{"id":"76fbab0d-f0cd-49b7-af98-20c49571a697","arxiv_id":"2607.08408","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Track2Map jointly optimizes camera poses and deformable 3D Gaussian maps online from surgical stereo video via track-anchored deformation and motion-gated pose updates.","lead":"Track2Map builds online 3D maps of deforming surgical tissue from stereo video while estimating camera motion, even without robot kinematics. It uses dense 2D point tracks to separate camera movement from tissue deformation, improving reconstruction when pose data is missing or noisy.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.5","headline":"Strongest claim rests on StereoMIS alone; motion-gate and tracker assumptions are untested outside that dataset.","rationale":"The reader correctly flags the motion gate as the weakest assumption and keeps CONDITIONAL for single-dataset evaluation, missing variance, and non-real-time runtime. That is the right load-bearing concern: the strongest claim is an empirical systems result whose mechanisms (track-anchored deformation + flow-dispersion gating) are only validated where the gate’s assumptions hold. No internal inconsistency in the math; the risk is external validity and unmeasured gate correctness. My concrete test would settle whether the gate is actually detecting camera motion or merely correlating with StereoMIS statistics. No stronger objection (e.g., broken equations or inverted baselines) appears in the manuscript, so the verdict should stay CONDITIONAL rather than move to REJECT or ACCEPT.","tokens_in":9372,"tokens_out":528,"duration_ms":5483,"concrete_test":"On StereoMIS (and ideally one additional public sequence with known kinematics), label frames as camera-static vs. moving from ground-truth pose; report gate precision/recall at τ=0.38 and recompute Tab. 1–2 after (a) forcing gt=1 always and (b) replacing CoTracker3 with a weaker tracker. If gate F1 is low or PSNR/ATE collapse under (a)/(b) on non-StereoMIS data, the multi-regime robustness claim weakens.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (robust reconstruction and pose under clean/noisy/absent priors) is supported only by StereoMIS tables (Tab. 1–3, Fig. 3–4). The load-bearing mechanism is the motion gate gt = I[σθ,t ≤ τ] with τ = 0.38 (Sec. 2.1) plus CoTracker3 tracks lifted by stereo depth. The paper itself lists failure modes—optical-axis motion in tubular anatomy, large coherent tissue motion, tracker failure, heavy tool occlusion, weak texture—yet provides no quantitative gate accuracy, no per-sequence static/moving frame labels, and no second dataset. If StereoMIS happens to have intermittent lateral camera motion and reliable tracks, the reported PSNR/ATE gains (e.g., no-pose stereo PSNR 27.58, ATE 0.0285 m) may not transfer; the ablations (Tab. 3) only show that gating helps on the same data, not that the gate is correct.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"Track2Map is an online stereo 3D Gaussian Splatting pipeline for deformable surgical scenes that jointly refines camera pose and a sparse-anchor deformation field from video. Dense 2D tracks (CoTracker3) are lifted with stereo depth to (i) initialize anchor deformations and (ii) gate pose updates via the circular standard deviation of track flow directions, freezing pose when motion is judged local rather than camera-driven. The system is designed to run with clean, noisy, or absent pose priors. On StereoMIS it reports higher reconstruction metrics than endoscopic SLAM baselines in the no-prior regime and than Online-endo-track under clean/noisy priors, plus lower ATE/RPE than EndoGSLAM-H and Endo3R, with ablations of pose optimization, gating, and deformation.","tokens_in":9674,"tokens_out":1272,"duration_ms":19289,"significance":"If the StereoMIS gains transfer, the work is practically useful: many RAMIS settings lack reliable kinematics, and a single pipeline that degrades gracefully from clean priors to pure vision-only SLAM is valuable for AR overlays and instrument–tissue analysis. Strengths include multi-regime evaluation (none / light / heavy / clean pose), comparison to both SLAM and prior-dependent 3DGS methods, pose ATE/RPE, a component ablation (Table 3), public code, and an explicit limitations section. The track-anchored deformation init and motion gate are concrete engineering contributions relative to unconstrained joint pose–deformation optimization.","major_comments":[{"comment":"The central claim of robust reconstruction and pose under clean, noisy, and absent priors is supported only on StereoMIS (Tables 1–3, Figs. 3–4). No second dataset, no per-sequence variance/error bars, and no cross-procedure split are reported. For a journal claim of regime-agnostic robustness this is load-bearing: gains may reflect StereoMIS motion statistics (intermittent lateral camera motion, trackable texture) rather than general RAMIS behavior. At minimum, results on another public stereo/deformable set (or a held-out StereoMIS procedure with different motion) and