移除冠脉造影图像中重叠起搏器以优化主要血管的分割

An optimized segmentation of main vessel in coronary angiography images via removing the overlapping pacemaker

冠状动脉造影(CAG)作为一种经典的血管影像模态,在临床上广泛应用于心血管疾病的诊断。对于基于CAG的计算机辅助诊断系统而言,准确的血管分割至关重要。但是,对安装心脏起搏器患者进行CAG检查时,所获得的图像中往往存在起搏器的干扰。在这种情况下,血管分割将变得更为困难。为了去除起搏器的干扰以优化血管分割的结果,本文提出了一种基于伪CAG图像生成并与待分割CAG图像配准后移除起搏器干扰的方法。首先,通过CAG视频中的部分图像序列生成仅包含起搏器的伪CAG图像;然后,利用一个图像局部特征描述子配准伪CAG图像和待分割CAG图像中起搏器的相对位置;最后,综合配准结果和起搏器、主要血管的分割结果,移除重叠起搏器的干扰,优化主要血管的分割。在临床采集的带起搏器干扰的11幅CAG图像上评估所提出的方法,戴斯系数的平均优化率达到12.03%,这表明本文方法可有效移除重叠起搏器,从而达到优化主要血管分割的目的。综上,本文研究结果或可助力基于CAG的计算机辅助诊断系统,进一步提高其诊断准确性和效率。

Coronary angiography(CAG) as a typical imaging modality for the diagnosis of coronary diseases hasbeen widely employed in clinical practices. For CAG-based computer-aided diagnosis systems, accurate vessel segmentation plays a fundamental role. However, patients with bradycardia usually have a pacemaker which frequently interferes the vessel segmentation. In this case, the segmentation of vessels will be hard. To mitigate interferences of pacemakers and then extract main vessels more effectively in CAG images, we propose an approach. At first, a pseudo CAG(pCAG) image is generated through a part of a CAG sequence, in which the pacemaker exists. Then, a local feature descriptor is employed to register the relative location of pacemaker between the pCAG image and the target CAG image.Finally, combining the registration result and segmentation results of main vessels and pacemaker, interferences of pacemaker are removed and the segmentation of main vessels is improved. The proposed method is evaluated based on 11CAG images with pacemakers acquired in clinical practices. An optimization ratio of the Dice coefficient is 12.04%, which demonstrates that our method can remove overlapping pacemakers and achieve the improvement of main vessel segmentation in CAG images.Our method can further become a helpful component in a CAG-based computer-aided diagnosis system, improving its diagnosis accuracy and efficiency.