{"id":"49af36a7-36ba-425f-9f48-72a572af445d","arxiv_id":"2502.01006","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A Geant4 simulation shows a dual-panel PET with 40 ps time-of-flight matches a 300 mm non-TOF ring PET in resolution, and 30 ps TOF offsets a drop to 20% detection efficiency.","lead":"This simulation study tests whether a flat two-panel PET scanner with extremely fast time-of-flight detectors can image as well as a conventional ring scanner. It finds that 40 picosecond timing matches a ring scanner's resolution, and 30 picosecond timing can recover image quality when detection efficiency drops to 20%.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central 40 ps and 20% RDE thresholds depend on the unvalidated assumptions of Gaussian TOF smearing and random event thinning in Section II.B; a test with realistic BGO Cherenkov timing spectra is needed before the feasibility claim is accepted.","rationale":"The reader's weakest assumption is exactly the load-bearing concern: the simulation represents detector timing as a single Gaussian, omits optical transport, and models reduced efficiency as uniform random thinning. The paper's strongest claims—the 40 ps requirement for parity with a non-TOF ring and the 20% RDE noise equivalence at 30 ps—are outputs of that idealized model. The manuscript itself flags the crucial gap in the Discussion: BGO timing histograms have long tails due to variable Cherenkov photon counts and scintillation contamination, and the simulation assumes that Cherenkov/scintillation event identification is always correct. This makes the central feasibility thresholds conditional on detector behavior that has not yet been demonstrated for the proposed BGO-integrated multi-anode MCP-PMT. I agree with the reader's conditional verdict; the concern does not invalidate the simulation study as an exploratory parametric analysis, but it prevents the quantitative thresholds from being treated as robust predictions. A concrete test using measured BGO timing spectra and a physically motivated correlation between efficiency and timing precision would settle whether the thresholds survive. No change to the reader's verdict is needed because it already conditions the feasibility claim on future validation.","tokens_in":14360,"tokens_out":2738,"duration_ms":30410,"concrete_test":"Replace the Gaussian TOF smearing in the existing list-mode pipeline with timing offsets drawn from a measured BGO Cherenkov timing spectrum (e.g., the tailed spectra of Kratochwil et al. 2020 or Gonzalez-Montoro et al. 2022), and implement RDE not as random thinning but as rejection of events whose generated Cherenkov photon count is below trigger threshold, with each event's timestamp taken from the first detected Cherenkov photon. Recompute Fig. 6 (%P2V vs TOF resolution) and Fig. 8 (%SD vs RDE) for these realistic timing and efficiency models. If either the 40 ps equivalence point or the 20%-RDE noise equivalence shifts by more than 10% in TOF or RDE, the headline thresholds are artifacts of the Gaussian/random-thinning idealization.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claims—40 ps TOF to match ring PET resolution and 20% RDE at 30 ps matching 100% RDE at 40 ps—depend on Section II.B's assumptions that each coincidence's TOF error is a single Gaussian and that reduced efficiency can be modeled by randomly deleting list-mode events. Both assumptions are favorable to the dual-panel concept and are acknowledged as unvalidated in the Discussion: the BGO timing histogram has a long tail from variable Cherenkov photon counts and scintillation contamination, and perfect Cherenkov/scintillation event identification is assumed. If real BGO/MCP-PMT detectors trigger on few Cherenkov photons, the timing error per event will be non-Gaussian and correlated with the same efficiency reduction that RDE is meant to model; fewer detected photons means both fewer events and worse, tail-heavy timing. Under that coupling, RDE=20% would no longer be equivalent to 100% efficiency at 40 ps, and the stated 40 ps/20% thresholds could shift. The paper's feasibility conclusion is therefore conditional on an idealized timing model rather than on demonstrated detector behavior.