AutoStent - An autonomous design assistant for aneurysm repair

• Applicant: Professor Dr.-Ing. Alexander Popp
• Subject Area: Mechanics ; Applied Mechanics, Statics and Dynamics
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 546792806

Wider research context: In the past decades, computer-aided and numerical methods have played a larger and larger role in engineering design. Starting from CAD software that simply replicated drawing boards, more and more features were added that assisted the design engineer. These include, in particular, methods of shape and topology optimization that not only implement the ideas of the design engineer, but also make suggestions for an improved design. Within these methods, what remains, however, is for the design engineer to select the initial design and the parameterization. This means that all creative tasks are still centered around the design engineer, meaning the human. Objectives: The proposed project, AutoStent, includes artificial intelligence in the design process by devising a creative rational agent that can aid the design engineer also in making design decisions. Methods: The rational agent is based on a combination of two main components. The first component is a design generator that follows a “sketchpad” approach and guarantees vast design freedom to the rational agent. Here, the rational agent can freely draw spline representations. The second component is a deep reinforcement learning (DRL) algorithm that allows for individualized designs. DRL is a machine learning approach (next to, e.g., supervised, unsupervised, or active learning) that is based on a trial-and-error interaction of an agent with an environment. While not necessarily superior to classic optimization algorithms (such as gradient-based approaches) for one single optimization problem, we expect DRL techniques to thrive when recurrent design tasks arise. The resulting rational agent can be seen as a first step to fully autonomous design. Innovation: The novel design method will then be applied to stent grafts in endovascular aneurysm repair (EVAR). Even though there are more than 200,000 cases of abdominal aortic aneurysms (AAA) in Germany every year, the stent graft designs available to patients are still mostly off-the-shelf. We expect a large impact on the care for AAA patients if customized, patient-specific stents become available. Stent grafts are an ideal development and benchmarking use case, since they involve complex, possibly competing and patient-specific design requirements that challenge human intuition, but at the same time they still exhibit manageable geometrical complexity due to the potential dimensional reduction of the metal stent wires to one-dimensional entities.

• DFG Programme: Research Grants
• International Connection: Austria
• Partner Organisation: Fonds zur Förderung der wissenschaftlichen Forschung (FWF)
• Cooperation Partner: Professorin Dr.-Ing. Stefanie Elgeti