{"id":"059a1337-e1db-4893-9d04-8a071ec973da","arxiv_id":"2412.11593","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"An offline stochastic resampling algorithm, adapted from neutron spectroscopy, corrected pileup-distorted microdosimetric spectra from a diamond detector in carbon-ion beams, reducing mean-value differences to a low-flux reference from about 3-6% to about 0.3-1%.","lead":"This paper tests a computer method that simulates and removes pulse pileup in microdosimetric measurements of clinical carbon-ion beams, using data from a diamond detector at the MedAustron facility. If it works, clinics and researchers could measure radiation quality spectra at realistic dose rates without specialized digital electronics.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Correction accuracy is evaluated against low-flux spectra that are admittedly not pileup-free; without a kHz validation for the 120 MeV/u setups, fitted ατ and the <1% residuals may be biased.","rationale":"The paper is a feasibility study of a known resampling algorithm applied to solid-state microdosimeters. The method is clearly described, the pulse shape is measured rather than assumed, and the forward simulation visibly reproduces the high-flux double-sum peak. However, the central quantitative claim—that iterative correction reduces differences to under 1%—compares corrected high-flux spectra to low-flux spectra that Section 3 explicitly calls \"not entirely pileup-free.\" The only independent low-flux-versus-very-low-flux check is for one 238.6 MeV/u configuration. This is exactly the weakest link: the reference endpoint is both the fitting target for ατ and the validation target for correction. If the 120 MeV/u low-flux spectra contain residual pileup, the reported accuracy is relative to the wrong ground truth. I do not regard this as fatal: the 238.6 MeV/u check suggests the 10%-rate endpoint can be close to pileup-free, and the estimated Reff for 120 MeV/u is lower, so residual pileup may be smaller. But the paper presents no direct evidence for that, and the reader's conditional verdict appropriately demands it. A single additional kHz measurement for the 120 MeV/u configurations would settle the issue. No data or code are provided, and no Monte Carlo uncertainty is quoted, but these are secondary to the endpoint validation.","tokens_in":13171,"tokens_out":5548,"duration_ms":46550,"concrete_test":"Measure a kHz-rate (or lower) spectrum for the 120 MeV/u plateau and Bragg-peak configurations with the same detector, gain, and readout chain, and compare it bin-wise to the low-flux 4.3 × 10^6 s^-1 spectrum using the paper's metric P(Δk_i)^2 and yF/yD. If the difference is comparable to the 238.6 MeV/u check, the endpoint is adequate and the correction stands; if non-negligible, re-fit ατ against the kHz reference and re-compute Table 1 to quantify the bias.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing assumption is that the low-flux spectra used as the pileup-free endpoint are themselves free of non-negligible pileup. Section 3 states they are \"not entirely pileup-free\" and cites a single very-low-flux kHz check for 238.6 MeV/u (P(Δk_i)^2 ≈ 0.02); no such check is presented for the 120 MeV/u plateau or Bragg-peak configurations, which use a different gain, different energy cutoff, and different Reff. If residual pileup at 4.3 × 10^6 s^-1 is not negligible there, then (i) the fitted ατ (0.1875/0.3) is biased, because the forward simulation starts from an already-contaminated spectrum, and (ii) the reported reductions to (0.28 ± 0.08)% (yF) and (0.97 ± 0.66)% (yD) measure agreement with a contaminated reference rather than with the true pileup-free spectrum. The paper's central claim of accurate offline correction therefore rests on an unvalidated endpoint for three of the four presented setups.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper adapts a stochastic resampling algorithm from Langen et al. to simulate and correct pulse pileup in microdosimetric pulse-height spectra acquired with a diamond microdosimeter in clinical carbon-ion beams at MedAustron. The method models pulse arrivals as a Poisson process, uses a measured semi-Gaussian pulse shape and a fitted pileup magnitude ατ, and corrects high-flux spectra by subtracting a simulated pileup component either in a single step or iteratively. The authors compare high-flux (4.3×10^7 s^-1) and low-flux (4.3×10^6 s^-1) measurements for four configurations (120 MeV/u plateau and Bragg peak; 120 MeV/u and 238.6 MeV/u plateau