{"id":"d392ddbf-636f-494e-9709-ea11353f9f0b","arxiv_id":"2506.22563","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"PlanTUS is an open-source heuristic planning tool that visualizes feasible transcranial ultrasound transducer placements on an individual's scalp for targeting a chosen brain region.","lead":"PlanTUS is a new open-source software that helps researchers plan where to place ultrasound transducers on a person's scalp for brain stimulation. It quickly maps feasible positions based on the individual's brain anatomy, without running slow acoustic simulations.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The conclusion that PlanTUS placements 'principally result in a very good overlap' rests on hand-picked examples; the simplified straight-beam heuristic is never quantitatively compared with full-wave simulated focus, so its ranking could be wrong.","rationale":"The reader correctly identifies the straight-line beam assumption as a weak point, but the more load-bearing issue is that the paper never quantitatively links heuristic rankings to full-wave simulated focus overlap. PlanTUS could be perfectly reasonable as a heuristic screening tool even if its rankings are imperfect; the problem is that the paper's conclusion states a stronger empirical claim ('principally result in a very good overlap') while reporting only a few favorable examples. The authors are transparent that PlanTUS outputs need acoustic simulation validation, and the open-source, externally compatible design is a genuine strength. However, the absence of a systematic comparison means the central claim is currently unverified rather than disproven. A concrete simulation-based rank-correlation check would settle whether the heuristic proxies are fit for purpose. Since the reader already assigned CONDITIONAL, I recommend no change to that verdict: the paper should be accepted only if the claim is softened or the systematic validation is added.","tokens_in":10465,"tokens_out":4458,"duration_ms":56153,"concrete_test":"Reanalyze the Figure 2A use cases: for each target, sample every scalp vertex that satisfies PlanTUS's depth and tilt constraints (or at least 50 randomly selected feasible vertices per target), run k-Plan full-wave simulations for each, and compute the Dice or volume overlap between the -6 dB (50% peak-pressure) isosurface and the target mask. Then rank positions by PlanTUS's composite heuristic score and compare with the simulated-overlap ranking (e.g., Spearman rho). If the PlanTUS-selected position falls outside the top 10% of simulated overlap, or rho < 0.5, the 'principally very good overlap' conclusion must be restricted or replaced.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim in §4 is that PlanTUS-selected transducer placements 'principally result in a very good overlap between the simulated acoustic focus and brain target.' For that to hold, the heuristic metrics of §2.2—especially the idealized straight beam normal to the skin and the simplified focus overlay—must rank scalp positions by true acoustic focus accuracy. The paper never establishes this link. §3.2 shows four qualitative examples, each with a single hand-picked placement, and even the amygdala case is acknowledged to have strong skin–skull curvature deviation while the simulated focus 'still overlaps.' Because the heuristic uses ray geometry only and ignores refraction, phase aberration, and skull attenuation, the direction and magnitude of its ranking errors are unknown. If the idealized beam misses the target center due to skull refraction while the actual focus lands on target, or vice versa, the heuristic could systematically favor positions with poor simulated overlap. The claimed generalization is therefore not supported by the reported evidence, which is purely illustrative and potentially selection-biased.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript introduces PlanTUS, an open-source heuristic tool for prospective planning of transcranial ultrasound transducer placements. Using an individual T1-weighted MRI and a target mask, together with transducer-specific parameters (focal depths, aperture, tilt limits, FWHM values), PlanTUS computes scalp-surface metrics such as target distance, intersection of an idealized straight beam with the target, required transducer tilt, skin-to-skull angular deviation, and skull thickness. It visualizes these in Connectome Workbench, exports placements to acoustic simulation and neuronavigation software, and demonstrates example use cases for nucleus accumbens, thalamus, amygdala, subgenual anterior cingulate cortex, and ventromedial prefrontal cortex, including template and HCP datasets. The authors claim that PlanTUS-selected placements 'principally result in a very good overlap between the simulated acoustic focus and brain target.'","tokens_in":10641,"tokens_out":2886,"duration_ms":38191,"significance":"If the central claim is supported, PlanTUS would fill a practical gap: it is fast, fully open-source, built on widely used tools (SimNIBS, FSL, FreeSurfer, Connectome Workbench), and designed to narrow the large space of possible transducer positions down to a tractable set for full-wave simulation. The compatibility with k-Plan, k-Wave, BabelBrain, Localite, and BrainSight is a concrete strength, and the feasibility-check use case on template data is a useful contribution for study planning. However, the evidence for the key claim is currently only qualitative and selected; the heuristic is explicitly acknowledged to be a simplification, and no systematic quantitative validation is provided. The significance of the tool is therefore contingent on either tightening the claim to 'heuristic pre-selection requiring simulation validation' or adding a quantitative evaluation against full-wave acoustic simulations.","major_comments":[{"comment":"The conclusion states that 'PlanTUS-selected transducer placements principally result in a very good overlap between the simulated acoustic focus and brain target,' but §3.2 supports this only with four hand-picked examples and visual inspection. There is no quantitative outcome measure such as Dice overlap, centroid distance, or percentage of the simulated pressure field inside the target mask, and no comparison with alternative or randomly selected placements. As written, the central claim is not supported by the reported evidence; please either provide a systematic quantitative evaluation across targets, participants, and candidate positions, or revise the conclusion to the more modest claim that PlanTUS identifies promising positions that require simulation-based validation.","section":"§3.2 and §4 (Conclusion)"},{"comment":"The heuristic in §2.2 uses an idealized straight beam trajectory perpendicular to the skin surface, ignoring refraction, phase aberration, and skull attenuation. The manuscript never establishes that this ray-geometric metric ranks transducer positions correctly relative to the full-wave simulated acoustic focus. The amygdala case in §3.2 is the only place where a deviation between skin and skull curvature is acknowledged, and even there the conclusion is qualitative ('the simulated acoustic focus still overlaps with the target'). The direction and magnitude of ranking errors introduced by the straight-beam assumption are unknown. Please include a quantitative comparison between the heuristic metrics (e.g., beam-target intersection, required tilt) and the simulated focus-target overlap for a range of positions, including positions where the heuristic predicts a miss or a large tilt.","section":"§2.2 and §3.2"},{"comment":"The 'simplified, ideal acoustic focus' shown in the right panels of Figure 2A is derived from the same idealized geometry used for the heuristic metrics, so its overlay on the target mask cannot serve as independent evidence for the accuracy of the heuristic. Only the k-Plan simulation results (Figure 1D and the simulated pressure fields in Figure 2A) can provide such evidence. The text should clearly distinguish the ideal-focus overlay from the simulated acoustic pressure field, and the quantitative validation requested above should be based on the simulated fields.","section":"§3.1 and §3.2 (simplified focus overlay)"}],"minor_comments":[{"comment":"The left metric maps in Figure 2A are informative, but please ensure that all color bars are visible, labeled with units, and use consistent scales across rows so that comparisons between targets are not misleading.","section":"§3.2 (Figure 2A)"},{"comment":"The code availability statement would be strengthened by providing a versioned release, a DOI, or a specific commit hash for reproducibility, in addition to the project URL.","section":"§2.3 (Code availability)"},{"comment":"In the reference list, the k-Wave entry is duplicated: 'Bradley E. Treeby, Treeby, B. E., Ben Cox, & Cox, B. T.' should be reformatted to a single standard citation.","section":"References"},{"comment":"The statement that sonication of the ventromedial prefrontal cortex is 'not possible with a transducer that has no lateral steering capabilities' is conditional on the specific transducer parameters used; please make that conditionality explicit in the main text as well as in the figure caption.","section":"§3.2 (fourth row)"},{"comment":"Minor typographical and formatting issues include inconsistent spacing in the abstract ('PlanTUS:' and missing spaces around references) and the absence of line numbers in the submitted PDF; these are presentation-only issues.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope and the open-source tool is useful, but the central claim of 'very good overlap' is currently supported only by illustrative examples. I would be comfortable with acceptance after a quantitative validation or a clearly softened conclusion. No integrity concerns; the manuscript appropriately states that outputs need acoustic-simulation validation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"PlanTUS is a practical, open-source heuristic tool for prospectively planning TUS transducer placements. What is genuinely new is the integration of several geometric metrics—target distance, idealized beam-target intersection, required transducer tilt, skin-skull curvature deviation, skull thickness—into a single interactive interface that exports directly to acoustic simulation and neuronavigation software. That fills a real gap: groups can screen feasibility and narrow candidate positions before committing to expensive full-wave simulations.\n\nThe paper is well written, the examples illustrate the workflow nicely, and the authors are appropriately cautious in framing the tool as a heuristic whose outputs 'need to be validated' with simulation. The open-source code and website are a bonus.