{"id":"6bfb6d59-8aee-4d5c-98fc-3265dcc8a3a1","arxiv_id":"2504.15824","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A flexible YAG scintillating sheet and camera system provides linear, relative real-time 2D dose maps in non-homogeneous ultra-high dose rate electron beams, including on curved ex vivo subjects, but only as relative dosimetry with case-specific calibration.","lead":"This paper tests a flexible scintillating coating paired with a camera as a real-time 2D dosimeter for ultra-high dose rate electron beams used in FLASH radiotherapy research. It shows the coating responds linearly with number of pulses and with inverse-square distance, and can be molded onto curved animal subjects, though calibration and absolute dosimetry remain unsolved.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reported linearity may be partly manufactured by a ramp-up correction fitted to the same data, while the history-dependent sensitization (Sec. 4.2) is left unmodeled; the quantitative linearity and robustness claims are not independently established.","rationale":"The reader's weakest assumption identified the ramp-up correction and the history-dependent sensitization as the key fragility, and I agree that this is the central load-bearing issue. My stress-test sharpens that concern: the correction is not merely difficult to generalize, it is fitted to a subset of the same data used to demonstrate linearity, so the reported R² values may be partly tautological. The per-pulse residual errors of 19.5–22.5% at low pulse counts, combined with the ascending-order irradiation sequence and the <5-minute inter-irradiation intervals in Section 4.2, create a genuine confound between dose linearity and cumulative sensitization. The ex vivo linearity measurement on the rat head applies the same characterization-derived correction without independent calibration, so it does not break the circularity. However, the paper is explicitly framed as a feasibility study, and its central qualitative claim—that a flexible YAG sheet with a triggered camera can provide relative 2D dosimetry on curved surfaces in UHDR beams—is supported by the images and the consistency with radiochromic film trends. The concern affects the strength of the quantitative linearity and robustness claims, not the basic feasibility conclusion. Therefore the appropriate verdict remains CONDITIONAL; my read does not change the reader's verdict. I mark agreement as 'partial' because the reader emphasized generalizability of the correction, while I emphasize the in-sample fitting/validation circularity as the sharper failure mode.","tokens_in":14407,"tokens_out":4764,"duration_ms":49926,"concrete_test":"Perform a strict cross-validation on the existing dose-linearity data: build k(x) using only the 40-, 60-, and 80-pulse traces from the characterization set, then apply it to the held-out 5-, 15-, 30-, 50-, 70-, and 90-pulse traces. Report the held-out R² and the per-pulse residual at each pulse count. If the held-out per-pulse deviation at low counts exceeds 10%, or if the R² drops below 0.99, the reported linearity is not independent of the calibration. In addition, repeat the pulse-count series in randomized order with inter-irradiation intervals of at least 10 minutes; if the per-pulse signal still varies with pulse count, the ascending-order sensitization confound is confirmed and the correction is not portable across irradiation histories.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim is that the scintillating sheet–camera system responds linearly with delivered pulse count (R²>0.998 without bolus, R²>0.995 with bolus) and is robust to setup variation. The most load-bearing assumption is that the ramp-up correction k(x) removes a physical transient rather than imposing the reported linearity. In Section 2.1.1, k(x) is defined as the inverse of the average of six normalized time traces from deliveries of 40–90 pulses, i.e., it is derived from a subset of the same dose-linearity dataset that it is later used to validate. Since k(x) normalizes each pulse to the final-pulse signal, summing the corrected frame signals for an N-pulse delivery yields approximately N times a constant whenever the pulse-to-pulse shape matches the calibration average. The R² for total signal versus pulse count is therefore largely insensitive to per-pulse errors. The data in Figure 4 (lower panel) show exactly such errors: average signal per pulse was 19.5% (C-blue) and 22.5% (Basler) lower at low pulse counts, and the paper reports saturation-induced over- and under-correction for the C-blue camera. Section 4.2 further documents a history-dependent sensitization that is not included in k(x): the linearity experiments were run in ascending pulse order, and the signal per pulse increased when subsequent irradiations were less than 5 minutes apart. This creates a confound between dose linearity and cumulative sensitization. The ex vivo linearity results (R²=0.997 and 0.998) used the same characterization-based correction on a curved surface without an independent calibration, so they do not resolve the concern. Until k(x) is validated on independent data, or the linearity experiment is repeated in randomized order with controlled time intervals, the quantitative linearity and robustness claims do not establish a setup-independent linear dose response.