from a second campaign) and report that forward simulation reproduces the high-flux spectra from the low-flux spectra, and that iterative and single-step corrections reduce the relative differences in mean lineal energies to sub-percent levels.","tokens_in":13314,"tokens_out":6337,"duration_ms":52207,"significance":"If the result holds, the paper offers a practical offline alternative to online pileup rejection for solid-state microdosimeters at clinical ion-beam rates, addressing a genuine need where conventional pileup rejection is insufficient. The strongest evidence is the forward simulation: starting from measured low-flux spectra, the algorithm with a measured pulse shape and a single fitted pileup magnitude reproduces the measured high-flux spectra to within a few percent in all four configurations, which is a nontrivial test of the pileup model. The paper is also transparent about the main limitation, namely that ατ cannot be predicted from particle rates via Eq. (2) and must be established empirically for each setup. The contribution is incremental, being an adaptation of an existing algorithm, but the experimental dataset and the fidelity of the forward simulation make it a useful feasibility study for the microdosimetry community. The main weaknesses are the incomplete validation of the low-flux endpoint for three of the four setups and the partially circular evaluation of the iterative correction.","major_comments":[{"comment":"The low-flux spectra at 4.3×10^6 s^-1 are adopted as the pileup-free endpoint for all four setups, yet the only validation against very-low-flux (kHz) data is reported for the 238.6 MeV/u plateau case (P(Δk_i)^2≈0.02, Fig. 3A). The 120 MeV/u plateau and Bragg-peak spectra (Fig. 4) were measured with a different amplifier gain, different lower cutoffs (y0=48 and 99.6 keV/µm versus 31.8 keV/µm), and different effective rates, so the statement that the low-flux spectra are 'not entirely pileup-free' leaves open the possibility that residual pileup at 4.3×10^6 s^-1 is non-negligible for those setups. Because the fitted ατ (0.1875 for Fig. 4) is determined by simulating pileup from these low-flux spectra, any residual pileup biases both the simulation and the evaluation, and the reported reduction to (0.28±0.08)% in yF and (0.97±0.66)% in yD would measure agreement with a contaminated reference. Please provide a kHz-rate validation for the 120 MeV/u setups, or a quantitative upper bound on the residual pileup probability using Eq. (1) with the established ατ and τ.","section":"Section 3, low-flux endpoint"},{"comment":"The evaluation is partially circular for the iterative correction: the algorithm is iterated until convergence with the measured low-flux spectrum, and Table 1 then reports the difference between the corrected spectrum and that same low-flux reference. The non-circular evidence is the forward simulation, which uses the same fitted ατ to add pileup to the low-flux spectrum and reproduces the measured high-flux spectrum to within 0.02–0.72% in yF and 0.72–2.60% in yD; this should be presented as the primary validation of the pileup model. The single-step correction, which uses ατ predetermined from the forward simulation rather than from matching the corrected spectrum to the low-flux target, is a fairer test of the correction accuracy and should be reported alongside the iterative results as the main quantitative claim.","section":"Section 4 and Table 1"},{"comment":"The paper acknowledges that the rate-based estimate of ατ from Eq. (2) (0.126 for 120 MeV/u and 0.254 for 238.6 MeV/u) does not match the values needed for correction (0.1875/0.3 and 0.6, respectively). Since ατ is therefore a fitted parameter per setup, the method as demonstrated requires a low-flux reference spectrum for calibration, which is in tension with the abstract's claim that the method is useful 'in situations where a direct pileup-free measurement is currently not practicable.' Please clarify how ατ would be established in practice for a new clinical setup without such a reference, or explicitly reframe the contribution as a proof-of-principle requiring a one-time calibration.","section":"Section 4, Eq. (2)"}],"minor_comments":[{"comment":"In the paragraph on internal measurements, the text states that 'the low flux setting is thus approximately 4.3·10^7 s^-1', but Figure 3 and Section 3 indicate the low-flux rate is 4.3·10^6 s^-1; please correct this typo.","section":"Section 2, Materials and Methods"},{"comment":"The notation P(Δk_i)^2 is used to quantify the