Timmins Lucas H, Molony David S, Eshtehardi Parham, Rasoul-Arzrumly Emad, Lam Adrian, Hung Olivia Y, McDaniel Michael C, Oshinski John N, Giddens Don P, Samady Habib
Wallace H. Coulter Department of Biomedical Engineering, Georgia Institute of Technology and Emory University School of Medicine, 313 Ferst Drive, Room 2127, Atlanta, GA, 30332, USA.
Division of Cardiology, Department of Medicine, Emory University School of Medicine, 1365 Clifton Road, NE Suite F606, Atlanta, GA, 30322, USA.
Int J Cardiovasc Imaging. 2017 Jan;33(1):13-24. doi: 10.1007/s10554-016-0969-y. Epub 2016 Nov 14.
The goal of this study was to evaluate the accuracy of a novel algorithm that circumferentially co-registers serial virtual histology-intravascular ultrasound (VH-IVUS) data for the focal assessment of coronary atherosclerosis progression. Thirty-three patients with an abnormal non-invasive cardiac stress test or stable angina underwent baseline and follow-up (6 or 12 months) invasive evaluation that included acquisition of VH-IVUS image data. Baseline and follow-up image pairs (n = 4194) were automatically co-registered in the circumferential direction via a multi-variate cross-correlation algorithm. Algorithm stability and accuracy were assessed by comparing results from multiple iterations of the algorithm (iteration 1 vs. iteration 2) and against values determined manually by two expert VH-IVUS readers (algorithm vs. two expert readers). Furthermore, focal plaque progression values were compared between the algorithm and expert readers following co-registration by the independently determined angles. Strong agreement in circumferential co-registration angles were observed across multiple iterations of the algorithm (stability) and between the algorithm and expert readers (accuracy; all concordance correlation coefficients >0.98). Furthermore, circumferential co-registration angles determined by the algorithm were not statistically when compared to values determined by two expert readers (p = 0. 99). Bland-Altman analysis indicated minimal bias when comparing focal VH-IVUS defined plaque progression in corresponding sectors following circumferential co-registration between the algorithm and expert readers. Finally, average differences in changes in total plaque and constituent areas between the algorithm and readers were within the average range of difference between readers (interobserver variability). We present a stable and validated algorithm to automatically circumferentially co-register serial VH-IVUS imaging data for the focal quantification of coronary atherosclerosis progression.
本研究的目的是评估一种新型算法的准确性，该算法可沿圆周方向对连续的虚拟组织学血管内超声（VH-IVUS）数据进行共同配准，以对冠状动脉粥样硬化进展进行局灶性评估。33例无创心脏负荷试验异常或稳定型心绞痛患者接受了基线和随访（6或12个月）的侵入性评估，包括获取VH-IVUS图像数据。通过多变量互相关算法在圆周方向上自动对基线和随访图像对（n = 4194）进行共同配准。通过比较算法的多次迭代结果（迭代1与迭代2）以及与两名VH-IVUS专家读者手动确定的值（算法与两名专家读者）来评估算法的稳定性和准确性。此外，在通过独立确定的角度进行共同配准后，比较算法和专家读者之间的局灶性斑块进展值。在算法的多次迭代（稳定性）之间以及算法与专家读者之间（准确性；所有一致性相关系数>0.98）观察到圆周共同配准角度的高度一致性。此外，与两名专家读者确定的值相比，算法确定的圆周共同配准角度无统计学差异（p = 0.99）。Bland-Altman分析表明，在比较算法和专家读者在圆周共同配准后相应扇区中由局灶性VH-IVUS定义的斑块进展时，偏差最小。最后，算法与读者之间总斑块和组成区域变化的平均差异在读者之间差异的平均范围内（观察者间变异性）。我们提出了一种稳定且经过验证的算法，用于自动沿圆周方向对连续的VH-IVUS成像数据进行共同配准，以对局灶性冠状动脉粥样硬化进展进行定量分析。