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript characterizes a flexible 2D scintillating-coating camera system for real-time relative dosimetry in an ultra-high dose rate (UHDR) electron beam from a converted Varian Trilogy linac. The authors investigate triggered versus free-running acquisition, apply a ramp-up correction to the pulse-by-pulse signal, and report linearity with delivered pulse count (R²>0.998 without bolus, R²>0.995 with bolus), inverse-square dependence on SSD (R²>0.963), robustness to camera setup variations within 5%, and preservation of linearity in three ex vivo preclinical geometries (rat brain, mouse hindlimb, whole-body mouse). They also document challenges including signal saturation in the C-blue camera, reduced signal-to-background ratio under ambient light, blind spots on curved surfaces, and a history-dependent sensitization of the scintillating sheet. The paper explicitly limits the system to relative dosimetry and notes that recalibration is required when the setup changes.","tokens_in":14636,"tokens_out":3001,"duration_ms":29512,"significance":"If the quantitative claims are sustained, this is a useful feasibility demonstration of real-time 2D scintillation dosimetry for non-homogeneous UHDR preclinical beams, addressing an unmet need in FLASH radiobiology. The work is careful in several respects: it identifies the dead-time problem in free-running acquisition, uses triggered acquisition with an extended integration window, provides side-by-side comparison with radiochromic film, and reports realistic ex vivo scenarios with curved surfaces. The supplementary material, including the time-trace data and sensitization plot, supports transparency. However, the central linearity and robustness claims are weakened by two confounds—the ramp-up correction derived from the same dataset and a cumulative sensitization effect—so the quantitative R² values and 'within 5%' robustness statement are not independently established as stated.","major_comments":[{"comment":"The ramp-up correction k(x) is constructed from an average of six normalized time traces from deliveries of 40-90 pulses, which is a subset of the same dose-linearity dataset that Figure 4 uses to claim linearity with pulse count. Since k(x) normalizes each pulse to the final-pulse signal, summing corrected frames for an N-pulse delivery yields approximately N times a constant whenever the pulse-to-pulse shape matches the calibration average. Consequently, the R²>0.998 linearity in total signal versus pulse count is largely insensitive to per-pulse deviations, and the lower panel of Figure 4 shows such deviations: the average signal per pulse is lower by 19.5% (C-blue) and 22.5% (Basler) at low pulse counts. The linearity claim therefore does not demonstrate accurate per-pulse dose proportionality; it only shows that the integrated corrected signal scales with pulse number after a correction that is fitted to the same data. To make the claim load-bearing, the authors should validate the correction on independent irradiations (e.g., a separate calibration run or comparison with film at low pulse counts) or explicitly restrict the claim to total integrated signal with the caveat that per-pulse accuracy is limited below 20-40 pulses.","section":"Section 2.1.1, Equations (1)-(2) and Figure 4"},{"comment":"The history-dependent sensitization described in Section 4.2 is a direct confound for the dose-linearity experiment. The linearity measurements were performed in ascending pulse order (1 to 90 pulses) with irradiations often less than 5 minutes apart, and Figure A6 shows that the signal per pulse increases when subsequent irradiations are close in time, which the authors attribute to filling of OSL-like traps. This means that the reported increase in total signal with pulse count could be partly due to cumulative sensitization rather than an intrinsic linear dose response, and the same confound affects the ex vivo linearity results where the ramp-up correction from the characterization was reused. The authors propose a plausible mechanism but do not quantify its contribution or test it with a randomized-order control in the main dataset. I request either a controlled experiment (e.g., interleaved pulse counts with long resting intervals) or an explicit sensitivity analysis showing that the linearity R² values are not driven by this effect.","section":"Section 4.2 and Figure A6"},{"comment":"The robustness claim of 'within 5% for camera setup differences' is based on data from which one C-blue outlier, deviating 25.8% from the average, was excluded. No predefined exclusion criterion is given, so the claim is not falsifiable as stated. The authors should report all data points, state the exclusion rule, or show that the conclusion is unchanged with and without the outlier. In addition, the text acknowledges that the assumption of identical dose per pixel across rotations is unlikely to hold because of the Gaussian beam profile and backscatter variations, so the 5% figure conflates setup variability with delivery variability; this limitation should be stated alongside the robustness claim.","section":"Section 3.3, Figure A4"},{"comment":"The saturation of