sequence-level statistics are needed.","section":null},{"comment":"Sec. 2.1 defines the motion gate gt = I[σθ,t ≤ τ] with τ = 0.38 from circular std of track flow directions, and freezes pose when gt = 0. This mechanism is central to drift reduction and to the no-prior / noisy-prior claims, yet the paper reports no gate accuracy, no static/moving frame labels, no sensitivity of PSNR/ATE to τ, and no breakdown of how often the gate fires per sequence. The Limitations section lists plausible failure modes (optical-axis motion, coherent tissue motion, occlusion, weak texture) without quantifying them. A short validation of gate decisions against kinematics-derived motion labels (or a τ sweep) is required to show the gate is correct, not merely helpful on this data (Table 3).","section":null},{"comment":"Table 1 monocular Track2Map (Depth Anything V2) already beats prior SLAM methods by a large margin (PSNR 26.70 vs EndoFlow-SLAM 21.96), and stereo adds a further gain. The manuscript does not isolate how much of the no-prior improvement comes from modern stereo/monocular depth and CoTracker3 versus the proposed gating and track-anchored deformation. A controlled comparison that holds depth and tracker fixed while swapping only the mapping/pose modules (or reports EndoGSLAM/EndoFlow with the same depth) would make the contribution attribution clearer.","section":null},{"comment":"The system is described as online SLAM but runs at ~6 s/frame (Limitations), with keyframe processing recommended. For journal positioning as an online SLAM method, either (i) report wall-clock breakdown and a keyframe-only accuracy/latency trade-off, or (ii) soften the real-time/online framing to “incremental / sequential” so claims match the implementation.","section":null}],"minor_comments":[{"comment":"Fig. 2 caption and body: STIR is cited as both [17] and [18]; unify the STIR2024 / STIRC2024 naming and reference keys.","section":null},{"comment":"Eq. (7) writes δξ ∈ SE(3) while Eq. (3) and the text treat δξ as a 6D se(3) increment; fix the group vs algebra notation.","section":null},{"comment":"Loss weights in Sec. 3.1 list wdepth, wcolor, wpose_prior but Sec. 2.3 uses wphoto, wgeo, wdef, wtrk, wreg; provide a single consistent weight table and values used for all regimes.","section":null},{"comment":"Table 1: EndoFlow-SLAM SSIM is printed as “0.590.27” (missing separator between SSIM and LPIPS).","section":null},{"comment":"Fig. 3 compares trajectories on three sequences but does not state units, alignment method (e.g., Sim(3)/SE(3)), or which frames are gated static; a short caption note would aid reproducibility.","section":null},{"comment":"Clarify whether tool masks (Fig. 1) are required at test time or optional, and how masked pixels enter Lphoto/Lgeo when masks are absent.","section":null}],"recommendation":"major_revision","confidential_remarks":"Conference-length MICCAI-style manuscript submitted in journal-referee form; contribution is solid engineering but evaluation breadth is the main gap. I would accept after expanded validation of the gate and at least one additional dataset or strong sequence-level analysis—not a reject. Code link is a plus for reproducibility."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The useful takeaway is that Track2Map is a working online stereo deformable 3DGS pipeline that can run with clean, noisy, or no pose priors, and it beats both SLAM baselines and a strong pose-prior method (Online-endo-track) on StereoMIS reconstruction and ATE.\n\nWhat is actually new is not 3DGS or CoTracker3, but the coupling: dense tracks are lifted to 3D anchors that initialize sparse deformation offsets, and the same tracks supply a simple circular-std flow gate that freezes pose when motion looks local rather than camera-global. That is a concrete engineering answer to the intermittent-camera problem in RAMIS. The multi-regime tables are the strength—no-pose stereo PSNR 27.58 / SSIM 0.745, heavy-noise still ~27.5, ATE 0.0285 m vs EndoGSLAM/Endo3R, plus an ablation that removes pose opt, gating, and deformation one at a time. Code is promised. Citations look appropriate; they sit on Online-endo-track, CoTracker3, and the usual endoscopic SLAM set without inventing a parallel literature.\n\nSoft spots are real but proportionate. Everything is StereoMIS; no second dataset, no run variance, no gate accuracy or static/moving labels. The authors themselves list the gate’s failure modes (optical-axis motion, coherent tissue motion, occlusion, weak texture). Runtime is ~6 s/frame, so keyframe-only in practice. Free parameters (τ=0.38, loss weights, α/N) are set empirically. None of that collapses the central claim on the data they show; it just means transfer is unproven.