Gennaro G, Ballaminut A, Contento G
Veneto Institute of Oncology (IOV), IRCCS, Padua, Italy.
CyberQual srl, Gorizia, Italy.
Eur Radiol. 2017 Sep;27(9):3776-3787. doi: 10.1007/s00330-017-4735-x. Epub 2017 Jan 27.
This study aims to illustrate a multiparametric automatic method for monitoring long-term reproducibility of digital mammography systems, and its application on a large scale.
Twenty-five digital mammography systems employed within a regional screening programme were controlled weekly using the same type of phantom, whose images were analysed by an automatic software tool. To assess system reproducibility levels, 15 image quality indices (IQIs) were extracted and compared with the corresponding indices previously determined by a baseline procedure. The coefficients of variation (COVs) of the IQIs were used to assess the overall variability.
A total of 2553 phantom images were collected from the 25 digital mammography systems from March 2013 to December 2014. Most of the systems showed excellent image quality reproducibility over the surveillance interval, with mean variability below 5%. Variability of each IQI was 5%, with the exception of one index associated with the smallest phantom objects (0.25 mm), which was below 10%.
The method applied for reproducibility tests-multi-detail phantoms, cloud automatic software tool to measure multiple image quality indices and statistical process control-was proven to be effective and applicable on a large scale and to any type of digital mammography system.
• Reproducibility of mammography image quality should be monitored by appropriate quality controls. • Use of automatic software tools allows image quality evaluation by multiple indices. • System reproducibility can be assessed comparing current index value with baseline data. • Overall system reproducibility of modern digital mammography systems is excellent. • The method proposed and applied is cost-effective and easily scalable.
本研究旨在阐述一种用于监测数字乳腺摄影系统长期再现性的多参数自动方法及其大规模应用。
在一项区域筛查计划中使用的25台数字乳腺摄影系统每周使用同类型的体模进行控制，其图像由自动软件工具进行分析。为评估系统再现性水平，提取了15个图像质量指标（IQIs）并与先前通过基线程序确定的相应指标进行比较。IQIs的变异系数（COVs）用于评估总体变异性。
2013年3月至2014年12月期间，从25台数字乳腺摄影系统共收集了2553幅体模图像。大多数系统在监测期间显示出优异的图像质量再现性，平均变异性低于5%。每个IQI的变异性均为5%，但与最小体模物体（0.25毫米）相关的一个指标除外，该指标低于10%。
用于再现性测试的方法——多细节体模、用于测量多个图像质量指标的云自动软件工具和统计过程控制——被证明是有效且可大规模应用于任何类型的数字乳腺摄影系统。
• 乳腺摄影图像质量的再现性应通过适当的质量控制进行监测。• 使用自动软件工具可通过多个指标进行图像质量评估。• 通过将当前指标值与基线数据进行比较可评估系统再现性。• 现代数字乳腺摄影系统的总体系统再现性优异。• 所提出和应用的方法具有成本效益且易于扩展。
