基于深度学习的单能成像用于能量积分探测器CT钙化性冠状动脉狭窄评估
本研究开发并评估DIAMOND深度学习框架,从常规能量积分探测器(EID)CT生成虚拟单能图像(VMI),以减少钙化斑块所致开花伪影并改善冠状动脉CT血管成像(CCTA)狭窄评估。在23例受试者中,DIAMOND使平均直径狭窄百分比(PDS)从EID CT的35.65%降至25.19%(P < .05),接近超高分辨率光子计数探测器(PCD)CT的24.27%(P < .05)。
要点
- DIAMOND框架可从单能EID CT生成高keV VMI,减少开花伪影并改善管腔显示。
- 平均PDS由EID CT的35.65%降至DIAMOND的25.19%(P < .05),接近PCD CT的24.27%(P < .05)。
- DIAMOND使26个病变中11个(42%)的CAD-RADS分类发生重分类,主要位于轻至中度狭窄范围。
核心结论
- 本研究证明DIAMOND可从单能EID CT生成高keV VMI,提供伪影减少的冠状动脉成像并改善重度钙化斑块的狭窄量化,效果与PCD CT相当且无需硬件升级。
研究设计与样本
- 研究于2022年8月至2023年9月使用回顾性与前瞻性影像数据,DIAMOND基于简化U-Net架构在10例超高分辨率PCD CT CCTA回顾性数据集上训练,输入为70-keV PCD VMI(能量等效于120-kV单能EID CT),目标为100-keV PCD VMI;训练后模型应用于前瞻性重度钙化斑块受试者数据集,受试者接受120 kV EID CT后同日行PCD CT。
关键影像表现/征象
- DIAMOND减少开花伪影并改善管腔可视化,图像质量与PCD CT相似。
- 在23例受试者(平均年龄69岁±8[SD],18例男性)中,平均PDS从EID CT的35.65%降至DIAMOND的25.19%(P < .05),接近超高分辨率PCD CT的24.27%(P < .05)。
- 与EID CT相比,DIAMOND使26个病变中11个(42%)发生CAD-RADS重分类,主要位于轻至中度狭窄范围。
诊断标准与关键数值
- 以PDS量化狭窄,DIAMOND使平均PDS从EID CT的35.65%降至25.19%(P < .05),接近PCD CT的24.27%(P < .05)。
- DIAMOND使26个病变中11个(42%)发生CAD-RADS重分类,主要位于轻至中度狭窄范围。
临床应用与阅片要点
- DIAMOND处理时间约为每轴位层面0.21秒(标准图形处理器)。
- DIAMOND可从单能EID CT生成高keV VMI,减少开花伪影并改善重度钙化斑块的狭窄量化,效果与PCD CT相当且无需硬件升级。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
Purpose To develop and evaluate a Deep learnIng-bAsed MONoenergetic imaging at Different energies (DIAMOND) framework for generating virtual monoenergetic images (VMIs) from conventional energy-integrating detector (EID) CT, aiming to reduce blooming artifacts and improve stenosis assessment in coronary CT angiography (CCTA) with heavily calcified plaques. Materials and Methods This study (August 2022-September 2023) used a combination of retrospective and prospective imaging data. DIAMOND was trained using a simplified U-Net architecture on a retrospective dataset of 10 CCTA examinations performed with ultrahigh-resolution (UHR) photon-counting detector (PCD) CT; 70-keV PCD VMIs (energy equivalent to 120-kV single-energy EID CT) served as inputs and 100-keV PCD VMIs as targets. The trained model was then applied to a prospective dataset of participants with heavily calcified plaques who underwent EID CT at 120 kV followed by same-day PCD CT. Percent diameter stenosis (PDS) was quantified for a phantom and participants by using commercial software and compared across EID CT, DIAMOND, and PCD CT using Bland-Altman analysis. Changes in stenosis severity categorization based on PDS were evaluated. Results DIAMOND reduced blooming artifacts and improved lumen visualization, with image quality resembling PCD CT. In 23 participants (mean age, 69 years ± 8 [SD]; 18 male), average PDS decreased from 35.65% (EID CT) to 25.19% (DIAMOND, P < .05), approaching 24.27% with UHR PCD CT (P < .05). Relative to EID CT, DIAMOND led to Coronary Artery Disease Reporting and Data System reclassification in 11 of 26 (42%) lesions, mainly in mild to moderate stenosis ranges. Processing time was approximately 0.21 second per axial section on a standard graphics processing unit. Conclusion This study demonstrated the feasibility of using DIAMOND to generate high-kiloelectron volt VMIs from single-energy EID CT, providing artifact-reduced coronary imaging and improved stenosis quantification for heavily calcified plaques comparable to PCD CT without hardware upgrades. Keywords: Coronary CT Angiography, Coronary Artery Stenosis, Energy-Integrating Detector CT, Photon-Counting Detector CT, Deep Learning, CT-Photon Counting, Angiography, Coronary Arteries Supplemental material is available for this article. © RSNA, 2026.
引用
- Chang S, Koons EK, Gong H, et al. Deep Learning-based Monoenergetic Imaging for Calcified Coronary Stenosis Assessment at Energy-integrating Detector CT. Radiol Cardiothorac Imaging. 2026. PMID: 41854396. doi:10.1148/ryct.250230
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。