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a Geant4 simulation study of a dual-panel PET geometry with ultrafast time-of-flight (TOF) detectors based on BGO crystals coupled to microchannel-plate PMTs. The authors simulate imaging performance for TOF resolutions from 30 to 90 ps and relative detection efficiencies (RDE) from 10% to 100%, using MLEM and backprojection, and compare against a non-TOF ring PET reference. The main claims are that a 40 ps TOF resolution is sufficient for the dual-panel PET to match the spatial resolution of the 300-mm-diameter ring PET at the same detection efficiency, that the image noise at 30 ps TOF with 20% RDE is similar to that at 40 ps TOF with complete efficiency, and that MLEM outperforms backprojection. The authors conclude that the proposed dual-panel PET is feasible.","tokens_in":14633,"tokens_out":5659,"duration_ms":51790,"significance":"If the central quantitative thresholds are robust, this work is significant for the design of open-geometry PET systems in applications such as PET-guided surgery and in-beam particle therapy monitoring. The simulation methodology is clearly described and internally consistent, and the authors disclose major simplifications up front: no optical transport, a Gaussian timing model, uniform random thinning for efficiency, and a reconstruction kernel matched to the simulated timing blur. The key numbers (40 ps, 20% RDE) are outputs of a well-defined forward simulation rather than fitted parameters. However, the paper's feasibility conclusion is conditional on idealized timing and efficiency models that are favorable to the dual-panel concept, so the claims are not yet established against realistic detector behavior.","major_comments":[{"comment":"The TOF error is modeled as a single Gaussian smearing of the arrival-time difference, and Section IV acknowledges that real BGO timing histograms show long tails due to variable Cherenkov photon statistics and scintillation contamination. Because the 40 ps spatial-resolution threshold and the 20% RDE noise-equivalence threshold are outputs of this Gaussian model, the central feasibility claim is not robust to realistic timing distributions. Please add a sensitivity study using a measured or literature-based timing spectrum (e.g., the multi-Gaussian or tailed model of Ref. [44]) and report how the %P2V and %SD curves shift.","section":"Section II.B"},{"comment":"The relative detection efficiency is implemented by uniformly random thinning of list-mode events. In a real detector, detection efficiency and timing quality are coupled: events with few detected Cherenkov photons are both more likely to be lost and to have worse, tail-heavy timing. Random thinning breaks this correlation in a way that is favorable to the dual-panel concept. The claim that 20% RDE at 30 ps TOF matches 100% RDE at 40 ps TOF should be tested with a model in which timing uncertainty worsens as the number of detected photons decreases, so that the efficiency-timing coupling is represented.","section":"Section II.B (RDE modeling)"},{"comment":"The reconstruction uses a TOF kernel whose Gaussian FWHM and bin size exactly match the simulated timing blur, and the detector response kernel is also matched. This kernel matching is a form of circularity that gives an upper bound on achievable image quality; any real system would have model mismatch from timing tails and misidentified events. The authors should state this explicitly and, ideally, quantify the degradation by running at least one reconstruction with a mismatched TOF kernel (e.g., a wider Gaussian or a tailed kernel) to show the sensitivity of the reported thresholds.","section":"Section II.D, Eq. (2)"}],"minor_comments":[{"comment":"The caption states that the %SD at 30 ps TOF with 20% RDE (6.9%) was 'better than' the %SD at 40 ps with complete efficiency (6.8%); numerically, 6.9% is slightly worse, so the wording should be 'comparable to' as used in the abstract.","section":"Fig. 8 caption"},{"comment":"The sentence 'In the case of half the efficiency, we showed that the TOF resolution should be 30 ps...' is inconsistent with the reported result, which compares 20% RDE with 100% RDE; 'half the efficiency' should be corrected to 'one-fifth the efficiency' or rephrased to match the actual RDE values.","section":"Section V (Conclusion)"},{"comment":"The sentence describing the timing calculation is repetitive: 'the interaction time at the detection position' and 'the interaction position with the maximum energy deposition' refer to the same quantity; clarify that the detection position is the crystal with maximum energy deposition and that the photon travel time to the crystal bottom is added to that interaction time.","section":"Section II.B"},{"comment":"The iteration-selection criteria for the spatial resolution and noise studies could be stated more explicitly; in particular, 'the %P2V was fixed at a value when %SD was 10% in the non-TOF ring PET' should make clear that the reference is the ring PET, not the dual-panel PET.","section":"Section II.E.1"},{"comment":"The claim that rods of a given diameter were 'resolved' is based on visual inspection of Fig. 5; consider providing a quantitative rod-visibility metric (e.g., contrast or peak-to-valley for each rod size) to support the resolution claims.","section":"Section III.A"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for physics.med-ph and the simulation methodology is sound as far as it goes. The central feasibility claim is plausible but conditional on the Gaussian timing and random-thinning assumptions, which are favorable to the proposed geometry. The requested sensitivity analyses (realistic timing spectra and efficiency-timing coupling) are feasible within the manuscript's scope, so I recommend major revision rather than rejection. The authors should also correct the numerical inconsistency in the Fig. 8 caption and the wording in the Conclusion."