{"id":"ff717021-d67d-4361-a9f9-3918a7e65f43","arxiv_id":"2501.04145","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Voxel-wise ortho-positronium lifetime imaging is feasible at 4 mm voxel size on a clinical LAFOV PET/CT scanner using 124I, with clear sample distinction when samples are separated.","lead":"This phantom study shows that voxel-wise positronium lifetime images can be obtained with a commercial long-axial field-of-view PET scanner using the isotope iodine-124. The result suggests that a tissue-microenvironment imaging signal, previously limited to specialized or low-resolution systems, could be measured on scanners already used in clinics.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Clinical feasibility claim rests on an activity concentration 3-20x higher than the thyroid-cancer values the authors themselves cite; the paper includes no measurement at those concentrations.","rationale":"The paper is a credible phantom feasibility study with a clear empirical result and openly deposited data. The whole-tube fits are precise, and the 4-mm voxel demonstration is a genuine step. The reader's conditional verdict is appropriate. I do not see a reason to reject the paper; the central claim is merely stronger than the evidence supports. The most load-bearing piece of that mismatch is the activity concentration: the authors' own Discussion admits that the used 232 kBq/ml exceeds the 10-70 kBq/ml expected in differentiated thyroid cancer metastases by a factor of 3-20, yet the Conclusion asserts clinical viability with respect to activity concentration. Even if TOF localization were perfect, the count statistics at 10 kBq/ml and 4-mm voxels in 15 minutes would likely be insufficient for the demonstrated precision. The reader identified TOF localization as the weakest assumption and mentioned the concentration mismatch in the rationale, but did not make it the primary attack. I therefore mark agreement as partial. My proposed check is a straightforward subsampling of the existing list-mode data, which would quantitatively settle whether the demonstrated voxel-level separation survives at clinically realistic concentrations. Since the paper is already CONDITIONAL, my read does not change the verdict; it sharpens the condition that must be satisfied before the clinical claim can be accepted.","tokens_in":15107,"tokens_out":6947,"duration_ms":70887,"concrete_test":"Using the deposited 3-gamma-event list-mode data from the separated-tube scan, randomly subsample events to effective concentrations of 70 kBq/ml and 10 kBq/ml (fractions 70/232 and 10/232) while preserving the 15-min live time; rerun the complete Bayesian voxel-fit pipeline (Eq. 1) for the 10.0, 7.1, and 4.0 mm voxels. Report the fraction of voxels passing the <20% background-error inclusion criterion and the tau3 bias/uncertainty per material. If the 4-mm voxel fits at 10-70 kBq/ml no longer separate the four sample lifetimes with errors comparable to Fig. 7, the 'activity concentration' component of the clinical-feasibility conclusion is not supported. An analytic fallback is to derive the count-rate scaling from the measured 3-gamma-event rate and compute the minimum scan time needed to match the demonstrated per-voxel precision at 10 kBq/ml.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The Conclusion (Sec. 4) claims feasibility 'with respect to... activity concentration', but the only measured concentrations are roughly 232 kBq/ml (Tab. 1: 1.12-1.44 MBq in ~5 ml). The Discussion explicitly states this 'is still somewhat higher than one could expect in a thyroid cancer patient' and cites Ref. [45] reporting 10-70 kBq/ml in differentiated thyroid cancer metastases. That is a factor of 3-20 above the cited clinical range. For a fixed scan time, 3-gamma-event counts per 4-mm voxel scale linearly with activity concentration, so a 15-min acquisition at 10 kBq/ml would have about 23x fewer counts (at 70 kBq/ml, about 3.3x fewer). The 4-mm voxel fits at 232 kBq/ml already show roughly 10% relative uncertainty and require a <20% background-error threshold for voxel inclusion; scaling by sqrt(counts) gives roughly 18-45% relative errors at clinical concentrations, with many voxels plausibly failing the inclusion criterion. No acquisition, subsampling, or simulation at 10-70 kBq/ml is reported. Thus the activity-concentration clause of the central claim is not demonstrated by the data, independently of the TOF-localization limitation. The whole-tube fits' <1.76% precision does not transfer to 4-mm voxels at clinical concentrations.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript reports phantom measurements of ortho-positronium (oPs) lifetime imaging on a commercial long-axial field-of-view PET/CT scanner (Biograph Vision Quadra) using 124I. Four tubes with different mixtures of Amberlite XAD4 and water, each spiked with 1.12–1.44 MBq of [124I]NaI (≈232 kBq/ml), were scanned in two configurations: separated by several centimeters (15 min) and taped together (40 min). The authors fit a Bayesian three-lifetime model to time-difference distributions of three-photon events localized by time-of-flight, obtaining whole-tube oPs lifetimes of 1.82–2.52 ns with <1.76% relative uncertainty, and voxel-wise lifetime maps at 10.0, 7.1, and 4.0 mm voxel sizes. The separated-tube images show visual differences in τ3 across tubes, while the taped-together images do not allow clear spatial distinction. The paper concludes that oPs lifetime imaging is feasible on a commercial PET/CT under clinically viable conditions with 124I.","tokens_in":15440,"tokens_out":10671,"duration_ms":97472,"significance":"If fully supported, this result would be significant: it would show that voxel-wise oPs lifetime imaging is possible on an unmodified clinical PET/CT scanner using a clinically available isotope (124I), without custom hardware, and the XAD4 phantom provides a simple recipe for intercomparison studies. Strengths include the high precision of the whole-tube fits, the systematic exploration of three voxel sizes, the use of Bayesian fitting with explicit priors, and the availability of evaluated data on Zenodo. The authors also honestly report the main limitations in the Discussion (activity concentration higher than clinical uptake, TOF localization as limiting factor, and poor separation in the taped-tube setup). However, the central conclusion overstates what is demonstrated: the scanned activity concentration is 3–20 times higher than the clinical values cited by the authors themselves, and the voxel-level distinction between similar samples is not statistically established. The clinical-feasibility claim therefore needs revision or additional supporting evidence.","major_comments":[{"comment":"The conclusion that oPs lifetime imaging is feasible 'with respect to ... activity concentration' is not supported by the data. The phantom activities in Table 1 (1.12–1.44 MBq in ~4.5–5.5 ml, ≈232 kBq/ml) are a factor of 3–20 higher than the 10–70 kBq/ml reported for differentiated thyroid cancer metastases in Ref. [45], which the Discussion itself cites as the clinically expected range. Since the number of 3γ events per voxel scales linearly with activity concentration, a 15-min acquisition at 10 kBq/ml would have roughly an order of magnitude fewer counts per voxel than the present data, and even at 70 kBq/ml about 3× fewer. At the current 4-mm voxel level the relative uncertainty is already ~10%; scaling by the square root of counts implies ~18–45% relative errors at clinical concentrations, with many voxels failing the 20% background-error inclusion criterion. No measurement, subsampling, or simulation at 10–70 kBq/ml is presented, so the activity-concentration clause of the central claim is not demonstrated.","section":"§3 Discussion / §4 Conclusions / Table 1"},{"comment":"The effective spatial resolution of the voxel-wise oPs lifetime images is not characterized and is likely far coarser than the 4-mm voxel size. Each 3γ event is localized along the LOR solely by TOF of the two annihilation photons; with a time bin width of 133 ps, the single-event localization uncertainty is on the order of c·Δt/2 ≈ 20 mm, several times the smallest voxel. The authors acknowledge in §3 that 'the limiting factor ... is likely the localization of the 3γE with TOF' and that events from tube walls or air may be present. This implies that 4-mm voxel lifetimes are averages over a much larger effective volume and are contaminated by neighboring materials—consistent with the failure of the taped-tube setup in Fig. 6. The abstract's claim that 'even with 4.0^3 mm^3 voxels the samples are clearly distinguishable' is only demonstrated for separated tubes with large interior regions; the authors should quantify the point-spread function of the lifetime images (e.g., edge profiles across the tube boundaries) or restrict the spatial-resolution claims accordingly.","section":"§1 Materials and methods / §3 Discussion"},{"comment":"The taped-together configuration is the closest analog in this study to the arrangement of tissues in clinical imaging, and it does not work: the authors state that the slices in Fig. 6 'do not allow for a clear spatial distinction of the four tubes' and describe the smaller voxel sizes as at the 'limit' of the scanner. Yet the Conclusion states that oPs lifetime imaging is 'feasible ... under clinically viable conditions' without mentioning this failure. Since clinical imaging typically involves adjacent structures with different lifetime properties, the feasibility claim should be qualified, or the method should be demonstrated in at least one geometry with adjacent regions of distinct lifetimes (e.g., larger voxels, longer acquisition, or post-reconstruction processing) before the general clinical-feasibility conclusion is drawn.","section":"§3 Discussion / §4 Conclusions / Fig. 6"},{"comment":"The voxel-level precision reported in Table 2 does not support the assertion that the four samples are 'clearly