Computer simulation of the human heart is a promising tool to better understand the heart's inner workings. Every patient-specific simulation involves many steps that form a pipeline of complex data processing tasks. In particular, several preprocessing stages are needed to extract geometry and anatomical information from medical images before any simulation can take place. When a simulation is done, several manual stages are typically also needed to extract and visualize the most important result in a form that is understandable by for instance clinicians. Many of these stages are very time-consuming and labour-intensive, such that the inefficiency and inconvenience seriously limit the applicability of patient-specific heart simulations in real life.
The main idea of this master project is to apply parallel programming, performance optimisation and proper use of high-performance computing systems, for creating more efficient and streamlined data processing pipelines needed for patient-specific heart simulations. The ambition is to help researchers and clinicians setting up, executing and analyzing thousands of such simulations.
Two specific topics will be investigated by the master student:
- The student will implement parallel algorithms for mapping anatomical coordinates suitable for different cardiac geometries. There exist sequential implementations of such coordinate mapping procedures. The student will re-program the algorithms while parallelizing the involved computations. Additionally, the speed will be optimised and assessed through performance and scalability studies.
- The student will enable a post-processing procedure that automatically extracts the most important information (such as in certain regions of the heart muscle wall) from the 3D simulation results, and visualizes it in a suitable manner understandable by both researchers and medical practitioners. Parallel processing and performance optimizations will also be applied in this topic. In addition, a friendly graphical user interface will be developed.
As a starting point of the project, the student will be introduced to an existing, "low-performance" data processing pipeline. Under the guidance of the supervisors, the student will improve the two stages mentioned above. When a prototype of the improved pipeline is ready, it will be tested in realistic settings, such that user feedback can guide the student's work on further improvements. The final result is expected to be integrated into existing, parallel simulation tools, and demonstrated on large cohorts of realistic heart geometries.
Required background knowledge/skills of the student:
- Basic knowledge in C programming
- Interest in learning basic skills in parallel programming and code optimization
- Interest in biomedical computing
- Courage in learning new knowledge in the intersection of high-performance computing and cardiac modeling/simulation
References:
- Pankewitz, L., et al. (2024). A universal biventricular coordinate system incorporating valve annuli: Validation in congenital heart disease. Medical image analysis, 93, 103091.
- Bayer, J., et al. (2018). Universal ventricular coordinates: A generic framework for describing position within the heart and transferring data. Medical image analysis, 45, 83-93.
- Unleash the power of ParaView.