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this as a well-scoped parametric simulation study, not as a detector feasibility proof. The genuinely new pieces are the simultaneous sweep of TOF resolution (30–90 ps) and relative detection efficiency (RDE) in a dual-panel geometry, and the direct comparison of MLEM with backprojection at 30 ps. The headline thresholds—40 ps to match a 300 mm non-TOF ring in resolution, and 20% RDE at 30 ps matching full efficiency at 40 ps in noise—are useful targets, but they are clearly outputs of an idealized model.\n\nThe paper earns credit for being transparent. Section II.B lays out the timing model as a single Gaussian plus a measured 20.1 ps DOI contribution, and efficiency loss is modeled as random thinning of list-mode events. The authors explicitly say they omit optical transport and assume perfect rejection of scintillation-only triggers. In the Discussion they acknowledge that real BGO timing histograms have long tails and that identification failures could shift the results. That is the right amount of disclosure for a simulation-first paper.\n\nThe stress-test concern lands, but it doesn't sink the paper. If actual Cherenkov timing is non-Gaussian, or if efficiency loss and timing jitter are correlated (fewer detected photons → worse timing), the 40 ps/20% numbers will move. The authors don't test that coupling, and they say so. A follow-up with optical transport or measured BGO timing spectra would be the natural way to firm up the feasibility claim.\n\nThe only other soft spot is the reference scanner: a non-TOF 5 mm BGO ring is not a modern clinical system. That's a fair choice for isolating TOF benefit, but readers should not read 'similar spatial resolution' as 'similar overall image quality.' The authors actually point this out in the Discussion, so it's a minor issue.\n\nThis is for PET instrumentation researchers and anyone designing open-geometry TOF scanners. Overall, the central argument holds within its stated scope. The simulation is reproducible, the parameter space is explored systematically, and the conclusions are appropriately hedged. I'd send this to a competent referee; it gives the community concrete numbers to design against.","headline":"A transparent dual-panel TOF-PET simulation that gives concrete thresholds; treat them as conditional until detector-level timing is tested.","tokens_in":15188,"tokens_out":3547,"would_cite":true,"duration_ms":31885,"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":"A dual-panel PET built from two flat BGO panels matches a 300-mm ring scanner's spatial resolution at 40 ps time-of-flight, and at 30 ps it keeps image noise low even when only 20% of coincidence events are retained.","keywords":["positron emission tomography","time-of-flight PET","dual-panel PET","Cherenkov radiation","MCP-PMT","BGO scintillator","image reconstruction","Monte Carlo simulation"],"falsifier":"Build a BGO-window multi-anode MCP-PMT panel and measure its coincidence time distribution and Cherenkov-triggered event fraction under realistic optical conditions: if the timing histogram has non-Gaussian tails wider than the simulated Gaussian smearing, or if fewer than roughly one in five coincidences is Cherenkov-triggered at the operating threshold, the simulated 40 ps and 20%-efficiency breakpoints would not transfer to hardware. A simulation with full optical photon transport, which the paper explicitly leaves out, would be a direct numerical check.","tokens_in":14163,"feed_emoji":"⚛️","tokens_out":8662,"duration_ms":79702,"temperature":0.7,"pith_summary":"Using Monte Carlo simulation, this paper asks whether a positron emission tomography scanner built from two flat detector panels instead of a ring can match conventional imaging performance if the detectors time photon arrivals with extreme precision. The answer it argues is yes: with 40 ps time-of-flight (TOF) resolution, the dual-panel geometry reaches the spatial resolution of a non-TOF ring PET of the same scale and efficiency, and at 30 ps it beats it. Because Cherenkov-based timing in BGO detectors collects very few photons, the authors also model reduced detection efficiency by randomly discarding coincidence events; at 30 ps, retaining only 20% of events gives image noise equivalent to a 40 ps system with full efficiency. They