\n\nThe soft spot is the one the stress-test flags. The paper's conclusion states that PlanTUS-selected placements 'principally result in a very good overlap between the simulated acoustic focus and brain target.' That claim is supported by four hand-picked examples, each with a single selected position, and no quantitative error analysis, no baseline comparison, and no systematic sweep over positions. The heuristic uses a straight beam normal to the skin, ignoring refraction, phase aberration, and skull attenuation, so the ranking of scalp positions it produces could diverge from what full-wave simulation would say. The amygdala example itself highlights a strong skin–skull curvature deviation, yet the simulated focus still overlaps; that is reassuring for that case, but it tells you nothing about how often or where the heuristic misleads. Without a systematic study—multiple subjects, multiple targets, comparing overlap metrics for PlanTUS-selected positions against alternative or random positions—the central claim is not established.\n\nThe self-referential validation is not a circularity problem, since no parameters are fitted; it simply means the tool's output is only as good as its assumed beam model. A versioned release with test data would also aid reproducibility.\n\nWho is this for? TUS researchers who want a fast pre-filter for placement planning, and study designers checking whether a target is reachable with given hardware. It deserves peer review—the niche is important and the integration is useful—but only if the authors add proper validation. I would cite it as a tool reference, not as evidence of near-optimal focus-target overlap. For a reading group, it could spark a useful discussion about heuristic pre-screening versus full-wave simulation, but the evidence presented is too thin for that discussion to be decisive.","headline":"A genuinely useful open-source TUS planning heuristic, but the 'very good overlap' claim needs quantitative validation before acceptance.","tokens_in":11133,"tokens_out":3082,"would_cite":true,"duration_ms":34626,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"PlanTUS offers a fast, open-source way to pick ultrasound transducer placements for brain targets.","keywords":["transcranial ultrasound stimulation","transducer placement planning","heuristic optimization","skull anatomy","acoustic simulation","neuronavigation","open-source software","focused ultrasound"],"falsifier":"Run full-wave acoustic simulations (for example with k-Wave or BabelBrain) for both top-ranked and explicitly rejected scalp positions across a cohort of twenty or more subjects and several target regions, and test whether the simulated focus-target overlap is consistently higher for the top-ranked positions; a systematic failure of that ordering would refute the heuristic's predictive value.","tokens_in":10272,"feed_emoji":"🧠","tokens_out":10942,"duration_ms":101391,"temperature":0.7,"pith_summary":"The paper introduces PlanTUS, an open-source software tool that addresses a practical question in transcranial ultrasound stimulation: where on a given person's scalp a transducer should be placed to focus on a chosen brain region. Starting from a T1-weighted MRI and a target mask, PlanTUS segments the head into skin and skull surfaces and colors every scalp point with heuristic scores—distance to the target, how much of an idealized straight beam would intersect the target, the tilt needed to aim at the target, and the angular mismatch between skin and skull. These maps let users interactively pick the few most promising spots, which can then be checked with full acoustic simulations and exported to neuronavigation software. The authors demonstrate the workflow on several deep targets and state that PlanTUS-selected placements principally result in a very good overlap between the simulated acoustic focus and the brain target.","feed_headline":"Find scalp spots where ultrasound reaches a brain target with PlanTUS","feed_subtitle":"Ranks transducer spots by target reach, required tilt, and skull thickness before full acoustic simulation.","key_machinery":"The carrying object is the idealized straight-line beam trajectory. For each scalp point, PlanTUS casts a ray perpendicular to the local skin surface; the ray's intersection length with the target mask and the tilt needed to redirect it through the target center are the primary ranking variables. Skull-thickness maps and the angular deviation between skin and skull normals act as warnings about likely aberration, and the transducer's minimum and maximum focal depth define the reachable region on the scalp. These geometric surrogates are what allow the tool to rank positions in seconds.","core_discovery":"The central claim is that simple geometric quantities measured from the scalp surface can pre-screen transducer placements well enough that expensive acoustic simulations only need to be run for a handful of candidates. PlanTUS computes, for each scalp point, the distance to the target center, the length of intersection between the target mask and a straight ray launched perpendicular to the skin, the tilt angle required to make that ray cross the target center, and the local angular deviation between skin and skull surfaces, and it marks avoidance zones around the ears, eyes, nose, and air-filled sinuses. The paper argues that placements chosen from these maps yield very good overlap between the simulated acoustic focus and the target, illustrating this with examples for the nucleus accumbens, thalamus, amygdala, subgenual anterior cingulate cortex, and ventromedial prefrontal cortex. It further claims the tool detects when a target is not reachable