Semi-automated Atrial Segmentation Division Method for Regional and Quantitative Analysis

Download as PDF

Motivation
Standardised segmentation methods allow quantitative comparisons between different regions of the heart. Nevertheless, due to anatomic variability, technical implications related to angle, plane and landmarks selection, and lack of naming consensus for each segment, comparisons between patients' data and different imaging modalities are challenging. In the particular case of the ventricles, a standard nomenclature and segmentation has already been adopted and has made it possible to perform regional analysis of wall motion and myocardial perfusion.

However, there is no such standard division method defined for the atria until now. In addition, this also implies that the segmentation of the atria is done manually, which makes it dependent on the operator's judgment and skills.

Bild

Student Project
For this project, you will first conduct a literature review of methods in the fields of computer vision and graph theory that can be used to divide 3D images. Then you will look for segmentation approaches that have already been proposed in the clinics to divide the atria. You will then apply these methods to regionally label images of a patient's atria derived from magnetic resonance and electro- anatomical mapping systems. Finally, you will quantify regional distribution differences between electrophysiological parameters such as fibrotic tissue and low voltage areas.

Skills

  • Motivation and interest to work on clinically relevant issues
  • Experience in Python and Matlab is desirable
  • Basic knowledge of cardiac anatomy and physiology is an advantage.