distinguishable' at 4-mm voxel size. The single-voxel fits have uncertainties of 0.12–0.23 ns; the differences between T1 and T2 (2.56 vs 2.37 ns) and between T2 and T3 (2.37 vs 2.30 ns) are smaller than the combined uncertainties, so these pairs are not significantly separated. Only the extremes (T1 vs T4) differ by more than approximately two standard deviations. To substantiate a voxel-wise distinguishability claim, the authors should report the distribution of τ3 across voxels within each tube (e.g., mean and standard deviation over many voxels), perform pairwise statistical comparisons, or show error maps rather than a single central voxel and qualitative gray-scale images.","section":"Table 2 / §2 Results"}],"minor_comments":[{"comment":"The sentence 'a central voxels have good count statistics' should read 'central voxels have good count statistics'.","section":"Abstract and §4"},{"comment":"'Deminalized water' in Table 1 should be 'Demineralized water'; also, the text in §1 refers to adding gelatine to 'T2', whereas Table 1 identifies the gelatine sample as T3.","section":"Table 1 and §1"},{"comment":"The voxel inclusion criterion ('relative error in the background region of less than 20%') is not defined; please specify how this relative error is computed and over which region.","section":"§1 Materials and methods"},{"comment":"The prior for τ3 is informative (N(1.78 ns, 0.8 ns)) and all fitted voxel lifetimes lie within ~1σ of this prior; a sensitivity analysis with a broader or flat prior would help confirm that the voxel-level differences are data-driven rather than prior-dominated.","section":"Eq. (1)"},{"comment":"The number of 3γ events per voxel for each voxel size and acquisition is not reported; count maps would aid the reader in assessing the statistical reliability of the voxel fits.","section":"§1 Materials and methods"},{"comment":"The fixed values τ1=125 ps and τ2=388 ps should be justified with a reference, as these parameters may vary with material and could bias the fitted τ3 in voxels where the direct-annihilation fraction is large.","section":"§1 Materials and methods"}],"recommendation":"major_revision","confidential_remarks":"For the editor: The manuscript is a brief report with useful phantom data and an honest discussion of limitations, but the abstract and conclusions go beyond what the measurements show. The main issues are (i) the activity concentration is 3–20× higher than the cited clinical range, and (ii) the taped-together geometry, which is the most clinically relevant arrangement, fails to distinguish regions. The fit methodology is largely carried over from the authors' previous work (Refs. [24,25]), so the incremental novelty is the voxel-wise application with 124I; this should be acknowledged more clearly. I recommend major revision rather than rejection, because the whole-tube results are sound and the voxel-wise question could be addressed with additional analysis or simulation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a real increment. Prior work from the same group measured whole-sample oPs lifetimes on the Quadra; this paper shows voxel-wise tau3 images at 4 mm voxels from a commercial LAFOV PET/CT using 124I, with separated phantoms clearly distinguishable. The whole-tube fits are precise (<1.8% relative error), the Bayesian fitting is described in enough detail to reproduce, and the evaluated data are on Zenodo. That is worth a serious referee.\n\nThe soft spots are real but contained. The conclusion claims feasibility \"with respect to... activity concentration,\" but the phantom activities are around 232 kBq/ml, a factor of 3-20 above the 10-70 kBq/ml the authors themselves cite from thyroid cancer metastases. They admit this in the Discussion, so the conclusion overstates what the data show. At clinical concentrations, per-voxel counts drop, and the 10% relative uncertainties at 4 mm would grow roughly by the square root of the count ratio—likely 20-50% or worse, with many voxels failing the background-error inclusion criterion. No measurement or simulation at those concentrations is reported. The claim should be conditional, or the conclusion softened.\n\nSecond, the TOF localization of three-gamma events is the stated limiting factor, but its bias is not quantified. Events from tube walls and air are acknowledged. That matters because 4 mm voxels are near the spatial resolution of the scanner. Third, the taped-together setup—meant to mimic realistic tumor geometry—does not allow clear spatial distinction, which is honest but further weakens the broad feasibility claim.\n\nNone of this kills the paper. The separated-tube demonstration is clean and the count-statistics story is credible. It is a step toward translation, not a validated clinical tool. A serious referee should engage with it. Recommended action: peer review, with requests to either measure at 10-70 kBq/ml or simulate the expected precision, to quantify TOF localization bias, and to temper the conclusion wording.","headline":"A credible phantom demonstration of 4 mm voxel oPs lifetime imaging on a commercial LAFOV PET/CT, but the activity-concentration clause of the feasibility claim outruns the data.","tokens_in":16037,"tokens_out":2094,"would_cite":true,"duration_ms":21359,"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":"Voxel-wise positronium lifetime imaging is feasible on a commercial PET/CT scanner with 124I down to 4 mm voxels.","keywords":["Positronium lifetime imaging","Long axial field-of-view PET/CT","124I","ortho-positronium","time-of-flight localization","voxel-wise lifetime fitting","Bayesian fitting","Amberlite XAD4 phantom"],"falsifier":"Scan a two-material phantom whose oPs lifetimes differ by about 1 ns with a sharp interface, using the same $^{124}$I activity and 4 mm voxels, and compare the central-voxel lifetime of each material with its bulk lifetime from a single-tube fit; if the central-voxel value shifts by more than the quoted statistical uncertainty when the neighbouring material is swapped, then TOF mislocalization contaminates the voxel result and the feasibility claim weakens.","tokens_in":14924,"feed_emoji":"⚛️","tokens_out":13000,"duration_ms":96607,"temperature":0.7,"pith_summary":"The paper aims to show that voxel-wise ortho-positronium (oPs) lifetime imaging is feasible on a commercial long-axial field-of-view PET/CT scanner under conditions close to clinical routine. Using $^{124}$I as the source and the scanner's ability to detect its 602.7 keV prompt gamma, the authors locate three-gamma annihilation events by time-of-flight and fit the oPs lifetime per voxel with a Bayesian model. In phantom measurements with four Amberlite XAD4/water samples, they recover bulk lifetimes from 1.82 ns to 2.52 ns and, for well-separated samples, distinguish the samples even at 4 mm voxel size. The claim matters because oPs lifetime is sensitive to oxygen, pH, and tissue pathology, so a voxel-wise lifetime map could add diagnostic information to an ordinary PET scan without new hardware.","feed_headline":"Positronium lifetime imaging works at 4 mm on a clinical PET/CT","feed_subtitle":"Using 124I, a clinical LAFOV scanner maps oPs lifetimes voxel by voxel, opening routine tissue-microenvironment imaging.","key_machinery":"The enabling mechanism is the selection of three-gamma events (3γE) from $^{124}$I decay: $^{124}$I emits a 602.73 keV prompt gamma with 62.9% branching ratio, and the scanner's energy resolution separates it from the two 511 keV annihilation photons. Each 3γE is located in space by the time-of-flight difference between the two annihilation photons, no tomographic reconstruction is used, and the time difference between the annihilation photons and the prompt gamma is binned into a time-difference distribution (TDD) per voxel. The oPs lifetime $\\tau_3$ is extracted by a Bayesian fit of a Gaussian convolved with three lifetime components (para-positronium, direct annihilation, and oPs) to the TDD, with fixed pPs and direct lifetimes (125 ps and 388 ps) and priors as specified in the paper.","core_discovery":"The central claim is that positronium lifetime imaging — a three-dimensional image whose voxel values are the ortho-positronium lifetime $\\tau_3$ — is feasible with a commercial PET/CT scanner (the Biograph Vision Quadra) using $^{124}$I at activity concentrations, scan times, and voxel sizes compatible with clinical use. The evidence is a set of phantom scans in which four sample tubes with different oPs lifetimes ($2.52 \\pm 0.03$, $2.37 \\pm 0.03$, $2.27 \\pm 0.04$, and $1.82 \\pm 0.02$ ns for the whole samples) were imaged both separated and taped together. With the samples separated, even $4.0 \\times 4.0 \\times 4.0$ mm$^3$ voxels yield clearly distinguishable lifetime values in the central voxels, with relative uncertainties around 10%; with the samples in contact, the spatial distinction is harder and the authors attribute the limitation mainly to the TOF-based localization accuracy of three-gamma events rather than to counting statistics. The paper concludes that diagnostic-level oPs lifetime imaging using $^{124}$I-based compounds is achievable on this class of scanner.","pith_inferences":["A natural next test is to lower the activity toward the 10-70 kBq/ml range reported in thyroid cancer metastases to see whether voxel-level precision degrades gracefully; the current phantom used about 232 kBq/ml.","If TOF localization is the bottleneck, the contamination between adjacent materials should scale with the ratio of TOF uncertainty to voxel size, a prediction that could be checked with existing data by varying the voxel size and measuring boundary sharpness.","The fixed τ1 and τ2 values (125 ps and 388 ps) may be a subtle bias source: for a homogeneous phantom with a known lifetime, a joint fit of all three lifetimes would reveal whether fixing them distorts τ3 in tissue-like materials.","The method's success with well-separated sources raises the possibility that a simple