Assessment, validation, and optimization of elastic volume-to-volume and volume-to-slice image registration algorithms for the visualization of inverse treatment planning data in MRI interventional sequences for the percutaneous 192Ir-HDR-brachytherapy of

磁共振成像（MRI）引导下经皮高剂量率近距离放射治疗192铱（192 Ir-HDR-BT）有助于精确破坏不可切除的肝脏肿瘤。在192 Ir-HDR-BT治疗过程中，将一个或多个具有封闭尖端的柔性生物相容性导管（所谓的施用器）永久植入肿瘤内。由于192 Ir源周围的剂量梯度陡峭，一个或多个施源器的最佳几何配置对于在肿瘤内应用规定的细胞毒性剂量和避免相邻局部结构的风险至关重要。反向治疗计划有助于在植入前定义最佳几何施用器配置。它使放射治疗师能够在192 Ir-HDR-BT治疗前一天采集的高质量MRI数据集中定义虚拟施源器、轮廓靶病变和风险结构，例如Gd-EOB-DTPA增强T1加权、脂肪饱和3D梯度回波序列（T1 w fs 3D-GRE）。将这种结构信息转移到MRI引导的192 Ir-HDR-BT的工作流程中将提高该方法的安全性和效率。这可以通过可视化在施用器植入前即刻采集的MRI计划数据中的信息和在肿瘤靶向期间采集用于引导的MRI介入序列中的信息来实现。可视化的基本先决条件是实施准确和鲁棒的体积到体积和体积到切片图像配准算法1。将逆向治疗计划的结构信息（特别是优化的施源器位置）转换为MRI计划数据（Gd-EOB-DTPA增强T1 w fs 3D-GRE）和2.将信息从计划数据传输到MRI介入序列（Gd-EOB-DTPA增强T1 w fs单层[2D]-GRE）。图像配准算法必须补偿肝脏、目标病变和危险结构的弹性变形。它们还必须产生稳健的结果，而不管MRI介入序列的图像采集参数、信号/对比度噪声比和相对数据稀疏性的差异。目前，对于肝脏MRI数据集的体积到切片配准场景，还没有鲁棒的方法。因此，本项目的主要目的是利用Insight分割和配准工具包（ITK）对肝脏容积和介入性单层MRI数据集的弹性容积到容积和容积到切片图像配准算法进行评估、验证和必要时优化。评估应包括准确性和稳健性方面的最低要求的定义、确认不同图像配准算法性能的适用距离尺寸，尤其是体积到切片场景的图像配准策略的调查、规范和实施。

Magnetic resonance imaging (MRI) guided percutaneous high-dose-rate-brachytherapy with 192Iridium (192Ir-HDR-BT) facilitates the precise destruction of irresectable liver tumors. One or more flexible, biocompatible catheters with closed tips (so-called applicators) are percutaneously implanted within the tumors in the course of a 192Ir-HDR-BT session. Due to the steep dose gradient around the 192Ir-source the optimal geometric configuration of one or several applicators is crucial for the application of the prescribed cytotoxic dose within the tumor and sparing of adjacently localized structures at risk. Inverse treatment planning facilitates the definition of the optimal geometric applicator configuration before the implantation. It enables the radiotherapist to define virtual applicators, contour target lesions and structures at risk in high quality MRI datasets such as Gd-EOB-DTPA enhanced T1 weighted, fat saturated 3D-gradientecho sequences (T1w fs 3D-GRE) acquired the day before the 192Ir-HDR-BT treatment. The transfer of this structural information into the workflow of MRI-guided 192Ir-HDR-BT would increase the safety and efficiency of this method. This could be accomplished by visualization of the information in the MRI planning data acquired immediately before applicator implantation and in the MRI interventional sequence acquired for guidance during tumor targeting. The fundamental prerequisite for the visualization is the implementation of accurate and robust volume-to-volume and volume-to-slice image registration algorithms 1. to transform the structural information of the inverse treatment planning (especially the optimized applicator positions) into the MRI planning data (Gd-EOB-DTPA enhanced T1w fs 3D-GRE) and 2. to transfer the information from the planning data to the MRI interventional sequence (Gd-EOB-DTPA enhanced T1w fs single-slice [2D]-GRE). The image registration algorithms must compensate the elastic deformation of the liver, target lesions and structures at risk. They must also yield robust results in spite of differences in image acquisition parameters, signal- / contrast-to-noise-ratios and relative data sparsity of the MRI interventional sequence. Currently there is no robust approach for the volume-to-slice registration scenario of liver MRI datasets. The primary aim of this project thus is the assessment, validation and if necessary optimization of elastic volume-to-volume and volume-to-slice image registration algorithms of volume- and interventional single-slice MRI datasets of the liver utilizing the Insight Segmentation and Registration Toolkit (ITK). The assessment shall include the definition of the minimum requirement with regards to accuracy and robustness, applicable distance dimensions to validate the performance of different image registration algorithms, and especially the investigation, specification and implementation of an image registration strategy for the volume-to-slice scenario.