treatment planning may find it relevant for estimating tumor spread beyond visible MRI margins.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Distance-transform labels from BraTS annotations are unvalidated proxies whose match to biological infiltration is unproven.","rationale":"The reader already isolated the identical weakest assumption. No additional internal inconsistency (e.g., loss-function derivation or architectural flaw) is visible from the provided abstract that would supersede this label-validity issue.","tokens_in":1665,"tokens_out":277,"duration_ms":16219,"concrete_test":"On any BraTS cases that also possess post-treatment follow-up MRI showing recurrence locations outside the original enhancing tumor, compute the spatial overlap (Dice or Hausdorff) between the high-risk zone predicted by InfiltrNet versus each baseline and the actual recurrence mask; if the reported advantage disappears or loses significance, the label-generation step fails to capture biological infiltration.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline claim is that InfiltrNet outperforms baselines at predicting three-zone infiltration risk maps. This claim presupposes that the distance-transform procedure applied to standard BraTS visible-tumor annotations yields zones that meaningfully represent true infiltration risk. The abstract states only that the strategy is “reproducible”; it supplies no correlation with histology, biopsy margins, or longitudinal recurrence data. If the generated labels are primarily geometric rather than biologically grounded, then superior Dice or boundary metrics on those labels do not establish improved infiltration-risk prediction.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper introduces InfiltrNet, a dual-branch CNN-Transformer architecture combining a CNN encoder with a Swin Transformer encoder via cross-attention fusion modules to predict three-zone infiltration risk maps from multimodal MRI. It proposes a reproducible label generation strategy using distance transforms on standard BraTS annotations, trains with combined Dice-CrossEntropy and boundary-aware losses plus auxiliary supervision, and claims outperformance over five baselines on BraTS 2020 and BraTS 2025, with supporting GradCAM++ and occlusion explainability analysis.","tokens_in":1772,"tokens_out":482,"duration_ms":32890,"significance":"If the distance-transform labels prove biologically meaningful, the work could advance beyond visible-tumor segmentation toward clinically actionable infiltration-risk maps for surgical and radiation planning. The dual-branch fusion and auxiliary heads are incremental but the reproducible proxy-label strategy and public-data experiments are positive contributions; however, the unvalidated nature of the labels substantially limits the potential impact even if empirical gains hold.","major_comments":[{"comment":"Abstract / label generation paragraph: The headline claim that InfiltrNet outperforms baselines at predicting infiltration risk presupposes that distance-transform zones derived from BraTS visible-tumor annotations meaningfully represent true biological infiltration. The text states only that the strategy is 'reproducible' and supplies no correlation with histology, biopsy margins, or longitudinal recurrence data; if the labels are primarily geometric, superior Dice or boundary metrics on them do not establish improved infiltration-risk prediction.","section":"Abstract / label generation paragraph"},{"comment":"Abstract: The claim of outperformance over five established baselines is stated without any quantitative metrics, error bars, statistical tests, or baseline implementation details, rendering the central empirical result unverifiable from the provided text and undermining assessment of the architecture's contribution.","section":"Abstract"}],"minor_comments":[{"comment":"Clarify whether BraTS 2025 refers to a released dataset or a projected one, and ensure all dataset versions and preprocessing steps are fully specified for reproducibility.","section":"Abstract"},{"comment":"The description of the cross-attention fusion modules would benefit from an explicit equation or diagram in the methods section to aid implementation.","section":null}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the detailed and constructive review. We address each major comment below with honest responses based on the current manuscript content.","responses":[{"response":"We agree that the distance-transform labels function as a reproducible geometric proxy derived from standard BraTS annotations rather than direct biological measurements. The manuscript positions this as an initial step toward infiltration-risk mapping using publicly available data, with the model demonstrating improved prediction of these zones. We