difference between normalized low-flux and very-low-flux spectra, but the definition of Δk_i is not given; please define it explicitly (e.g., bin-wise count difference).","section":"Section 3, Figure 3A"},{"comment":"Table 1 reports relative differences to several decimal places without any statement of counting-statistics or systematic uncertainties; please add at least a rough uncertainty estimate for the mean values or explain why they are omitted.","section":"Table 1"},{"comment":"Equation (3) describes the semi-Gaussian pulse shape as V_out(t) = (V_max/n!) (t/T)^n exp(-t/T), but with this normalization the maximum of the pulse is V_max n^n e^{-n}/n!, not V_max; please clarify whether V_max denotes the peak amplitude or a scaling constant.","section":"Equation (3)"},{"comment":"The sentence 'For a different amplifier gain and particle energy, new ατ needed to be established' would benefit from a specification of which of the four spectra share the same amplifier gain and which require recalibration, since Figures 4 and 5 appear to come from two different measurement campaigns.","section":"Section 4"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the scope of Physics in Medicine and Biology. In revision, the authors should de-emphasize the circular iterative-correction evaluation in favor of the forward-simulation and single-step results, and they should address the missing low-flux-endpoint validation for the 120 MeV/u setups with either new kHz-rate data or a quantitative residual-pileup bound. The latter point is the main barrier to accepting the quantitative accuracy claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a solid feasibility demonstration, not a new method. The stochastic resampling algorithm is Langen et al.'s; the contribution is adapting it to a diamond microdosimeter in clinical carbon-ion beams and testing it on four measurement pairs at MedAustron. That is a legitimate cross-application with real new data. The paper is clearly written, the methods are described in enough detail to reproduce, and the authors are honest about the main limitation: alpha*tau must be fitted per setup because the Poisson rate-based estimate from Eq. (2) does not give usable values. They show that with careful choice of alpha*tau the resampling can reproduce the pileup in the high-flux spectra from the low-flux spectra, and that applying the correction reduces the average relative differences in yF and yD from a few percent to below a percent. Using one alpha*tau for both plateau and Bragg peak at 120 MeV/u is a nice touch.\n\nThe soft spots are real but not fatal. First, the evaluation is circular in a mild sense: the same low-flux spectra used to pick alpha*tau and define convergence are the reference against which success is judged. A true independent check would be a kHz-rate measurement for the 120 MeV/u setups, not just the single 238.6 MeV/u validation reported. The paper actually says the low-flux spectra are 'not entirely pileup-free,' so the endpoint is not pristine. Second, alpha*tau is a free parameter fitted per setup; the predictive content is limited, since a new beam or gain requires re-fitting. Third, there are no error bars on the corrected spectra, so the quoted sub-percent residuals are hard to interpret. None of these kill the paper; they make it a conditional demonstration rather than a validated procedure.\n\nWho should read this: people working on microdosimetry for ion therapy, detector physicists dealing with pileup in spectroscopy, and anyone planning QA measurements with solid-state microdosimeters. It deserves a serious referee—the experiment is well executed and the writing is transparent. A referee should ask for the kHz validation for the 120 MeV/u case, a sensitivity analysis on alpha*tau, and a clearer statement of how much the result depends on the chosen low-flux endpoint.","headline":"A solid feasibility study of offline pileup correction for diamond microdosimeters; the method works in the four demonstrated cases but rests on a partially validated reference endpoint and a fitted pileup parameter.","tokens_in":13931,"tokens_out":2558,"would_cite":true,"duration_ms":21993,"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":"Offline stochastic resampling corrects pulse pileup in microdosimetric spectra at clinical ion-beam rates, reducing mean lineal energy differences to below one percent.","keywords":["microdosimetry","microdosimetric spectrum","pulse pileup","pileup correction","stochastic resampling","solid-state microdosimeter","clinical