{"id":"997e2d34-5fed-4631-b252-44dd9a244170","arxiv_id":"2607.02816","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"In ISO/TS 10974 Tier 3 MRI safety assessment of a DBS system, elevated BMI cut safe B1+ by 19–31% while FDTD and FEM solvers agreed closely and dielectric sweeps failed to reproduce the obese heating tail.","lead":"Patient body size (BMI), not the choice of electromagnetic solver, drives most of the uncertainty in computer predictions of MRI heating around implanted devices. That finding argues for treating high-BMI anatomy as a primary variable in safety labeling rather than relying on a few standard body models.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"Trajectory adaptation across models may confound the BMI-dominance claim by mixing geometry changes with habitus.","rationale":"The Reader correctly flags n=50 tail sampling and the nonlinear B1+–ΔT map as limiting CI precision; that is real but secondary. Even with infinite trajectories, the comparison remains only as clean as the trajectory-matching protocol. The paper’s design already shows solver equivalence (TOST within experimental margin) and dielectric sweeps failing to move p95, so those limbs of the hierarchy are solid. The load-bearing soft spot is whether the large phenotype shifts are pure habitus effects or partly artifacts of re-routing leads to fit larger bodies—the very geometric degree of freedom the Introduction cites as able to change heating by up to three-fold. A fixed-centerline (or purely scaled-centerline) re-analysis is a concrete, doable check that would either lock in or qualify the headline claim. Verdict stays CONDITIONAL; confidence would rise if the reduction survives the isolation test and the promised trajectory artifacts are released.","tokens_in":20426,"tokens_out":696,"duration_ms":7582,"concrete_test":"Recompute p95(ΔT) and Maximum Allowable B1+ for Duke vs Fats (and Ella vs Ella BMI 30) using one fixed set of 50 full-system centerlines: either (i) rigidly place the same polyline coordinates into each model after only global rigid registration of the head/IPG landmarks, or (ii) apply the same Sx/Sy morph factors used for Morphed-HBM to the trajectories themselves without further CAD re-routing. If the BMI-driven B1+ reduction falls below ~10% (or loses statistical separation from the Duke–HBM solver gap), the habitus-dominance ranking is confounded by trajectory adaptation.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central ranking—that body habitus dominates solver choice, dielectric uncertainty, and sex—rests on comparing p95(ΔT) and Maximum Allowable B1+ across models using n=50 “matched” full-system trajectories per phenotype (Methods II-D.3; Results III-E–F). Those trajectories are not fixed paths in a common coordinate system: they are CAD-adapted to each anatomy so that cranial entry, extracranial routing, and IPG pocket remain “anatomically valid,” with only “minor adjustments” allowed. Because RF coupling for elongated leads is highly sensitive to extracranial path geometry and orientation relative to Etan (Introduction; Eq. 1), any systematic re-routing forced by larger torso/shoulder dimensions in Fats or Ella BMI 30 can raise the heating tail even if bulk habitus were held fixed. The paper’s own morphing experiment (Strategy B) scales the whole body while reusing adapted paths and still recovers the obese p95, but that does not isolate whether the native Fats/Ella-BMI30 elevations are driven by coil-loading/body size versus the adaptation step itself. The Limitations section acknowledges “minor geometric variation across models” but does not quantify how much of the 19–31% B1+ reduction survives when the identical centerline geometry is rigidly embedded (or only isometrically scaled) into each model. Without that isolation, the dominance claim over solver/dielectric/sex is only partially secured.","agreement_with_reader":"partial"},"referee_report":{"model":"grok-4.5","summary":"The manuscript reports a cross-platform ISO/TS 10974 Tier 3 evaluation of RF-induced heating for a DBS system at 1.5 T, comparing FDTD (Sim4Life) and FEM (ANSYS HFSS) across phantom fields, transfer-function (TF) validation, and in vivo assessments in standard and diverse anatomies (Duke, HBM, Glenn, Fats, Ella, Ella BMI 30; >250 trajectories). The central claim is that body habitus (elevated BMI) is the dominant source of predictive uncertainty, exceeding solver choice, dielectric-property assumptions, and sex: FDTD and FEM Maximum Allowable B1+ limits converge near 2.6–3.0 µT in standard models, while elevated BMI reduces safe B1+ by 19–31% in both sexes; geometric morphing approximates the obese limit, whereas dielectric sweeps do not.","tokens_in":20821,"tokens_out":1551,"duration_ms":17899,"significance":"If the ranking holds, the work has clear regulatory and practical impact: it argues that Tier 3 labeling resources should prioritize anthropometric diversity (including elevated-BMI female models or validated morphing surrogates) over exhaustive dielectric sweeps or solver-class debates. Strengths include a like-for-like FDTD–FEM comparison through the full TF workflow, pre-specified TOST equivalence margins derived from measured experimental uncertainty (±2.56 °C), bootstrap CIs on p95 and B1+ limits, independent phantom heating validation (R≥0.90), cross-sex BMI corroboration, and an explicit commitment to share analysis code, trajectory geometries, and solver-agnostic intermediate fields. These elements make the study more auditable than typical AIMD heating papers.","major_comments":[{"comment":"Methods II-D.3 and Results III-E–F: the dominance ranking (habitus ≫ solver/dielectric/sex) rests on p95(ΔT) and Maximum Allowable B1+ from n=50 “matched” full-system trajectories that are CAD-adapted per anatomy so cranial entry, extracranial routing, and IPG pocket remain “anatomically valid,” with only “minor adjustments.” Because Eq. (1) is highly sensitive to extracranial path geometry relative to Etan (Introduction; also the paper’s own lead-routing literature), systematic re-routing forced by larger torso/shoulder dimensions in Fats or Ella BMI 30 can elevate the heating tail even if bulk habitus were fixed. The Limitations section notes “minor geometric variation” but does not quantify how much of the 19–31% B1+ reduction survives when identical centerlines are rigidly embedded (or only isometrically scaled) into each model. Strategy B (morphing) recovers the obese p95 but still","section":"Methods II-D.3; Results III-E–F; Limitations"},{"comment":"Methods II-F.1 and Limitations: p95 and derived Maximum Allowable B1+ are estimated from n=50 trajectories per model. The paper correctly notes that this limits bootstrap CI precision, especially under the nonlinear B1+–ΔT map used in Eq. (4). The reported 19–31% BMI-driven reductions and the solver/sex equivalence conclusions are therefore tail-sensitive. Either enlarge the trajectory sets for the key phenotype contrasts (Duke/Fats, Ella/Ella BMI 30, Duke/HBM) or report a sensitivity analysis (e.g., leave-k-out or progressive subsample stability of p95 and B1+) so readers can judge whether the quantitative ranking is robust to upper-tail sampling.","section":"Methods II-F.1; Limitations; Results III-C–F"},{"comment":"Results III-B and II-C.4: TF predictions show strong correlation with measurements (R≥0.90) but absolute RMSE of ~5.5–6.0 °C, while the clinical endpoint is p95(ΔT)≤2 °C and the TOST margin is ±2.56 °C. Probe-offset sensitivity alone spans 16.2 °C. The ranking and equivalence claims are relative and may still hold, but the manuscript should more explicitly separate (i) relative cross-solver/phenotype ranking from (ii) absolute accuracy of Maximum Allowable B1+ as a labeling number, and state how experimental uncertainty propagates into the reported µT limits (beyond the ΔT margin already used for TOST).","section":"Results III-B; Methods II-C.4; Section II-F"}],"minor_comments":[{"comment":"Table I lists HBM age/BMI as “–”; if approximate values are known from the ANSYS library documentation, stating them would help readers compare HBM to Duke (BMI 22.4) when interpreting the small B1+ offset in Fig. 5.","section":"Table I"},{"comment":"Fig. 2 summary table and MAPD definitions are clear; consider adding the phantom-only mask boundary on the field maps so readers can see the evaluation domain used for MAPD.","section":"Fig. 2; Methods II-A.3"},{"comment":"Eq. (1)–(2): state units and whether TF is normalized (e.g., per unit length) so