A Fully Automated Segment-Specific CCTA Plaque Quantification Workflow using Deep Learning

Q: What are the key findings of this study?

DVPlaque achieved high per-patient reliability for total, calcified, and noncalcified plaque volumes (ICCs 0.962, 0.985, 0.932) while reducing analysis time vs experts (4.24 vs 45.86 min, p<0.0001).

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Coronary artery disease (n=271). Intervention: DVPlaque (fully automated deep learning CCTA plaque quantification workflow) vs. Expert manual plaque measurement (Autoplaque 3.0). Primary outcome: Per-patient interclass correlation coefficient (ICC) and agreement for total plaque (TP) volume between DVPlaque and expert measurement (ICC 0.962, 95% CI 0.941-0.977, p=<0.001).

Q: What does this research mean for the field?

A fully automated deep learning workflow accurately and efficiently quantifies coronary artery plaque volume across patient, vessel, and segment levels with high agreement with expert manual quantification. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Q: What question did this study set out to answer?

To evaluate the efficiency and reliability of a fully automated deep learning workflow (DVPlaque) for quantifying segment-specific coronary artery plaque on coronary computed tomography angiography (CCTA).

Hongsen Wang; Xiaofei Hou; Zifan Liu; Jielin Yang; Juan Shen; Zhe Tang; Jinyao Zhang; Tao Chen; Junjie Yang; Tong Chen

doi:10.1007/s10278-026-02478-x

November 5, 2026Journal of Digital ImagingOpen Access

A Fully Automated Segment-Specific CCTA Plaque Quantification Workflow using Deep Learning

View Full Paper

Ask AI

Bookmark

Share

Key result

Automated DVPlaque matches expert CCTA plaque quantification reliability while reducing analysis time by ~91%.

ICC 0.962
95% CI 0.941-0.977
P<0.001
n=271

Why the study?

Current CCTA plaque quantification workflows either require manual adjustments that limit consistency and efficiency or lack segment-specific precision, prompting evaluation of a fully automated deep learning-based method.

Does the fully automated deep learning DVPlaque workflow reliably and efficiently quantify segment-specific coronary plaque volumes on CCTA compared to expert manual measurements in patients with ≥50% coronary stenosis?

Key Points

To evaluate the efficiency and reliability of a fully automated deep learning workflow (DVPlaque) for quantifying segment-specific coronary artery plaque on coronary computed tomography angiography (CCTA).
Analyzed 271 CCTA scans to quantify total plaque (TP), calcified plaque (CP), noncalcified plaque (NCP), and low-density NCP using the automated DVPlaque model compared against expert manual quantification.
Assessed measurement reliability and agreement across patient, vessel, and segment levels using intraclass correlation coefficients (ICC) and Bland–Altman analyses.
DVPlaque completed measurements significantly faster than expert analysis (p < 0.0001), with the proximal left anterior descending artery (LAD) exhibiting the highest plaque prevalence (97.0%).
On a per-patient level, ICCs were 0.962 for TP, 0.985 for CP, and 0.932 for NCP, with mean biases and 95% limits of agreement of −25.39 [−261.51, 210.73] mm³, 11.88 [−52.78, 76.55] mm³, and −37.27 [−255.88, 181.33] mm³, respectively.
Per-vessel ICCs were all > 0.90 for TP, CP, and NCP, and per-segment ICCs were > 0.90 across all segments except for mid-LAD NCP (0.887).

Study Design

Type

Observational (n=271)

Blinding

Single-blind

Multicenter

No

Structured PICO

Does the fully automated deep learning DVPlaque workflow reliably and efficiently quantify segment-specific coronary plaque volumes on CCTA compared to expert manual measurements in patients with ≥50% coronary stenosis?

P

Population

271 adults (mean age 61.8 years, 25.5% female) with at least one ≥50% coronary artery stenosis on CCTA and free of prior stents, bypass grafts, or chronic total occlusions.

I

Intervention

DVPlaque (DEEPVESSEL Plaque, version 1.1), a fully automated deep learning-based multistage CCTA analysis workflow for segment-specific coronary plaque detection and volume quantification (total plaque, calcified plaque, noncalcified plaque, and low-density noncalcified plaque) without manual correction.

C

Comparator

Semiautomated quantitative plaque analysis (Autoplaque 3.0) with manual contour and centerline correction performed independently by two blinded senior physicians with ≥5 years of CCTA reading experience (mean of the two experts used as reference standard).

