Quantification of bone tissue growth and adaptation in longitudinal murine studies

本研究的目的是基于Sheffield registration Toolkit version 2 （ShIRTv2）开发一种新的弹性配准算法，用于小鼠模型骨生长的研究。该算法将提供生长位移图，并将改进目前忽略生长的骨重塑测量。背景：简要总结，以便分配给合适的内部审稿人。传统上，骨增强药物或其他干预措施的效果是通过在每个时间点牺牲一些动物来研究小鼠模型，然后用微计算机断层扫描（microCT）对解剖骨进行骨三维组织形态学分析。在由Bellantuono教授Viceconti教授和Dall'Ara博士领导的NC3R项目中，我们最近引入了一种新的方法，即使用体内微ct来跟踪同一只小鼠骨组织随时间的演变。目前的方法假设，除了要测量的适应性的局部小变化外，骨骼在两个不同的时间点保持不变。因此，使用刚性图像配准对同一动物在不同时间点的多幅3D图像进行对齐，然后使用布尔算子计算组织自适应[Lambers, 2011；陆,2015]。不幸的是，大多数研究使用的是10-15周大的老鼠，但这是长骨仍在生长的年龄。因此，通过干预产生的组织适应叠加在骨的正常生长上。研究计划：包括第一年要进行的实验细节，6个月和12个月的预期可交付成果清单，以及第2年和第3年项目预期方向的总体大纲（如果预计获得博士学位）。为了解决这个问题，我们需要开发一种弹性配准算法，它本质上同时计算可以描述生长成分的仿射缩放和量化实际组织适应的局部位移缩放。由于这些三维图像是以高分辨率（80亿体素）获取的，这使得弹性配准和图像插值的计算量很大，因此问题进一步复杂化。在这个项目中，我们的目标是优化标准的衬衫弹性配准算法，以a)有效地注册如此大的数据集，b)准确地将增长与适应分开。我们将探索多分辨率技术和能够同时解决这两个问题的全局-局部方法的使用。开发的方法将在一系列数字幻影上进行准确性测试，并在同一时间对同一只老鼠进行重复扫描的集合（零生长精度，至少5只老鼠）。然后，它们将应用于NC3R项目期间收集的整个实验结果队列（不同组的小鼠野生型，卵巢切除和PTH治疗），以及任何后续项目（MULTISIM项目，目前正在收集另外20只小鼠的纵向数据）。由此产生的配准算法将被打包为高通量大数据分析管道，供insineo生物研究人员用于分析他们的骨骼研究结果。

The goal of the study is to develop a novel elastic registration algorithm is developed, based on the Sheffield Registration Toolkit version 2 (ShIRTv2), for the investigation of the bone growth in mice models. The algorithm will provide growth displacement map and will improve current measurements of bone remodeling which ignore growth.Background: Brief summary to allow assignment to appropriate internal reviewers in field. Traditionally the effect of bone enhancing drugs or other interventions are investigated in murine models by sacrificing a number of animals at each time point, and then analysing bone 3D histomorphometry with micro-Computed Tomography (microCT) on dissected bones. Within an NC3R project, led by Prof Bellantuono Prof Viceconti and Dr Dall'Ara, we recently introduced new methodology whereby in vivo microCT is used to follow the evolution of bone tissue in the same mouse over time. Current methods make the assumption that except for small local changes that are the adaptation to be measured, the bone remains the same at two distinct time points. Thus, rigid image registration is used to align multiple 3D images of the same animal at different time points, and then Boolean operators are used to compute the tissue adaptation [Lambers, 2011; Lu, 2015]. Unfortunately, most studies use 10-15 week old mice but this is an age where the long bones are still growing. Thus, the tissue adaptation produced by the intervention, is superimposed on the normal growth of the bone.Research Plan: Including detail of experiments to be undertaken in the first year and a list of expected deliverables at 6 and 12 months plus a general outline of expected direction of project in years 2 and 3 if a PhD is anticipated. To address this problem we need to develop an elastic registration algorithm that essentially simultaneously computes an affine scaling that can describe the growth component, and a local displacement scaling that quantifies the actual tissue adaptation. The problem is further complicated by the fact that these 3D images are acquired at a high resolution (8 billion voxels), which makes the elastic registration and image interpolation computationally intensive. In this project we aim to optimise the standard ShIRT elastic registration algorithm to a) register such large datasets efficiently, and b) accurately separate growth from adaptation. We will explore the use of both multi-resolution techniques and global-local methods that are able to tackle both problems at the same time. The developed methods will be tested for accuracy on a series of digital phantoms, and on a collection of repeated scans performed on the same mouse at the same time (zero-growth accuracy, on at least 5 mice). They will then be applied to the entire cohort of experimental results collected during the NC3R project (different groups of mice wild type, ovariectomized and treated with PTH), and from any follow-up projects (MULTISIM project, longitudinal data are currently being collected on another 20 mice). The resulting registration algorithm will be packaged as a high-throughput big data analytics pipeline to be used by Insigneo biological researchers to analyse their bone research results.