that C is unambiguously dimensioned; Supplementary Fig. S2 is referenced but the main text could note the TF magnitude scale briefly.","section":"Methods II-C.1"},{"comment":"The abstract and Conclusion state “sex at matched BMI had no significant effect” and TOST equivalence; also report the Wilcoxon p-value and TOST CI already in Results III-D in the abstract-level summary for balance with the BMI percentages.","section":"Abstract; Results III-D"},{"comment":"Minor wording: “Body Habitus Dominates Solver Choice as a Source of Uncertainty” is slightly ambiguous (habitus dominates solver choice, or habitus is the dominant source among listed factors). The abstract’s final sentence is clearer; align the title phrasing if space allows.","section":"Title; Abstract"},{"comment":"References [18]–[19] and FDA AccessData access dates are fine; ensure arXiv/version consistency for the preliminary abstract [41] once the full paper is finalized.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The study is a solid, well-scoped contribution for TBME or similar; the main risk is overclaiming BMI dominance without isolating trajectory adaptation. If the authors add the isolation analysis (or a tight bound) and stabilize the p95 sampling story, this could move to accept/minor revision quickly. Scope fit is good for biomedical engineering / MRI safety; not a novelty concern relative to the conference abstract [41], which is properly disclosed."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The useful takeaway is simple: when you run the full ISO/TS 10974 Tier 3 TF chain on the same DBS system, FDTD and FEM give practically the same Maximum Allowable B1+ in standard anatomies (~2.6–3.0 µT), while elevating BMI drops that limit 19–31% in both sexes and dielectric ±50% sweeps do not recover the obese tail. Geometric morphing does. That ranking is what the paper actually adds.\n\nWhat they did well is the controlled design. Phantom MAPD under 6%, experimental TF validation with R ≥ 0.90 for both solvers, pre-specified TOST margins from measured uncertainty (±2.56 °C), bootstrap CIs, and the same direction of effect in Fats and Ella BMI 30. Solver equivalence is treated as an equivalence problem rather than a difference hunt, which is the right framing. The morphing vs dielectric contrast is clean and practically useful for groups that lack native high-BMI models. Citations track the prior BMI/body-composition and TF literature without pretending the heating effect itself is new.\n\nSoft spots, in proportion. Absolute RMSE is ~5.5–6 °C against a 2 °C safety threshold, so absolute heating numbers are noisier than the relative ranking. Everything is 1.5 T and one DBS system. n = 50 trajectories per model is thin for p95 tails under the quadratic B1+ map; the authors say so. The stress-test point on CAD-adapted trajectories is fair: paths are not rigidly fixed, so some of the obese elevation could be re-routing forced by larger torsos rather than pure coil-loading. Morphing still recovers the native obese p95 while reusing adapted paths, which softens the confound, but they never quantify how much of the 19–31% survives with identical centerlines. That is a real gap, not a fatal one; the multi-model, multi-sex consistency still supports treating habitus as primary.\n\nThis is for people who write or review AIMD MRI labeling evidence and for anyone building Tier 3 pipelines. It deserves a serious referee. I would cite the ranking and the morphing result. Engage.","headline":"Solid Tier-3 head-to-head: habitus (BMI) outranks solver class and dielectric sweeps for DBS B1+ limits at 1.5 T; trajectory adaptation is a real but secondary soft spot, not a collapse of the claim.","tokens_in":21467,"tokens_out":555,"would_cite":true,"duration_ms":5298,"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":"Patient body size, not which simulation software you pick, drives the uncertainty in MRI heating safety limits for implants.","keywords":["MRI safety","RF-induced heating","active implantable medical devices","deep brain stimulation","ISO/TS 10974","transfer function","anthropometric variability","body habitus"],"falsifier":"Repeat the full Tier 3 workflow with several hundred independent trajectories drawn from native high-BMI