{"id":"4b3fa039-957d-4d6e-9e7d-518774812306","arxiv_id":"2606.00491","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":4.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"RAMP combines anatomically constrained spatial perturbations, CT intensity transforms, and stochastic multi-corruption during training, raising mean corrupted Dice from 0.610 to 0.753 and shrinking the clean-to-corrupted gap from 0.264 to 0.064 on a five-organ benchmark versus nnU-Net.","lead":"The paper introduces RAMP, a training augmentation pipeline that mixes spatial, intensity, and multi-corruption degradations to make CT organ segmentation models more stable under realistic clinical image problems. A smart reader might care because unreliable medical AI can fail silently in hospitals when scans are noisy or low-contrast.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Evaluation uses the same corruption family as training augmentations, so reduced gap shows in-distribution robustness rather than generalization to unseen clinical degradations.","rationale":"The reader's weakest assumption directly identifies the same load-bearing point. Full-text elaboration of the corruption definitions would not remove the concern unless the authors also supply an explicit held-out corruption protocol or real clinical validation set, which the abstract gives no indication of.","tokens_in":1802,"tokens_out":329,"duration_ms":16299,"concrete_test":"Add one held-out corruption family (metal artifacts with Hounsfield values >3000 or 3 mm slice-gap motion blur) to both the five-organ and Abdomen1K test sets at the same severity levels used in the original benchmarks; recompute mean corrupted Dice and robustness gap for RAMP vs. nnU-Net. If the gap reduction shrinks by >50% relative to the reported numbers, the generalization claim does not hold.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The headline result (Dice 0.610→0.753, gap 0.264→0.064 on the five-organ benchmark; similar on Abdomen1K) is measured on test images generated from the identical stochastic multi-corruption composition rules used inside RAMP. Because the paper does not report results on held-out corruption families (e.g., metal streak, motion blur, or scanner-specific ring artifacts absent from the augmentation set), the observed robustness improvement is consistent with memorization of the training corruption distribution rather than acquisition of invariance to the broader heterogeneous conditions claimed in the abstract.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript proposes Robustness via Augmented Multi-corruption Pipeline (RAMP), an augmentation framework combining anatomically constrained spatial perturbations, CT intensity transformations, and stochastic multi-corruption composition to train CT segmentation models. It reports that RAMP yields the strongest corrupted-image performance and smallest clean-to-corrupted robustness gap on two benchmarks, improving mean corrupted Dice from 0.610 to 0.753 (gap reduced from 0.264 to 0.064) on a five-organ noisy evaluation benchmark and from 0.633 to 0.789 (gap from 0.290 to 0.070) on Abdomen1K, relative to an nnU-Net baseline.","tokens_in":1915,"tokens_out":465,"duration_ms":19175,"significance":"If the robustness gains generalize beyond the specific corruptions used in training, the approach could provide a practical pre-deployment method for mitigating segmentation collapse under heterogeneous clinical imaging conditions. The reported numeric improvements in corrupted Dice and robustness gap are substantial and directly address a known deployment limitation of CT segmentation systems.","major_comments":[{"comment":"The evaluation uses test images generated from the identical stochastic multi-corruption composition rules employed inside RAMP training. This measures in-distribution robustness rather than generalization to the broader heterogeneous clinical degradations claimed in the abstract (e.g., metal streak, motion blur, or scanner-specific ring artifacts absent from the augmentation set). Without held-out corruption families, the central claim that RAMP improves reliability under 'clinically plausible image degradation' is not fully supported. (Abstract; evaluation benchmarks description)","section":"Evaluation"},{"comment":"Exact corruption parameters, stochastic composition rules, and any post-hoc selection of corruption strengths are not reported, nor are statistical significance tests for the Dice improvements. These omissions prevent assessment of whether the gains are robust or reproducible. (Methods section)","section":"Methods"}],"minor_comments":[{"comment":"The abstract refers to a 'five-organ noisy evaluation benchmark' without specifying its construction details, relation to public datasets, or how the clean vs. corrupted splits were formed.