{"id":"c1b920c9-0a97-490a-a366-e552ec40349d","arxiv_id":"2608.06746","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"An orthogonally stacked xDetector TOF-PET module with faster scintillator, improved SiPM, HF readout, and position-based timing correction reaches 108.6 ps FWHM paired CTR.","lead":"A PET detector prototype built from crosswise-stacked thin scintillators reached a coincidence time resolution of 108.6 picoseconds in paired-detector calculations, close to the 100 ps target for time-of-flight PET. The result suggests the design could improve image sharpness in PET scanners without sacrificing detection efficiency.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 108.6 ps headline relies on DOI positions known to 3.15 mm, but only ~6 mm longitudinal resolution is demonstrated; the paper's own two-position recalculation gives 110.7 ps.","rationale":"The reader's weakest assumption correctly identifies the most load-bearing concern: the DOI correction uses externally imposed positions with 3.15 mm spacing while the detector's demonstrated longitudinal resolution is ~6 mm, so the corrected 108.6 ps CTR is not achievable without a signal-based DOI estimator at that resolution. The paper itself provides the key evidence for this concern in Section 4, where the two-position recalculation yields 110.7 ± 2.0 ps. I considered whether the shift-by-fitted-means procedure could artificially narrow the histogram, but that effect would slightly widen the distribution because the sample-mean shift adds estimation noise, so it is not the critical issue. The uncorrected measurements and the quadrature propagation leading to 113.5 ps are credible and represent a useful result. The conditional verdict remains appropriate: the paper should clearly separate the idealized, upper-bound 108.6 ps value from the realistic 110.7 ps projection, or demonstrate a DOI estimator with the orthogonal channels. No change to the reader's verdict is needed.","tokens_in":9259,"tokens_out":4368,"duration_ms":41438,"concrete_test":"Acquire the four-position dataset and implement a signal-based DOI estimator using the orthogonal channels (e.g., charge ratio Ch1/(Ch1+Ch3) or timing difference) from the recorded waveforms. Quantify the RMS error of the estimated DOI against the known manual positions. If the RMS error substantially exceeds 3.15 mm (e.g., ~6 mm), recompute the CTR after timing correction using the estimated DOI bins; if the paired CTR shifts toward or above 110.7 ps, the 108.6 ps headline is not achievable in real operation. Alternatively, re-analyze the existing four-position data by merging adjacent positions into two 6.3 mm bins, as the authors did in Section 4, and confirm the resulting paired CTR of 110.7 ± 2.0 ps.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The DOI-based timing correction in Section 2.3 was performed by manually sliding the xDetector to four known positions separated by 3.15 mm, not by estimating interaction depth from detector signals. The paper explicitly states: 'the interaction position was controlled by manually moving the xDetector instead of estimating it from the signals of channels aligned orthogonally to the longitudinal axis.' Thus the 108.6 ps paired CTR assumes the detector knows DOI to 3.15 mm. But the xDetector's demonstrated longitudinal resolution is ~6 mm, as acknowledged in Section 4: 'the longitudinal resolution of the xDetector was reported to be ~6 mm.' With only two positions (~6 mm bins), the corrected CTR degrades to 110.7 ± 2.0 ps, as the authors themselves calculate in Section 4. Therefore the abstract's headline value of 108.6 ps is an idealized upper bound conditional on an unrealized DOI estimator, not an achieved system CTR. The raw measurements and the uncorrected 113.5 ps result remain credible; the central claim should be reframed to present the two-position, realistic-DOI value as the primary system projection.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports an experimental study of a TOF-PET detector module called xDetector, which consists of orthogonally stacked groups of thin LYSO scintillators read out by SiPMs. The authors optimize the scintillator decay time, SiPM PDE, and readout electronics, and then apply a depth-of-interaction (DOI) based timing correction. Measured in coincidence with a reference detector, the best corrected CTR is 105.3 ± 0.6 ps FWHM, and by quadrature error propagation the authors estimate a paired-xDetector CTR of 108.6 ± 1.9 ps FWHM. They also compare against conventional single detectors with 12.6 mm and 20.0 mm scintillators and report an average CTR improvement of 10.3% over the 20 mm device. The abstract claims that the xDetector offers potential to reach a 100 ps system-level CTR.","tokens_in":42,"tokens_out":2944,"duration_ms":32815,"significance":"If the raw measurements are taken at face value, the work provides a credible, well-characterized demonstration of a TOF-PET detector with sub-110 ps CTR while retaining a 20-mm-equivalent detection