further show that iterative MLEM reconstruction outperforms simple backprojection even at 30 ps, and that TOF improves contrast in a NEMA-style phantom with 40% relative efficiency. The practical stake is that open, flat detector geometries, useful for surgery guidance and beam monitoring, could become clinically viable if roughly 30 to 40 ps detectors can be built.","feed_headline":"Dual-panel PET needs 40 ps timing to match ring scanners","feed_subtitle":"Simulation: at 30 ps, keeping just 20% of events holds noise to full-efficiency 40 ps levels.","key_machinery":"The load-bearing mechanism is time-of-flight localization along each line of response, folded into list-mode MLEM through a TOF kernel that spreads each event around the most likely source position. The simulation assumes Cherenkov photons supply the timing signal while scintillation light supplies position and energy, and it models TOF resolution as a Gaussian smearing of the arrival-time difference; section II.B also introduces the relative detection efficiency (RDE), a single parameter that randomly thins coincidence data to represent the fraction of events triggered by detected Cherenkov photons. RDE is what lets the study trade timing precision against light-collection loss and identify the boundary where TOF gain cancels efficiency loss.","core_discovery":"The paper's central claim is that ultrafast TOF measurement substitutes for the angular coverage that a ring geometry provides. In simulated acquisitions of two 137 by 137 mm panels of 5 mm BGO pixels separated by 300 mm, 40 ps TOF resolution produced peak-to-valley contrast of 97.6% on 4 mm rods, statistically comparable to the 92.3% of a 300 mm-diameter non-TOF ring PET, while 30 ps reached 99.4%. Because BGO Cherenkov radiation yields only a few photons and the fraction of events in which both detectors trigger on Cherenkov light is partial, the authors thinned list-mode data to define a relative detection efficiency; with 30 ps TOF and 20% efficiency the image noise was 6.9%, essentially equal to the 6.8% noise of a 40 ps system with full efficiency. The authors conclude the dual-panel geometry is feasible and that the benefit of TOF outweighs the loss of efficiency down to roughly 20% retained events.","pith_inferences":["If the 20%-efficiency result carries to hardware, Cherenkov photon collection in BGO need not be near-total for a panel scanner to be useful; a system retaining one in five coincidences could still yield acceptable noise.","The paper compares against a 300 mm ring, but clinical whole-body rings are much larger; scaling that comparison suggests the panel design could match even a roughly 700 to 800 mm ring with relaxed TOF requirements, a direction the discussion only touches on.","A direct test of the single-Gaussian timing assumption, by replacing it with the multi-Gaussian tails measured for BGO Cherenkov and scintillation events and adding optical photon transport, would firm up whether 40 ps is the true threshold.","Misidentification of scintillation-triggered events as Cherenkov-triggered ones is assumed to be perfectly rejected; imperfect event classification would degrade both timing and efficiency and likely require a TOF resolution even better than 30 ps."],"forward_implications":["A 40 ps dual-panel scanner matches the spatial resolution of a 300 mm non-TOF ring PET with identical detector efficiency, so ultrafast TOF removes the need for full angular sampling.","At 30 ps TOF, image noise with only 20% of coincidence events retained equals the noise of a 40 ps system with full efficiency, meaning TOF can absorb large Cherenkov detection losses.","MLEM reconstruction remains worthwhile at 30 ps: it gives higher rod contrast than backprojection at the same noise level using only a few iterations.","In a NEMA-style image-quality phantom with 40% relative efficiency, faster TOF improves contrast at fixed background variability and cuts vertical artifacts.","Better than 50 ps TOF resolves 3 mm rods in the panel geometry, the same rod resolution as the non-TOF ring reference."],"supporting_citations":[{"why":"provides the 30 ps coincidence time resolution benchmark from Cherenkov-radiator-integrated MCP-PMT pairs that sets the TOF targets","marker":"[5]"},{"why":"demonstrated reconstruction-free cross-sectional imaging with an ultrafast TOF detector pair, motivating the panel geometry","marker":"[7]"},{"why":"establishes BGO as a hybrid scintillator and Cherenkov radiator, the basis for the timing and scintillation model","marker":"[22]"},{"why":"supplies the experimental figure of only about 17 Cherenkov photons from BGO, motivating the relative-efficiency parameterization","marker":"[23]"},{"why":"provides measured Cherenkov-based coincidence timing and event-rate fractions in BGO detectors, supporting the 