Motion compensation for transcatheter aortic valve implantation

先进的医学图像数据可视化工具对于确保经导管主动脉瓣植入（TAVI）期间的安全导航至关重要。本研究计划的主要目标是提出一种基于图像的呼吸运动补偿方法，用于TAVI期间的导航引导。TAVI正在成为传统心脏瓣膜手术的一种有吸引力的替代方案，减少了手术时间，提高了患者的康复。目前TAVI是在单视图X线血管造影指导下进行的，不能提供任何深度信息。心脏病专家必须评估瓣膜放置的最佳性，并在完美的时机部署瓣膜，所有这些都是在复杂的运动下进行的。自瓣膜植入术早期以来，影像学的主导作用已被认识到。X射线血管造影套件的最新进展现在允许在空间中自由移动单个放射源，以从任何任意视图获取图像。这种被称为X线旋转血管造影的技术有望在介入前获得多幅图像，进行完整的3D重建。首先，在干预开始时，使用X线旋转血管造影术在3D中重建呼吸暂停患者的主动脉。在干预期间，将研究基于图谱的3D重建和3D -2D配准，以在考虑呼吸运动的X线血管造影上生成可靠且准确的主动脉覆盖。这种方法将结合患者特定和人群特定的运动信息。从单个X线血管造影序列中，先前计算并在患者参考框架中获得的3D体积将通过连续的非刚性3D- 2D注册来更新。基于动态时间翘曲的一致性保持方法将被考虑用于建模和学习循环呼吸运动模式在二维和之后的三维。动态时间翘曲被广泛应用于语音识别领域，用于测量两个时间序列之间的相似度，有助于将实际运动模式与通用运动图谱中的运动模式匹配起来。运动图谱，基于我们之前的单面随机运动补偿的工作，将从连续的配准中学习，以可复制的方式从X射线血管造影（患者特异性）中表示空间运动。此外，术前运动图谱将从模拟数据和真实患者数据集构建，以更好地概括复杂的心血管运动模式（人群特异性）。本研究计划的长期目标是在TAVI期间生成一个用于多模态图像融合的在线导航系统，将多种术前成像模式动态叠加在实时术中成像上，如经食管超声心动图的x线血管造影。

Advanced visualization tools of medical image data are paramount to ensure safe navigation during Transcatheter Aortic Valve Implantations (TAVI). The main goal of this research program is to propose a methodology for image-based breathing motion compensation for navigation guidance during TAVI. TAVI is becoming an appealing alternative to traditional open­-heart valve surgery, reducing operating time and improving patient recovery. TAVI is currently performed under single view X-­ray angiography guidance, which does not provide any depth information. Cardiologists must evaluate the optimality of valve placement and deploy the valve with perfect timing, all under a complex motion. The predominant role of imaging has been recognized since the early stages of valve implantation. Recent advances in X-­ray angiography suites nowadays allow moving a single radiographic source freely in space to acquire images from any arbitrary views. This technique, called X­-ray rotational angiography is promising to acquire multiple images for a full 3D reconstruction prior to the intervention. First, at the beginning of the intervention, the aorta of the patient in apnea will be reconstructed in 3D using X­-ray rotational angiography. During the intervention, atlas-­based 3D reconstruction and 3D­-2D registration will be investigated to produce robust and accurate overlay of the aorta over X­-ray angiography considering breathing motion. This approach will incorporate both patient­-specific and population-­specific motion information. From a single X­-ray angiography sequence, the 3D volume computed previously and acquired in the patient’s reference frame, will be updated by successive non­-rigid 3D-­2D registration. A consistency preserving approach based on dynamic time warping will be considered for modelling and learning the cyclic respiratory motion patterns in 2D and afterwards in 3D. Dynamic time warping is widely used in the speech recognition community for measuring similarity between two temporal sequences and would contribute to match actual motion pattern with motion patterns from generic motion atlas. A motion atlas, based on our previous work on monoplane stochastic motion compensation, will be learned from successive registration to represent the spatial motion in a reproducible manner from X-­ray angiography (patient­-specific). Furthermore, preoperative motion atlases will be constructed from both simulated data and from real patient datasets to better generalize the complex cardiovascular motion pattern (population-­specific). The long-term objective of this research program is to generate an online navigation system for multimodality image fusion during TAVI, with dynamic overlay of several preoperative imaging modalities on real-time per operative imaging such as X-ray angiography of transoesophagal echocardiography.