spatial-regularization or smoothing step could rescue the taped-together case, potentially making the distinction of adjacent lesions possible without hardware changes."],"forward_implications":["Existing LAFOV PET/CT systems could produce oPs lifetime maps without hardware changes, turning a standard 124I scan into a tissue-microenvironment imaging study.","The demonstrated whole-sample precision (relative τ3 uncertainty below 1.8% at ~232 kBq/ml in 15 minutes) indicates that clinically realistic activities suffice for voxel-level fitting.","The ability to resolve lifetime differences around 0.1 ns at 4 mm voxels suggests that oxygen- or pH-related contrast could be mapped at the scale of tumor heterogeneity.","The main obstacle to imaging adjacent regions with different lifetimes is TOF localization accuracy, so improved coincidence timing or statistical image reconstruction should directly improve the lifetime maps."],"supporting_citations":[{"why":"Supplies the singles-mode data acquisition and the Bayesian TDD fitting procedure used for the lifetime fits.","marker":"[25]"},{"why":"Prior in vivo oPs lifetime measurements with LAFOV PET/CT that this work extends to voxel level.","marker":"[24]"},{"why":"Shows that Amberlite XAD4 samples can be prepared with tunable oPs lifetimes, providing the phantom design.","marker":"[36]"},{"why":"Establishes 124I as a feasible isotope for positron annihilation lifetime measurement with TOF-PET detectors.","marker":"[32]"},{"why":"Provides the nuclear data for 124I, including the 602.73 keV prompt gamma and 62.9% branching ratio that enable 3γE selection.","marker":"[37]"},{"why":"Characterizes the Biograph Vision Quadra scanner used in the measurements.","marker":"[26]"},{"why":"Reports activity concentrations in thyroid cancer patients, the clinical reference range against which the phantom activity is compared.","marker":"[45]"}],"fun_headline_variants":["Clinical PET maps positronium lifetimes at 4 mm","LAFOV PET sees tissue microenvironments via positronium","124I enables voxel-wise positronium lifetime on a PET/CT","Positronium lifetime imaging reaches clinical viability","4 mm positronium lifetime maps on a commercial PET/CT"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the time-of-flight localization of each three-gamma event places it in the correct 4 mm voxel; if events are mislocalized by more than a voxel, the measured voxel lifetime is a mixture of neighboring materials' lifetimes rather than the true local value.","fun_headline_variants_meta":{"raw":{"variants":["Clinical PET maps positronium lifetimes at 4 mm","LAFOV PET sees tissue microenvironments via positronium","124I enables voxel-wise positronium lifetime on a PET/CT","Positronium lifetime imaging reaches clinical viability","4 mm positronium lifetime maps on a commercial PET/CT"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000733,"raw_usage":{"total_tokens":3400,"prompt_tokens":1190,"completion_tokens":2210,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":806,"completion_tokens_details":{"reasoning_tokens":2123}},"tokens_in":806,"tokens_out":2210,"duration_ms":14935,"temperature":1.0,"reasoning_tokens":2123,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:39:55.513763+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Scan a two-material phantom whose oPs lifetimes differ by about 1 ns with a sharp interface, using the same $^{124}$I activity and 4 mm voxels, and compare the central-voxel lifetime of each material with its bulk lifetime from a single-tube fit; if the central-voxel value shifts by more than the quoted statistical uncertainty when the neighbouring material is swapped, then TOF mislocalization contaminates the voxel result and the feasibility claim weakens.","supporting_citations":[{"cited_title":"Acta Physica Polonica","cited_arxiv_id":null,"evidence_quote":"Shows that Amberlite XAD4 samples can be prepared with tunable oPs lifetimes, providing the phantom design."},{"cited_title":"Applied Physics Express 16(11), 116001 (2023) https://doi.org/10.35848/ 1882-0786/ad047c","cited_arxiv_id":null,"evidence_quote":"Establishes 124I as a feasible isotope for positron annihilation lifetime measurement with TOF-PET detectors."},{"cited_title":"Nuclear Data Sheets 109(7), 1655–1877 (2008) https://doi.org/10.1016/j.nds.2008.06.001","cited_arxiv_id":null,"evidence_quote":"Provides the nuclear data for 124I, including the 602.73 keV prompt gamma and 62.9% branching ratio that enable 3γE selection."},{"cited_title":"Journal of nuclear medicine 63, 476–484 (2022) https://doi.org/10","cited_arxiv_id":null,"evidence_quote":"Characterizes the Biograph Vision Quadra scanner used in the measurements."},{"cited_title":"Journal of Nuclear Medicine 49(6), 1017–1023 (2008) https://doi.org/10.2967/jnumed.107.047159","cited_arxiv_id":null,"evidence_quote":"Reports activity concentrations in thyroid cancer patients, the clinical reference range against which the phantom activity is compared."}],"review_version":1}