AI-supported computer-aided design assistant system for heart surgery (heartCAAS)

From an engineering perspective, the human heart can be understood as a muscle displacement pump. After a myocardial infarction, the infarcted region of the heart muscle may become non-contractile and dilate to a ventricular aneurysm (Figure 1 left). These changes in contractility and shape impair the ability of the heart muscle to pump blood. To restore the pumping function in these patients, surgical ventricular reconstruction (SVR) was developed, a surgery, where the scarred region of the heart muscle is removed to restore the normal shape, volume, and contraction of the heart (Figure 1 middle and right). For effective surgery, optimal treatment planning is essential.

• The goal of this project is to develop a multi-component, computer-aided assistance system to optimize the individual design of the heart as muscle displacement pump in SVR
• This includes a virtual surgery model, fluid and solid models to estimate the hemodynamics and tissue mechanics, and a neural network to predict the patient symptoms after surgery

The virtual surgery model serves to perform the surgery in silico, providing the option to test different treatment variations, as follows:

1. Estimate the zero-pressure state of the heart with finite element analysis
2. Virtually remove the scarred myocardial wall with the software Geomagic Freeform Plus as haptic interface
3. Virtual sew the myocardial wall by using position-based dynamics
4. Model the post-operative contraction via graph neural network trained on retrospective data

• To model the ventricular mechanics, cardiac computed tomography data is used to segment the heart anatomy and contraction
• The blood flow is modeled with computational fluid dynamics and the tissue mechanics with finite element analysis
• Major mechanical target parameters are energetic efficiency, ventricular washout and global longitudinal strain

• To estimate the surgical outcome in terms of improvement of patient symptoms measured in the New York Heart Association (NYHA) class, a neural network will be trained 

[1] Goubergrits L, Vellguth K, Obermeier L, Schlief A, Tautz L, Bruening J, Lamecker H, Szengel A, Nemchyna O, Knosalla C, Kuehne T, Solowjowa N. CT-Based Analysis of Left Ventricular Hemodynamics Using Statistical Shape Modeling and Computational Fluid Dynamics. Front Cardiovasc Med. 2022; 9:901902. doi: 10.3389/fcvm.2022.901902.

[2] Solowjowa N, Nemchyna O, Hrytsyna Y, Meyer A, Hennig F, Falk V, Knosalla C. Surgical Restoration of Antero-Apical Left Ventricular Aneurysms: Cardiac Computed Tomography for Therapy Planning. Frontiers in Cardiovascular Medicine. 2022; 9.763073. doi: 10.3389/fcvm.2022.763073.

[3] Nemchyna O, Solowjowa N, Dandel M, Hrytsyna Y, Stein J, Knierim J, Schoenrath F, Hennig F, Falk V, Knosalla C. Predictive Value of Two-Dimensional Speckle-Tracking Echocardiography in Patients Undergoing Surgical Ventricular Restoration. Front Cardiovasc Med. 2022; 9:824467. doi: 10.3389/fcvm.2022.824467.

[4] Obermeier L, Wiegand M, Kuehne T, Falk V, Knosalla C, Solowjowa N, Goubergrits L, Vellguth K. Impact of surgical ventricular restoration on intracardiac hemodynamics: An in-silico study using CCT data. Comput Biol Med. 2025 Jun;192(Pt B):110227. doi: 10.1016/j.compbiomed.2025.110227.