文献速递心血管CT

深度学习重建用于存在起源异常、支架或旁路移植物的患者冠状动脉CT血管成像

Deep learning reconstruction for coronary CT angiography in patients with origin anomaly, stent or bypass graft.
PubMed 文献 · Radiol Med · 2024-07-18

本回顾性研究开发并验证了用于冠状动脉CT血管成像自动重建的深度学习模型,训练集和验证集分别来自6063例和1962例患者,并在812例独立外部测试集(357例有起源异常或血运重建,455例无)上评估。自动重建在自体血管、起源异常、支架和旁路移植患者中的成功率分别为100%(455/455)、97%(97/100)、100%(152/152)和76.2%(80/105);深度学习重建的图像质量评分在所有亚组均显著高于手动重建(所有p < 0.001),总体后处理时间从手动法的465秒降至深度学习的11秒(p < 0.001)。

要点

  • 旁路移植患者自动重建成功率76.2%(80/105),低于自体血管、起源异常和支架患者。
  • 深度学习重建图像质量评分在所有亚组均高于手动重建(所有p < 0.001)。
  • 深度学习重建总体后处理时间11秒,手动法465秒(p < 0.001)。

核心结论

  • 所开发的深度学习模型能够对旁路移植、支架和起源异常患者进行准确的自动冠状动脉CT血管成像重建,显著缩短后处理时间并改善临床工作流程。

研究设计与样本

  • 回顾性研究,开发并验证用于冠状动脉CT血管成像自动重建的深度学习模型,训练集6063例患者、验证集1962例患者,独立外部测试集812例患者(357例有起源异常或血运重建,455例无)。

诊断标准与关键数值

  • 自动重建成功率:自体血管100%(455/455)、起源异常97%(97/100)、支架100%(152/152)、旁路移植76.2%(80/105)。
  • 图像质量评分(5分制)深度学习重建 vs 手动重建:自体血管4分 vs 3分、起源异常4分 vs 4分、支架4分 vs 3分、旁路移植4分 vs 3分,所有p < 0.001。
  • 总体后处理时间:深度学习重建11秒 vs 手动法465秒,p < 0.001。

临床应用与阅片要点

  • 外部测试集包括100例起源异常、152例支架和105例旁路移植患者,可参考各亚组的自动重建成功率。
  • 旁路移植患者自动重建成功率为76.2%(80/105),解读该亚组结果时需结合此成功率。

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 validate a deep learning (DL)-model for automatic reconstruction for coronary CT angiography (CCTA) in patients with origin anomaly, stent or bypass graft.

MATERIAL AND METHODS: In this retrospective study, a DL model for automatic CCTA reconstruction was developed with training and validation sets from 6063 and 1962 patients. The algorithm was evaluated on an independent external test set of 812 patients (357 with origin anomaly or revascularization, 455 without). The image quality of DL reconstruction and manual reconstruction (using dedicated cardiac reconstruction software provided by CT vendors) was compared using a 5-point scale. The successful reconstruction rates and post-processing time for two methods were recorded.

RESULTS: In the external test set, 812 patients (mean age, 64.0 ± 11.6, 100 with origin anomalies, 152 with stents, 105 with bypass grafts) were evaluated. The successful rates for automatic reconstruction were 100% (455/455), 97% (97/100), 100% (152/152), and 76.2% (80/105) in patients with native vessel, origin anomaly, stent, and bypass graft, respectively. The image quality scores were significantly higher for DL reconstruction than those for manual approach in all subgroups (4 vs. 3 for native vessel, 4 vs. 4 for origin anomaly, 4 vs. 3 for stent and 4 vs. 3 for bypass graft, all p < 0.001). The overall post-processing time was remarkably reduced for DL reconstruction compared to manual method (11 s vs. 465 s, p < 0.001).

CONCLUSIONS: The developed DL model enabled accurate automatic CCTA reconstruction of bypass graft, stent and origin anomaly. It significantly reduced post-processing time and improved clinical workflow.

来源:Radiol Med,查看原文。版权归原作者所有。

引用

  1. Yu L, Yu Y, Li M, et al. Deep learning reconstruction for coronary CT angiography in patients with origin anomaly, stent or bypass graft. Radiol Med. 2024. PMID: 39023665. doi:10.1007/s11547-024-01846-3

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。