ion beams","lineal energy"],"falsifier":"Measure each of the four setups again at a true kHz-rate pileup-free reference and compare the corrected high-flux spectra to it; if the relative differences in $\\bar{y}_F$ and $\\bar{y}_D$ exceed the claimed $(0.28 \\pm 0.08)\\%$ and $(0.97 \\pm 0.66)\\%$, or if the low-flux spectra differ from the kHz reference by more than the correction's residual, the endpoint assumption is false and the correction is biased.","tokens_in":12901,"feed_emoji":"⚛️","tokens_out":6782,"duration_ms":52578,"temperature":0.7,"pith_summary":"This paper argues that pulse pileup, which distorts microdosimetric spectra recorded with solid-state detectors in clinical ion beams, can be corrected offline with a stochastic resampling algorithm rather than prevented in hardware. The algorithm simulates the pileup process from a low-flux reference spectrum, determines a pileup vector, and subtracts it from the measured high-flux spectrum iteratively. In four measurements with carbon ions at clinical rates, the correction reduced the average relative differences in the mean lineal energies $\\bar{y}_F$ and $\\bar{y}_D$ from $(3.57 \\pm 1.92)\\%$ and $(5.80 \\pm 2.73)\\%$ to $(0.28 \\pm 0.08)\\%$ and $(0.97 \\pm 0.66)\\%$. If this holds, microdosimetric characterization can be performed under true clinical beam conditions using conventional analog readout, supporting more accurate radiobiological modeling and quality assurance.","feed_headline":"Resampling fixes pileup in ion-beam microdosimetry","feed_subtitle":"Post-processing cuts average differences in mean lineal energies from 3.6% and 5.8% to 0.28% and 0.97%.","key_machinery":"The load-bearing object is the stochastic resampling algorithm, which treats the pulse train as a Poisson process with a pileup magnitude $\\alpha\\tau$, where $\\tau$ is the pileup resolution time fixed by the shaping amplifier's unit pulse shape and $\\alpha$ is an effective particle rate fitted per setup. It samples pulse heights from the measured spectrum, superposes scaled unit pulse shapes separated by random time intervals, and builds a simulated pileup spectrum; subtracting the bin-wise difference between the input and simulated spectrum produces a corrected spectrum, and the process is repeated until convergence. The unit pulse shape was measured from oscilloscope waveforms, giving a Gaussian pulse with shaping time $T = 1.7\\,\\mu\\text{s}$ and $\\tau = 7.5\\,\\mu\\text{s}$.","core_discovery":"The central claim is that a resampling algorithm, originally developed for neutron spectra from tissue-equivalent proportional counters, can realistically recreate and then correct the pulse pileup seen in solid-state microdosimeters exposed to clinical ion beams. The paper demonstrates this by showing that pileup simulated from low-flux spectra reproduces the double-sum peaks and tail distortions of high-flux measured spectra, and that the iterative correction brings high-flux spectra into agreement with low-flux reference spectra. Quantitatively, the average relative difference between corrected and low-flux spectra falls to $(0.28 \\pm 0.08)\\%$ for the frequency-mean lineal energy $\\bar{y}_F$ and $(0.97 \\pm 0.66)\\%$ for the dose-mean lineal energy $\\bar{y}_D$, starting from uncorrected differences of $(3.57 \\pm 1.92)\\%$ and $(5.80 \\pm 2.73)\\%$.","pith_inferences":["A testable extension would apply the resampling correction to silicon or SiC microdosimeters; because the method only needs a unit pulse shape and a fitted $\\alpha\\tau$, it should transfer whenever the shaping response is stable.","The need to fit $\\alpha$ per setup implies the method would benefit from an independent measurement of arrival-time statistics; comparing fitted $\\alpha\\tau$ values with beam-monitor timing data could expose how bunch structure biases the Poisson assumption.","Replacing the low-flux reference with a truly pileup-free kHz reference would let the same algorithm estimate residual pileup in the low-flux endpoint, potentially explaining the remaining $\\sim1\\%$ disagreement in $\\bar{y}_D$.","The residual after correction sets a floor on achievable accuracy; tracking how that residual varies with spectral shape could guide whether a single-step or iterative correction should be trusted for a given radiation quality."],"forward_implications":["Once $\\alpha\\tau$ and the unit pulse shape are established for a readout setup, high-flux spectra can be corrected in a single step, allowing accurate microdosimetry at clinical dose rates with