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive comments, which help strengthen the manuscript. We address each major point below and will revise accordingly to improve clarity, reproducibility, and the scope of the claims.","responses":[{"response":"We agree that the reported benchmarks evaluate robustness to the same corruption families used in training and thus constitute in-distribution evaluation. While the corruptions were selected to reflect common clinical issues (noise, resolution loss, contrast/intensity variation), this does not demonstrate generalization to entirely unseen artifact types. In the revision we will add a new experiment section using held-out corruption families (metal streak artifacts, motion blur, and ring artifacts) that are not part of the RAMP training distribution. We will also revise the abstract and discussion to state more precisely that the gains apply to the clinically motivated degradations included in the augmentation pipeline.","revision_made":"yes","referee_comment":"[Evaluation] The evaluation uses test images generated from the identical stochastic multi-corruption composition rules employed inside RAMP training. This measures in-distribution robustness rather than generalization to the broader heterogeneous clinical degradations claimed in the abstract (e.g., metal streak, motion blur, or scanner-specific ring artifacts absent from the augmentation set). Without held-out corruption families, the central claim that RAMP improves reliability under 'clinically plausible image degradation' is not fully supported. (Abstract; evaluation benchmarks description)"},{"response":"We accept this criticism. The revised manuscript will include a detailed supplementary appendix that specifies all corruption parameters (noise variances, blur kernel sizes, intensity shift ranges, etc.), the exact stochastic composition probabilities and ordering rules, and the procedure used to select corruption strengths. We will also add statistical significance testing (paired Wilcoxon signed-rank tests with Bonferroni correction) for all reported Dice improvements and include the resulting p-values in the main results tables.","revision_made":"yes","referee_comment":"[Methods] Exact corruption parameters, stochastic composition rules, and any post-hoc selection of corruption strengths are not reported, nor are statistical significance tests for the Dice improvements. These omissions prevent assessment of whether the gains are robust or reproducible. (Methods section)"}],"tokens_in":1470,"tokens_out":462,"duration_ms":16787,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"RAMP is a training-time augmentation recipe that mixes anatomically constrained spatial changes, intensity shifts, and stochastic multi-corruption for CT segmentation models. On the reported benchmarks it lifts mean corrupted Dice from 0.610 to 0.753 on the five-organ set and from 0.633 to 0.789 on Abdomen1K, while shrinking the clean-to-corrupted gap by roughly 75 percent relative to nnU-Net. Those numbers are the concrete result.\n\nThe paper does one thing cleanly: it assembles standard augmentation families into a single, named procedure and measures the effect on two fixed benchmarks. That supplies a practical data point for anyone already running nnU-Net-style training who wants a drop-in robustness tweak.\n\nThe soft spot is exactly the one flagged in the stress-test note. The corrupted test images appear to be drawn from the same corruption families and stochastic rules used inside RAMP, so the smaller gap demonstrates robustness inside the training distribution rather than to new degradations such as metal streak, motion blur, or scanner ring artifacts that were never seen during augmentation. The abstract calls the corruptions “clinically plausible,” but without held-out families the claim that the method prepares models for heterogeneous deployment conditions rests on an untested assumption.\n\nMethods details on exact parameter ranges, whether corruption strengths were tuned after seeing test performance, and statistical testing are not visible in the abstract, though the full text presumably supplies them. No new math or derivation is offered; the work is an empirical comparison.