at all with a given transducer's focal depth and steering capabilities, information that is useful before equipment is purchased or a study is designed.","pith_inferences":["A systematic validation study that compares PlanTUS's top-ranked and bottom-ranked placements against full-wave acoustic simulations across many subjects and targets would tell how often the heuristic ordering matches true focus-target overlap.","The same per-scalp-point scoring could be extended from single transducers to multi-transducer arrays by combining individual position costs with an objective for constructive interference at the target.","Template-based feasibility maps are a conservative screen rather than a verdict: because the standard head is larger than typical heads, an 'unreachable' label in standard space should not be treated as definitive for an individual."],"forward_implications":["Researchers can quickly screen a cohort's MRI scans for whether a target is reachable with a given transducer before acquiring hardware or starting a study.","The exported placements can be loaded directly into acoustic simulation and neuronavigation software, so the heuristic choice becomes the starting point for rigorous wave-based validation and MR-guided delivery.","For targets or transducer configurations that are not feasible, PlanTUS can reveal that before any simulation effort is spent, as in the ventromedial prefrontal cortex example with no lateral steering.","Maps over template or group-average heads allow feasibility checks very early in study planning, with the caveat that standard-space heads may underestimate reachability relative to individual anatomy."],"supporting_citations":[{"why":"Review that establishes acoustic simulation as the gold standard against which PlanTUS-selected placements are checked.","marker":"(Angla et al., 2023)"},{"why":"Practical guide that enumerates the anatomical and hardware constraints PlanTUS encodes.","marker":"(Murphy et al., 2025)"},{"why":"Supplies the Charm segmentation pipeline that extracts skin and skull surfaces from T1 MRI.","marker":"(Puonti et al., 2020)"},{"why":"Provides the SimNIBS head-modeling framework within which the segmentation runs.","marker":"(Thielscher et al., 2015)"},{"why":"Provides the Connectome Workbench viewer that PlanTUS uses to display and interact with scalp-surface metrics.","marker":"(Marcus et al., 2013)"},{"why":"BabelBrain is one of the acoustic simulation packages that can import PlanTUS outputs for validation.","marker":"(Pichardo, 2023)"},{"why":"k-Wave is the open-source wave-simulation toolbox that can consume PlanTUS placements.","marker":"(Bradley E. Treeby et al., 2010)"},{"why":"Earlier method for optimizing single-element transducer placement that PlanTUS is designed to make faster and more accessible.","marker":"(Park et al., 2019)"}],"fun_headline_variants":["Plan ultrasound spots by geometry, not full simulations","Pre-screen TUS transducer spots with PlanTUS heuristics","Skip costly acoustic sims: PlanTUS ranks feasible transducer sites","Heuristic tool finds reachable brain targets for focused ultrasound"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The tool's rankings rest on treating the ultrasound beam as a straight ray entering the skull perpendicular to the skin, ignoring the bending and slowing caused by skull and brain tissue; if that idealized path misrepresents the real acoustic path for a given skull region or target, the heuristic may rank placements incorrectly.","fun_headline_variants_meta":{"raw":{"variants":["Plan ultrasound spots by geometry, not full simulations","Pre-screen TUS transducer spots with PlanTUS heuristics","Skip costly acoustic sims: PlanTUS ranks feasible transducer sites","Heuristic tool finds reachable brain targets for focused ultrasound"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000175,"raw_usage":{"total_tokens":1263,"prompt_tokens":903,"completion_tokens":360,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":519,"completion_tokens_details":{"reasoning_tokens":292}},"tokens_in":519,"tokens_out":360,"duration_ms":4463,"temperature":1.0,"reasoning_tokens":292,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:02:59.239811+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run full-wave acoustic simulations (for example with k-Wave or BabelBrain) for both top-ranked and explicitly rejected scalp positions across a cohort of twenty or more subjects and several target regions, and test whether the simulated focus-target overlap is consistently higher for the top-ranked positions; a systematic failure of that ordering would refute the heuristic's predictive value.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Review that establishes acoustic simulation as the gold standard against which PlanTUS-selected placements are checked."},{"cited_title":"B., Siebner, H","cited_arxiv_id":null,"evidence_quote":"Supplies the Charm segmentation pipeline that extracts skin and skull surfaces from T1 MRI."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the SimNIBS head-modeling framework within which the segmentation runs."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"BabelBrain is one of the acoustic simulation packages that can import PlanTUS outputs for validation."},{"cited_title":"Y., Pahk, K","cited_arxiv_id":null,"evidence_quote":"Earlier method for optimizing single-element transducer placement that PlanTUS is designed to make faster and more accessible."}],"review_version":1}