will revise the abstract, introduction, and a new limitations paragraph to explicitly describe the labels as a proxy strategy, clarify that no histological or recurrence correlation is provided, and temper the claims accordingly while retaining the reproducibility contribution.","revision_made":"partial","referee_comment":"[Abstract / label generation paragraph] Abstract / label generation paragraph: The headline claim that InfiltrNet outperforms baselines at predicting infiltration risk presupposes that distance-transform zones derived from BraTS visible-tumor annotations meaningfully represent true biological infiltration. The text states only that the strategy is 'reproducible' and supplies no correlation with histology, biopsy margins, or longitudinal recurrence data; if the labels are primarily geometric, superior Dice or boundary metrics on them do not establish improved infiltration-risk prediction."},{"response":"The abstract summarizes the outperformance without specific numbers due to length constraints, while the results section contains the full quantitative comparisons, error bars, and implementation details for the five baselines. We will revise the abstract to include key aggregate metrics (e.g., mean Dice improvements) and a brief note on statistical testing to make the central claim verifiable from the abstract alone.","revision_made":"yes","referee_comment":"[Abstract] Abstract: The claim of outperformance over five established baselines is stated without any quantitative metrics, error bars, statistical tests, or baseline implementation details, rendering the central empirical result unverifiable from the provided text and undermining assessment of the architecture's contribution."}],"tokens_in":1350,"tokens_out":446,"duration_ms":24785,"standing_objections":["Direct correlation of the distance-transform labels with histology, biopsy margins, or longitudinal recurrence data is absent from the manuscript and cannot be supplied without new experiments or external datasets."]},"desk_editor":{"model":"grok-4.3","letter":"The main thing to know is that the paper puts forward a dual-branch CNN-Transformer architecture with cross-attention fusion to predict three-zone infiltration risk maps, yet the labels come from distance transforms on standard BraTS annotations without evidence they match actual infiltration biology.\n\nThe new pieces are the specific combination of CNN and Swin Transformer encoders joined by cross-attention modules, plus the distance-transform procedure to generate the three risk zones. Those elements are not present in the cited prior work. The design also includes auxiliary supervision heads and a combined Dice-CrossEntropy plus boundary loss, which are reasonable choices for this task.\n\nThe paper targets a genuine clinical need in glioma planning, where knowing peritumoral risk matters for surgery and radiotherapy. The explainability analysis with GradCAM++ and occlusion sensitivity is a straightforward addition that could help readers see what the model uses.\n\nThe soft spots are clear. The abstract states that InfiltrNet outperforms five baselines on BraTS 2020 and 2025, but it gives no Dice scores, no error bars, no statistical tests, and no implementation details on the baselines. That leaves the central performance claim uncheckable from the text. More importantly, the label strategy is described only as reproducible; the paper supplies no histology correlation, biopsy margin data, or longitudinal recurrence evidence to show the zones reflect true infiltration rather than simple geometry. The stress-test concern holds: if the labels are mainly distance-based artifacts, then good metrics on them do not prove better risk prediction. The rest of the setup uses public data and standard losses, so there is no circularity or other obvious flaw.\n\nThis work is for researchers in medical image analysis who focus on brain tumor segmentation extensions. A reader looking for architecture variants or label-generation ideas could extract value from the fusion modules and distance approach. It deserves peer review because the architecture is concrete and the clinical motivation is real; a referee can request the missing numbers and any external label checks. I would send it out rather than desk reject.","headline":"InfiltrNet proposes a dual-branch CNN-Swin Transformer with cross-attention for three-zone infiltration maps from BraTS data, but the distance-transform labels lack any biological validation and the abstract shows no metrics.","tokens_in":2242,"tokens_out":493,"would_cite":false,"duration_ms":27967,"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":"InfiltrNet, a dual-branch CNN-Transformer, predicts three-zone infiltration risk maps from multimodal MRI and outperforms five baselines on BraTS 2020 and