\n\nThis paper is for medical-imaging practitioners who need a concrete augmentation recipe for CT segmentation. A reader already working on deployment robustness would extract usable numbers and a pipeline description. It is coherent on its own terms and has enough empirical content to merit peer review, provided referees press for out-of-distribution corruption results.","headline":"RAMP packages existing augmentations into a named pipeline and shows clear in-distribution gains on corrupted CT test sets, but the evaluation does not test generalization to unseen clinical degradations.","tokens_in":2428,"tokens_out":451,"would_cite":false,"duration_ms":15413,"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":"RAMP multi-corruption augmentation narrows the clean-to-corrupted Dice gap in CT segmentation from 0.26 to 0.06.","keywords":["CT segmentation","robustness","data augmentation","medical imaging","image corruption","deep learning","pre-deployment testing","organ segmentation"],"falsifier":"A RAMP-trained model showing large Dice drops on real clinical CT scans that contain degradation types not included in the augmentation set would indicate the central claim does not hold.","tokens_in":2709,"feed_emoji":"🩺","tokens_out":726,"duration_ms":20394,"temperature":0.7,"pith_summary":"The paper introduces RAMP, a training method that applies multiple image degradations such as noise, resolution loss, and artifacts to CT scans while learning organ segmentations. The goal is to prepare models for the inconsistent quality of real clinical scans instead of relying only on clean benchmark data. Readers would care because segmentation reliability matters for consistent automated analysis in radiology where scanner settings and patient factors produce varied image conditions. Results across two benchmarks show higher scores on degraded test images and a much smaller performance drop from clean to corrupted cases.","feed_headline":"Augmentation shrinks CT segmentation robustness gap by 75 percent","feed_subtitle":"RAMP training with realistic corruptions lifts corrupted Dice from 0.61 to 0.75 while narrowing the clean-corrupted difference.","key_machinery":"Robustness via Augmented Multi-corruption Pipeline (RAMP), which uses stochastic composition of spatial, intensity, and artifact corruptions to simulate heterogeneous clinical conditions during training.","core_discovery":"RAMP combines anatomically constrained spatial perturbations, CT intensity transformations, and stochastic multi-corruption composition to expose models to clinically plausible image degradation during training. Across two CT segmentation evaluation settings, RAMP achieved the strongest corrupted-image performance and the smallest clean-to-corrupted robustness gap. In the five-organ noisy evaluation benchmark, RAMP improved mean corrupted Dice from 0.610 to 0.753 and reduced the robustness gap from 0.264 to 0.064 compared with the nnU-Net baseline. In Abdomen1K, RAMP improved mean corrupted Dice from 0.633 to 0.789 and reduced the robustness gap from 0.290 to 0.070.","pith_inferences":["Similar augmentation strategies could be adapted for MRI or ultrasound segmentation tasks where image quality also varies.","Deployed systems might benefit from periodic re-training with site-specific corruption profiles drawn from local scanner data.","Combining RAMP with ensemble or uncertainty methods could further stabilize outputs when facing conditions outside the training corruptions."],"forward_implications":["In the five-organ noisy benchmark, mean corrupted Dice rose from 0.610 to 0.753.","The robustness gap dropped from 0.264 to 0.064 in that setting.","Similar gains occurred in the Abdomen1K dataset with gap reduction from 0.290 to 0.070.","Models avoid severe segmentation collapse under strong degradation even if not topping clean-image scores.","Multi-corruption augmentation serves as a practical pre-deployment reliability strategy for heterogeneous clinical environments."],"fun_headline_variants":["RAMP cuts CT segmentation robustness gap from 0.264 to 0.064","RAMP improves corrupted Dice to 0.753 in five organ CT benchmark","RAMP improves corrupted Dice to 0.789 in Abdomen1K benchmark","RAMP reduces CT robustness gap to 0.064 on noisy images"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The specific set of corruptions and their stochastic composition rules used in RAMP are representative of the heterogeneous imaging conditions that will actually appear at deployment time.","fun_headline_variants_meta":{"raw":{"variants":["RAMP cuts CT segmentation robustness gap from 0.264 to 0.064","RAMP improves corrupted Dice to 0.753 in five organ CT benchmark","RAMP improves corrupted Dice to 0.789 in Abdomen1K benchmark","RAMP reduces CT robustness gap to 0.064 on noisy images"]},"model":"grok-4.3","cost_usd":0.01142,"raw_usage":{"total_tokens":5063,"prompt_tokens":775,"num_sources_used":0,"completion_tokens":83,"cost_in_usd_ticks":114199500,"prompt_tokens_details":{"text_tokens":775,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":4205,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":775,"tokens_out":83,"duration_ms":26876,"temperature":1.0,"reasoning_tokens":4205,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-28T19:03:32.263010+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A RAMP-trained model showing large Dice drops on real clinical CT scans that contain degradation types not included in the augmentation set would indicate the central claim does not hold.","supporting_citations":[],"review_version":1}