thickness, which is a meaningful step toward the 100 ps TOF-PET goal. The strengths of the paper include direct measurements of the component improvements (faster LYSO, improved SiPM PDE, high-frequency readout), explicit Gaussian fitting with reported uncertainties, and a direct comparison against conventional detectors under the same conditions. The key limitation is that the DOI-based timing correction is performed with externally imposed interaction positions rather than with a signal-derived DOI estimator, and the paper itself acknowledges that the demonstrated longitudinal resolution of the xDetector is only ~6 mm. With that realistic resolution, the corrected CTR degrades to 110.7 ± 2.0 ps. The central claim should therefore be reframed so that the headline value reflects the achievable system projection, not the idealized 3.15 mm DOI case.","major_comments":[{"comment":"The headline result of 108.6 ± 1.9 ps paired CTR is presented in the abstract and conclusion as an achieved system-level value, but it is conditional on knowing the interaction DOI to 3.15 mm. As stated in Section 2.3, the interaction position was controlled by manually sliding the xDetector rather than estimated from detector signals. Section 4 then acknowledges that the demonstrated longitudinal resolution of the xDetector is ~6 mm and reports that with two positions the corrected CTR becomes 110.7 ± 2.0 ps. The 108.6 ps value is therefore an idealized upper-bound projection, not a realistic system CTR. The abstract and conclusions should present the two-position value (110.7 ± 2.0 ps) as the primary projection, or clearly label the 108.6 ps value as conditional on an as-yet-undemonstrated DOI estimator.","section":"Abstract and Section 3.2 (Figure 8)"},{"comment":"The DOI-based timing correction is an in-sample calibration rather than a predictive correction. For each position, the time-difference histogram is shifted by the mean value obtained from a Gaussian fit to that same histogram; the shifted histograms are then superimposed and re-fitted. This procedure removes position-dependent mean offsets by construction and does not validate the correction on independent data or against DOI values estimated from the detector signals. The small observed improvement (~2-3 ps) should be interpreted with this circularity in mind. I recommend either demonstrating the correction with a signal-based DOI estimate or reporting the uncorrected and realistic-two-position results as the central evidence.","section":"Section 2.3"}],"minor_comments":[{"comment":"In the reference for Pagano et al. 2024, the author name appears as 'Cate J W' but should be 'Cates J W' to match the other citations in the text.","section":"References"},{"comment":"The citation 'van Suluis et al. 2019' contains a typo; the correct name is 'van Sluis' as in the reference list.","section":"Introduction"},{"comment":"The text states that the shifted histograms are referenced to 'an arbitrary reference value.' While this does not affect the FWHM, it would be clearer to specify that only the relative positions of the shifts matter, and that the fitted FWHM is invariant under a constant offset.","section":"Section 2.3"},{"comment":"The report of 'the average CTR of the xDetector over all channels was 110.0 ps FWHM' would be clearer if the standard deviation or range across channels were also given, since Figure 6 presumably contains that information.","section":"Section 3.1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of physics in medicine and biology. The main concern is that the headline number (108.6 ps) overstates what is actually demonstrated, because the DOI correction relies on externally imposed positions and an in-sample fit. The authors do disclose the ~6 mm longitudinal resolution and provide a two-position estimate, which is commendable, but the abstract and conclusion should lead with that more honest number. The paper would likely be acceptable after a revision that reframes the claims and clarifies the in-sample nature of the correction."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe thing to know: this group has pushed their xDetector concept from 187 ps to a measured 107.8 ps against a reference, and after a manual DOI correction they quote 105.3 ps, which propagates to 108.6 ps paired. The raw timing is credible; the quadrature math checks out. The new part is the combination of a faster LYSO, higher-PDE SiPM, HF readout, and a demonstration of position-based timing shifts. That is real engineering progress.\n\nThe soft spot is exactly what the stress test flags. The 108.6 headline assumes four DOI bins at 3.15 mm, but the interaction depth was set by sliding the detector, not estimated from signals. The paper admits the detector's demonstrated longitudinal resolution is ~6 mm, and when they recalculate with two bins the corrected CTR becomes 110.7 ps. Worse, the correction itself is in-sample: each position's histogram is shifted by the mean of its own Gaussian fit, so some improvement is built in by construction. The ~3 ps gain is small enough that it is plausible, but it is not a predictive DOI correction.