BraTS 2025.","keywords":["brain tumor infiltration","infiltration risk prediction","CNN-Transformer","BraTS dataset","multimodal MRI","dual-branch architecture","distance transform labels","glioma segmentation"],"falsifier":"Direct comparison of the model's three-zone risk maps against histopathological sampling of peritumoral tissue from the same patients would show mismatch if the predicted zones do not align with actual infiltration patterns.","tokens_in":2567,"feed_emoji":"🧠","tokens_out":653,"duration_ms":26278,"temperature":0.7,"pith_summary":"Existing deep learning methods segment only the visible tumor on MRI but leave infiltration risk in surrounding tissue unaddressed. InfiltrNet combines a CNN encoder and a Swin Transformer encoder via cross-attention fusion to output three-zone risk maps. Labels are created by applying distance transforms to standard BraTS annotations. Training employs a combined Dice-CrossEntropy and boundary-aware loss with auxiliary heads. On BraTS 2020 and 2025 the model exceeds five baselines, and GradCAM++ plus occlusion analysis shows attention on peritumoral regions.","feed_headline":"Dual-branch model outperforms baselines in predicting tumor infiltration risk","feed_subtitle":"InfiltrNet fuses CNN and Swin Transformer features to output three-zone risk maps from standard BraTS annotations on two datasets.","key_machinery":"Dual-branch architecture that merges a CNN encoder with a Swin Transformer encoder through cross-attention fusion modules, paired with a distance-transform label generation method.","core_discovery":"InfiltrNet outperforms five established baselines on BraTS 2020 and BraTS 2025 for predicting three-zone infiltration risk maps from multimodal MRI. The architecture fuses CNN and Swin Transformer features through cross-attention, uses distance-transform labels derived from standard annotations, and trains with a composite loss plus auxiliary supervision. Explainability maps confirm focus on clinically relevant peritumoral tissue.","pith_inferences":["The same dual-branch fusion could be tested on other infiltrative tumors such as glioblastoma variants or metastatic lesions where margin definition is similarly uncertain.","Integration with intraoperative ultrasound or fluorescence imaging might allow real-time update of the risk maps during surgery.","If the distance-transform assumption holds across scanners, the approach could reduce the need for specialized infiltration-specific training data."],"forward_implications":["Risk maps can inform resection margins and radiation field design beyond visible tumor boundaries.","The label generation method allows any existing BraTS-annotated dataset to be reused for infiltration studies without new manual labeling.","Auxiliary supervision heads improve boundary accuracy in the final decoder output.","GradCAM++ and occlusion results indicate the model learns clinically relevant peritumoral features rather than spurious image artifacts."],"fun_headline_variants":["InfiltrNet outperforms baselines in three-zone infiltration risk prediction","Dual-branch architecture fuses CNN and Swin Transformer for MRI risk maps","InfiltrNet generates infiltration risk maps from BraTS annotations","Model predicts tumor infiltration beyond visible margins on two BraTS datasets","Explainability confirms focus on peritumoral tissue in infiltration prediction"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The distance-transform strategy applied to standard BraTS annotations produces labels that accurately represent true biological infiltration risk zones.","fun_headline_variants_meta":{"raw":{"variants":["InfiltrNet outperforms baselines in three-zone infiltration risk prediction","Dual-branch architecture fuses CNN and Swin Transformer for MRI risk maps","InfiltrNet generates infiltration risk maps from BraTS annotations","Model predicts tumor infiltration beyond visible margins on two BraTS datasets","Explainability confirms focus on peritumoral tissue in infiltration prediction"]},"model":"grok-4.3","cost_usd":0.007856,"raw_usage":{"total_tokens":3569,"prompt_tokens":638,"num_sources_used":0,"completion_tokens":83,"cost_in_usd_ticks":78562000,"prompt_tokens_details":{"text_tokens":638,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2848,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":638,"tokens_out":83,"duration_ms":27522,"temperature":1.0,"reasoning_tokens":2848,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-01T00:52:52.057044+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Direct comparison of the model's three-zone risk maps against histopathological sampling of peritumoral tissue from the same patients would show mismatch if the predicted zones do not align with actual infiltration patterns.","supporting_citations":[],"review_version":2}