the C-blue camera during the construction of the ramp-up correction causes over-correction for 5 and 90 pulses and under-correction for 40 pulses, as the authors themselves note in Section 3.1. This directly affects the C-blue linearity results in Figure 4a, which are reported with R²>0.999. The authors should either re-analyze the C-blue data with a non-saturated calibration trace or state which of the reported linearity metrics are affected by the saturation and why the R² values remain meaningful despite the known correction error.","section":"Section 3.1, Figure 3c"}],"minor_comments":[{"comment":"There are several typographical errors: 'prelcinical' in the abstract, 'investigation' used as a verb in Section 2.2, 'T able 1' in the manuscript text, and 'V erdi' in the CRediT statement. These should be corrected.","section":"Abstract and Section 2.2.1"},{"comment":"The frame rates are reported as '554.9 and 1000.0 kHz' for the C-blue and Basler cameras; given the camera specifications (frame rates of 1594.7 and 751.9 Hz), these values should be in Hz, not kHz.","section":"Section 2.2.1"},{"comment":"The last three data points in the inverse-square plot for the C-blue camera deviate from linearity, and the text attributes this to a change in camera-sheet distance with SSD. Since these points are part of the reported R²>0.963, the authors should clarify whether the R² is computed over all data points or only over the linear region, and if the latter, state the range explicitly.","section":"Section 3.2 and Figure 6a"},{"comment":"The text refers to 'supplementary Figure A1', 'A2', and 'A3', but the figures are located in an appendix after the main text and are not labeled as supplementary in the appendix header. Please harmonize the referencing (e.g., 'Appendix A, Figure A1') and ensure all cited figures are present and legible.","section":"Section 3.3 and Appendix A"},{"comment":"The claim that the decay time of the Ce-doped YAG sheet exceeds the inter-pulse time is unexpected because bulk YAG:Ce has a ~100 ns decay; the authors propose an interaction with the silicon matrix as a hypothesis. This is an important observation for the dosimetric applicability, and a quantitative estimate of the effective decay constant from the time traces would strengthen the discussion.","section":"Section 4.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid empirical characterization with valuable negative results (triggering, saturation, sensitization, bolus issues), but the headline quantitative claims—linearity R² values and 5% robustness—are not independently supported because of the self-referential ramp-up correction and the history-dependent sensitization. I would encourage the editor to request a revised manuscript that either provides an independent validation of the correction or narrows the claims to the integrated-signal regime where the correction is valid. The work is not 'reject' material; the issues are fixable within the scope of the manuscript."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this is a genuinely useful feasibility study: it shows a flexible YAG scintillator sheet plus a triggered camera can track pulse-by-pulse signal on curved ex vivo subjects in a non-homogeneous UHDR electron beam, with film comparison. Second, the quantitative linearity claim (R² > 0.998) is partly built into the analysis. The ramp-up correction k(x) is derived by averaging six time traces from 40–90 pulse deliveries, then applied back to the same dataset. Since k(x) normalizes each pulse to the final-pulse signal, summing corrected frames for N pulses will look linear whenever the pulse-to-pulse shape matches that average. That doesn't falsify the system, but it means the R² isn't independent evidence of linearity.\n\nWhat the paper does well: the authors are unusually candid. They report the per-pulse signal deviations (up to 22.5% at low pulse counts), the C-blue saturation leading to over- and under-correction, the history-dependent sensitization in Section 4.2, and they explicitly restrict the system to relative dosimetry without bolus. The ex vivo applications—rat brain, hindlimb, whole-body mouse—show practical challenges (blind spots, light scattering through bolus, field size limits) that people in the field will want to know.\n\nSoft spots, in proportion. The ramp-up correction is calibrated and validated on the same dataset; no independent test is shown. The linearity experiment was run in ascending pulse order, and the sensitization section documents that signal per pulse increased for irradiations less than 5 minutes apart—so dose linearity and cumulative sensitization are confounded. The robustness claim (within 5%) excludes one outlier without a pre-defined criterion. No code or raw data are provided, so the pipeline can't be independently checked. These are real issues, but they are addressable and they don't sink the qualitative feasibility conclusion.\n\nWho gets value: anyone setting up real-time 2D dosimetry on a converted linac for preclinical FLASH work. The practical setup advice and the honest error discussion are worth the read. The quantitative claims should be treated as provisional until the correction is validated on independent data.\n\nRecommendation: yes, send this to peer review. A good referee can push for a randomized pulse-order experiment, a pre-registered outlier rule, and data release. That would turn a solid feasibility paper into a referenceable one.","headline":"A useful, honest feasibility study for flexible 2D scintillation dosimetry in UHDR beams, but the headline linearity numbers are softer than they look because the correction factor is fit to the same data.","tokens_in":15395,"tokens_out":2730,"would_cite":true,"duration_ms":24329,"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 flexible scintillating sheet paired with a CMOS camera can map ultra-high-dose-rate electron beams in real time, preserving linearity with pulse count on flat and curved surfaces.","keywords":["FLASH radiotherapy","UHDR dosimetry","2D dosimetry","scintillating sheet","real-time dosimetry","preclinical electron beams","YAG:Ce","CMOS camera"],"falsifier":"Deliver the same total number of pulses in randomized sequences (for example, 5, 15, 40, and 90 pulses in different orders) to a fresh sheet at fixed geometry, with the camera gain lowered to avoid saturation; if the corrected signal per pulse varies by more than the claimed few percent across sequences, the ramp-up correction is history-dependent and the reported linearity does not generalize. Monitoring the first-pulse signal over time while varying inter-irradiation gaps would directly test the sensitization mechanism.","tokens_in":14144,"feed_emoji":"⚡","tokens_out":6451,"duration_ms":57166,"temperature":0.7,"pith_summary":"This paper aims to establish that a flexible scintillating sheet, read out by a triggered CMOS camera, can serve as a real-time two-dimensional dosimeter for non-homogeneous ultra-high dose rate (UHDR) electron beams in preclinical FLASH radiotherapy. The authors show that the light signal grows linearly with the number of delivered pulses on flat surfaces ($R^2>0.998$), remains linear when up to 3 cm of transparent bolus is added ($R^2>0.995$), follows the inverse-square law with source-to-surface distance ($R^2>0.963$), and stays linear on curved ex vivo rat and mouse setups ($R^2 \\ge 0.997$). They also demonstrate that triggered acquisition avoids missed pulses and that setup-induced variations stay within 5%. The stated value is an unmet need: biological FLASH experiments require per-pulse, spatially resolved dose information in beams that are not flat, and this system is presented as a workable relative-dosimetry answer.","feed_headline":"Bendable scintillator tracks FLASH electron dose in 2D","feed_subtitle":"The YAG coating stays linear with pulse count and dose rate on flat and curved surfaces for preclinical FLASH setups.","key_machinery":"The central object is a flexible sheet of cerium-doped yttrium aluminium garnet (YAG:Ce) crystals embedded in silicone (about 30% particle loading), which emits light in proportion to the radiation dose rate, plus a global-shutter CMOS camera that records the light patterns. The mechanism that makes pulse-resolved dosimetry work in this paper is triggered acquisition: the camera is synchronized to the linac's monitor-chamber trigger so every pulse is captured, avoiding the dead-time missed pulses of free-running mode. Because the sheet's decay and afterglow exceed the inter-pulse interval, the paper introduces a ramp-up correction $k(x)$, built as the inverse of an averaged normalized time trace, to convert the rising per-frame signal into a per-pulse signal. A projective transformation maps the camera view to a beam's-eye view. The linear response with pulse number, bolus thickness, and inverse-square SSD is what the machinery is used to establish.","core_discovery":"The central claim is that a camera-read flexible YAG:Ce scintillating coating provides pulse-resolved two-dimensional dose maps in a converted clinical linac's UHDR electron beam, with a response that is linear in delivered pulses and independent of average dose rate over the tested range. The paper demonstrates this in a flat-field characterization and then in three ex vivo preclinical geometries—rat brain, mouse hindlimb, and whole-body mouse—where the sheet is molded to curved surfaces and linearity is preserved. The authors further claim the system is robust to camera set-up differences (within 5%) and repeatable (within 2% across irradiations), and they identify the practical boundary: with small field sizes the light output is too low for usable maps, and a single camera leaves blind spots on curved surfaces. Their conclusion is not that the system is ready for absolute dosimetry, but that it is feasible for relative real-time 2D dosimetry in non-homogeneous UHDR beams, provided it is recalibrated per setup.","pith_inferences":["If the proposed trap-related sensitization is confirmed, a pre-irradiation filling protocol or per-pulse self-normalization could remove the need for a fixed ramp-up correction and move the system toward absolute dosimetry.","The authors' suggested multi-camera arrangement could not only remove blind spots but also allow 3D surface dose reconstruction on contoured subjects, extending the method beyond a single beam's-eye view.","With machine-log timestamps added to triggered frames, the system could recover the lost time information and report average dose rate per irradiation, closing the gap identified in the paper.","The same sheet-and-camera combination could plausibly be adapted to other pulsed UHDR modalities, such as pencil-beam scanning protons or very-high-energy electrons, where the trigger and decay trade-off is adjusted for those pulse structures."],"forward_implications":["Preclinical FLASH experiments can obtain real-time, spatially resolved relative dose maps of non-homogeneous UHDR electron beams, instead of relying only on passive film or point dosimeters.","Triggered acquisition resolves individual linac pulses, so dose per pulse and its spatial variation can be read directly during an irradiation.","Because the signal stays linear with bolus thickness up to 3 cm and follows the inverse-square law, the system can track dose buildup and beam divergence in realistic setups.","The flexible sheet's conformity to curved anatomy makes surface dosimetry possible for rat brain, mouse hindlimb, and whole-body mouse irradiations, with stability within a few percent.","For small fields (below about $0.7\\times 0.7$ cm² with the tested cameras), the light output is too low for reliable maps, so field-size constraints must be respected."],"supporting_citations":[{"why":"Supplies the prior characterization of the same sheet-camera approach in UHDR electron beams and the free-running comparison this work extends.","marker":"[23]"},{"why":"Provides the scintillating-sheet plus charge-monitor dosimetry method for UHDR very-high-energy electrons whose calibration sensitivity motivates this work's per-setup recalibration.","marker":"[22]"},{"why":"Shows a YAG-based point scintillator working in UHDR electron beams, supporting the material choice for the sheets.","marker":"[27]"},{"why":"Documents optically stimulated luminescence in Ce-doped YAG, the mechanism invoked to explain the observed temporal sensitization of the sheet.","marker":"[45]"},{"why":"Provides the reference decay-property context for Ce-doped garnet phosphors, used when interpreting the unexpectedly long afterglow that forces the ramp-up correction.","marker":"[41]"}],"fun_headline_variants":["Flexible scintillator enables real-time 2D FLASH dosimetry","Bendable YAG coating maps dose in curved FLASH beams","Pulse-linear scintillator sheets for UHDR electron beams","Ex vivo FLASH dose tracking with flexible 2D scintillator","Scintillating sheet reveals 2D dose in non-uniform FLASH"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The ramp-up correction assumes that every pulse produces the same light output for a given dose, independent of how many pulses were delivered before it; the paper's own data show deviations of up to roughly 20% for short irradiations and saturation artifacts for one camera, so the generality of the linearity claims rests on this assumption holding across setups.","fun_headline_variants_meta":{"raw":{"variants":["Flexible scintillator enables real-time 2D FLASH dosimetry","Bendable YAG coating maps dose in curved FLASH beams","Pulse-linear scintillator sheets for UHDR electron beams","Ex vivo FLASH dose tracking with flexible 2D scintillator","Scintillating sheet reveals 2D dose in non-uniform FLASH"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000285,"raw_usage":{"total_tokens":1766,"prompt_tokens":1122,"completion_tokens":644,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":738,"completion_tokens_details":{"reasoning_tokens":551}},"tokens_in":738,"tokens_out":644,"duration_ms":5835,"temperature":1.0,"reasoning_tokens":551,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:15:57.925752+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Deliver the same total number of pulses in randomized sequences (for example, 5, 15, 40, and 90 pulses in different orders) to a fresh sheet at fixed geometry, with the camera gain lowered to avoid saturation; if the corrected signal per pulse varies by more than the claimed few percent across sequences, the ramp-up correction is history-dependent and the reported linearity does not generalize. Monitoring the first-pulse signal over time while varying inter-irradiation gaps would directly test the sensitization mechanism.","supporting_citations":[{"cited_title":"A dose rate independent 2D Ce-doped YAG scintillating dosimetry system for time resolved beam monitoring in ultra-high dose rate electron “FLASH” radiation therapy","cited_arxiv_id":null,"evidence_quote":"Supplies the prior characterization of the same sheet-camera approach in UHDR electron beams and the free-running comparison this work extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows a YAG-based point scintillator working in UHDR electron beams, supporting the material choice for the sheets."},{"cited_title":"Infrared Stimulated Luminescence of Ce3+ Doped YAG Crystals","cited_arxiv_id":null,"evidence_quote":"Documents optically stimulated luminescence in Ce-doped YAG, the mechanism invoked to explain the observed temporal sensitization of the sheet."},{"cited_title":"Ce 3+-Doped garnet phosphors: composition modification, luminescence properties and applications","cited_arxiv_id":null,"evidence_quote":"Provides the reference decay-property context for Ce-doped garnet phosphors, used when interpreting the unexpectedly long afterglow that forces the ramp-up correction."}],"review_version":1}