patient imaging for both sexes; if the BMI-driven reduction in Maximum Allowable B1+ shrinks below the solver-to-solver difference, the dominance claim fails.","tokens_in":21284,"feed_emoji":"🧲","tokens_out":711,"duration_ms":5743,"temperature":0.7,"pith_summary":"Safety labels that decide whether patients with deep brain stimulators and similar implants can enter an MRI scanner rest on computational predictions of radiofrequency heating. Those predictions have usually been run with one solver family and one or two standard-BMI body models, so it has been unclear whether the software choice, the assumed tissue properties, or the patient's anatomy contributes most to the uncertainty. This paper runs the full ISO Tier 3 transfer-function workflow on both an FDTD solver and an FEM solver for a deep-brain-stimulation system at 1.5 T, then repeats the calculation across more than 250 realistic lead trajectories in standard, elderly, and elevated-BMI male and female models. In ordinary body shapes the two solvers give nearly the same safe B1+ limits (about 2.6–3.0 µT). Raising BMI, however, cuts those limits by 19–31 % in both sexes, while age and sex at matched BMI do not. Simple geometric scaling of a standard model recovers the obese limit; sweeping tissue dielectric constants by ±50 % does not. The authors therefore conclude that body habitus is the dominant uncertainty source and that safety assessments must treat elevated-BMI anatomies as a primary variable.","feed_headline":"Body size, not solver choice, sets MRI implant heating risk","feed_subtitle":"Elevated BMI cuts safe B1+ limits 19–31% while FDTD and FEM agree within uncertainty","key_machinery":"The Maximum Allowable B1+ limit: the largest B1+ such that the 95th-percentile temperature rise across clinically realistic trajectories stays ≤2 °C, obtained by quadratically rescaling the transfer-function heating distribution measured at a reference exposure of 4.9 µT.","core_discovery":"Body habitus dominates solver choice, dielectric assumptions, and sex as the source of predictive uncertainty in ISO/TS 10974 Tier 3 MRI safety assessment of elongated-lead implants. FDTD and FEM implementations produce equivalent Maximum Allowable B1+ limits near 2.6–3.0 µT in standard anatomies, whereas elevated-BMI models of both sexes lower those limits by 19–31 %; geometric morphing reproduces the obese limit while dielectric sweeps do not.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["Body habitus, not solver, drives MRI AIMD heating uncertainty","Elevated BMI lowers safe B1+ limits 19–31% in Tier 3 tests","Patient size tops solver choice for implant MRI safety risk","BMI dominates age, sex, solvers in AIMD heating predictions","Anatomical diversity sets MRI implant B1+ limits more than software"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The ranking of risk sources rests on 95th-percentile heating values estimated from only fifty adapted trajectories per body model, a sample that the paper itself notes leaves the upper-tail confidence intervals imprecise.","fun_headline_variants_meta":{"raw":{"variants":["Body habitus, not solver, drives MRI AIMD heating uncertainty","Elevated BMI lowers safe B1+ limits 19–31% in Tier 3 tests","Patient size tops solver choice for implant MRI safety risk","BMI dominates age, sex, solvers in AIMD heating predictions","Anatomical diversity sets MRI implant B1+ limits more than software"]},"model":"grok-4.5","effort":"low","cost_usd":0.004548,"raw_usage":{"total_tokens":1416,"prompt_tokens":891,"num_sources_used":0,"completion_tokens":96,"cost_in_usd_ticks":45480000,"prompt_tokens_details":{"text_tokens":891,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":429,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":891,"tokens_out":96,"duration_ms":3533,"temperature":1.0,"reasoning_tokens":429,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-12T06:51:14.701571+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Repeat the full Tier 3 workflow with several hundred independent trajectories drawn from native high-BMI patient imaging for both sexes; if the BMI-driven reduction in Maximum Allowable B1+ shrinks below the solver-to-solver difference, the dominance claim fails.","supporting_citations":[],"review_version":1}