Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Machine learning-based end-to-end mapping of cardiac metabolism
Hyperpolarized MRI allows studying cardiac metabolism. Quantification of cardiac metabolism based on advanced kinetic models is desired. As an alternative to conventional parameter fitting, machine learning-based end-to-end mapping can be used to predict both metabolic and perfusion parameters.
Keywords: Magnetic Resonance Imaging, Machine Learning, End-to-End Mapping, Kinetic Modeling
Hyperpolarized 13C MRI can be used to monitor
cardiac metabolism. Based on kinetic models
describing the metabolic exchange, conversion
rates can be estimated and thus, cardiac
metabolism quantified in order to detect myocardial infarction.
Advanced models combining perfusion and metabolism allow for modeling the required input function into the kinetic model. As an alternative to conventional parameter estimation through numerical fitting approaches, machine learning-based approaches can be used to predict the
desired parameters by means of model-based
training. Especially for the multi-parameter
combined kinetic model, an end-to-end mapping
of metabolic and perfusion parameters
might be a more reliable and computationally
less expensive choice compared to classical parameter fitting.
Hyperpolarized 13C MRI can be used to monitor cardiac metabolism. Based on kinetic models describing the metabolic exchange, conversion rates can be estimated and thus, cardiac metabolism quantified in order to detect myocardial infarction. Advanced models combining perfusion and metabolism allow for modeling the required input function into the kinetic model. As an alternative to conventional parameter estimation through numerical fitting approaches, machine learning-based approaches can be used to predict the desired parameters by means of model-based training. Especially for the multi-parameter combined kinetic model, an end-to-end mapping of metabolic and perfusion parameters might be a more reliable and computationally less expensive choice compared to classical parameter fitting.
The idea of this project is to use a machine learning-based approach to quantify rate constants describing cardiac metabolism.
In particular, the goal is to develop a model-based training for end-to-end mapping of first-order metabolic rate constants and perfusion indices.
Based on realistic synthetic data, a convolutional neural network (CNN) will be trained, tested and validated.
In addition, the trained CNN will be applied to in-vivo data.
The idea of this project is to use a machine learning-based approach to quantify rate constants describing cardiac metabolism. In particular, the goal is to develop a model-based training for end-to-end mapping of first-order metabolic rate constants and perfusion indices. Based on realistic synthetic data, a convolutional neural network (CNN) will be trained, tested and validated. In addition, the trained CNN will be applied to in-vivo data.
Supervisors: Julia Traechtler (traechtler@biomed.ee.ethz.ch);
Supervising Professor: Prof. Dr. Sebastian Kozerke (kozerke@biomed.ee.ethz.ch)
Supervisors: Julia Traechtler (traechtler@biomed.ee.ethz.ch); Supervising Professor: Prof. Dr. Sebastian Kozerke (kozerke@biomed.ee.ethz.ch)