standard analog electronics.","The correction removes double-sum peaks and spectral broadening, so the frequency-mean and dose-mean lineal energies of corrected spectra match low-flux measurements to within about one percent.","For repeated quality-assurance measurements under fixed conditions, correction parameters can be pre-determined, enabling short, high-statistics acquisitions that would otherwise be unusable due to pileup.","The same pileup magnitude corrects spectra at different depths along the depth-dose curve for a fixed particle energy and readout chain, covering both plateau and Bragg-peak measurements."],"supporting_citations":[{"why":"Supplies the stochastic resampling algorithm for pileup simulation and iterative correction that this paper adapts to solid-state microdosimeters.","marker":"Langen et al. 2002"},{"why":"Provides the Poisson-statistics derivation of pileup probability given rate and resolution time, the basis for the pileup magnitude $\\alpha\\tau$.","marker":"Knoll 2000"},{"why":"Documents the single-crystal diamond microdosimeter used for the measurements.","marker":"Verona et al. 2018"},{"why":"Establishes the diamond-detector microdosimetric characterization of clinical carbon-ion beams that this work extends.","marker":"Magrin et al. 2020"},{"why":"Supplies the carbon-edge calibration method used to assign lineal energy to the measured pulse-height spectra.","marker":"Meouchi et al. 2022"},{"why":"Defines the microdosimetric quantities $\\bar{y}_F$, $\\bar{y}_D$, and $yd(y)$ used to evaluate correction accuracy.","marker":"Braby et al. 2023"},{"why":"Provides evidence of non-Poissonian bunch structure in clinical beams, supporting the paper's limitation that $\\alpha$ must be determined empirically.","marker":"Data et al. 2024"}],"fun_headline_variants":["Offline pileup correction improves microdosimetric accuracy","Resampling tames pileup in clinical ion-beam microdosimetry","From neutron to ion: a resampling fix for microdosimetry pileup","Cut microdosimetry errors tenfold with offline pileup correction","Simulate and correct pileup offline for sharper ion-beam microdosimetry"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The correction relies on the assumption that spectra recorded at one-tenth of the clinical particle rate are effectively free of pileup and can serve as the target for all four setups, even though the paper validates this only for one carbon-ion case and calls the low-rate spectra 'not entirely pileup-free'.","fun_headline_variants_meta":{"raw":{"variants":["Offline pileup correction improves microdosimetric accuracy","Resampling tames pileup in clinical ion-beam microdosimetry","From neutron to ion: a resampling fix for microdosimetry pileup","Cut microdosimetry errors tenfold with offline pileup correction","Simulate and correct pileup offline for sharper ion-beam microdosimetry"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000625,"raw_usage":{"total_tokens":2929,"prompt_tokens":1018,"completion_tokens":1911,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":634,"completion_tokens_details":{"reasoning_tokens":1821}},"tokens_in":634,"tokens_out":1911,"duration_ms":13515,"temperature":1.0,"reasoning_tokens":1821,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:47:04.607512+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure each of the four setups again at a true kHz-rate pileup-free reference and compare the corrected high-flux spectra to it; if the relative differences in $\\bar{y}_F$ and $\\bar{y}_D$ exceed the claimed $(0.28 \\pm 0.08)\\%$ and $(0.97 \\pm 0.66)\\%$, or if the low-flux spectra differ from the kHz reference by more than the correction's residual, the endpoint assumption is false and the correction is biased.","supporting_citations":[{"cited_title":"M., Binns, P","cited_arxiv_id":null,"evidence_quote":"Supplies the stochastic resampling algorithm for pileup simulation and iterative correction that this paper adapts to solid-state microdosimeters."},{"cited_title":"T., Rosenfeld, A., Verona, C., Verona Rinati, G., Palmans, H","cited_arxiv_id":null,"evidence_quote":"Supplies the carbon-edge calibration method used to assign lineal energy to the measured pulse-height spectra."},{"cited_title":"M., Mas Milian, F., Abujami, M., Bersani, D., Cerello, P., Donetti, M., Mazinani, M","cited_arxiv_id":null,"evidence_quote":"Provides evidence of non-Poissonian bunch structure in clinical beams, supporting the paper's limitation that $\\alpha$ must be determined empirically."}],"review_version":1}