Image-based 3D-printing for biomedical applications

动机：增材制造（“3D 打印”）领域最近取得了相当大的发展，包括开发使用粉末床激光熔合直接在金属合金中进行打印的技术。金属合金增材制造已成为研究人员和制造商常规使用的方法，商业系统也已广泛使用。这些技术增加了精准医疗的潜力，包括患者特定的组件和手术导板。对于生物医学应用，增材制造由先进成像（CT 和 MRI）驱动，因为所需的几何数据通常来自 3D 体积图像。将图像转换为组件设计源数据的基本过程有机会得到显着改进。同样，3D 成像在验证和验证生物医学应用制造组件的形状和机械性能方面可以发挥重要作用。目标：在本项目中，我们将探索用于图像重建和分割的新颖算法（包括机器学习和深度学习）的实现，以提高生物医学3D打印的效率和准确性。我们还将开发过程监控技术，作为下一代质量管理的一部分。我们开发的技术将应用于一系列研究应用，包括兽医部件、多孔支架和“智能”植入物。我们相信，开发用于设计和测试的集成“管道”将显着扩大基于图像的 3D 打印的影响。方法：我们这个项目的研究方法集中在三个主题：（1）图像采集和分割的优化，使用双能成像、迭代重建和基于模型的深度学习； (2) 基于图像的有限元设计和优化，并通过动态3D无损测试和嵌入式传感器进行验证； (3) 基于图像的验证和工艺改进，在激光粉末床融合过程中使用定量 3D 计算机断层扫描和过程中光学成像。影响：该程序集成了最近开发的技术和新计算算法（例如使用神经网络的深度学习）的多个方面，为生物医学工程中基于图像的设计和制造创建了一个新的软件工具平台。该计划的预期进展包括从 3D 图像体积中提取几何数据的新工具，以及验证有限元模型（即机械性能、药物洗脱和热传输）的新硬件软件技术。该项目利用了自然科学和工程学多个领域的专业知识，包括：成像科学、生物力学工程、先进制造和软件工程。拟议的计划将带来在研究和开发中具有广泛应用的新颖的软件实用程序和技术。

Motivation: There has been considerable recent development in the area of additive manufacturing ("3D-printing"), including the development of techniques for printing directly in metal alloys, using powder-bed laser fusion. Additive manufacturing in metal alloys has become routinely available to researchers and manufacturers, with commercial systems widely available. These techniques have increased the potential for precision medicine, including patient-specific components and surgical guides. For biomedical applications, additive manufacturing is driven by advanced imaging (CT and MRI), as the required geometric data is typically derived from 3D volume images. There is an opportunity for significant improvement in the fundamental processes by which images are converted into source data for component design. Similarly, 3D imaging can serve an important role in verifying and validating the shape and mechanical performance of fabricated components for biomedical applications. Objectives: In this program, we will explore the implementation of novel algorithms (including machine learning and deep learning) for image reconstruction and segmentation, with the goal of improving the efficiency and accuracy of biomedical 3D printing. We will also develop techniques for process monitoring, as part of next-generation quality management. The techniques that we develop will be applied to a range of research applications, including veterinary components, porous scaffolds, and "smart" implants. We believe that the development of an integrated "pipeline" for design and testing will significantly expand the impact of image-based 3D printing. Approach: Our research approach for this program is focused on three themes: (1) optimization of image acquisition and segmentation, using dual-energy imaging, iterative reconstruction, and model-based deep-learning; (2) Image-based finite-element design and optimization, validated with dynamic 3D non-destructive testing and embedded sensors; (3) image-based validation and process improvement, using quantitative 3D computed tomography and in-process optical imaging during laser powder-bed fusion. Impact: This program integrates several aspects of recently developed technology and new computing algorithms (such as deep learning using neural networks), creating a new platform of software tools for image-based design and fabrication in biomedical engineering. Anticipated advances from this program include new tools to extract geometry data from 3D image volumes, and new hardware-software techniques to verify FE models (i.e. mechanical properties, drug elution, and thermal transport). This project takes advantage of expertise in several areas of the natural sciences and engineering, including: imaging science, biomechanical engineering, advanced manufacturing, and software engineering. The proposed program will lead to novel software utilities and techniques with broad applications in research and development.