\n\nThis is for people building surgical AR, interaction analysis, or autonomy who currently choke on missing/noisy kinematics. It is not a theory paper and will not reorganize general 3D vision. I would bring it to reading group if the group cares about endoscopic mapping; I would cite the no-pose / heavy-noise numbers when discussing prior-free deformable GS-SLAM. A serious editor should send it to referees—systems contribution is clear enough, evidence is multi-regime and ablated, and the limitations are already stated honestly. Expect requests for another dataset or gate diagnostics, not a desk reject.","headline":"Solid systems paper: track-anchored deformation + motion-gated pose lets online deformable GS-SLAM work without clean kinematics, with clear StereoMIS gains; single-dataset and unvalidated gate are the real limits.","tokens_in":10323,"tokens_out":573,"would_cite":true,"duration_ms":5757,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Dense 2D tracks let surgical video rebuild deformable 3D anatomy online even when camera poses are missing or noisy.","keywords":["deformable SLAM","3D Gaussian Splatting","robot-assisted surgery","stereo endoscopy","motion-aware pose optimization","point tracking","online reconstruction"],"falsifier":"Sequences in which large coherent tissue motion or pure optical-axis camera motion produces low flow-direction dispersion while the endoscope is actually static (or the reverse), causing the gate to open or close incorrectly and measurably raising trajectory error or reconstruction distortion relative to the reported StereoMIS numbers.","tokens_in":10250,"feed_emoji":"🔬","tokens_out":588,"duration_ms":6295,"temperature":0.7,"pith_summary":"Most dense 3D reconstruction for robot-assisted surgery still needs clean camera trajectories, often from robot kinematics, and runs offline. Track2Map instead jointly optimizes camera motion and a deformable 3D Gaussian map frame by frame from stereo surgical video. Dense 2D point tracks initialize tissue deformation and, through a simple motion gate on flow-direction consistency, tell the system when the endoscope is still so that local tissue motion is not wrongly blamed on the camera. On the StereoMIS benchmark the method improves reconstruction quality and trajectory accuracy over competing SLAM systems and over non-SLAM pipelines that receive pose priors, remaining stable under clean, noisy, or absent pose initialization. The practical stake is metric, drift-resistant online maps that do not require reliable robot kinematics.","feed_headline":"Surgical SLAM rebuilds soft tissue without clean camera poses","feed_subtitle":"Dense 2D tracks gate pose updates so maps stay metric when kinematics are missing or noisy","key_machinery":"Motion-aware pose gating from track statistics: the circular standard deviation of 2D track flow directions decides whether the camera is moving; only then is pose refined jointly with the scene, while static periods freeze pose and rely on track-lifted 3D anchors to drive deformation.","core_discovery":"Track2Map shows that dense 2D point tracks can both initialize sparse deformation anchors and gate camera-pose updates, yielding an online deformable 3D Gaussian SLAM system that produces higher-quality reconstructions and more accurate trajectories on StereoMIS than prior endoscopic SLAM and than pose-prior-dependent online Gaussian pipelines, under clean, noisy, or missing pose initialization.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["Track2Map: online deformable SLAM without pose priors","Dense 2D tracks gate pose for surgical Gaussian SLAM","Online soft-tissue maps when camera trajectories are missing","Track-anchored deformation enables pose-free RAMIS SLAM","Surgical video alone drives joint pose and deformable Gaussians"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The system assumes that coherent, low-dispersion flow directions mean the camera is moving and that high dispersion means only tissue or tools are moving, so it freezes pose whenever the flow directions are too scattered.","fun_headline_variants_meta":{"raw":{"variants":["Track2Map: online deformable SLAM without pose priors","Dense 2D tracks gate pose for surgical Gaussian SLAM","Online soft-tissue maps when camera trajectories are missing","Track-anchored deformation enables pose-free RAMIS SLAM","Surgical video alone drives joint pose and deformable Gaussians"]},"model":"grok-4.5","effort":"low","cost_usd":0.003886,"raw_usage":{"total_tokens":1227,"prompt_tokens":773,"num_sources_used":0,"completion_tokens":66,"cost_in_usd_ticks":38860000,"prompt_tokens_details":{"text_tokens":773,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":388,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":773,"tokens_out":66,"duration_ms":4084,"temperature":1.0,"reasoning_tokens":388,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-10T07:54:23.173058+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Sequences in which large coherent tissue motion or pure optical-axis camera motion produces low flow-direction dispersion while the endoscope is actually static (or the reverse), causing the gate to open or close incorrectly and measurably raising trajectory error or reconstruction distortion relative to the reported StereoMIS numbers.","supporting_citations":[],"review_version":1}