40% RDE value","marker":"[24]"},{"why":"gives the fraction of coincidences triggered by Cherenkov photons under dual-ended BGO readout, another basis for the RDE values","marker":"[28]"},{"why":"the prior simulation study of a multi-panel limited-angle TOF PET that this work extends with BGO and ultrafast timing","marker":"[14]"},{"why":"the Monte Carlo toolkit used to simulate all detector geometries and acquisitions","marker":"[34]"},{"why":"supplies the in-house list-mode MLEM reconstruction code with detector response and TOF modeling","marker":"[37]"},{"why":"provides the TOF response-function model used in the projector and backprojector","marker":"[39]"}],"fun_headline_variants":["Flat PET panels rival rings with 40 ps timing","Ultrafast TOF lets planar PET geometry match ring quality","Simulation: dual-panel PET with 40 ps TOF rivals ring PET","30 ps TOF: 20% efficiency matches full 40 ps noise","Panel PET: TOF compensates for missing angular coverage"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that real detector timing can be represented as a single Gaussian blur and that lost detection efficiency acts purely as random deletion of coincidence events; if actual BGO-based detectors show timing tails, misidentified events, or efficiency losses that are correlated with timing or position, the 40 ps requirement and the 20% noise-equivalence threshold would shift.","fun_headline_variants_meta":{"raw":{"variants":["Flat PET panels rival rings with 40 ps timing","Ultrafast TOF lets planar PET geometry match ring quality","Simulation: dual-panel PET with 40 ps TOF rivals ring PET","30 ps TOF: 20% efficiency matches full 40 ps noise","Panel PET: TOF compensates for missing angular coverage"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000456,"raw_usage":{"total_tokens":2375,"prompt_tokens":1117,"completion_tokens":1258,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":733,"completion_tokens_details":{"reasoning_tokens":1168}},"tokens_in":733,"tokens_out":1258,"duration_ms":11787,"temperature":1.0,"reasoning_tokens":1168,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T16:54:03.505909+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Build a BGO-window multi-anode MCP-PMT panel and measure its coincidence time distribution and Cherenkov-triggered event fraction under realistic optical conditions: if the timing histogram has non-Gaussian tails wider than the simulated Gaussian smearing, or if fewer than roughly one in five coincidences is Cherenkov-triggered at the operating threshold, the simulated 40 ps and 20%-efficiency breakpoints would not transfer to hardware. A simulation with full optical photon transport, which the paper explicitly leaves out, would be a direct numerical check.","supporting_citations":[{"cited_title":"Coincidence time resolution of 30 ps FWHM using a pair of Cherenkov -radiator-integrated MCP-PMTs,","cited_arxiv_id":null,"evidence_quote":"provides the 30 ps coincidence time resolution benchmark from Cherenkov-radiator-integrated MCP-PMT pairs that sets the TOF targets"},{"cited_title":"Ultrafast timing enables reconstruction-free positron emission imaging,","cited_arxiv_id":null,"evidence_quote":"demonstrated reconstruction-free cross-sectional imaging with an ultrafast TOF detector pair, motivating the panel geometry"},{"cited_title":"Cherenkov radiation –based coincidence time resolution measurements in BGO scintillators,","cited_arxiv_id":null,"evidence_quote":"provides measured Cherenkov-based coincidence timing and event-rate fractions in BGO detectors, supporting the 40% RDE value"},{"cited_title":"Dual -ended readout of bismuth germanate to improve timing resolution in time-of-flight PET,","cited_arxiv_id":null,"evidence_quote":"gives the fraction of coincidences triggered by Cherenkov photons under dual-ended BGO readout, another basis for the RDE values"},{"cited_title":"Multipanel limited angle PET system with 50 ps FWHM coincidence time resolution: a simulation study,","cited_arxiv_id":null,"evidence_quote":"the prior simulation study of a multi-panel limited-angle TOF PET that this work extends with BGO and ultrafast timing"},{"cited_title":"Geant4 developments and applications,","cited_arxiv_id":null,"evidence_quote":"the Monte Carlo toolkit used to simulate all detector geometries and acquisitions"},{"cited_title":"Design study of a brain-dedicated time- of-flight PET system with a hemispherical detector arrangement,","cited_arxiv_id":null,"evidence_quote":"supplies the in-house list-mode MLEM reconstruction code with detector response and TOF modeling"},{"cited_title":"High-resolution image reconstruction method for time - of-flight positron emission tomography,","cited_arxiv_id":null,"evidence_quote":"provides the TOF response-function model used in the projector and backprojector"}],"review_version":1}