\n\nThat said, the paper is honest. The discussion includes the two-position result and the limitation statement is explicit. The uncorrected 113.5 ps paired value is a useful system projection. The comparison with a 20 mm single-ended detector shows a 10% CTR gain from the geometry, which is the actual contribution.\n\nCitation pattern is fine. Self-citation to their own 2025 concept paper is appropriate, and the literature on side/double-sided readout is covered.\n\nWho is this for: people building TOF-PET detector modules, especially those chasing 100 ps. It deserves a serious referee. A revision should move the headline to the realistic two-position value or show a real DOI estimator. I'd send it to review, and require that reframing.","headline":"Credible component-optimization result, but the 108 ps headline depends on an unrealized DOI estimator; the realistic number is 110.7 ps.","tokens_in":10041,"tokens_out":1686,"would_cite":true,"duration_ms":16140,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["87.57.uk"],"model":"deepseek-v4-flash","headline":"Orthogonally stacked scintillator detectors reach 108.6 ps coincidence timing in TOF-PET.","keywords":["positron emission tomography","time-of-flight PET","coincidence time resolution","depth of interaction","xDetector","silicon photomultiplier","LYSO scintillator","DOI-based timing correction"],"falsifier":"Measure the xDetector's CTR under uniform irradiation with interaction depth estimated internally from the channel light-sharing ratios instead of set by manual translation. If the internally estimated DOI has about 6 mm resolution, the paired CTR should land near 110.7 ± 2.0 ps; a measured paired CTR above about 111 ps would indicate that the 108.6 ps four-position result is not achievable as a system-level value.","tokens_in":8958,"feed_emoji":"⏱️","tokens_out":8676,"duration_ms":74256,"temperature":0.7,"pith_summary":"The paper claims that a PET detector geometry called xDetector can overcome the usual trade-off between timing resolution and detection efficiency. Two groups of thin scintillator crystals are stacked orthogonally and read by separate silicon photomultipliers, keeping the module thick enough for efficient 511 keV gamma detection while reducing photon transport time spread and encoding interaction depth in the light-sharing pattern. After switching to faster crystals, higher-photon-detection-efficiency SiPMs, high-frequency readout electronics, and applying depth-of-interaction timing correction, the authors measured 105.3 ± 0.6 ps FWHM in coincidence with a reference detector and calculated 108.6 ± 1.9 ps FWHM for two xDetectors facing each other. That is an average 10.3% improvement over a conventional 20.0 mm single-ended LYSO detector and brings system-level time-of-flight PET close to the 100 ps FWHM target.","feed_headline":"Stacked crystal detectors reach 108.6 ps PET timing","feed_subtitle":"Depth-interaction timing correction pushes time-of-flight PET toward the 100 ps resolution goal.","key_machinery":"The mechanism is the xDetector: a module in which two groups of three thin LYSO scintillator crystals, each group coupled to its own SiPM, are stacked orthogonally with their bare sides facing each other through an air gap, so that interaction depth is encoded in how scintillation light is shared between channels. The paper's timing correction then shifts the coincidence-time histogram for each depth position to a common reference before fitting, removing the position-dependent optical path length variation (photon transport time spread) that otherwise broadens the CTR. Supporting that correction are the component choices: faster 36.4 ns decay-time LYSO crystals, SiPMs with roughly 10% higher PDE at 420 nm, and high-frequency readout built on two cascaded BGA616 amplifiers.","core_discovery":"The central discovery is that the timing penalty of long scintillators can be removed without sacrificing detection efficiency by replacing a single 20 mm crystal with two orthogonally stacked detector groups read out at their ends, using the light-sharing pattern to sense depth of interaction. With optimized components, the best single-channel CTR of the xDetector in coincidence with a reference detector was 107.8 ± 1.2 ps FWHM at 63 V bias. Applying DOI-based timing correction over four positions spaced 3.15 mm along the longitudinal axis improved the measured CTR to 105.3 ± 0.6 ps FWHM; error propagation for two xDetectors in coincidence gives 108.6 ± 1.9 ps FWHM, versus 113.5 ± 2.7 ps without correction. The authors conclude that the xDetector offers a realistic route toward the 100 ps FWHM system-level CTR goal, with the remaining gap coming mainly from finite longitudinal position resolution.","pith_inferences":["Editorial inference: since the four-position correction was applied with the interaction position set manually rather than estimated by the detector, the headline 108.6 ps paired CTR should be read as the upper bound of what is achievable; with the detector's own roughly 6 mm longitudinal resolution, the realistic paired CTR is closer to the two-position value of 110.7 ± 2.0 ps.","Editorial inference: combining the xDetector geometry with further DOI-resolution improvements, such as finer channel segmentation or signal-shape-based depth estimation, is the most direct path below 100 ps, because the paper's own analysis shows the residual spread is dominated by position resolution rather than by single-photon timing.","Editorial inference: a testable prediction is that an xDetector with continuously estimated DOI from channel ratios will land between 108.6 and 110.7 ps paired CTR; if it lands above about 111 ps, the manual-position result overstates the achievable system performance."],"forward_implications":["The paired xDetector CTR of 108.6 ± 1.9 ps FWHM places a detector built from clinical-thickness scintillators within reach of the 100 ps system-level TOF-PET target.","Because the xDetector keeps scintillator thickness and detection efficiency while improving CTR by about 10% over a conventional 20.0 mm single-ended detector, it offers an alternative to side- or dual-ended readout that uses fewer SiPMs.","The SNR gain of TOF-PET scales roughly as the inverse square root of CTR, so moving from about 200 ps to about 108 ps would translate into a meaningful improvement in image signal-to-noise ratio, shorter scan times, or reduced dose.","The two-position DOI correction result of 110.7 ± 2.0 ps paired CTR shows that improving longitudinal resolution alone should push the concept below 108 ps and closer to 100 ps.","The authors identify packing fraction, assembly reproducibility, and the number of SiPMs as practical issues for translating the measured module CTR into a full PET ring."],"supporting_citations":[{"why":"Introduced the xDetector concept and reported 187.4 ps CTR, the baseline this study improves on.","marker":"Onishi and Ota 2025"},{"why":"Provides the analytical CTR formula based on decay time, rise time, SPTR, PTS, PDE, LTE, and light yield that motivates the component optimizations.","marker":"Gundacker et al. 2020"},{"why":"Showed that high-frequency SiPM readout improves CTR, motivating the BGA616-based readout used here.","marker":"Gundacker et al. 2019"},{"why":"Supplies the DOI estimation and timing correction idea behind the four-position correction.","marker":"Loignon-Houle et al. 2021"},{"why":"Demonstrated a side-readout TOF-PET detector with sub-100 ps CTR, the performance benchmark the xDetector approaches with fewer SiPMs.","marker":"Cates and Levin 2018"},{"why":"Reports sub-100 ps CTR using thin scintillators and the FastIC ASIC, providing a comparison for the side-readout result cited in the discussion.","marker":"Mariscal-Castilla et al. 2024"},{"why":"Documents roughly 200 ps CTR of a current commercial TOF-PET system, defining the clinical baseline the paper targets.","marker":"van Sluis et al. 2019"},{"why":"Gives the SNR gain relation and TOF-PET physics context that make 100 ps CTR clinically desirable.","marker":"Schaart 2021"}],"fun_headline_variants":["Stacked crystals and DOI correction yield 108.6 ps timing","108.6 ps CTR achieved via orthogonally stacked detectors","DOI timing correction pushes PET to 108.6 ps resolution","Orthogonal stacking + DOI: 108.6 ps coincidence time"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that interaction depth is known to about 3.15 mm in real operation: in the experiment the xDetector was manually slid to set the interaction position and the time shifts were derived from Gaussian fits to the same data, while the detector's intrinsic longitudinal resolution is about 6 mm, and recalculating with only two positions raises the paired CTR to 110.7 ± 2.0 ps.","fun_headline_variants_meta":{"raw":{"variants":["Stacked crystals and DOI correction yield 108.6 ps timing","108.6 ps CTR achieved via orthogonally stacked detectors","DOI timing correction pushes PET to 108.6 ps resolution","Orthogonal stacking + DOI: 108.6 ps coincidence time"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000262,"raw_usage":{"total_tokens":1695,"prompt_tokens":1139,"completion_tokens":556,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":755,"completion_tokens_details":{"reasoning_tokens":483}},"tokens_in":755,"tokens_out":556,"duration_ms":5378,"temperature":1.0,"reasoning_tokens":483,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:24:55.135588+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the xDetector's CTR under uniform irradiation with interaction depth estimated internally from the channel light-sharing ratios instead of set by manual translation. If the internally estimated DOI has about 6 mm resolution, the paired CTR should land near 110.7 ± 2.0 ps; a measured paired CTR above about 111 ps would indicate that the 108.6 ps four-position result is not achievable as a system-level value.","supporting_citations":[],"review_version":1}