Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Image-Guided Cardiac Strain Analysis Using Machine Learning Approaches
Application of machine learning techniques for cardiac strain mapping using cardiovascular magnetic resonance (CMR) imaging.
Keywords: Cardiac Mechanics, Computational Modeling, Finite Element Method, Magnetic resonance Imaging, MRI, Cardiac Imaging, Machine Learning
CMR imaging allows the non-invasive assessment of cardiac function. Analysis of cine CMR images can be used to evaluate global pump performance of the heart. However, in various cardiac diseases, local contractile dysfunction develops, hence requires regional contractile function to be diagnosed [1]. Using tagged CMR images [2], it is possible to quantify myocardial deformation at a regional level. Various approaches for post-processing of CMR cine and tagged images exist but present with limitations.
The objective of the present project is to explore machine learning (ML) techniques for mapping cardiac strain from CMR cine and tagged images. To validate these ML approaches, finite element models (FEM) of the heart will be generated, processed and compared to state-of-the-art finite element digital image correlation methods [3].
CMR imaging allows the non-invasive assessment of cardiac function. Analysis of cine CMR images can be used to evaluate global pump performance of the heart. However, in various cardiac diseases, local contractile dysfunction develops, hence requires regional contractile function to be diagnosed [1]. Using tagged CMR images [2], it is possible to quantify myocardial deformation at a regional level. Various approaches for post-processing of CMR cine and tagged images exist but present with limitations. The objective of the present project is to explore machine learning (ML) techniques for mapping cardiac strain from CMR cine and tagged images. To validate these ML approaches, finite element models (FEM) of the heart will be generated, processed and compared to state-of-the-art finite element digital image correlation methods [3].
Using FEM-based synthetic tagged images of the heart, it is possible to compare image correlation strain results with the ground truth obtained by mechanical analysis. Following ideas of machine learning to map mechanical stiffness of tissue [4], the aim of the present project is to implement, tune and validate the performance of an ML network for tissue strain estimation from synthetic tagged data.
The project involves:
- Working on an interdisciplinary field and gaining insight into new approaches for biomechanical modeling of the heart.
- Learning about synthetic tagged CMR image generation using FEM methods.
- Acquiring background knowledge about various ML techniques.
- Conceiving and testing of ML techniques for tissue strain mapping.
**References**
[1] T. P. Abraham, and R. A. Nishimura. “Myocardial Strain: Can We Finally Measure Contractility?” Journal of the American College of Cardiology 37, no. 3 (March 1, 2001): 731–34.
[2] E. A. Zerhouni, et al.. “Human Heart: Tagging with MR Imaging--a Method for Noninvasive Assessment of Myocardial Motion”. Radiology 169, no. 1 (October 1988): 59–63.
[3] M. Genet, et al.. “Finite Element Digital Image Correlation for Cardiac Strain Analysis from 3D Whole-Heart Tagging”. ISMRM 24rd Annual Meeting and Exhibition 2016, May 2016, Singapore, Singapore. 2016.
[4] M.C.Murphy, et al.. “Artificial Neural Networks for Stiffness Estimation in Magnetic Resonance Elastography”. Magnetic Resonance in Medicine, 2017.
Using FEM-based synthetic tagged images of the heart, it is possible to compare image correlation strain results with the ground truth obtained by mechanical analysis. Following ideas of machine learning to map mechanical stiffness of tissue [4], the aim of the present project is to implement, tune and validate the performance of an ML network for tissue strain estimation from synthetic tagged data.
The project involves:
- Working on an interdisciplinary field and gaining insight into new approaches for biomechanical modeling of the heart. - Learning about synthetic tagged CMR image generation using FEM methods. - Acquiring background knowledge about various ML techniques. - Conceiving and testing of ML techniques for tissue strain mapping.
**References**
[1] T. P. Abraham, and R. A. Nishimura. “Myocardial Strain: Can We Finally Measure Contractility?” Journal of the American College of Cardiology 37, no. 3 (March 1, 2001): 731–34.
[2] E. A. Zerhouni, et al.. “Human Heart: Tagging with MR Imaging--a Method for Noninvasive Assessment of Myocardial Motion”. Radiology 169, no. 1 (October 1988): 59–63.
[3] M. Genet, et al.. “Finite Element Digital Image Correlation for Cardiac Strain Analysis from 3D Whole-Heart Tagging”. ISMRM 24rd Annual Meeting and Exhibition 2016, May 2016, Singapore, Singapore. 2016.
[4] M.C.Murphy, et al.. “Artificial Neural Networks for Stiffness Estimation in Magnetic Resonance Elastography”. Magnetic Resonance in Medicine, 2017.
Supervisors:Ezgi Berberoglu (berberoglu@biomed.ee.ethz.ch), Javier Montoya (montoya@biomed.ee.ethz.ch)
Professors: Sebastian Kozerke (ETH Zurich), Martin Genet (Ecole Polytechique Paris)
Supervisors:Ezgi Berberoglu (berberoglu@biomed.ee.ethz.ch), Javier Montoya (montoya@biomed.ee.ethz.ch)
Professors: Sebastian Kozerke (ETH Zurich), Martin Genet (Ecole Polytechique Paris)