Project Details
Projekt
AI-supported computer-aided design assistant system for heart surgery (heartCAAS)
Subject Area
Mechanics
Term
since 2025
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 564319982
From an engineering perspective, the human heart is a dynamic muscle displacement pump. Dysfunctions of this pump caused by structural heart diseases can have life-threatening consequences with poor clinical prognosis for patients. The heart diseases often cause remodelling and impaired contraction of the left ventricle. A common treatment option for the heart is surgical ventricular restoration. In the open-heart surgery, the dilated region of the left ventricle is identified and removed to restore the shape as well as pumping function of the heart. The treatment procedure is a multifactorial patient-individual design optimization task, which is affected by size, shape, and contractility of the diseased heart as non-optimal pump. In contrast to pure engineering systems, the challenge of design optimization of the heart is the inability of an incremental, iterative optimization at all. However, optimal treatment planning is essential for the patient. To date, only rudimentary planning tools exist for surgical ventricular restoration. No advanced and physics-based models exist. In this project, we therefore aim at combining our clinical and engineering expert knowledge to develop a computer-aided design assistant system that allows advanced treatment planning for heart restoration. Such an assistant system requires various model components, including reduced, physics-based, and data-driven models. Central component will be a virtual surgery model, which mimics the clinical procedure of the surgery step by step. This includes identification and removal of scarred heart wall region, suture modelling to close the opened heart, and estimating the post-operative contractile motion. To target clinically feasible computation times, these steps of the surgery will be modelled using state-of-the-art methods (graph neural networks and position-based dynamics). In addition to shape optimization according to clinical guidelines, we strive to provide design recommendations by analyzing heart mechanics. Therefore, previously established workflows are used to compute intraventricular hemodynamics and myocardial tissue mechanics using computational fluid dynamics and finite element method simulations, respectively. Finally, a supervised neural network based on computed tomography and treatment data will be trained aiming to predict the mid-term surgical outcome by means of improved patient symptoms classified by New-York Heart Association class. At the end of this project a workflow for virtual treatment planning combining data-driven methods and conventional analysis of biomechanics to go beyond the state of the art in system design of the heart surgery will be established. Thus, we aim to establish a benchmark process allowing treatment optimization.
DFG Programme
Priority Programmes
Co-Investigator
Dr. Natalia Solowjowa