O

Outcome

Reliability and reproducibility of plaque volume quantification (total plaque [TP], calcified plaque [CP], noncalcified plaque [NCP], and low-density noncalcified plaque [LD-NCP]) at the per-patient, per-vessel, and per-segment levels, measured via intraclass correlation coefficient (ICC) and Bland-Altman bias/limits of agreement.surrogate

Main Result

Effect estimate: ICC 0.962 (95% CI 0.941-0.977)

p-value: p=<0.001

The fully automated DVPlaque deep learning workflow achieves expert-level reliability for quantifying calcified and noncalcified coronary plaque volumes down to the segment level while reducing analysis time by over 90% (4.24 vs 45.86 minutes).

Limitations

External validation was derived from a single center, limiting generalizability.
Retrospective design restricted to patients with ≥50% stenosis and adequate image quality may introduce selection bias.
Excluded patients with prior coronary stents, bypass grafts, or chronic total occlusions.
Moderate-to-poor consistency in measuring low-density noncalcified plaque (LD-NCP) volumes at the segment and LCX vessel levels.
Fixed Hounsfield unit (HU) thresholds for plaque components may limit cross-platform consistency.
Automated detection of high-risk plaque features was not assessed.
Lack of direct comparison with intravascular ultrasound (IVUS) or optical coherence tomography (OCT) limits evaluation of absolute accuracy.
No clinical outcome follow-up to evaluate incremental prognostic value.
Fixed Hounsfield unit (HU) thresholds for plaque classification may limit cross-platform consistency.
Automated detection of high-risk plaque features (e.g., positive remodeling, napkin-ring sign, spotty calcification) was not assessed.
Lack of direct comparison with intravascular ultrasound (IVUS) or optical coherence tomography (OCT) limits the evaluation of absolute accuracy.
Moderate to poor agreement for low-density noncalcified plaque (LD-NCP) at the per-segment level and in the left circumflex artery (LCX).
Retrospective single-center validation without prospective clinical outcome assessment.

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

Supports automated CCTA plaque quantification; extends segment-specific validation but leaves prospective outcome studies open.

Abstract

Coronary computed tomography angiography (CCTA) remains essential for diagnosing coronary artery disease, while the focus has moved from stenosis to plaque burden. Current workflows either need manual adjustments that limit consistency and efficiency, or they lack segment-specific precision. This study evaluated the effectiveness and reliability of a fully automated deep learning–based plaque quantification method using CCTA. DVPlaque workflow, a fully automated deep learning model, was developed to quantify segment-specific plaque volume. A total of 271 CCTA scans were included. Plaque quantification was performed by the expert and DVPlaque model, respectively. Reliability was measured through interclass correlation coefficient (ICC) analysis and Bland–Altman plots. DVPlaque workflow cost less time than expert measurement ( p < 0.0001). Among these segments of major coronary arteries, proximal left ascending artery (LAD) had the highest prevalence of plaque (97.0%). Among these coronary arteries, the volumes of total plaque (TP), calcified plaque (CP), noncalcified plaque (NCP), and low-density NCP were the highest in LAD (all p -values < 0.05). On a per-patient level, the ICCs for TP, CP, and NCP volume were 0.962, 0.985, and 0.932. The corresponding biases and 95% limits of agreement were − 25.39 [ − 261.51, 210.73] mm 3 , 11.88 [ − 52.78, 76.55] mm 3 , and − 37.27 [ − 255.88, 181.33] mm 3 . For TP, CP, and NCP, per-vessel ICCs were all > 0.90, and per-segment ICCs were > 0.90 except for mid-LAD NCP (0.887). DVPlaque scheme was capable of quantifying plaque volume with excellent efficiency. The automated quantification of CP and NCP exhibited strong reliability even on a per-segment level, facilitating clinical application.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Wang et al. (2026) conducted an observational in Coronary artery disease (n=271). DVPlaque (fully automated deep learning CCTA plaque quantification workflow) vs. Expert manual plaque measurement (Autoplaque 3.0) was evaluated on Per-patient interclass correlation coefficient (ICC) and agreement for total plaque (TP) volume between DVPlaque and expert measurement (ICC 0.962, 95% CI 0.941-0.977, p=<0.001). DVPlaque achieved high per-patient reliability for total, calcified, and noncalcified plaque volumes (ICCs 0.962, 0.985, 0.932) while reducing analysis time vs experts (4.24 vs 45.86 min, p<0.0001).

synapsesocial.com/papers/6aefbd4950d2071ffda4bbb9 https://doi.org/10.1007/s10278-026-02478-x

View Full Paper

Ask AI

Bookmark

Share

A Fully Automated Segment-Specific CCTA Plaque Quantification Workflow using Deep Learning

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

A Fully Automated Segment-Specific CCTA Plaque Quantification Workflow using Deep Learning

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading