TRACE: Two-Stage Detector-Response Estimation With Angular Cosine Expansion for Ring Artifact Correction in Photon-Counting CT
Jigang Duan, Heran Wang, Ligen Shi, Zheng Sun, Ping Yang, Xing Zhao
cs.CV
Quick verdict
Unsupervised fixed + DCT sinogram decomposition for PCD-CT ring artifacts; convincing measured-data results, yet thin statistics and implementation details.
Summary
The paper proposes TRACE, an unsupervised sinogram-decomposition method for correcting ring artifacts caused by detector-response nonuniformity in photon-counting CT. Stripes are modeled as a per-detector fixed bias plus K low-order, non-DC DCT components along the view axis, capturing slow angular amplitude variation that fixed-bias models miss. The ideal sinogram is represented by a learnable analysis-synthesis network with a band-stop residual branch. Two-stage optimization first fits fixed stripes only, then jointly refines with dynamic DCT components initialized at zero; L1 data consistency, sorting-based detector smoothness, angular curvature, and an angular-gradient soft-orthogonality penalty constrain the split. All parameters are fitted to the measured sinogram without paired data. On measured data from an in-house PCD-CT (QRM mouse phantom, porcine trotter; 720 views, 10 keV threshold bin), the K=1...9 sweep supports K=7, and TRACE reaches PRAC 0.98% with MTF50 3.31 and MTF10 5.77 lp/mm versus 3.27/5.54 uncorrected, while VBSC attains higher PRAC (1.86%) at sharply reduced MTF50 (2.16). The fixed-versus-fixed+dynamic ablation shows dynamic terms remove residual arc artifacts. Main limitations: quantitative evaluation is confined to one phantom region from single scans without uncertainty; loss weights, architecture, optimizer, and runtime are unreported and no code is released; validation is 2D, single-threshold, on one in-house scanner.
Review complete
Assessment at a glance
5
risk flags identified
● high 0
● medium 2
● low 3
5 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- Residual response-related stripes in measured PCD-CT data vary slowly with projection angle and cannot be adequately captured by fixed per-detector bias models.
- A truncated non-DC DCT expansion with K≪V coefficients per detector element compactly models smooth angular stripe variation while controlling model complexity.
- Two-stage optimization plus an angular-gradient soft orthogonality constraint reduces leakage of object structures into the artifact estimate.
- On measured QRM phantom data, TRACE attains PRAC 0.98% with MTF50 3.31 and MTF10 5.77 lp/mm (uncorrected 3.27/5.54), whereas VBSC's higher PRAC (1.86%) comes with MTF50 reduced to 2.16 lp/mm.
- Dynamic DCT modeling removes arc-shaped artifacts that fixed-bias correction alone leaves in porcine trotter reconstructions (Fig. 11).
Strengths
Measured-data-only evaluation with full acquisition disclosure
Section III explicitly rejects synthetic stripes for biasing model comparison and reports complete acquisition parameters (Hamamatsu L10101 source, DECTRIS EIGER2 1M-W, 720 views/360°, 10/30 keV thresholds with the 10 keV bin used, 300 µA, 1 s/view, FBP Ram-Lak to 2048×2048), which strengthens internal validity.
Principled, compact stripe parameterization
Modeling SA as b·1ᵀ + ΦK Aᵀ (Eq. 9) with the DC basis excluded guarantees zero angular mean of the dynamic component (ΦKᵀ1=0), avoiding redundancy with the fixed bias, and K≪V keeps the artifact branch degrees of freedom far below per-view alternatives.
Novel subspace-targeted anti-leakage constraint
Multi-angle empirical support
Two-stage warm start is well motivated
Weaknesses
No statistical variability or repeated scans
Operational configuration unreported
New regularization terms are not individually ablated
No ground-truth verification of the artifact estimate
Narrow external validity
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
This method paper targets ring artifacts in photon-counting detector CT (PCD-CT) that survive conventional calibration because residual detector-response errors vary with scan conditions and projection angle (Sec. I). The authors claim three contributions (Sec. I-C): (1) a two-stage unsupervised sinogram decomposition combining a learnable analysis-synthesis IS representation with progressive fixed-then-dynamic stripe estimation, fitted per scan without paired data; (2) a fixed-dynamic response-error model using a truncated non-DC DCT basis with K≪V coefficients per detector element; (3) an angular-gradient soft orthogonality constraint reducing structure leakage into the artifact estimate. The scope is deliberately narrow — 2D fan-beam correction at a single energy threshold — and this is stated honestly in the conclusion.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 5 strengths & 5 weaknesses, analyzed in detail
- 5 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
Does TRACE need training data or a pretrained network?
No. TRACE is unsupervised: its learnable analysis-synthesis network and all stripe parameters (fixed biases plus K low-order DCT coefficients per detector element) are optimized directly on the measured sinogram of each scan, with no paired artifact-free reference or flat-field acquisition.
What data was TRACE evaluated on?
Only measured data from the authors' in-house photon-counting CT system (Hamamatsu L10101 X-ray source, DECTRIS EIGER2 1M-W detector): a QRM mouse phantom scanned at 100 kVp and a porcine trotter specimen at 80 kVp, 720 views over 360°, using counts from the 10 keV threshold. No public datasets were used.
How well does TRACE perform compared to other ring-removal methods?
On the QRM phantom's bone-equivalent region, TRACE achieved PRAC 0.98% with MTF50 3.31 and MTF10 5.77 lp/mm versus 3.27/5.54 uncorrected. VBSC scored higher PRAC (1.86%) but blurred edges (MTF50 2.16). The paper describes TRACE's PRAC as comparable to RAS and higher than TV-L1aG, DDR, and INR, with the best edge preservation among tested methods.
Is code or data available for TRACE?
The paper contains no code or data availability statement; neither implementation nor the measured datasets are stated to be released.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Numeric panel values in Figs. 5, 7, and 10 are interleaved with axis ticks in the PDF extraction; per-method attribution of some bar/curve values (e.g., PRAC for RAS, DDR, INR) is inferred from in-text statements rather than panel order, so only text-confirmed numbers (TRACE 0.98%/3.31/5.77; VBSC 1.86%/2.16/3.72; uncorrected 3.27/5.54) are cited in this review.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.